saving b/c im scared

engine/core/src/console.zig

@@ -9,14 +9,14 @@ pub const Console = struct {
     pub var NeonObjectTable: core.EngineObjectVTable = core.EngineObjectVTable.from(@This());
 
     allocator: std.mem.Allocator,
-    stringArena: std.heap.ArenaAllocator,
+    arena: std.heap.ArenaAllocator,
 
     commandMap: std.StringHashMapUnmanaged(ConsoleCommand) = .{},
 
     pub fn create(allocator: std.mem.Allocator) !*@This() {
         const self = try allocator.create(@This());
         self.* = .{
-            .stringArena = std.heap.ArenaAllocator.init(allocator),
+            .arena = std.heap.ArenaAllocator.init(allocator),
             .allocator = allocator,
         };
 
@@ -27,7 +27,7 @@ pub const Console = struct {
 
     pub fn addConsoleCommand(self: *@This(), funcName: []const u8, func: ConsoleFunc) !void {
         try self.commandMap.put(self.allocator, funcName, .{
-            .command = try self.stringArena.allocator().dupe(u8, funcName),
+            .command = try self.arena.allocator().dupe(u8, funcName),
             .func = func,
         });
     }
@@ -55,7 +55,7 @@ pub const Console = struct {
     }
 
     pub fn destroy(self: *@This()) void {
-        self.stringArena.deinit();
+        self.arena.deinit();
         self.commandMap.deinit(self.allocator);
         self.allocator.destroy(self);
     }

engine/core/tests/tests.zig

@@ -63,6 +63,21 @@ fn test_consoleCommands() !void {
     try core.console.addCommand("echo", ConsoleCommands.echo);
     try core.console.addCommand("foo", ConsoleCommands.foo);
 
+    const T = struct {
+        var ScreenResolution: u32 = undefined;
+
+        pub fn F(args: []const u8) void {
+            core.engine_log("Setting screen resolution to = {s}", .{args});
+            ScreenResolution = std.fmt.parseInt(u32, args, 10) catch return;
+        }
+    };
+
+    T.ScreenResolution = core.configVar(u32, "Some.bullshit.ScreenResolution", 420);
+    try core.assert(T.ScreenResolution == 420);
+    try core.console.addCommand("ScreenResolution", T.F);
+    core.console.evaluate("ScreenResolution 1");
+    try core.assert(T.ScreenResolution == 1);
+
     core.console.evaluate("echo lmfao");
     core.console.evaluate("foo lmfao");
 }

engine/imgui/src/sgpu/backend_impl.zig

@@ -5,6 +5,8 @@ pub const Impl = struct {
     drawData: [*c]ig.DrawData = undefined,
     rendererDebug: bool = true,
 
+    dtAverage: f64 = 0.0,
+
     pub var NeonObjectTable: core.EngineObjectVTable = core.EngineObjectVTable.from(@This());
 
     pub const Settings = struct {
@@ -55,7 +57,7 @@ pub const Impl = struct {
             //const renderpass = cmd.beginGPURenderPass(color_target_infos: [*c]const GPUColorTargetInfo, num_color_targets: u32, depth_stencil_target_info: [*c]const GPUDepthStencilTargetInfo);
 
             var targetInfo: gpu.GPUColorTargetInfo = std.mem.zeroes(gpu.GPUColorTargetInfo);
-            targetInfo.texture = rend.context().swapchainTexture.?;
+            targetInfo.texture = rend.context().state.swapchainTexture.?;
             targetInfo.clear_color = .{ .r = 0.0, .g = 0.0, .b = 0.0, .a = 0.0 };
             targetInfo.load_op = .loadopLoad;
             targetInfo.store_op = .storeopStore;
@@ -71,14 +73,19 @@ pub const Impl = struct {
         }
     }
 
-    pub fn preTick(self: *@This(), _: f64) core.EngineDataEventError!void {
+    pub fn preTick(self: *@This(), dt: f64) core.EngineDataEventError!void {
         c.Imgui_SDL3_NewFrame();
         c.igNewFrame();
 
+        core.rollingAverage(&self.dtAverage, dt, 50);
+
         if (ig.begin("renderer debug", &self.rendererDebug, .{})) {
             _ = ig.sliderFloat("directional light yaw", &rend.context().directionalLightYaw, 0, 360, null, .{});
             _ = ig.sliderFloat("ortho near", &rend.context().shadowOrthoNear, 0, 200, null, .{});
             _ = ig.sliderFloat("ortho far", &rend.context().shadowOrthoFar, 0, 6000, null, .{});
+            if (self.dtAverage > 0.000001) {
+                ig.textFmt("frameTime: {d:.2}ms ({d:.2}fps)", .{ self.dtAverage * 1000, @as(u32, @intFromFloat(1 / self.dtAverage)) }) catch return error.UnknownStatePanic;
+            }
             ig.end();
         }
     }

engine/rend/shaders/lit_mesh.frag.hlsl

@@ -1,9 +1,14 @@
-Texture2D<float4> Texture : register(t0, space2);
+Texture2D<float4> Texture0 : register(t0, space2);
-SamplerState Sampler : register(s0, space2);
+SamplerState Sampler0 : register(s0, space2);
 
-Texture2D<float> DirectionalShadowDepthMap : register(t1, space2);
+Texture2D<float4> Texture1 : register(t1, space2);
-SamplerState DirectionalShadowSampler : register(s1, space2);
+SamplerState Sampler1 : register(s1, space2);
 
+Texture2D<float4> Texture2 : register(t2, space2);
+SamplerState Sampler2 : register(s2, space2);
+
+Texture2D<float> DirectionalShadowDepthMap : register(t3, space2);
+SamplerState DirectionalShadowSampler : register(s3, space2);
 
 cbuffer Uniforms : register(b0, space3)
 {
@@ -15,6 +20,16 @@ cbuffer Uniforms : register(b0, space3)
     float time;
 };
 
+struct Scene
+{
+    float4x4 Model;
+
+    uint textureMode;    // 0 = regular triple, 
+                        // 0x10 = video yuv,
+};
+
+StructuredBuffer<Scene> scene: register(t0, space0);
+
 // from learnopengl.com
 float3 BlinnPhong(float3 normal, float3 fragPos, float3 lightPos, float3 lightColor)
 {
@@ -157,18 +172,68 @@ float3 DirectionalLight(float3 normal, float3 fragPos, float4  DirectionalShadow
 }
 
 
+
 float4 main(
     float4 ScreenPosition: SV_POSITION,
 
     float2 UV : TEXCOORD0,
     float3 WorldPos: TEXCOORD1,
     float3 Normal: TEXCOORD2,
-    float4 DirectionalShadowFragPos: TEXCOORD3
+    float4 DirectionalShadowFragPos: TEXCOORD3,
+    uint Instance: TEXCOORD4
 ) : SV_Target0
 {
-    float4 s = Texture.Sample(Sampler, UV);
+
+
+    float4 s; 
     float alpha = s.w;
 
+    int texMode = scene[Instance].textureMode;
+    if (texMode == 0x10) // YUV
+    {
+        float3 offset = float3(-0.0625, -0.5, -0.5);
+        float3 Rcoeff = float3(1.164,  0.000,  1.793);
+        float3 Gcoeff = float3(1.164, -0.213, -0.533);
+        float3 Bcoeff = float3(1.164,  2.112,  0.000);
+
+        float3 yuv, rgb;
+
+        yuv.x =Texture0.Sample(Sampler0, UV).r;
+        yuv.y =Texture1.Sample(Sampler1, UV).r;
+        yuv.z =Texture2.Sample(Sampler2, UV).r;
+
+        yuv += offset;
+        rgb.r = dot(yuv, Rcoeff);
+        rgb.g = dot(yuv, Gcoeff);
+        rgb.b = dot(yuv, Bcoeff);
+
+        // "	gl_FragColor = vec4(rgb, 1.0);\n"
+        // "uniform sampler2D samp0;\n"
+        // "uniform sampler2D samp1;\n"
+        // "uniform sampler2D samp2;\n"
+        // "const vec3 offset = vec3(-0.0625, -0.5, -0.5);\n"
+        // "const vec3 Rcoeff = vec3(1.164,  0.000,  1.793);\n"
+        // "const vec3 Gcoeff = vec3(1.164, -0.213, -0.533);\n"
+        // "const vec3 Bcoeff = vec3(1.164,  2.112,  0.000);\n"
+        // "void main() {\n"
+        // "	vec2 tcoord;\n"
+        // "	vec3 yuv, rgb;\n"
+        // "	tcoord = gl_TexCoord[0].xy;\n"
+        // "	yuv.x = texture2D(samp0, tcoord).r;\n"
+        // "	yuv.y = texture2D(samp1, tcoord).r;\n"
+        // "	yuv.z = texture2D(samp2, tcoord).r;\n"
+        // "	yuv += offset;\n"
+        // "	rgb.r = dot(yuv, Rcoeff);\n"
+        // "	rgb.g = dot(yuv, Gcoeff);\n"
+        // "	rgb.b = dot(yuv, Bcoeff);\n"
+        // "	gl_FragColor = vec4(rgb, 1.0);\n"
+        // "}\n";
+    }
+    else if(texMode == 0)
+    {
+        s = Texture0.Sample(Sampler0, UV);
+    }
+
     if(alpha < 0.01)
     {
         discard;

engine/rend/shaders/lit_mesh.frag.json

@@ -6,7 +6,7 @@
         }
     ],
     "types" : {
-        "_14" : {
+        "_15" : {
             "name" : "type.Uniforms",
             "members" : [
                 {
@@ -40,6 +40,40 @@
                     "offset" : 72
                 }
             ]
+        },
+        "_18" : {
+            "name" : "Scene",
+            "members" : [
+                {
+                    "name" : "Model",
+                    "type" : "mat4",
+                    "offset" : 0,
+                    "matrix_stride" : 16,
+                    "row_major" : true
+                },
+                {
+                    "name" : "textureMode",
+                    "type" : "uint",
+                    "offset" : 64
+                }
+            ]
+        },
+        "_17" : {
+            "name" : "type.StructuredBuffer.Scene",
+            "members" : [
+                {
+                    "name" : "_m0",
+                    "type" : "_18",
+                    "array" : [
+                        0
+                    ],
+                    "array_size_is_literal" : [
+                        true
+                    ],
+                    "offset" : 0,
+                    "array_stride" : 80
+                }
+            ]
         }
     },
     "inputs" : [
@@ -62,6 +96,11 @@
             "type" : "vec4",
             "name" : "in.var.TEXCOORD3",
             "location" : 3
+        },
+        {
+            "type" : "uint",
+            "name" : "in.var.TEXCOORD4",
+            "location" : 4
         }
     ],
     "outputs" : [
@@ -74,7 +113,7 @@
     "separate_images" : [
         {
             "type" : "texture2D",
-            "name" : "Texture",
+            "name" : "Texture0",
             "set" : 2,
             "binding" : 0
         },
@@ -82,13 +121,13 @@
             "type" : "texture2D",
             "name" : "DirectionalShadowDepthMap",
             "set" : 2,
-            "binding" : 1
+            "binding" : 3
         }
     ],
     "separate_samplers" : [
         {
             "type" : "sampler",
-            "name" : "Sampler",
+            "name" : "Sampler0",
             "set" : 2,
             "binding" : 0
         },
@@ -96,12 +135,22 @@
             "type" : "sampler",
             "name" : "DirectionalShadowSampler",
             "set" : 2,
-            "binding" : 1
+            "binding" : 3
+        }
+    ],
+    "ssbos" : [
+        {
+            "type" : "_17",
+            "name" : "scene",
+            "readonly" : true,
+            "block_size" : 0,
+            "set" : 0,
+            "binding" : 0
         }
     ],
     "ubos" : [
         {
-            "type" : "_14",
+            "type" : "_15",
             "name" : "type.Uniforms",
             "block_size" : 76,
             "set" : 3,

engine/rend/shaders/meshes.vert.hlsl

@@ -18,6 +18,7 @@ struct Output
     float3 WorldPos : TEXCOORD1;
     float3 Normal: TEXCOORD2;
     float4 DirectionalShadowFragPos: TEXCOORD3;
+    uint Instance: TEXCOORD4;
 
     float4 Position : SV_Position;
 };
@@ -25,6 +26,9 @@ struct Output
 struct Scene
 {
     float4x4 Model;
+
+    uint textureMode; // 0 = regular triple, 
+                     // 0x10 = video yuv,
 };
 
 StructuredBuffer<Scene> scene: register(t0, space0);
@@ -51,6 +55,7 @@ Output main(Input input)
     output.WorldPos = WorldPos.xyz;
     output.Normal = input.Normal;
     output.DirectionalShadowFragPos = mul(ShadowMapProjection, WorldPos);
+    output.Instance = input.Instance;
 
     return output;
 }

engine/rend/shaders/meshes.vert.json

@@ -6,7 +6,7 @@
         }
     ],
     "types" : {
-        "_12" : {
+        "_13" : {
             "name" : "Scene",
             "members" : [
                 {
@@ -15,15 +15,20 @@
                     "offset" : 0,
                     "matrix_stride" : 16,
                     "row_major" : true
+                },
+                {
+                    "name" : "textureMode",
+                    "type" : "uint",
+                    "offset" : 64
                 }
             ]
         },
-        "_11" : {
+        "_12" : {
             "name" : "type.StructuredBuffer.Scene",
             "members" : [
                 {
                     "name" : "_m0",
-                    "type" : "_12",
+                    "type" : "_13",
                     "array" : [
                         0
                     ],
@@ -31,11 +36,11 @@
                         true
                     ],
                     "offset" : 0,
-                    "array_stride" : 64
+                    "array_stride" : 80
                 }
             ]
         },
-        "_14" : {
+        "_15" : {
             "name" : "type.Uniforms",
             "members" : [
                 {
@@ -97,11 +102,16 @@
             "type" : "vec4",
             "name" : "out.var.TEXCOORD3",
             "location" : 3
+        },
+        {
+            "type" : "uint",
+            "name" : "out.var.TEXCOORD4",
+            "location" : 4
         }
     ],
     "ssbos" : [
         {
-            "type" : "_11",
+            "type" : "_12",
             "name" : "scene",
             "readonly" : true,
             "block_size" : 0,
@@ -111,7 +121,7 @@
     ],
     "ubos" : [
         {
-            "type" : "_14",
+            "type" : "_15",
             "name" : "type.Uniforms",
             "block_size" : 132,
             "set" : 1,

engine/rend/src/meshes/MeshComponent.zig

@@ -9,6 +9,13 @@ texture: ?*Texture = null,
 
 entity: core.Entity = undefined,
 
+textureMode: u32 = 0,
+// DEBUG for-fun function, completely override what happens when the renderer tries to render this mesh.
+cmrf: ?*rend.renderer.CustomMeshRenderFunc = null,
+
+pub const textureMode_default = 0;
+pub const textureMode_videoYUV = 0x10;
+
 pub var BaseContainer: *MeshSet = undefined;
 pub const ComponentName = "Mesh";
 

engine/rend/src/sgpu/SkyboxSystem.zig

@@ -85,7 +85,7 @@ pub fn destroy(self: *@This()) void {
 
 pub fn render(self: *@This(), cmd: *gpu.GPUCommandBuffer) void {
     const targetInfo: gpu.GPUColorTargetInfo = std.mem.zeroInit(gpu.GPUColorTargetInfo, .{
-        .texture = rend.context().swapchainTexture.?,
+        .texture = rend.context().state.swapchainTexture.?,
         .clear_color = self.skyboxColor,
         .load_op = .loadopClear,
         .store_op = .storeopStore,
@@ -101,10 +101,10 @@ pub fn render(self: *@This(), cmd: *gpu.GPUCommandBuffer) void {
         }
     }
 
+    const renderpass = cmd.beginGPURenderPass(&targetInfo, 1, null);
     if (self.skyboxTexture) |skyboxTexture| {
         if (self.skyboxMesh) |skyboxMesh| {
             const ctx = rend.context();
-            const renderpass = cmd.beginGPURenderPass(&targetInfo, 1, null);
 
             self.uploadSkyboxUniforms(cmd);
 
@@ -115,9 +115,9 @@ pub fn render(self: *@This(), cmd: *gpu.GPUCommandBuffer) void {
             renderpass.bindGPUGraphicsPipeline(self.skyboxPipeline);
             renderpass.bindGPUFragmentSamplers(0, &.{ .texture = skyboxTexture.texture, .sampler = self.sampler }, 1);
             renderpass.drawGPUIndexedPrimitives(skyboxMesh.index.size, 1, skyboxMesh.index.start, @intCast(skyboxMesh.vertex.start), 0);
-            renderpass.endGPURenderPass();
         }
     }
+    renderpass.endGPURenderPass();
 }
 
 pub fn uploadSkyboxUniforms(self: *@This(), cmd: *gpu.GPUCommandBuffer) void {

engine/rend/src/sgpu/mesh-pool.zig

@@ -69,14 +69,14 @@ pub const MeshPool = struct {
             switch (u) {
                 .new => |new| {
                     var name = new.name;
-                    core.engine_log("uploading mesh => {s}", .{name.utf8()});
+                    // core.engine_log("uploading mesh => {s}", .{name.utf8()});
 
                     const indexSpan = try self.addTransfer(copyPass, self.indexBuffer, u32, &self.indexSpans, new.indices);
                     const vertexSpan = try self.addTransfer(copyPass, self.vertexBuffer, meshes.MeshVertex, &self.vertexSpans, new.vertices);
 
-                    core.graphics_log("uploading {d} vertices and {d} indices", .{ vertexSpan.size, indexSpan.size });
+                    // core.graphics_log("uploading {d} vertices and {d} indices", .{ vertexSpan.size, indexSpan.size });
 
-                    core.engine_log("install mesh by name {d}", .{name.handle()});
+                    // core.engine_log("install mesh by name {d}", .{name.handle()});
                     try self.installedMeshes.put(self.allocator, name.handle(), .{
                         .vertex = vertexSpan,
                         .index = indexSpan,
@@ -131,7 +131,7 @@ pub const MeshPool = struct {
             false,
         );
 
-        core.engine_log("uploading {s} buffer span {d}[{d}] ({d} bytes)", .{ @typeName(T), newSpan.start, newSpan.size, newSpan.size * @sizeOf(T) });
+        // core.engine_log("uploading {s} buffer span {d}[{d}] ({d} bytes)", .{ @typeName(T), newSpan.start, newSpan.size, newSpan.size * @sizeOf(T) });
 
         return newSpan;
     }

engine/rend/src/sgpu/renderer.zig

@@ -63,10 +63,14 @@ pub const Renderer = struct {
     shadowDepthDebugOutput: *gpu.GPUTexture = undefined,
     shadowCastingPipeline: *gpu.GPUGraphicsPipeline = undefined,
 
+    shadowDepthTextureSlot: u32 = 3,
+
     skyboxSystem: *SkyboxSystem = undefined,
 
-    // transients DO NOT TOUCH
+    // transients DO NOT TOUCH NORMALLY, just a way to let me make the renderer more modular and flexible for exploring
-    swapchainTexture: ?*gpu.GPUTexture = undefined,
+    // only valid when rendering
+    state: RendererState = .{},
+    // swapchainTexture: ?*gpu.GPUTexture = undefined,
 
     pub var NeonObjectTable: core.EngineObjectVTable = core.EngineObjectVTable.from(@This());
 
@@ -112,6 +116,7 @@ pub const Renderer = struct {
 
         try self.createSamplers();
 
+        // todo... make these embedded
         try assets.load(
             assets.MakeImportRefOptions(
                 "Texture",
@@ -133,6 +138,16 @@ pub const Renderer = struct {
 
         try self.createDirectionalShadowsPipeline();
         self.skyboxSystem = try self.createRendererEngineObject(SkyboxSystem);
+
+        _ = try core.fs().installFileBytesMount("embedded:plane.obj", @constCast(&plane_obj), true);
+
+        try assets.load(
+            assets.MakeImportRefOptions(
+                "Mesh",
+                "m_plane",
+                .{ .path = "embedded:plane.obj" },
+            ),
+        );
     }
 
     pub fn createDirectionalShadowsPipeline(self: *@This()) !void {
@@ -366,7 +381,7 @@ pub const Renderer = struct {
 
             // iterate over MeshComponents
             for (0..uploadCount) |i| {
-                // const object = &container.dense.items[i].value;
+                const object = &container.dense.items[i].value;
                 const objectId = container.dense.items[i].sparseIndex;
                 var transform = core.zm.identity();
 
@@ -374,6 +389,7 @@ pub const Renderer = struct {
                     transform = repr.transform;
                 }
                 uploadMapped[i].Model = @bitCast(transform);
+                uploadMapped[i].textureMode = object.textureMode;
             }
         }
 
@@ -506,7 +522,7 @@ pub const Renderer = struct {
     pub fn tick(self: *@This(), dt: f64) void {
         self.totalTime += dt;
 
-        const cmd = self.device.acquireGPUCommandBuffer();
+        self.state.cmd = self.device.acquireGPUCommandBuffer();
         self.frameUploads();
 
         // mesh pre-iteration
@@ -526,10 +542,10 @@ pub const Renderer = struct {
         }
 
         for (self.preDraws.items) |interface| {
-            interface.func(interface.ptr, cmd);
+            interface.func(interface.ptr, self.state.cmd.?);
         }
 
-        self.draw(cmd);
+        self.draw();
     }
 
     // orthographic view projection from the sun towards the center of the thing
@@ -584,9 +600,11 @@ pub const Renderer = struct {
         renderpass.endGPURenderPass();
     }
 
-    pub fn draw(self: *@This(), cmd: *gpu.GPUCommandBuffer) void {
+    pub fn draw(self: *@This()) void {
-        if (cmd.waitAndAcquireGPUSwapchainTexture(self.window, @ptrCast(&self.swapchainTexture), null, null)) {
+        const cmd = self.state.cmd.?;
-            if (self.swapchainTexture == null) {
+
+        if (cmd.waitAndAcquireGPUSwapchainTexture(self.window, @ptrCast(&self.state.swapchainTexture), null, null)) {
+            if (self.state.swapchainTexture == null) {
                 _ = cmd.submitGPUCommandBuffer();
                 return;
             }
@@ -602,7 +620,7 @@ pub const Renderer = struct {
 
             // can cache this
             const targetInfo: gpu.GPUColorTargetInfo = std.mem.zeroInit(gpu.GPUColorTargetInfo, .{
-                .texture = self.swapchainTexture.?,
+                .texture = self.state.swapchainTexture.?,
                 .load_op = .loadopLoad,
                 .store_op = .storeopStore,
             });
@@ -617,7 +635,8 @@ pub const Renderer = struct {
                 .clear_stencil = 0,
             });
 
-            const renderpass = cmd.beginGPURenderPass(&targetInfo, 1, &depthTarget);
+            self.state.pass = cmd.beginGPURenderPass(&targetInfo, 1, &depthTarget);
+            const renderpass = self.state.pass.?;
             //renderpass.bindGPUGraphicsPipeline(self.testPipeline);
 
             renderpass.bindGPUVertexStorageBuffers(0, &self.ssboScene, 1);
@@ -626,7 +645,7 @@ pub const Renderer = struct {
 
             // render the meshes
             renderpass.bindGPUGraphicsPipeline(self.meshPipe);
-            renderpass.bindGPUFragmentSamplers(1, &.{ .texture = self.shadowDepthTexture, .sampler = self.blockySampler }, 1);
+            renderpass.bindGPUFragmentSamplers(self.shadowDepthTextureSlot, &.{ .texture = self.shadowDepthTexture, .sampler = self.blockySampler }, 1);
             // todo move into materials system
 
             renderpass.setGPUScissor(&self.scissor);
@@ -696,7 +715,7 @@ const assets = @import("assets");
 const core = @import("core");
 const platform = @import("platform");
 const sdl3 = @import("sdl3");
-const gpu = sdl3.gpu;
+pub const gpu = sdl3.gpu;
 
 const meshes_vert = @import("meshes.vert");
 const lit_mesh_frag = @import("lit_mesh.frag");
@@ -766,4 +785,14 @@ pub fn getTexture(name: *core.Name) ?*rend.Texture {
     return gRenderer.textureList.map.get(name.handle());
 }
 
+pub const RendererState = struct {
+    copyPass: ?*gpu.GPURenderPass = null,
+    pass: ?*gpu.GPURenderPass = null,
+    cmd: ?*gpu.GPUCommandBuffer = null,
+    swapchainTexture: ?*gpu.GPUTexture = null,
+};
+
+pub const CustomMeshRenderFunc = *const fn (mesh: *rend.MeshComponent, renderState: *RendererState) void;
 pub const GPUTextureType = gpu.GPUTexture;
+
+const skybox_mesh_obj align(8) = @embedFile("embedded/plane.obj").*;

extras/videoplayer/build.zig

@@ -0,0 +1,22 @@
+const std = @import("std");
+
+pub fn build(b: *std.Build) void {
+    const target = b.standardTargetOptions(.{});
+    const optimize = b.standardOptimizeOption(.{});
+
+    const mod = b.addModule("videoplayer", .{
+        .target = target,
+        .optimize = optimize,
+        .root_source_file = b.path("src/videoplayer.zig"),
+    });
+
+    {
+        const dep = b.dependency("Backlog", .{ .target = target, .optimize = optimize });
+        mod.addImport("Backlog", dep.module("Backlog"));
+    }
+
+    {
+        const dep = b.dependency("theorafile", .{ .target = target, .optimize = optimize });
+        mod.addImport("theorafile", dep.module("theorafile"));
+    }
+}

extras/videoplayer/build.zig.zon

@@ -0,0 +1,12 @@
+.{
+    .name = .gameExtras,
+    .version = "0.0.0",
+    .dependencies = .{
+        .Backlog = .{ .path = "../../" },
+        .theorafile = .{ .path = "../../lib/theoratest" },
+    },
+    .paths = .{
+        "",
+    },
+    .fingerprint = 0x993cf5f438b86d4a,
+}

extras/videoplayer/readme.md

@@ -0,0 +1,5 @@
+# extras
+
+These are sample implementations of various common functionality that's meant to be in your game not really part of the engine.
+
+To get access to these extras

extras/videoplayer/src/videoplayer.zig

@@ -0,0 +1,247 @@
+// video player implemented for sgpu,
+// uses theora/ogg playback from file.
+
+allocator: std.mem.Allocator,
+
+videoFile: OggTheora_File = undefined,
+width: c_int = undefined,
+height: c_int = undefined,
+uvWidth: c_int = undefined,
+uvHeight: c_int = undefined,
+fps: f64 = undefined,
+fmt: c.th_pixel_fmt = undefined,
+
+playbackStarted: bool = false,
+
+currentFrame: c_int = 0,
+currentTime: f64 = 0.0,
+
+showDebug: bool = true,
+
+device: *gpu.GPUDevice = undefined,
+// TH_PF_420,
+// /**Currently reserved.*/
+// TH_PF_RSVD,
+// /**Chroma decimation by 2 in the X direction (4:2:2).
+//    The Cb and Cr chroma planes are half the width of the luma plane, but full
+//     height.*/
+// TH_PF_422,
+// /**No chroma decimation (4:4:4).
+//    The Cb and Cr chroma planes are full width and full height.*/
+// TH_PF_444,
+// /**The total number of currently defined pixel formats.*/
+// TH_PF_NFORMATS
+
+yTexture: *gpu.GPUTexture = undefined,
+uTexture: *gpu.GPUTexture = undefined,
+vTexture: *gpu.GPUTexture = undefined,
+
+gpuTransferBuffer: *gpu.GPUTransferBuffer = undefined,
+
+fn log(comptime fmt: []const u8, args: anytype) void {
+    core.engine_log("[Videoplayer] " ++ fmt, args);
+}
+pub fn create(allocator: std.mem.Allocator) !*@This() {
+    const self = try allocator.create(@This());
+    self.* = .{
+        .allocator = allocator,
+    };
+
+    log("initialized", .{});
+
+    self.device = rend.context().device;
+
+    return self;
+}
+
+pub fn startPlayback(self: *@This(), filePath: [*c]const u8) !void {
+    if (c.tf_fopen(filePath, &self.videoFile) < 0) {
+        return error.UnableToOpenTestFile;
+    }
+
+    c.tf_videoinfo(&self.videoFile, &self.width, &self.height, &self.fps, &self.fmt);
+    log("playing video {s}", .{filePath});
+    log("dimensions : {d}x{d} px {d}fps format={d}", .{ self.width, self.height, self.fps, self.fmt });
+
+    if (self.fmt == c.TH_PF_420) {
+        log("PF = 420", .{});
+        // /* Subsampled in both dimensions */
+        self.uvWidth = @divTrunc(self.width, 2);
+        self.uvHeight = @divTrunc(self.height, 2);
+    } else if (self.fmt == c.TH_PF_422) {
+        log("PF = 422", .{});
+        // /* Subsampled only horizontally */
+        self.uvWidth = @divTrunc(self.width, 2);
+        self.uvHeight = self.height;
+    } else {
+        // /* No subsampling at all... */
+        log("PF = ANY", .{});
+        self.uvWidth = self.width;
+        self.uvHeight = self.height;
+    }
+
+    try self.createTextures();
+
+    // self.frame = try self.allocator.alloc(u8, @intCast(self.width * self.height * 2));
+    self.playbackStarted = true;
+
+    const frame = self.acquireVideoFrameBuffer();
+
+    while (c.tf_readvideo(&self.videoFile, frame.ptr, 1) == 0) {}
+
+    const cmd = rend.context().device.acquireGPUCommandBuffer();
+    self.device.unmapGPUTransferBuffer(self.gpuTransferBuffer);
+    self.releaseAndUploadFrameBuffer(cmd);
+    _ = cmd.submitGPUCommandBuffer();
+}
+
+fn getFrameLength(self: @This()) c_int {
+    return self.width * self.height;
+}
+
+fn acquireVideoFrameBuffer(self: *@This()) []u8 {
+    var slice: []u8 = undefined;
+    slice.ptr = rend.context().device.mapGPUTransferBuffer(self.gpuTransferBuffer, false);
+    slice.len = @intCast(self.width * self.width * 2);
+    return slice;
+}
+
+fn releaseAndUploadFrameBuffer(self: *@This(), cmd: *gpu.GPUCommandBuffer) void {
+    const copyPass = cmd.beginGPUCopyPass();
+    // copy Y buffer
+    copyPass.uploadToGPUTexture(
+        &.{
+            .transfer_buffer = self.gpuTransferBuffer,
+            .offset = 0,
+            .pixels_per_row = @intCast(self.width),
+            .rows_per_layer = @intCast(self.height),
+        },
+        &std.mem.zeroInit(gpu.GPUTextureRegion, .{
+            .texture = self.yTexture,
+            .w = @as(u32, @intCast(self.width)),
+            .h = @as(u32, @intCast(self.height)),
+            .d = 1,
+        }),
+        true,
+    );
+
+    // copy u buffer
+    copyPass.uploadToGPUTexture(
+        &.{
+            .transfer_buffer = self.gpuTransferBuffer,
+            .offset = @intCast(self.getFrameLength()),
+            .pixels_per_row = @intCast(self.uvWidth),
+            .rows_per_layer = @intCast(self.uvHeight),
+        },
+        &std.mem.zeroInit(gpu.GPUTextureRegion, .{
+            .texture = self.yTexture,
+            .w = @as(u32, @intCast(self.uvWidth)),
+            .h = @as(u32, @intCast(self.uvHeight)),
+            .d = 1,
+        }),
+        true,
+    );
+
+    // copy v buffer
+    copyPass.uploadToGPUTexture(
+        &.{
+            .transfer_buffer = self.gpuTransferBuffer,
+            .offset = @intCast(self.getFrameLength() + self.uvWidth * self.uvHeight),
+            .pixels_per_row = @intCast(self.uvWidth),
+            .rows_per_layer = @intCast(self.uvHeight),
+        },
+        &std.mem.zeroInit(gpu.GPUTextureRegion, .{
+            .texture = self.yTexture,
+            .w = @as(u32, @intCast(self.uvWidth)),
+            .h = @as(u32, @intCast(self.uvHeight)),
+            .d = 1,
+        }),
+        true,
+    );
+
+    copyPass.endGPUCopyPass();
+}
+
+fn createTextureInner(width: c_int, height: c_int) !*gpu.GPUTexture {
+    const ctx = rend.context();
+
+    const gpuTexture = ctx.device.createGPUTexture(&std.mem.zeroInit(gpu.GPUTextureCreateInfo, .{
+        .type = .texturetype2d,
+        .format = .textureformatR8Unorm,
+        .usage = .{ .textureusageSampler = true },
+        .width = @as(u32, @intCast(width)),
+        .height = @as(u32, @intCast(height)),
+        .layer_count_or_depth = 1,
+        .num_levels = 1,
+    }));
+
+    return gpuTexture;
+}
+
+pub fn createTextures(self: *@This()) !void {
+    self.yTexture = try createTextureInner(self.width, self.height);
+    self.uTexture = try createTextureInner(self.uvWidth, self.uvHeight);
+    self.vTexture = try createTextureInner(self.uvWidth, self.uvHeight);
+
+    // create transfer Buffer
+    self.gpuTransferBuffer = self.device.createGPUTransferBuffer(&.{
+        .usage = .transferbufferusageUpload,
+        .size = @as(u32, @intCast(self.width * self.height * 2)),
+        .props = 0,
+    });
+}
+
+pub fn tick(self: *@This(), dt: f64) void {
+    if (!self.playbackStarted)
+        return;
+
+    if (c.tf_eos(&self.videoFile) != 0) {
+        c.tf_reset(&self.videoFile);
+    }
+
+    self.currentTime += dt;
+    const thisFrame: c_int = @intFromFloat(self.currentTime * self.fps);
+
+    // read the file's new frame
+    if (thisFrame > self.currentFrame) {
+        const frame = self.acquireVideoFrameBuffer();
+        const newFrame = c.tf_readvideo(&self.videoFile, frame.ptr, thisFrame - self.currentFrame);
+
+        if (newFrame != 0) {
+            const cmd = rend.context().device.acquireGPUCommandBuffer();
+            self.releaseAndUploadFrameBuffer(cmd);
+            _ = cmd.submitGPUCommandBuffer();
+            // kick off texture upload here
+        }
+
+        self.device.unmapGPUTransferBuffer(self.gpuTransferBuffer);
+
+        self.currentFrame = thisFrame;
+    }
+
+    if (ig.begin("videoPlayer", &self.showDebug, .{})) {
+        ig.textFmt("video frame: {d}", .{self.currentFrame}) catch unreachable;
+        ig.end();
+    }
+}
+
+// creates a video player entity
+pub fn createVideoPlayerEntity() !core.Entity {}
+
+pub fn destroy(self: *@This()) void {
+    core.engine_logs("[Videoplayer] destroying player");
+
+    if (self.playbackStarted) {}
+    self.allocator.destroy(self);
+}
+
+const std = @import("std");
+const backlog = @import("Backlog");
+const core = backlog.core;
+const rend = backlog.rend;
+const c = @import("theorafile").c;
+const gpu = rend.renderer.gpu;
+
+const OggTheora_File = c.OggTheora_File;
+
+const ig = backlog.imgui.api;

lib/theoratest/.gitignore

@@ -0,0 +1,6 @@
+.DS_Store
+xcuserdata/
+*.xcworkspace/
+
+bin/
+obj/

lib/theoratest/Makefile

@@ -0,0 +1,133 @@
+# Makefile for Theorafile
+# Written by Ethan "flibitijibibo" Lee
+
+# Detect cross targets
+TRIPLET=$(shell $(CC) -dumpmachine)
+WINDOWS_TARGET=0
+APPLE_TARGET=0
+ifeq ($(OS), Windows_NT) # cygwin/msys2
+	WINDOWS_TARGET=1
+endif
+ifneq (,$(findstring w64-mingw32,$(TRIPLET)))
+	WINDOWS_TARGET=1
+endif
+ifneq (,$(findstring w64-windows,$(TRIPLET)))
+	WINDOWS_TARGET=1
+endif
+ifneq (,$(findstring apple-darwin,$(TRIPLET)))
+	APPLE_TARGET=1
+endif
+ifneq (,$(findstring x86_64,$(TRIPLET)))
+	DEFINES += -DOC_X86_ASM -DOC_X86_64_ASM
+	TFSRC_ARCH_OPT = $(TFSRC_ARCH_X86)
+endif
+ifneq (,$(findstring i686,$(TRIPLET)))
+	DEFINES += -DOC_X86_ASM
+	TFSRC_ARCH_OPT = $(TFSRC_ARCH_X86)
+endif
+# this requires the compiler to support the ARM Neon, Media, and EDSP extensions
+ifneq (,$(findstring armv7,$(TRIPLET)))
+	DEFINES += -DOC_ARM_ASM -DOC_ARM_ASM_EDSP -DOC_ARM_ASM_MEDIA -DOC_ARM_ASM_NEON
+	TFSRC_ARCH_OPT = $(TFSRC_ARCH_ARM)
+endif
+ifneq (,$(findstring aarch64,$(TRIPLET)))
+	DEFINES += -DOC_ARM_ASM -DOC_ARM_ASM_EDSP -DOC_ARM_ASM_MEDIA -DOC_ARM_ASM_NEON
+	TFSRC_ARCH_OPT = $(TFSRC_ARCH_ARM)
+endif
+
+# Compiler
+ifeq ($(WINDOWS_TARGET),1)
+	TARGET = dll
+	LDFLAGS += -static-libgcc
+else ifeq ($(APPLE_TARGET),1)
+	CC += -mmacosx-version-min=10.9
+	TARGET = dylib
+	CFLAGS += -fpic -fPIC
+	LDFLAGS += -install_name @rpath/libtheorafile.dylib
+else
+	TARGET = so
+	CFLAGS += -fpic -fPIC
+endif
+
+LIB = libtheorafile.$(TARGET)
+
+CFLAGS += -O3
+
+SRCDIR = $(dir $(MAKEFILE_LIST))
+
+vpath %.c $(SRCDIR)
+
+# Includes
+INCLUDES = -I$(SRCDIR) -I$(SRCDIR)/lib -I$(SRCDIR)/lib/ogg -I$(SRCDIR)/lib/vorbis -I$(SRCDIR)/lib/theora
+
+# Source
+TFSRC = $(TFSRC_ARCH_GENERIC) $(TFSRC_ARCH_OPT)
+TFSRC_ARCH_GENERIC = \
+	theorafile.c \
+	lib/ogg/bitwise.c \
+	lib/ogg/framing.c \
+	lib/vorbis/analysis.c \
+	lib/vorbis/bitrate.c \
+	lib/vorbis/block.c \
+	lib/vorbis/codebook.c \
+	lib/vorbis/envelope.c \
+	lib/vorbis/floor0.c \
+	lib/vorbis/floor1.c \
+	lib/vorbis/vinfo.c \
+	lib/vorbis/lookup.c \
+	lib/vorbis/lpc.c \
+	lib/vorbis/lsp.c \
+	lib/vorbis/mapping0.c \
+	lib/vorbis/mdct.c \
+	lib/vorbis/psy.c \
+	lib/vorbis/registry.c \
+	lib/vorbis/res0.c \
+	lib/vorbis/sharedbook.c \
+	lib/vorbis/smallft.c \
+	lib/vorbis/synthesis.c \
+	lib/vorbis/window.c \
+	lib/theora/apiwrapper.c \
+	lib/theora/bitpack.c \
+	lib/theora/decapiwrapper.c \
+	lib/theora/decinfo.c \
+	lib/theora/decode.c \
+	lib/theora/dequant.c \
+	lib/theora/fragment.c \
+	lib/theora/huffdec.c \
+	lib/theora/idct.c \
+	lib/theora/tinfo.c \
+	lib/theora/internal.c \
+	lib/theora/quant.c \
+	lib/theora/state.c
+TFSRC_ARCH_X86 = \
+	lib/theora/x86/mmxfrag.c \
+	lib/theora/x86/mmxidct.c \
+	lib/theora/x86/mmxstate.c \
+	lib/theora/x86/sse2idct.c \
+	lib/theora/x86/x86cpu.c \
+	lib/theora/x86/x86state.c
+TFSRC_ARCH_ARM = \
+	lib/theora/arm-intrinsics/armcpu.c \
+	lib/theora/arm-intrinsics/armfrag.c \
+	lib/theora/arm-intrinsics/armidct.c \
+	lib/theora/arm-intrinsics/armloop.c \
+	lib/theora/arm-intrinsics/armstate.c
+
+# Targets
+.PHONY: lib all clean test
+lib: $(LIB)
+all: $(LIB) theorafile-test
+clean:
+	rm -f $(LIB) theorafile-test
+test: theorafile-test
+$(LIB): $(TFSRC)
+	$(CC) $(CFLAGS) -shared -o $@ $^ $(INCLUDES) $(DEFINES) -lm $(LDFLAGS)
+theorafile-test: $(TFSRC)
+	$(CC) $(CFLAGS) -g -o $@ sdl3test/sdl3test.c $(TFSRC) $(INCLUDES) $(DEFINES) -lSDL3 -lm
+lib/theora/arm/armfrag.o: lib/theora/arm/armopts-gnu.S
+.INTERMEDIATE: lib/theora/arm/armopts-gnu.S
+.SUFFIXES:
+%-gnu.S: %.s
+	lib/theora/arm/arm2gnu.pl < $< > $@
+%.o: %-gnu.S
+	$(CC) -c -o $@ -Ilib/theora/arm $<

lib/theoratest/README

@@ -0,0 +1,33 @@
+This is Theorafile, a library for quickly and easily decoding Ogg Theora videos.
+
+Project Website: https://github.com/FNA-XNA/Theorafile
+
+License
+-------
+Theorafile is released under the zlib license.
+
+libogg/libvorbis/libtheora are released under the BSD license.
+
+See the licenses/ folder for details.
+
+About Theorafile
+----------------
+Theorafile was written to be used for FNA's VideoPlayer. We access this library
+via Theorafile#, which you can find in the 'csharp/' directory.
+
+Dependencies
+------------
+Theorafile depends solely on the C runtime. libogg, libvorbis, and libtheoradec
+are statically linked into Theorafile.
+
+Theorafile's "sdl3test" test program requires SDL3.
+
+Building Theorafile
+-------------------
+For *nix platforms, just type `make` in the root directory!
+
+For Windows, see the 'visualc/' directory.
+
+For Xbox GDK, see the 'visualc-gdk/' directory.
+
+For iOS/tvOS and macOS universal binaries, see the 'Xcode/' directory.

lib/theoratest/build.zig

@@ -0,0 +1,126 @@
+const std = @import("std");
+
+pub fn build(b: *std.Build) void {
+    const target = b.standardTargetOptions(.{});
+    const optimize = b.standardOptimizeOption(.{});
+
+    const theorafile = b.addModule(
+        "theorafile",
+        .{
+            .optimize = optimize,
+            .target = target,
+            .root_source_file = b.path("theorafile.zig"),
+            .link_libc = true,
+        },
+    );
+
+    const sdl_dep = b.dependency("sdl", .{
+        .target = target,
+        .optimize = optimize,
+    });
+
+    const sdl_lib = sdl_dep.artifact("SDL3");
+    theorafile.linkLibrary(sdl_lib);
+    theorafile.addIncludePath(b.path("../sdl3/SDL/include"));
+    theorafile.addIncludePath(b.path("lib/ogg"));
+    theorafile.addIncludePath(b.path("lib/theora"));
+    theorafile.addIncludePath(b.path("lib/vorbis"));
+    theorafile.addIncludePath(b.path("lib"));
+    theorafile.addIncludePath(b.path("."));
+
+    theorafile.addCSourceFiles(.{
+        .files = &.{
+            "theorafile.c",
+            "lib/ogg/bitwise.c",
+            "lib/ogg/framing.c",
+            "lib/vorbis/analysis.c",
+            "lib/vorbis/bitrate.c",
+            "lib/vorbis/block.c",
+            "lib/vorbis/codebook.c",
+            "lib/vorbis/envelope.c",
+            "lib/vorbis/floor0.c",
+            "lib/vorbis/floor1.c",
+            "lib/vorbis/vinfo.c",
+            "lib/vorbis/lookup.c",
+            "lib/vorbis/lpc.c",
+            "lib/vorbis/lsp.c",
+            "lib/vorbis/mapping0.c",
+            "lib/vorbis/mdct.c",
+            "lib/vorbis/psy.c",
+            "lib/vorbis/registry.c",
+            "lib/vorbis/res0.c",
+            "lib/vorbis/sharedbook.c",
+            "lib/vorbis/smallft.c",
+            "lib/vorbis/synthesis.c",
+            "lib/vorbis/window.c",
+            "lib/theora/apiwrapper.c",
+            "lib/theora/bitpack.c",
+            "lib/theora/decapiwrapper.c",
+            "lib/theora/decinfo.c",
+            "lib/theora/decode.c",
+            "lib/theora/dequant.c",
+            "lib/theora/fragment.c",
+            "lib/theora/huffdec.c",
+            "lib/theora/idct.c",
+            "lib/theora/tinfo.c",
+            "lib/theora/internal.c",
+            "lib/theora/quant.c",
+            "lib/theora/state.c",
+        },
+        .flags = &.{
+            "-fno-sanitize=undefined",
+        },
+    });
+
+    theorafile.addCSourceFiles(.{
+        .files = &.{
+            "lib/theora/x86/mmxfrag.c",
+            "lib/theora/x86/mmxidct.c",
+            "lib/theora/x86/mmxstate.c",
+            "lib/theora/x86/sse2idct.c",
+            "lib/theora/x86/x86cpu.c",
+            "lib/theora/x86/x86state.c",
+        },
+        .flags = &.{
+            "-fno-sanitize=undefined",
+        },
+    });
+
+    // TFSRC_ARCH_X86 ="
+    // lib/theora/x86/mmxfrag.c \
+    // lib/theora/x86/mmxidct.c \
+    // lib/theora/x86/mmxstate.c \
+    // lib/theora/x86/sse2idct.c \
+    // lib/theora/x86/x86cpu.c \
+    // lib/theora/x86/x86state.c
+    // link theorafile
+
+    const test_step = b.step("test", "run unit tests for sdl3");
+    const tests = b.addExecutable(.{
+        .name = "test",
+        .target = target,
+        .optimize = optimize,
+        .link_libc = true,
+    });
+
+    tests.addCSourceFile(.{
+        .file = b.path("sdl3test/sdl3test.c"),
+        .flags = &.{
+            "-fno-sanitize=undefined",
+        },
+    });
+    tests.root_module.addImport("theorafile", theorafile);
+
+    tests.root_module.addIncludePath(b.path("../sdl3/SDL/include"));
+    tests.root_module.addIncludePath(b.path("lib/ogg"));
+    tests.root_module.addIncludePath(b.path("lib/theora"));
+    tests.root_module.addIncludePath(b.path("lib/vorbis"));
+    tests.root_module.addIncludePath(b.path("lib"));
+
+    tests.root_module.addIncludePath(b.path("."));
+    tests.root_module.addIncludePath(b.path("sdl3test"));
+
+    const runArtifact = b.addRunArtifact(tests);
+    test_step.dependOn(&runArtifact.step);
+    b.installArtifact(tests);
+}

lib/theoratest/build.zig.zon

@@ -0,0 +1,11 @@
+.{
+    .name = .theorafile,
+    .version = "0.0.0",
+    .dependencies = .{
+        .sdl = .{ .path = "../sdl3/SDL" },
+    },
+    .paths = .{
+        ""
+    },
+    .fingerprint = 0x5105e7c085faf2fb,
+}

lib/theoratest/c.zig

@@ -0,0 +1,3 @@
+pub const c = @cImport(
+    @cInclude("theorafile.h"),
+);

lib/theoratest/lib/README

@@ -0,0 +1,10 @@
+This is a local copy of the Xiph libs used by Theorafile, with all the unused
+files stripped out. Below are the notes for each library.
+
+libogg: Version 1.3.3
+	- os_types.h falls back to stdint for unrecognized platforms
+libvorbis: Version 1.3.6
+	- info.c has been renamed to vinfo.c
+libtheora: git commit 7180717276af1ebc7da15c83162d6c5d6203aabf
+	- info.c has been renamed to tinfo.c
+	- AArch64 NEON support from https://github.com/linnaea/theora

lib/theoratest/lib/ogg/ogg.h

@@ -0,0 +1,210 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2007             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: toplevel libogg include
+ last mod: $Id$
+
+ ********************************************************************/
+#ifndef _OGG_H
+#define _OGG_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stddef.h>
+#include <ogg/os_types.h>
+
+typedef struct {
+  void *iov_base;
+  size_t iov_len;
+} ogg_iovec_t;
+
+typedef struct {
+  long endbyte;
+  int  endbit;
+
+  unsigned char *buffer;
+  unsigned char *ptr;
+  long storage;
+} oggpack_buffer;
+
+/* ogg_page is used to encapsulate the data in one Ogg bitstream page *****/
+
+typedef struct {
+  unsigned char *header;
+  long header_len;
+  unsigned char *body;
+  long body_len;
+} ogg_page;
+
+/* ogg_stream_state contains the current encode/decode state of a logical
+   Ogg bitstream **********************************************************/
+
+typedef struct {
+  unsigned char   *body_data;    /* bytes from packet bodies */
+  long    body_storage;          /* storage elements allocated */
+  long    body_fill;             /* elements stored; fill mark */
+  long    body_returned;         /* elements of fill returned */
+
+
+  int     *lacing_vals;      /* The values that will go to the segment table */
+  ogg_int64_t *granule_vals; /* granulepos values for headers. Not compact
+                                this way, but it is simple coupled to the
+                                lacing fifo */
+  long    lacing_storage;
+  long    lacing_fill;
+  long    lacing_packet;
+  long    lacing_returned;
+
+  unsigned char    header[282];      /* working space for header encode */
+  int              header_fill;
+
+  int     e_o_s;          /* set when we have buffered the last packet in the
+                             logical bitstream */
+  int     b_o_s;          /* set after we've written the initial page
+                             of a logical bitstream */
+  long    serialno;
+  long    pageno;
+  ogg_int64_t  packetno;  /* sequence number for decode; the framing
+                             knows where there's a hole in the data,
+                             but we need coupling so that the codec
+                             (which is in a separate abstraction
+                             layer) also knows about the gap */
+  ogg_int64_t   granulepos;
+
+} ogg_stream_state;
+
+/* ogg_packet is used to encapsulate the data and metadata belonging
+   to a single raw Ogg/Vorbis packet *************************************/
+
+typedef struct {
+  unsigned char *packet;
+  long  bytes;
+  long  b_o_s;
+  long  e_o_s;
+
+  ogg_int64_t  granulepos;
+
+  ogg_int64_t  packetno;     /* sequence number for decode; the framing
+                                knows where there's a hole in the data,
+                                but we need coupling so that the codec
+                                (which is in a separate abstraction
+                                layer) also knows about the gap */
+} ogg_packet;
+
+typedef struct {
+  unsigned char *data;
+  int storage;
+  int fill;
+  int returned;
+
+  int unsynced;
+  int headerbytes;
+  int bodybytes;
+} ogg_sync_state;
+
+/* Ogg BITSTREAM PRIMITIVES: bitstream ************************/
+
+extern void  oggpack_writeinit(oggpack_buffer *b);
+extern int   oggpack_writecheck(oggpack_buffer *b);
+extern void  oggpack_writetrunc(oggpack_buffer *b,long bits);
+extern void  oggpack_writealign(oggpack_buffer *b);
+extern void  oggpack_writecopy(oggpack_buffer *b,void *source,long bits);
+extern void  oggpack_reset(oggpack_buffer *b);
+extern void  oggpack_writeclear(oggpack_buffer *b);
+extern void  oggpack_readinit(oggpack_buffer *b,unsigned char *buf,int bytes);
+extern void  oggpack_write(oggpack_buffer *b,unsigned long value,int bits);
+extern long  oggpack_look(oggpack_buffer *b,int bits);
+extern long  oggpack_look1(oggpack_buffer *b);
+extern void  oggpack_adv(oggpack_buffer *b,int bits);
+extern void  oggpack_adv1(oggpack_buffer *b);
+extern long  oggpack_read(oggpack_buffer *b,int bits);
+extern long  oggpack_read1(oggpack_buffer *b);
+extern long  oggpack_bytes(oggpack_buffer *b);
+extern long  oggpack_bits(oggpack_buffer *b);
+extern unsigned char *oggpack_get_buffer(oggpack_buffer *b);
+
+extern void  oggpackB_writeinit(oggpack_buffer *b);
+extern int   oggpackB_writecheck(oggpack_buffer *b);
+extern void  oggpackB_writetrunc(oggpack_buffer *b,long bits);
+extern void  oggpackB_writealign(oggpack_buffer *b);
+extern void  oggpackB_writecopy(oggpack_buffer *b,void *source,long bits);
+extern void  oggpackB_reset(oggpack_buffer *b);
+extern void  oggpackB_writeclear(oggpack_buffer *b);
+extern void  oggpackB_readinit(oggpack_buffer *b,unsigned char *buf,int bytes);
+extern void  oggpackB_write(oggpack_buffer *b,unsigned long value,int bits);
+extern long  oggpackB_look(oggpack_buffer *b,int bits);
+extern long  oggpackB_look1(oggpack_buffer *b);
+extern void  oggpackB_adv(oggpack_buffer *b,int bits);
+extern void  oggpackB_adv1(oggpack_buffer *b);
+extern long  oggpackB_read(oggpack_buffer *b,int bits);
+extern long  oggpackB_read1(oggpack_buffer *b);
+extern long  oggpackB_bytes(oggpack_buffer *b);
+extern long  oggpackB_bits(oggpack_buffer *b);
+extern unsigned char *oggpackB_get_buffer(oggpack_buffer *b);
+
+/* Ogg BITSTREAM PRIMITIVES: encoding **************************/
+
+extern int      ogg_stream_packetin(ogg_stream_state *os, ogg_packet *op);
+extern int      ogg_stream_iovecin(ogg_stream_state *os, ogg_iovec_t *iov,
+                                   int count, long e_o_s, ogg_int64_t granulepos);
+extern int      ogg_stream_pageout(ogg_stream_state *os, ogg_page *og);
+extern int      ogg_stream_pageout_fill(ogg_stream_state *os, ogg_page *og, int nfill);
+extern int      ogg_stream_flush(ogg_stream_state *os, ogg_page *og);
+extern int      ogg_stream_flush_fill(ogg_stream_state *os, ogg_page *og, int nfill);
+
+/* Ogg BITSTREAM PRIMITIVES: decoding **************************/
+
+extern int      ogg_sync_init(ogg_sync_state *oy);
+extern int      ogg_sync_clear(ogg_sync_state *oy);
+extern int      ogg_sync_reset(ogg_sync_state *oy);
+extern int      ogg_sync_destroy(ogg_sync_state *oy);
+extern int      ogg_sync_check(ogg_sync_state *oy);
+
+extern char    *ogg_sync_buffer(ogg_sync_state *oy, long size);
+extern int      ogg_sync_wrote(ogg_sync_state *oy, long bytes);
+extern long     ogg_sync_pageseek(ogg_sync_state *oy,ogg_page *og);
+extern int      ogg_sync_pageout(ogg_sync_state *oy, ogg_page *og);
+extern int      ogg_stream_pagein(ogg_stream_state *os, ogg_page *og);
+extern int      ogg_stream_packetout(ogg_stream_state *os,ogg_packet *op);
+extern int      ogg_stream_packetpeek(ogg_stream_state *os,ogg_packet *op);
+
+/* Ogg BITSTREAM PRIMITIVES: general ***************************/
+
+extern int      ogg_stream_init(ogg_stream_state *os,int serialno);
+extern int      ogg_stream_clear(ogg_stream_state *os);
+extern int      ogg_stream_reset(ogg_stream_state *os);
+extern int      ogg_stream_reset_serialno(ogg_stream_state *os,int serialno);
+extern int      ogg_stream_destroy(ogg_stream_state *os);
+extern int      ogg_stream_check(ogg_stream_state *os);
+extern int      ogg_stream_eos(ogg_stream_state *os);
+
+extern void     ogg_page_checksum_set(ogg_page *og);
+
+extern int      ogg_page_version(const ogg_page *og);
+extern int      ogg_page_continued(const ogg_page *og);
+extern int      ogg_page_bos(const ogg_page *og);
+extern int      ogg_page_eos(const ogg_page *og);
+extern ogg_int64_t  ogg_page_granulepos(const ogg_page *og);
+extern int      ogg_page_serialno(const ogg_page *og);
+extern long     ogg_page_pageno(const ogg_page *og);
+extern int      ogg_page_packets(const ogg_page *og);
+
+extern void     ogg_packet_clear(ogg_packet *op);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  /* _OGG_H */

lib/theoratest/lib/ogg/os_types.h

@@ -0,0 +1,155 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2002             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: #ifdef jail to whip a few platforms into the UNIX ideal.
+ last mod: $Id$
+
+ ********************************************************************/
+#ifndef _OS_TYPES_H
+#define _OS_TYPES_H
+
+/* make it easy on the folks that want to compile the libs with a
+   different malloc than stdlib */
+#define _ogg_malloc  malloc
+#define _ogg_calloc  calloc
+#define _ogg_realloc realloc
+#define _ogg_free    free
+
+#if defined(_WIN32)
+
+#  if defined(__CYGWIN__)
+#    include <stdint.h>
+     typedef int16_t ogg_int16_t;
+     typedef uint16_t ogg_uint16_t;
+     typedef int32_t ogg_int32_t;
+     typedef uint32_t ogg_uint32_t;
+     typedef int64_t ogg_int64_t;
+     typedef uint64_t ogg_uint64_t;
+#  elif defined(__MINGW32__)
+#    include <sys/types.h>
+     typedef short ogg_int16_t;
+     typedef unsigned short ogg_uint16_t;
+     typedef int ogg_int32_t;
+     typedef unsigned int ogg_uint32_t;
+     typedef long long ogg_int64_t;
+     typedef unsigned long long ogg_uint64_t;
+#  elif defined(__MWERKS__)
+     typedef long long ogg_int64_t;
+     typedef int ogg_int32_t;
+     typedef unsigned int ogg_uint32_t;
+     typedef short ogg_int16_t;
+     typedef unsigned short ogg_uint16_t;
+#  else
+#    if defined(_MSC_VER) && (_MSC_VER >= 1800) /* MSVC 2013 and newer */
+#      include <stdint.h>
+       typedef int16_t ogg_int16_t;
+       typedef uint16_t ogg_uint16_t;
+       typedef int32_t ogg_int32_t;
+       typedef uint32_t ogg_uint32_t;
+       typedef int64_t ogg_int64_t;
+       typedef uint64_t ogg_uint64_t;
+#    else
+       /* MSVC/Borland */
+       typedef __int64 ogg_int64_t;
+       typedef __int32 ogg_int32_t;
+       typedef unsigned __int32 ogg_uint32_t;
+       typedef __int16 ogg_int16_t;
+       typedef unsigned __int16 ogg_uint16_t;
+#    endif
+#  endif
+
+#elif (defined(__APPLE__) && defined(__MACH__)) /* MacOS X Framework build */
+
+#  include <inttypes.h>
+   typedef int16_t ogg_int16_t;
+   typedef uint16_t ogg_uint16_t;
+   typedef int32_t ogg_int32_t;
+   typedef uint32_t ogg_uint32_t;
+   typedef int64_t ogg_int64_t;
+
+#elif defined(__HAIKU__)
+
+  /* Haiku */
+#  include <sys/types.h>
+   typedef short ogg_int16_t;
+   typedef unsigned short ogg_uint16_t;
+   typedef int ogg_int32_t;
+   typedef unsigned int ogg_uint32_t;
+   typedef long long ogg_int64_t;
+
+#elif defined(__BEOS__)
+
+   /* Be */
+#  include <inttypes.h>
+   typedef int16_t ogg_int16_t;
+   typedef uint16_t ogg_uint16_t;
+   typedef int32_t ogg_int32_t;
+   typedef uint32_t ogg_uint32_t;
+   typedef int64_t ogg_int64_t;
+
+#elif defined (__EMX__)
+
+   /* OS/2 GCC */
+   typedef short ogg_int16_t;
+   typedef unsigned short ogg_uint16_t;
+   typedef int ogg_int32_t;
+   typedef unsigned int ogg_uint32_t;
+   typedef long long ogg_int64_t;
+
+#elif defined (DJGPP)
+
+   /* DJGPP */
+   typedef short ogg_int16_t;
+   typedef int ogg_int32_t;
+   typedef unsigned int ogg_uint32_t;
+   typedef long long ogg_int64_t;
+
+#elif defined(R5900)
+
+   /* PS2 EE */
+   typedef long ogg_int64_t;
+   typedef int ogg_int32_t;
+   typedef unsigned ogg_uint32_t;
+   typedef short ogg_int16_t;
+
+#elif defined(__SYMBIAN32__)
+
+   /* Symbian GCC */
+   typedef signed short ogg_int16_t;
+   typedef unsigned short ogg_uint16_t;
+   typedef signed int ogg_int32_t;
+   typedef unsigned int ogg_uint32_t;
+   typedef long long int ogg_int64_t;
+
+#elif defined(__TMS320C6X__)
+
+   /* TI C64x compiler */
+   typedef signed short ogg_int16_t;
+   typedef unsigned short ogg_uint16_t;
+   typedef signed int ogg_int32_t;
+   typedef unsigned int ogg_uint32_t;
+   typedef long long int ogg_int64_t;
+
+#else
+
+/* Theorafile change! We assume you have stdint... */
+#    include <stdint.h>
+     typedef int16_t ogg_int16_t;
+     typedef uint16_t ogg_uint16_t;
+     typedef int32_t ogg_int32_t;
+     typedef uint32_t ogg_uint32_t;
+     typedef int64_t ogg_int64_t;
+     typedef uint64_t ogg_uint64_t;
+
+#endif
+
+#endif  /* _OS_TYPES_H */

lib/theoratest/lib/theora/apiwrapper.c

@@ -0,0 +1,166 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <limits.h>
+#include "apiwrapper.h"
+
+
+
+const char *theora_version_string(void){
+  return th_version_string();
+}
+
+ogg_uint32_t theora_version_number(void){
+  return th_version_number();
+}
+
+void theora_info_init(theora_info *_ci){
+  memset(_ci,0,sizeof(*_ci));
+}
+
+void theora_info_clear(theora_info *_ci){
+  th_api_wrapper *api;
+  api=(th_api_wrapper *)_ci->codec_setup;
+  memset(_ci,0,sizeof(*_ci));
+  if(api!=NULL){
+    if(api->clear!=NULL)(*api->clear)(api);
+    _ogg_free(api);
+  }
+}
+
+void theora_clear(theora_state *_th){
+  /*Provide compatibility with mixed encoder and decoder shared lib versions.*/
+  if(_th->internal_decode!=NULL){
+    (*((oc_state_dispatch_vtable *)_th->internal_decode)->clear)(_th);
+  }
+  if(_th->internal_encode!=NULL){
+    (*((oc_state_dispatch_vtable *)_th->internal_encode)->clear)(_th);
+  }
+  if(_th->i!=NULL)theora_info_clear(_th->i);
+  memset(_th,0,sizeof(*_th));
+}
+
+int theora_control(theora_state *_th,int _req,void *_buf,size_t _buf_sz){
+  /*Provide compatibility with mixed encoder and decoder shared lib versions.*/
+  if(_th->internal_decode!=NULL){
+    return (*((oc_state_dispatch_vtable *)_th->internal_decode)->control)(_th,
+     _req,_buf,_buf_sz);
+  }
+  else if(_th->internal_encode!=NULL){
+    return (*((oc_state_dispatch_vtable *)_th->internal_encode)->control)(_th,
+     _req,_buf,_buf_sz);
+  }
+  else return TH_EINVAL;
+}
+
+ogg_int64_t theora_granule_frame(theora_state *_th,ogg_int64_t _gp){
+  /*Provide compatibility with mixed encoder and decoder shared lib versions.*/
+  if(_th->internal_decode!=NULL){
+    return (*((oc_state_dispatch_vtable *)_th->internal_decode)->granule_frame)(
+     _th,_gp);
+  }
+  else if(_th->internal_encode!=NULL){
+    return (*((oc_state_dispatch_vtable *)_th->internal_encode)->granule_frame)(
+     _th,_gp);
+  }
+  else return -1;
+}
+
+double theora_granule_time(theora_state *_th, ogg_int64_t _gp){
+  /*Provide compatibility with mixed encoder and decoder shared lib versions.*/
+  if(_th->internal_decode!=NULL){
+    return (*((oc_state_dispatch_vtable *)_th->internal_decode)->granule_time)(
+     _th,_gp);
+  }
+  else if(_th->internal_encode!=NULL){
+    return (*((oc_state_dispatch_vtable *)_th->internal_encode)->granule_time)(
+     _th,_gp);
+  }
+  else return -1;
+}
+
+void oc_theora_info2th_info(th_info *_info,const theora_info *_ci){
+  _info->version_major=_ci->version_major;
+  _info->version_minor=_ci->version_minor;
+  _info->version_subminor=_ci->version_subminor;
+  _info->frame_width=_ci->width;
+  _info->frame_height=_ci->height;
+  _info->pic_width=_ci->frame_width;
+  _info->pic_height=_ci->frame_height;
+  _info->pic_x=_ci->offset_x;
+  _info->pic_y=_ci->offset_y;
+  _info->fps_numerator=_ci->fps_numerator;
+  _info->fps_denominator=_ci->fps_denominator;
+  _info->aspect_numerator=_ci->aspect_numerator;
+  _info->aspect_denominator=_ci->aspect_denominator;
+  switch(_ci->colorspace){
+    case OC_CS_ITU_REC_470M:_info->colorspace=TH_CS_ITU_REC_470M;break;
+    case OC_CS_ITU_REC_470BG:_info->colorspace=TH_CS_ITU_REC_470BG;break;
+    default:_info->colorspace=TH_CS_UNSPECIFIED;break;
+  }
+  switch(_ci->pixelformat){
+    case OC_PF_420:_info->pixel_fmt=TH_PF_420;break;
+    case OC_PF_422:_info->pixel_fmt=TH_PF_422;break;
+    case OC_PF_444:_info->pixel_fmt=TH_PF_444;break;
+    default:_info->pixel_fmt=TH_PF_RSVD;
+  }
+  _info->target_bitrate=_ci->target_bitrate;
+  _info->quality=_ci->quality;
+  _info->keyframe_granule_shift=_ci->keyframe_frequency_force>0?
+   OC_MINI(31,oc_ilog(_ci->keyframe_frequency_force-1)):0;
+}
+
+int theora_packet_isheader(ogg_packet *_op){
+  return th_packet_isheader(_op);
+}
+
+int theora_packet_iskeyframe(ogg_packet *_op){
+  return th_packet_iskeyframe(_op);
+}
+
+int theora_granule_shift(theora_info *_ci){
+  /*This breaks when keyframe_frequency_force is not positive or is larger than
+     2**31 (if your int is more than 32 bits), but that's what the original
+     function does.*/
+  return oc_ilog(_ci->keyframe_frequency_force-1);
+}
+
+void theora_comment_init(theora_comment *_tc){
+  th_comment_init((th_comment *)_tc);
+}
+
+char *theora_comment_query(theora_comment *_tc,char *_tag,int _count){
+  return th_comment_query((th_comment *)_tc,_tag,_count);
+}
+
+int theora_comment_query_count(theora_comment *_tc,char *_tag){
+  return th_comment_query_count((th_comment *)_tc,_tag);
+}
+
+void theora_comment_clear(theora_comment *_tc){
+  th_comment_clear((th_comment *)_tc);
+}
+
+void theora_comment_add(theora_comment *_tc,char *_comment){
+  th_comment_add((th_comment *)_tc,_comment);
+}
+
+void theora_comment_add_tag(theora_comment *_tc, char *_tag, char *_value){
+  th_comment_add_tag((th_comment *)_tc,_tag,_value);
+}

lib/theoratest/lib/theora/apiwrapper.h

@@ -0,0 +1,54 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id: apiwrapper.h 13596 2007-08-23 20:05:38Z tterribe $
+
+ ********************************************************************/
+
+#if !defined(_apiwrapper_H)
+# define _apiwrapper_H (1)
+# include <ogg/ogg.h>
+# include <theora/theora.h>
+# include "theora/theoradec.h"
+# include "theora/theoraenc.h"
+# include "state.h"
+
+typedef struct th_api_wrapper th_api_wrapper;
+typedef struct th_api_info    th_api_info;
+
+/*Provide an entry point for the codec setup to clear itself in case we ever
+   want to break pieces off into a common base library shared by encoder and
+   decoder.
+  In addition, this makes several other pieces of the API wrapper cleaner.*/
+typedef void (*oc_setup_clear_func)(void *_ts);
+
+/*Generally only one of these pointers will be non-NULL in any given instance.
+  Technically we do not even really need this struct, since we should be able
+   to figure out which one from "context", but doing it this way makes sure we
+   don't flub it up.*/
+struct th_api_wrapper{
+  oc_setup_clear_func  clear;
+  th_setup_info       *setup;
+  th_dec_ctx          *decode;
+  th_enc_ctx          *encode;
+};
+
+struct th_api_info{
+  th_api_wrapper api;
+  theora_info    info;
+};
+
+
+void oc_theora_info2th_info(th_info *_info,const theora_info *_ci);
+
+#endif

lib/theoratest/lib/theora/arm-intrinsics/armcpu.c

@@ -0,0 +1,115 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2010                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+ CPU capability detection for ARM processors.
+
+ function:
+  last mod: $Id: cpu.c 17344 2010-07-21 01:42:18Z tterribe $
+
+ ********************************************************************/
+
+#include "armcpu.h"
+
+#if !defined(OC_ARM_ASM) || !defined(OC_ARM_ASM_NEON)
+ogg_uint32_t oc_cpu_flags_get(void){
+  return 0;
+}
+
+#elif defined(__aarch64__) || defined(_M_ARM64)
+ogg_uint32_t oc_cpu_flags_get(void) {
+  return OC_CPU_ARM_NEON;
+}
+
+#elif defined(_MSC_VER)
+/*For GetExceptionCode() and EXCEPTION_ILLEGAL_INSTRUCTION.*/
+# define WIN32_LEAN_AND_MEAN
+# define WIN32_EXTRA_LEAN
+# include <windows.h>
+
+ogg_uint32_t oc_cpu_flags_get(void){
+  ogg_uint32_t flags;
+  flags=0;
+  /*MSVC has no inline __asm support for ARM, but it does let you __emit
+     instructions via their assembled hex code.
+    All of these instructions should be essentially nops.*/
+# if defined(OC_ARM_ASM_NEON)
+  __try{
+#  if defined(__aarch64__) || defined(_M_ARM64)
+    /*MOV v0.16B,v0.16B*/
+    __emit(0x4EA01C00);
+#  else
+    /*VORR q0,q0,q0*/
+    __emit(0xF2200150);
+#  endif
+    flags|=OC_CPU_ARM_NEON;
+  }
+  __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){
+    /*Ignore exception.*/
+  }
+# endif
+  return flags;
+}
+
+#elif defined(__linux__)
+# include <stdio.h>
+# include <stdlib.h>
+# include <string.h>
+
+ogg_uint32_t oc_cpu_flags_get(void){
+  ogg_uint32_t  flags;
+  FILE         *fin;
+  flags=0;
+  /*Reading /proc/self/auxv would be easier, but that doesn't work reliably on
+     Android.
+    This also means that detection will fail in Scratchbox.*/
+  fin=fopen("/proc/cpuinfo","r");
+  if(fin!=NULL){
+    /*512 should be enough for anybody (it's even enough for all the flags that
+       x86 has accumulated... so far).*/
+    char buf[512];
+    while(fgets(buf,511,fin)!=NULL){
+      if(memcmp(buf,"Features",8)==0){
+        char *p;
+        p=strstr(buf," neon");
+        if(p!=NULL&&(p[5]==' '||p[5]=='\n'))flags|=OC_CPU_ARM_NEON;
+        p=strstr(buf," asimd");
+        if(p!=NULL&&(p[6]==' '||p[6]=='\n'))flags|=OC_CPU_ARM_NEON;
+      }
+    }
+    fclose(fin);
+  }
+  return flags;
+}
+
+#elif defined(__riscos__)
+#include <kernel.h>
+#include <swis.h>
+
+ogg_uint32_t oc_cpu_flags_get(void) {
+  ogg_uint32_t flags = 0;
+
+#if defined(OC_ARM_ASM_NEON)
+  ogg_uint32_t mvfr1;
+  test = _swix(VFPSupport_Features, _IN(0)|_OUT(2), 0, &mvfr1);
+  if (test==NULL && (mvfr1 & 0xFFF00)==0x11100)flags|=OC_CPU_ARM_NEON;
+#endif
+
+  return flags;
+}
+
+#else
+/*The feature registers which can tell us what the processor supports are
+   accessible in priveleged modes only, so we can't have a general user-space
+   detection method like on x86.*/
+# error "Configured to use ARM asm but no CPU detection method available for " \
+ "your platform.  Reconfigure with --disable-asm (or send patches)."
+#endif

lib/theoratest/lib/theora/arm-intrinsics/armcpu.h

@@ -0,0 +1,25 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2010                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+ function:
+    last mod: $Id: cpu.h 17344 2010-07-21 01:42:18Z tterribe $
+
+ ********************************************************************/
+
+#if !defined(_arm_armcpu_H)
+# define _arm_armcpu_H (1)
+#include "../internal.h"
+
+#define OC_CPU_ARM_NEON     (1<<12)
+
+ogg_uint32_t oc_cpu_flags_get(void);
+
+#endif

lib/theoratest/lib/theora/arm-intrinsics/armfrag.c

@@ -0,0 +1,93 @@
+#include "armfrag.h"
+
+#if defined(OC_ARM_ASM_NEON)
+#include <arm_neon.h>
+#include "../state.h"
+
+void oc_frag_copy_neon(unsigned char *_dst,const unsigned char *_src,int _ystride) {
+  for (int i = 0; i < 8; i++)
+    vst1_u8(&_dst[i * _ystride], vld1_u8(&_src[i * _ystride]));
+}
+
+void oc_frag_copy_list_neon(unsigned char *_dst_frame, const unsigned char *_src_frame,int _ystride,
+                            const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs) {
+  ptrdiff_t fragii;
+  for (fragii = 0; fragii < _nfragis; fragii++) {
+    ptrdiff_t frag_buf_off;
+    frag_buf_off = _frag_buf_offs[_fragis[fragii]];
+    oc_frag_copy_neon(_dst_frame + frag_buf_off,
+                      _src_frame + frag_buf_off, _ystride);
+  }
+}
+
+void oc_frag_recon_intra_neon(unsigned char *_dst,
+                              int _ystride, const int16_t _residue[64]) {
+  for (int i = 0; i < 8; i++)
+    vst1_u8(&_dst[i * _ystride],
+            vqmovun_s16(vaddq_s16(vld1q_s16(&_residue[i * 8]),
+                                  vdupq_n_s16(128))));
+}
+
+void oc_frag_recon_inter_neon(unsigned char *_dst, const unsigned char *_src,
+                              int _ystride,const int16_t _residue[64]) {
+  for (int i = 0; i < 8; i++)
+    vst1_u8(&_dst[i * _ystride],
+            vqmovun_s16(vaddq_s16(vld1q_s16(&_residue[i * 8]),
+                                  vreinterpretq_s16_u16(vmovl_u8(vld1_u8(&_src[i * _ystride]))))));
+}
+
+void oc_frag_recon_inter2_neon(unsigned char *_dst,const unsigned char *_src1, const unsigned char *_src2,
+                               int _ystride,const int16_t _residue[64]) {
+  for (int i = 0; i < 8; i++)
+    vst1_u8(&_dst[i * _ystride],
+            vqmovun_s16(vaddq_s16(vld1q_s16(&_residue[i * 8]),
+                                  vreinterpretq_s16_u16(vmovl_u8(vhadd_u8(vld1_u8(&_src1[i * _ystride]),
+                                                                          vld1_u8(&_src2[i * _ystride])))))));
+}
+
+void oc_state_frag_recon_neon(const oc_theora_state *_state,ptrdiff_t _fragi,
+ int _pli,int16_t _dct_coeffs[128],int _last_zzi,uint16_t _dc_quant){
+  unsigned char *dst;
+  ptrdiff_t      frag_buf_off;
+  int            ystride;
+  int            refi;
+  /*Apply the inverse transform.*/
+  /*Special case only having a DC component.*/
+  if(_last_zzi<2){
+    ogg_int16_t p;
+    int         ci;
+    /*We round this dequant product (and not any of the others) because there's
+       no iDCT rounding.*/
+    p=(ogg_int16_t)(_dct_coeffs[0]*(ogg_int32_t)_dc_quant+15>>5);
+    /*LOOP VECTORIZES.*/
+    /*Apparently GCC doesn't want to vectorize it. This is the only line changed.*/
+//    for(ci=0;ci<64;ci++)_dct_coeffs[64+ci]=p;
+    for(ci=8;ci<16;ci++) vst1q_s16(&_dct_coeffs[ci*8], vdupq_n_s16(p));
+  }
+  else{
+    /*First, dequantize the DC coefficient.*/
+    _dct_coeffs[0]=(ogg_int16_t)(_dct_coeffs[0]*(int)_dc_quant);
+    oc_idct8x8_neon(_dct_coeffs+64,_dct_coeffs,_last_zzi);
+  }
+  /*Fill in the target buffer.*/
+  frag_buf_off=_state->frag_buf_offs[_fragi];
+  refi=_state->frags[_fragi].refi;
+  ystride=_state->ref_ystride[_pli];
+  dst=_state->ref_frame_data[OC_FRAME_SELF]+frag_buf_off;
+  if(refi==OC_FRAME_SELF)oc_frag_recon_intra_neon(dst,ystride,_dct_coeffs+64);
+  else{
+    const unsigned char *ref;
+    int                  mvoffsets[2];
+    ref=_state->ref_frame_data[refi]+frag_buf_off;
+    if(oc_state_get_mv_offsets(_state,mvoffsets,_pli,
+     _state->frag_mvs[_fragi])>1){
+      oc_frag_recon_inter2_neon(
+       dst,ref+mvoffsets[0],ref+mvoffsets[1],ystride,_dct_coeffs+64);
+    }
+    else{
+      oc_frag_recon_inter_neon(dst,ref+mvoffsets[0],ystride,_dct_coeffs+64);
+    }
+  }
+}
+
+#endif

lib/theoratest/lib/theora/arm-intrinsics/armfrag.h

@@ -0,0 +1,28 @@
+#ifndef _arm_intrinsics_armfrag_H
+#define _arm_intrinsics_armfrag_H 1
+#include <stdint.h>
+#include <stddef.h>
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#ifdef OC_ARM_ASM_NEON
+typedef struct oc_theora_state oc_theora_state;
+void oc_idct8x8_neon(int16_t _y[64],int16_t _x[64],int _last_zzi);
+void oc_loop_filter_init_neon(signed char _bv[256],int _flimit);
+void oc_state_loop_filter_frag_rows_neon(const oc_theora_state *_state,
+                                         signed char *_bv,int _refi,int _pli,int _fragy0,int _fragy_end);
+void oc_frag_copy_neon(unsigned char *_dst,const unsigned char *_src,int _ystride);
+void oc_frag_copy_list_neon(unsigned char *_dst_frame, const unsigned char *_src_frame,int _ystride,
+                            const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs);
+void oc_frag_recon_intra_neon(unsigned char *_dst,
+                              int _ystride, const int16_t _residue[64]);
+void oc_frag_recon_inter_neon(unsigned char *_dst, const unsigned char *_src,
+                              int _ystride,const int16_t _residue[64]);
+void oc_frag_recon_inter2_neon(unsigned char *_dst,const unsigned char *_src1, const unsigned char *_src2,
+                               int _ystride,const int16_t _residue[64]);
+void oc_state_frag_recon_neon(const oc_theora_state *_state,ptrdiff_t _fragi,
+                              int _pli,int16_t _dct_coeffs[128],int _last_zzi,uint16_t _dc_quant);
+#endif
+
+#endif //_arm_intrinsics_armfrag_H

lib/theoratest/lib/theora/arm-intrinsics/armidct.c

@@ -0,0 +1,235 @@
+#include "armfrag.h"
+
+#if defined(OC_ARM_ASM_NEON)
+typedef int32_t ogg_int32_t;
+#include "../dct.h"
+#include "neon_a64_compat.h"
+#include "neon_transpose.h"
+
+static inline int16x8_t scale_nu_s16(int16x8_t a, uint16_t b) {
+    if (!(b & 1))
+        return vqdmulhq_n_s16(a, b >> 1);
+    if (b < 32768)
+        return vshrq_n_s16(vqdmulhq_n_s16(a, b), 1);
+
+    int32x4_t al = vmovl_s16(vget_low_s16(a));
+    int32x4_t ah = vmovl_high_s16(a);
+    al = vmulq_n_s32(al, b);
+    ah = vmulq_n_s32(ah, b);
+    return vuzp2q_s16(vreinterpretq_s16_s32(al), vreinterpretq_s16_s32(ah));
+}
+
+static void v_transpose4x8(const int16x4_t a[8], int16x8_t b[4]) {
+    int16x8x2_t s0 = vtrnq_s16(vcombine_s16(a[0], a[4]), vcombine_s16(a[1], a[5]));
+    int16x8x2_t s1 = vtrnq_s16(vcombine_s16(a[2], a[6]), vcombine_s16(a[3], a[7]));
+    /*  1  5  3  7 17 21 19 23  s0[0] = VTRN1.1N p0 p1 */
+    /*  2  6  4  8 18 22 20 24  s0[1] = VTRN2.1N p0 p1 */
+    /*  9 13 11 15 25 29 27 31  s1[0] = VTRN1.1N p2 p3 */
+    /* 10 14 12 16 26 30 28 32  s1[1] = VTRN2.1N p2 p3 */
+
+    int32x4x2_t t0 = vtrnq_s32(vreinterpretq_s32_s16(s0.val[0]), vreinterpretq_s32_s16(s1.val[0]));
+    int32x4x2_t t1 = vtrnq_s32(vreinterpretq_s32_s16(s0.val[1]), vreinterpretq_s32_s16(s1.val[1]));
+    /*  1  5  9 13 17 21 25 29  t0[0] = VTRN1.2N s0[0] s1[0] */
+    /*  2  6 10 14 18 22 26 30  t1[0] = VTRN1.2N s0[1] s1[1] */
+    /*  3  7 11 15 19 23 27 31  t0[1] = VTRN2.2N s0[0] s1[0] */
+    /*  4  8 12 16 20 24 28 32  t1[1] = VTRN2.2N s0[1] s1[1] */
+
+    b[0] = vreinterpretq_s16_s32(t0.val[0]);
+    b[1] = vreinterpretq_s16_s32(t1.val[0]);
+    b[2] = vreinterpretq_s16_s32(t0.val[1]);
+    b[3] = vreinterpretq_s16_s32(t1.val[1]);
+}
+
+static void idct8x8(const int16x8_t x[8], int16x8_t y[8]) {
+    int16x8_t t[8], r;
+
+    /*Stage 1:*/
+    /*0-1 butterfly.*/
+    t[0] = scale_nu_s16(vaddq_s16(x[0], x[4]), OC_C4S4);
+    t[1] = scale_nu_s16(vsubq_s16(x[0], x[4]), OC_C4S4);
+    /*2-3 rotation by 6pi/16.*/
+    t[2] = vsubq_s16(scale_nu_s16(x[2], OC_C6S2), scale_nu_s16(x[6], OC_C2S6));
+    t[3] = vaddq_s16(scale_nu_s16(x[2], OC_C2S6), scale_nu_s16(x[6], OC_C6S2));
+    /*4-7 rotation by 7pi/16.*/
+    t[4] = vsubq_s16(scale_nu_s16(x[1], OC_C7S1), scale_nu_s16(x[7], OC_C1S7));
+    t[7] = vaddq_s16(scale_nu_s16(x[1], OC_C1S7), scale_nu_s16(x[7], OC_C7S1));
+    /*5-6 rotation by 3pi/16.*/
+    t[5] = vsubq_s16(scale_nu_s16(x[5], OC_C3S5), scale_nu_s16(x[3], OC_C5S3));
+    t[6] = vaddq_s16(scale_nu_s16(x[5], OC_C5S3), scale_nu_s16(x[3], OC_C3S5));
+
+    /*Stage 2:*/
+    /*4-5 butterfly.*/
+    r = vaddq_s16(t[4], t[5]);
+    t[5] = scale_nu_s16(vsubq_s16(t[4], t[5]), OC_C4S4);
+    t[4] = r;
+    /*7-6 butterfly.*/
+    r = vaddq_s16(t[7], t[6]);
+    t[6] = scale_nu_s16(vsubq_s16(t[7], t[6]), OC_C4S4);
+    t[7] = r;
+
+    /*Stage 3:*/
+    /*0-3 butterfly.*/
+    r = vaddq_s16(t[0], t[3]);
+    t[3] = vsubq_s16(t[0], t[3]);
+    t[0] = r;
+    /*1-2 butterfly.*/
+    r = vaddq_s16(t[1], t[2]);
+    t[2] = vsubq_s16(t[1], t[2]);
+    t[1] = r;
+    /*6-5 butterfly.*/
+    r = vaddq_s16(t[6], t[5]);
+    t[5] = vsubq_s16(t[6], t[5]);
+    t[6] = r;
+
+    /*Stage 4:*/
+    /*0-7 butterfly.*/
+    y[0] = vaddq_s16(t[0], t[7]);
+    y[7] = vsubq_s16(t[0], t[7]);
+    /*1-6 butterfly.*/
+    y[1] = vaddq_s16(t[1], t[6]);
+    y[6] = vsubq_s16(t[1], t[6]);
+    /*2-5 butterfly.*/
+    y[2] = vaddq_s16(t[2], t[5]);
+    y[5] = vsubq_s16(t[2], t[5]);
+    /*3-4 butterfly.*/
+    y[3] = vaddq_s16(t[3], t[4]);
+    y[4] = vsubq_s16(t[3], t[4]);
+}
+
+// cos(n*pi/16) (resp. sin(m*pi/16)) scaled by 2147483648
+// GCC doesn't have vld1_s32_x4
+//static const int32_t I32_COSINE[8] = {
+//        2106220352,  418953276, // C1 C7
+//        1984016189,  821806413, // C2 C6
+//        1785567396, 1193077991, // C3 C5
+//        1518500250, 1518500250, // C4 C4
+//};
+static const int32_t I32_COSINE[8] = {
+        OC_C1S7 << 15, OC_C2S6 << 15, OC_C3S5 << 15, OC_C4S4 << 15,
+        OC_C7S1 << 15, OC_C6S2 << 15, OC_C5S3 << 15, OC_C4S4 << 15,
+};
+
+static void idct4x4_4x8(const int16x4_t x[4], int16x4_t y[8]) {
+    int32x4_t s[4];
+    int16x4_t t[8], r;
+    int32x2x4_t c = vld4_s32(I32_COSINE);
+
+    /*Stage 1:*/
+    t[0] = vmovn_s32(vqdmulhq_lane_s32(vmovl_s16(x[0]), c.val[3], 0)); // C4S4
+    t[1] = vmovn_s32(vqdmulhq_lane_s32(vmovl_s16(x[0]), c.val[3], 0)); // C4S4
+    t[2] = vmovn_s32(vqdmulhq_lane_s32(vmovl_s16(x[2]), c.val[1], 1)); // C6S2
+    t[3] = vmovn_s32(vqdmulhq_lane_s32(vmovl_s16(x[2]), c.val[1], 0)); // C2S6
+    s[0] = vqdmulhq_lane_s32(vmovl_s16(x[1]), c.val[0], 1); // C7S1
+    s[3] = vqdmulhq_lane_s32(vmovl_s16(x[1]), c.val[0], 0); // C1S7
+    s[1] = vqdmulhq_lane_s32(vmovl_s16(x[3]), c.val[2], 1); // C5S3
+    s[2] = vqdmulhq_lane_s32(vmovl_s16(x[3]), c.val[2], 0); // C3S5
+    s[1] = vnegq_s32(s[1]);
+
+    /*Stage 2:*/
+    t[5] = vmovn_s32(vqdmulhq_lane_s32(vsubq_s32(s[0], s[1]), c.val[3], 0)); // C4S4
+    t[4] = vmovn_s32(vaddq_s32(s[0], s[1]));
+    t[6] = vmovn_s32(vqdmulhq_lane_s32(vsubq_s32(s[3], s[2]), c.val[3], 0)); // C4S4
+    t[7] = vmovn_s32(vaddq_s32(s[3], s[2]));
+
+    /*Stage 3:*/
+    r = vadd_s16(t[0], t[3]);
+    t[3] = vsub_s16(t[0], t[3]);
+    t[0] = r;
+    r = vadd_s16(t[1], t[2]);
+    t[2] = vsub_s16(t[1], t[2]);
+    t[1] = r;
+    r = vadd_s16(t[6], t[5]);
+    t[5] = vsub_s16(t[6], t[5]);
+    t[6] = r;
+
+    /*Stage 4:*/
+    y[0] = vadd_s16(t[0], t[7]);
+    y[7] = vsub_s16(t[0], t[7]);
+    y[1] = vadd_s16(t[1], t[6]);
+    y[6] = vsub_s16(t[1], t[6]);
+    y[2] = vadd_s16(t[2], t[5]);
+    y[5] = vsub_s16(t[2], t[5]);
+    y[3] = vadd_s16(t[3], t[4]);
+    y[4] = vsub_s16(t[3], t[4]);
+}
+
+static void idct8x4_8x8(const int16x8_t x[4], int16x8_t y[8]) {
+    int16x8_t t[8], r;
+
+    /*Stage 1:*/
+    t[0] = scale_nu_s16(x[0], OC_C4S4);
+    t[1] = scale_nu_s16(x[0], OC_C4S4);
+    t[2] = scale_nu_s16(x[2], OC_C6S2);
+    t[3] = scale_nu_s16(x[2], OC_C2S6);
+    t[4] = scale_nu_s16(x[1], OC_C7S1);
+    t[7] = scale_nu_s16(x[1], OC_C1S7);
+    t[5] = vnegq_s16(scale_nu_s16(x[3], OC_C5S3));
+    t[6] = scale_nu_s16(x[3], OC_C3S5);
+
+    /*Stage 2:*/
+    r = vaddq_s16(t[4], t[5]);
+    t[5] = scale_nu_s16(vsubq_s16(t[4], t[5]), OC_C4S4);
+    t[4] = r;
+    r = vaddq_s16(t[7], t[6]);
+    t[6] = scale_nu_s16(vsubq_s16(t[7], t[6]), OC_C4S4);
+    t[7] = r;
+
+    /*Stage 3:*/
+    r = vaddq_s16(t[0], t[3]);
+    t[3] = vsubq_s16(t[0], t[3]);
+    t[0] = r;
+    r = vaddq_s16(t[1], t[2]);
+    t[2] = vsubq_s16(t[1], t[2]);
+    t[1] = r;
+    r = vaddq_s16(t[6], t[5]);
+    t[5] = vsubq_s16(t[6], t[5]);
+    t[6] = r;
+
+    /*Stage 4:*/
+    y[0] = vaddq_s16(t[0], t[7]);
+    y[7] = vsubq_s16(t[0], t[7]);
+    y[1] = vaddq_s16(t[1], t[6]);
+    y[6] = vsubq_s16(t[1], t[6]);
+    y[2] = vaddq_s16(t[2], t[5]);
+    y[5] = vsubq_s16(t[2], t[5]);
+    y[3] = vaddq_s16(t[3], t[4]);
+    y[4] = vsubq_s16(t[3], t[4]);
+}
+
+static void scale_final(const int16x8_t y[8], int16_t _y[64]) {
+    for (int i = 0; i < 8; i++)
+        vst1q_s16(&_y[i * 8], vrshrq_n_s16(y[i], 4));
+}
+
+static void oc_idct8x8_4(int16_t _y[64], int16_t _x[64]) {
+    int16x4_t x0[4], y0[8];
+    int16x8_t x1[4], y1[8];
+    for (int i = 0; i < 4; i++)
+        x0[i] = vld1_s16(&_x[i * 8]);
+    idct4x4_4x8(x0, y0);
+    v_transpose4x8(y0, x1);
+    idct8x4_8x8(x1, y1);
+    scale_final(y1, _y);
+    for (int i = 0; i < 4; i++)
+        vst1_s16(&_x[i * 8], vcreate_s16(0));
+}
+
+static void oc_idct8x8_8(int16_t _y[64], int16_t _x[64]) {
+    int16x8_t x[8], y[8];
+    for (int i = 0; i < 8; i++)
+        x[i] = vld1q_s16(&_x[i * 8]);
+    idct8x8(x, y);
+    v_transpose8x8(y, x);
+    idct8x8(x, y);
+    scale_final(y, _y);
+    for (int i = 0; i < 8; i++)
+        vst1q_s16(&_x[i * 8], veorq_s16(x[0], x[0]));
+}
+
+void oc_idct8x8_neon(int16_t _y[64],int16_t _x[64],int _last_zzi) {
+    if (_last_zzi <= 10)
+        oc_idct8x8_4(_y, _x);
+    else
+        oc_idct8x8_8(_y, _x);
+}
+#endif

lib/theoratest/lib/theora/arm-intrinsics/armint.h

@@ -0,0 +1,47 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2010                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id: x86int.h 17344 2010-07-21 01:42:18Z tterribe $
+
+ ********************************************************************/
+#if !defined(_arm_armint_H)
+# define _arm_armint_H (1)
+# include "../internal.h"
+
+# if defined(OC_ARM_ASM)
+#  define oc_state_accel_init oc_state_accel_init_arm
+#  if defined(__aarch64__) || defined(_M_ARM64)
+#   define oc_loop_filter_init(_state,...) oc_loop_filter_init_neon(__VA_ARGS__)
+#   define oc_state_loop_filter_frag_rows(...) oc_state_loop_filter_frag_rows_neon(__VA_ARGS__)
+#   define oc_frag_copy(_state,...) oc_frag_copy_neon(__VA_ARGS__)
+#   define oc_frag_copy_list(_state,...) oc_frag_copy_list_neon(__VA_ARGS__)
+
+#   define oc_state_frag_recon oc_state_frag_recon_neon
+#   define oc_idct8x8(_state,...) oc_idct8x8_neon(__VA_ARGS__)
+#   define oc_frag_recon_intra(_state,...) oc_frag_recon_intra_neon(__VA_ARGS__)
+#   define oc_frag_recon_inter(_state,...) oc_frag_recon_inter_neon(__VA_ARGS__)
+#   define oc_frag_recon_inter2(_state,...) oc_frag_recon_inter2_neon(__VA_ARGS__)
+#  else
+#   define OC_STATE_USE_VTABLE (1)
+#  endif
+# endif
+
+# include "../state.h"
+# include "armcpu.h"
+
+# if defined(OC_ARM_ASM)
+#  include "armfrag.h"
+void oc_state_accel_init_arm(oc_theora_state *_state);
+# endif
+
+#endif

lib/theoratest/lib/theora/arm-intrinsics/armloop.c

@@ -0,0 +1,130 @@
+#include "armint.h"
+
+#if defined(OC_ARM_ASM_NEON)
+#include <stddef.h>
+#include <string.h>
+#include <arm_neon.h>
+
+static inline uint8x8x2_t loop_filter_neon(uint8x8x4_t pix, uint16_t lim2u) {
+    uint16x8_t lim2 = vdupq_n_u16(lim2u);
+
+    int16x8_t r2 = vreinterpretq_s16_u16(vsubl_u8(pix.val[2], pix.val[1]));
+    int16x8_t r = vreinterpretq_s16_u16(vsubl_u8(pix.val[0], pix.val[3]));
+    r = vaddq_s16(r, r2);
+    r = vaddq_s16(r, vshlq_n_s16(r2, 1));
+    r = vrshrq_n_s16(r, 3);
+
+    uint16x8_t absR = vreinterpretq_u16_s16(vabsq_s16(r));
+    int16x8_t sgnR = vshrq_n_s16(r, 15);
+
+    uint16x8_t absF = vminq_u16(absR, vqsubq_u16(lim2, absR)); // |f| = MIN(|R|,MAX(2L-|R|, 0));
+    int16x8_t f = veorq_s16(vaddq_s16(sgnR, vreinterpretq_s16_u16(absF)), sgnR);
+
+    uint16x8_t p1 = vaddw_u8(vreinterpretq_u16_s16(f), pix.val[1]);
+    int16x8_t p2 = vsubq_s16(vreinterpretq_s16_u16(vmovl_u8(pix.val[2])), f);
+
+    uint8x8x2_t fr = {{vqmovun_s16(vreinterpretq_s16_u16(p1)), vqmovun_s16(p2)}};
+    return fr;
+}
+
+static void loop_filter_neon_h(unsigned char *_pix,int _ystride,uint16_t lim2) {
+    _pix -= 2;
+    uint8x8x4_t cols = {};
+    cols = vld4_lane_u8(&_pix[_ystride * 0], cols, 0);
+    cols = vld4_lane_u8(&_pix[_ystride * 1], cols, 1);
+    cols = vld4_lane_u8(&_pix[_ystride * 2], cols, 2);
+    cols = vld4_lane_u8(&_pix[_ystride * 3], cols, 3);
+    cols = vld4_lane_u8(&_pix[_ystride * 4], cols, 4);
+    cols = vld4_lane_u8(&_pix[_ystride * 5], cols, 5);
+    cols = vld4_lane_u8(&_pix[_ystride * 6], cols, 6);
+    cols = vld4_lane_u8(&_pix[_ystride * 7], cols, 7);
+
+    uint8x8x2_t fp = loop_filter_neon(cols, lim2);
+
+    _pix += 1;
+    vst2_lane_u8(&_pix[_ystride * 0], fp, 0);
+    vst2_lane_u8(&_pix[_ystride * 1], fp, 1);
+    vst2_lane_u8(&_pix[_ystride * 2], fp, 2);
+    vst2_lane_u8(&_pix[_ystride * 3], fp, 3);
+    vst2_lane_u8(&_pix[_ystride * 4], fp, 4);
+    vst2_lane_u8(&_pix[_ystride * 5], fp, 5);
+    vst2_lane_u8(&_pix[_ystride * 6], fp, 6);
+    vst2_lane_u8(&_pix[_ystride * 7], fp, 7);
+}
+
+static void loop_filter_neon_v(unsigned char *_pix,int _ystride,uint16_t lim2) {
+    _pix -= _ystride * 2;
+    uint8x8x4_t rows = {{
+            vld1_u8(&_pix[_ystride * 0]),
+            vld1_u8(&_pix[_ystride * 1]),
+            vld1_u8(&_pix[_ystride * 2]),
+            vld1_u8(&_pix[_ystride * 3])
+    }};
+
+    uint8x8x2_t fp = loop_filter_neon(rows, lim2);
+
+    vst1_u8(&_pix[_ystride * 1], fp.val[0]);
+    vst1_u8(&_pix[_ystride * 2], fp.val[1]);
+}
+
+/*Direct copy of the generic functions,
+  other than _bv being used to store 2*flimit */
+
+void oc_loop_filter_init_neon(signed char _bv[256],int _flimit) {
+    uint16_t lim2u = _flimit * 2;
+    memcpy(_bv, &lim2u, sizeof(uint16_t));
+}
+
+void oc_state_loop_filter_frag_rows_neon(const oc_theora_state *_state,
+                                         signed char *_bv,int _refi,int _pli,int _fragy0,int _fragy_end) {
+    const oc_fragment_plane *fplane;
+    const oc_fragment *frags;
+    const ptrdiff_t *frag_buf_offs;
+    unsigned char *ref_frame_data;
+    ptrdiff_t fragi_top;
+    ptrdiff_t fragi_bot;
+    ptrdiff_t fragi0;
+    ptrdiff_t fragi0_end;
+    int ystride;
+    int nhfrags;
+    uint16_t lim2;
+    fplane = _state->fplanes + _pli;
+    nhfrags = fplane->nhfrags;
+    fragi_top = fplane->froffset;
+    fragi_bot = fragi_top + fplane->nfrags;
+    fragi0 = fragi_top + _fragy0 * (ptrdiff_t) nhfrags;
+    fragi0_end = fragi_top + _fragy_end * (ptrdiff_t) nhfrags;
+    ystride = _state->ref_ystride[_pli];
+    frags = _state->frags;
+    frag_buf_offs = _state->frag_buf_offs;
+    ref_frame_data = _state->ref_frame_data[_refi];
+    memcpy(&lim2, _bv, sizeof(uint16_t));
+    /*The following loops are constructed somewhat non-intuitively on purpose.
+      The main idea is: if a block boundary has at least one coded fragment on
+       it, the filter is applied to it.
+      However, the order that the filters are applied in matters, and VP3 chose
+       the somewhat strange ordering used below.*/
+    while (fragi0 < fragi0_end) {
+        ptrdiff_t fragi;
+        ptrdiff_t fragi_end;
+        fragi = fragi0;
+        fragi_end = fragi + nhfrags;
+        while (fragi < fragi_end) {
+            if (frags[fragi].coded) {
+                unsigned char *ref;
+                ref = ref_frame_data + frag_buf_offs[fragi];
+                if (fragi > fragi0)loop_filter_neon_h(ref, ystride, lim2);
+                if (fragi0 > fragi_top)loop_filter_neon_v(ref, ystride, lim2);
+                if (fragi + 1 < fragi_end && !frags[fragi + 1].coded) {
+                    loop_filter_neon_h(ref + 8, ystride, lim2);
+                }
+                if (fragi + nhfrags < fragi_bot && !frags[fragi + nhfrags].coded) {
+                    loop_filter_neon_v(ref + (ystride << 3), ystride, lim2);
+                }
+            }
+            fragi++;
+        }
+        fragi0 += nhfrags;
+    }
+}
+#endif

lib/theoratest/lib/theora/arm-intrinsics/armstate.c

@@ -0,0 +1,42 @@
+#include "armint.h"
+
+static const unsigned char OC_FZIG_ZAG_NEON[128]={
+   0, 8, 1, 2, 9,16,24,17,
+  10, 3, 4,11,18,25,32,40,
+  33,26,19,12, 5, 6,13,20,
+  27,34,41,48,56,49,42,35,
+  28,21,14, 7,15,22,29,36,
+  43,50,57,58,51,44,37,30,
+  23,31,38,45,52,59,60,53,
+  46,39,47,54,61,62,55,63,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64
+};
+
+void oc_state_accel_init_arm(oc_theora_state *_state){
+  oc_state_accel_init_c(_state);
+  _state->cpu_flags=oc_cpu_flags_get();
+#if defined(OC_ARM_ASM_NEON)
+  if(_state->cpu_flags & OC_CPU_ARM_NEON){
+# if defined(OC_STATE_USE_VTABLE)
+    _state->opt_vtable.loop_filter_init=oc_loop_filter_init_neon;
+    _state->opt_vtable.state_loop_filter_frag_rows=oc_state_loop_filter_frag_rows_neon;
+    _state->opt_vtable.frag_copy=oc_frag_copy_neon;
+    _state->opt_vtable.frag_copy_list=oc_frag_copy_list_neon;
+
+    _state->opt_vtable.state_frag_recon=oc_state_frag_recon_neon;
+    _state->opt_vtable.idct8x8=oc_idct8x8_neon;
+    _state->opt_vtable.frag_recon_intra=oc_frag_recon_intra_neon;
+    _state->opt_vtable.frag_recon_inter=oc_frag_recon_inter_neon;
+    _state->opt_vtable.frag_recon_inter2=oc_frag_recon_inter2_neon;
+# endif
+    _state->opt_data.dct_fzig_zag=OC_FZIG_ZAG_NEON;
+  }
+#endif
+}

lib/theoratest/lib/theora/arm-intrinsics/neon_a64_compat.h

@@ -0,0 +1,34 @@
+#ifndef _neon_a64_compat_H
+#define _neon_a64_compat_H
+#include <arm_neon.h>
+
+#if !defined(__aarch64__) && !defined(_M_ARM64)
+# define vuzp1q_s64(a, b)  vcombine_s64(vget_low_s64(a), vget_low_s64(b))
+# define vuzp2q_s64(a, b)  vcombine_s64(vget_high_s64(a), vget_high_s64(b))
+# define vmovl_high_s16(a) vmovl_s16(vget_high_s16(a))
+# define vuzp2q_s16(a, b)  vuzpq_s16(a, b).val[1]
+
+#define vaddv_s32(p) vget_lane_s32(vpadd_s32(p, p), 0)
+
+static inline int32_t vaddvq_s32(int32x4_t p) {
+    int32x2_t r = vadd_s32(vget_high_s32(p), vget_low_s32(p));
+    return vaddv_s32(r);
+}
+
+static inline int32_t vaddlvq_s16(int16x8_t p) {
+    return vaddvq_s32(vpaddlq_s16(p));
+}
+
+#define vaddv_u32(p) vget_lane_u32(vpadd_u32(p, p), 0)
+
+static inline uint32_t vaddvq_u32(uint32x4_t p) {
+    uint32x2_t r = vadd_u32(vget_high_u32(p), vget_low_u32(p));
+    return vaddv_u32(r);
+}
+
+static inline uint32_t vaddlvq_u16(uint16x8_t p) {
+    return vaddvq_u32(vpaddlq_u16(p));
+}
+#endif
+
+#endif //_neon_a64_compat_H

lib/theoratest/lib/theora/arm-intrinsics/neon_transpose.h

@@ -0,0 +1,59 @@
+#ifndef _arm_neon_transpose_H
+#define _arm_neon_transpose_H
+#include "neon_a64_compat.h"
+
+static void v_transpose8x8(const int16x8_t a[8], int16x8_t b[8]) {
+    int16x8x2_t s0 = vtrnq_s16(a[0], a[1]);
+    int16x8x2_t s1 = vtrnq_s16(a[2], a[3]);
+    int16x8x2_t s2 = vtrnq_s16(a[4], a[5]);
+    int16x8x2_t s3 = vtrnq_s16(a[6], a[7]);
+    /*  1  9  3 11  5 13  7 15  s0[0] = VTRN1.1N p0 p1 */
+    /*  2 10  4 12  6 14  8 16  s0[1] = VTRN2.1N p0 p1 */
+    /* 17 25 19 27 21 29 23 31  s1[0] = VTRN1.1N p2 p3 */
+    /* 18 26 20 28 22 30 24 32  s1[1] = VTRN2.1N p2 p3 */
+    /* 33 41 35 43 37 45 39 47  s2[0] = VTRN1.1N p4 p5 */
+    /* 34 42 36 44 38 46 40 48  s2[1] = VTRN2.1N p4 p5 */
+    /* 49 57 51 59 53 61 55 63  s3[0] = VTRN1.1N p6 p7 */
+    /* 50 58 52 60 54 62 56 64  s3[1] = VTRN2.1N p6 p7 */
+
+    int32x4x2_t t0 = vtrnq_s32(vreinterpretq_s32_s16(s0.val[0]), vreinterpretq_s32_s16(s1.val[0]));
+    int32x4x2_t t1 = vtrnq_s32(vreinterpretq_s32_s16(s0.val[1]), vreinterpretq_s32_s16(s1.val[1]));
+    int32x4x2_t t2 = vtrnq_s32(vreinterpretq_s32_s16(s2.val[0]), vreinterpretq_s32_s16(s3.val[0]));
+    int32x4x2_t t3 = vtrnq_s32(vreinterpretq_s32_s16(s2.val[1]), vreinterpretq_s32_s16(s3.val[1]));
+    /*  1  9 17 25  5 13 21 29  t0[0] = VTRN1.2N s0[0] s1[0] */
+    /*  2 10 18 26  6 14 22 30  t1[0] = VTRN1.2N s0[1] s1[1] */
+    /*  3 11 19 27  7 15 23 31  t0[1] = VTRN2.2N s0[0] s1[0] */
+    /*  4 12 20 28  8 16 24 32  t1[1] = VTRN2.2N s0[1] s1[1] */
+    /* 33 41 49 57 37 45 53 61  t2[0] = VTRN1.2N s2[0] s3[0] */
+    /* 34 42 50 58 38 46 54 62  t3[0] = VTRN1.2N s2[1] s3[1] */
+    /* 35 43 51 59 39 47 55 63  t2[1] = VTRN2.2N s2[0] s3[0] */
+    /* 36 44 52 60 40 48 56 64  t3[1] = VTRN2.2N s2[1] s3[1] */
+
+    int64x2_t u0 = vuzp1q_s64(vreinterpretq_s64_s32(t0.val[0]), vreinterpretq_s64_s32(t2.val[0]));
+    int64x2_t u1 = vuzp1q_s64(vreinterpretq_s64_s32(t1.val[0]), vreinterpretq_s64_s32(t3.val[0]));
+    int64x2_t u2 = vuzp1q_s64(vreinterpretq_s64_s32(t0.val[1]), vreinterpretq_s64_s32(t2.val[1]));
+    int64x2_t u3 = vuzp1q_s64(vreinterpretq_s64_s32(t1.val[1]), vreinterpretq_s64_s32(t3.val[1]));
+    int64x2_t u4 = vuzp2q_s64(vreinterpretq_s64_s32(t0.val[0]), vreinterpretq_s64_s32(t2.val[0]));
+    int64x2_t u5 = vuzp2q_s64(vreinterpretq_s64_s32(t1.val[0]), vreinterpretq_s64_s32(t3.val[0]));
+    int64x2_t u6 = vuzp2q_s64(vreinterpretq_s64_s32(t0.val[1]), vreinterpretq_s64_s32(t2.val[1]));
+    int64x2_t u7 = vuzp2q_s64(vreinterpretq_s64_s32(t1.val[1]), vreinterpretq_s64_s32(t3.val[1]));
+    /*  1  9 17 25 33 41 49 57  u0 = VUZP1 t0[0] t2[0] */
+    /*  2 10 18 26 34 42 50 58  u1 = VUZP1 t1[0] t3[0] */
+    /*  3 11 19 27 35 43 51 59  u2 = VUZP1 t0[1] t2[1] */
+    /*  4 12 20 28 36 44 52 60  u3 = VUZP1 t1[1] t3[1] */
+    /*  5 13 21 29 37 45 53 61  u4 = VUZP2 t0[0] t2[0] */
+    /*  6 14 22 30 38 46 54 62  u5 = VUZP2 t1[0] t3[0] */
+    /*  7 15 23 31 39 47 55 63  u6 = VUZP2 t0[1] t2[1] */
+    /*  8 16 24 32 40 48 56 64  u7 = VUZP2 t1[1] t3[1] */
+
+    b[0] = vreinterpretq_s16_s64(u0);
+    b[1] = vreinterpretq_s16_s64(u1);
+    b[2] = vreinterpretq_s16_s64(u2);
+    b[3] = vreinterpretq_s16_s64(u3);
+    b[4] = vreinterpretq_s16_s64(u4);
+    b[5] = vreinterpretq_s16_s64(u5);
+    b[6] = vreinterpretq_s16_s64(u6);
+    b[7] = vreinterpretq_s16_s64(u7);
+}
+
+#endif

lib/theoratest/lib/theora/bitpack.c

@@ -0,0 +1,114 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggTheora SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function: packing variable sized words into an octet stream
+  last mod: $Id$
+
+ ********************************************************************/
+#include <string.h>
+#include <stdlib.h>
+#include "bitpack.h"
+
+/*We're 'MSb' endian; if we write a word but read individual bits,
+   then we'll read the MSb first.*/
+
+void oc_pack_readinit(oc_pack_buf *_b,unsigned char *_buf,long _bytes){
+  memset(_b,0,sizeof(*_b));
+  _b->ptr=_buf;
+  _b->stop=_buf+_bytes;
+}
+
+static oc_pb_window oc_pack_refill(oc_pack_buf *_b,int _bits){
+  const unsigned char *ptr;
+  const unsigned char *stop;
+  oc_pb_window         window;
+  int                  available;
+  unsigned             shift;
+  stop=_b->stop;
+  ptr=_b->ptr;
+  window=_b->window;
+  available=_b->bits;
+  shift=OC_PB_WINDOW_SIZE-available;
+  while(7<shift&&ptr<stop){
+    shift-=8;
+    window|=(oc_pb_window)*ptr++<<shift;
+  }
+  _b->ptr=ptr;
+  available=OC_PB_WINDOW_SIZE-shift;
+  if(_bits>available){
+    if(ptr>=stop){
+      _b->eof=1;
+      available=OC_LOTS_OF_BITS;
+    }
+    else window|=*ptr>>(available&7);
+  }
+  _b->bits=available;
+  return window;
+}
+
+int oc_pack_look1(oc_pack_buf *_b){
+  oc_pb_window window;
+  int          available;
+  window=_b->window;
+  available=_b->bits;
+  if(available<1)_b->window=window=oc_pack_refill(_b,1);
+  return window>>OC_PB_WINDOW_SIZE-1;
+}
+
+void oc_pack_adv1(oc_pack_buf *_b){
+  _b->window<<=1;
+  _b->bits--;
+}
+
+/*Here we assume that 0<=_bits&&_bits<=32.*/
+long oc_pack_read_c(oc_pack_buf *_b,int _bits){
+  oc_pb_window window;
+  int          available;
+  long         result;
+  window=_b->window;
+  available=_b->bits;
+  if(_bits==0)return 0;
+  if(available<_bits){
+    window=oc_pack_refill(_b,_bits);
+    available=_b->bits;
+  }
+  result=window>>OC_PB_WINDOW_SIZE-_bits;
+  available-=_bits;
+  window<<=1;
+  window<<=_bits-1;
+  _b->window=window;
+  _b->bits=available;
+  return result;
+}
+
+int oc_pack_read1_c(oc_pack_buf *_b){
+  oc_pb_window window;
+  int          available;
+  int          result;
+  window=_b->window;
+  available=_b->bits;
+  if(available<1){
+    window=oc_pack_refill(_b,1);
+    available=_b->bits;
+  }
+  result=window>>OC_PB_WINDOW_SIZE-1;
+  available--;
+  window<<=1;
+  _b->window=window;
+  _b->bits=available;
+  return result;
+}
+
+long oc_pack_bytes_left(oc_pack_buf *_b){
+  if(_b->eof)return -1;
+  return _b->stop-_b->ptr+(_b->bits>>3);
+}

lib/theoratest/lib/theora/bitpack.h

@@ -0,0 +1,71 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggTheora SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function: packing variable sized words into an octet stream
+  last mod: $Id: bitwise.c 7675 2004-09-01 00:34:39Z xiphmont $
+
+ ********************************************************************/
+#if !defined(_bitpack_H)
+# define _bitpack_H (1)
+# include <stddef.h>
+# include <limits.h>
+# include "internal.h"
+
+
+
+typedef size_t             oc_pb_window;
+typedef struct oc_pack_buf oc_pack_buf;
+
+
+
+# if !defined(oc_pack_read)
+#  define oc_pack_read oc_pack_read_c
+# endif
+# if !defined(oc_pack_read1)
+#  define oc_pack_read1 oc_pack_read1_c
+# endif
+# if !defined(oc_huff_token_decode)
+#  define oc_huff_token_decode oc_huff_token_decode_c
+# endif
+
+# define OC_PB_WINDOW_SIZE ((int)sizeof(oc_pb_window)*CHAR_BIT)
+/*This is meant to be a large, positive constant that can still be efficiently
+   loaded as an immediate (on platforms like ARM, for example).
+  Even relatively modest values like 100 would work fine.*/
+# define OC_LOTS_OF_BITS (0x40000000)
+
+
+
+struct oc_pack_buf{
+  const unsigned char *stop;
+  const unsigned char *ptr;
+  oc_pb_window         window;
+  int                  bits;
+  int                  eof;
+};
+
+void oc_pack_readinit(oc_pack_buf *_b,unsigned char *_buf,long _bytes);
+int oc_pack_look1(oc_pack_buf *_b);
+void oc_pack_adv1(oc_pack_buf *_b);
+/*Here we assume 0<=_bits&&_bits<=32.*/
+long oc_pack_read_c(oc_pack_buf *_b,int _bits);
+int oc_pack_read1_c(oc_pack_buf *_b);
+/* returns -1 for read beyond EOF, or the number of whole bytes available */
+long oc_pack_bytes_left(oc_pack_buf *_b);
+
+/*These two functions are implemented locally in huffdec.c*/
+/*Read in bits without advancing the bitptr.
+  Here we assume 0<=_bits&&_bits<=32.*/
+/*static int oc_pack_look(oc_pack_buf *_b,int _bits);*/
+/*static void oc_pack_adv(oc_pack_buf *_b,int _bits);*/
+
+#endif

lib/theoratest/lib/theora/codec.h

@@ -0,0 +1,606 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+  last mod: $Id: theora.h,v 1.8 2004/03/15 22:17:32 derf Exp $
+
+ ********************************************************************/
+
+/**\mainpage
+ *
+ * \section intro Introduction
+ *
+ * This is the documentation for the <tt>libtheora</tt> C API.
+ *
+ * The \c libtheora package is the current reference
+ * implementation for <a href="http://www.theora.org/">Theora</a>, a free,
+ * patent-unencumbered video codec.
+ * Theora is derived from On2's VP3 codec with additional features and
+ *  integration with Ogg multimedia formats by
+ *  <a href="http://www.xiph.org/">the Xiph.Org Foundation</a>.
+ * Complete documentation of the format itself is available in
+ * <a href="http://www.theora.org/doc/Theora.pdf">the Theora
+ *  specification</a>.
+ *
+ * \section Organization
+ *
+ * The functions documented here are divided between two
+ * separate libraries:
+ * - \c libtheoraenc contains the encoder interface,
+ *   described in \ref encfuncs.
+ * - \c libtheoradec contains the decoder interface,
+ *   described in \ref decfuncs, \n
+ *   and additional \ref basefuncs.
+ *
+ * New code should link to \c libtheoradec. If using encoder
+ * features, it must also link to \c libtheoraenc.
+ *
+ * During initial development, prior to the 1.0 release,
+ * \c libtheora exported a different \ref oldfuncs which
+ * combined both encode and decode functions.
+ * In general, legacy API symbols can be indentified
+ * by their \c theora_ or \c OC_ namespace prefixes.
+ * The current API uses \c th_ or \c TH_ instead.
+ *
+ * While deprecated, \c libtheoraenc and \c libtheoradec
+ * together export the legacy api as well at the one documented above.
+ * Likewise, the legacy \c libtheora included with this package
+ * exports the new 1.x API. Older code and build scripts can therefore
+ * but updated independently to the current scheme.
+ */
+
+/**\file
+ * The shared <tt>libtheoradec</tt> and <tt>libtheoraenc</tt> C API.
+ * You don't need to include this directly.*/
+
+#if !defined(_O_THEORA_CODEC_H_)
+# define _O_THEORA_CODEC_H_ (1)
+# include <ogg/ogg.h>
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+
+
+/**\name Return codes*/
+/*@{*/
+/**An invalid pointer was provided.*/
+#define TH_EFAULT     (-1)
+/**An invalid argument was provided.*/
+#define TH_EINVAL     (-10)
+/**The contents of the header were incomplete, invalid, or unexpected.*/
+#define TH_EBADHEADER (-20)
+/**The header does not belong to a Theora stream.*/
+#define TH_ENOTFORMAT (-21)
+/**The bitstream version is too high.*/
+#define TH_EVERSION   (-22)
+/**The specified function is not implemented.*/
+#define TH_EIMPL      (-23)
+/**There were errors in the video data packet.*/
+#define TH_EBADPACKET (-24)
+/**The decoded packet represented a dropped frame.
+   The player can continue to display the current frame, as the contents of the
+    decoded frame buffer have not changed.*/
+#define TH_DUPFRAME   (1)
+/*@}*/
+
+/**The currently defined color space tags.
+ * See <a href="http://www.theora.org/doc/Theora.pdf">the Theora
+ *  specification</a>, Chapter 4, for exact details on the meaning
+ *  of each of these color spaces.*/
+typedef enum{
+  /**The color space was not specified at the encoder.
+      It may be conveyed by an external means.*/
+  TH_CS_UNSPECIFIED,
+  /**A color space designed for NTSC content.*/
+  TH_CS_ITU_REC_470M,
+  /**A color space designed for PAL/SECAM content.*/
+  TH_CS_ITU_REC_470BG,
+  /**The total number of currently defined color spaces.*/
+  TH_CS_NSPACES
+}th_colorspace;
+
+/**The currently defined pixel format tags.
+ * See <a href="http://www.theora.org/doc/Theora.pdf">the Theora
+ *  specification</a>, Section 4.4, for details on the precise sample
+ *  locations.*/
+typedef enum{
+  /**Chroma decimation by 2 in both the X and Y directions (4:2:0).
+     The Cb and Cr chroma planes are half the width and half the
+      height of the luma plane.*/
+  TH_PF_420,
+  /**Currently reserved.*/
+  TH_PF_RSVD,
+  /**Chroma decimation by 2 in the X direction (4:2:2).
+     The Cb and Cr chroma planes are half the width of the luma plane, but full
+      height.*/
+  TH_PF_422,
+  /**No chroma decimation (4:4:4).
+     The Cb and Cr chroma planes are full width and full height.*/
+  TH_PF_444,
+  /**The total number of currently defined pixel formats.*/
+  TH_PF_NFORMATS
+}th_pixel_fmt;
+
+
+
+/**A buffer for a single color plane in an uncompressed image.
+ * This contains the image data in a left-to-right, top-down format.
+ * Each row of pixels is stored contiguously in memory, but successive
+ *  rows need not be.
+ * Use \a stride to compute the offset of the next row.
+ * The encoder accepts both positive \a stride values (top-down in memory)
+ *  and negative (bottom-up in memory).
+ * The decoder currently always generates images with positive strides.*/
+typedef struct{
+  /**The width of this plane.*/
+  int            width;
+  /**The height of this plane.*/
+  int            height;
+  /**The offset in bytes between successive rows.*/
+  int            stride;
+  /**A pointer to the beginning of the first row.*/
+  unsigned char *data;
+}th_img_plane;
+
+/**A complete image buffer for an uncompressed frame.
+ * The chroma planes may be decimated by a factor of two in either
+ *  direction, as indicated by th_info#pixel_fmt.
+ * The width and height of the Y' plane must be multiples of 16.
+ * They may need to be cropped for display, using the rectangle
+ *  specified by th_info#pic_x, th_info#pic_y, th_info#pic_width,
+ *  and th_info#pic_height.
+ * All samples are 8 bits.
+ * \note The term YUV often used to describe a colorspace is ambiguous.
+ * The exact parameters of the RGB to YUV conversion process aside, in
+ *  many contexts the U and V channels actually have opposite meanings.
+ * To avoid this confusion, we are explicit: the name of the color
+ *  channels are Y'CbCr, and they appear in that order, always.
+ * The prime symbol denotes that the Y channel is non-linear.
+ * Cb and Cr stand for "Chroma blue" and "Chroma red", respectively.*/
+typedef th_img_plane th_ycbcr_buffer[3];
+
+/**Theora bitstream information.
+ * This contains the basic playback parameters for a stream, and corresponds to
+ *  the initial 'info' header packet.
+ * To initialize an encoder, the application fills in this structure and
+ *  passes it to th_encode_alloc().
+ * A default encoding mode is chosen based on the values of the #quality and
+ *  #target_bitrate fields.
+ * On decode, it is filled in by th_decode_headerin(), and then passed to
+ *  th_decode_alloc().
+ *
+ * Encoded Theora frames must be a multiple of 16 in size;
+ *  this is what the #frame_width and #frame_height members represent.
+ * To handle arbitrary picture sizes, a crop rectangle is specified in the
+ *  #pic_x, #pic_y, #pic_width and #pic_height members.
+ *
+ * All frame buffers contain pointers to the full, padded frame.
+ * However, the current encoder <em>will not</em> reference pixels outside of
+ *  the cropped picture region, and the application does not need to fill them
+ *  in.
+ * The decoder <em>will</em> allocate storage for a full frame, but the
+ *  application <em>should not</em> rely on the padding containing sensible
+ *  data.
+ *
+ * It is also generally recommended that the offsets and sizes should still be
+ *  multiples of 2 to avoid chroma sampling shifts when chroma is sub-sampled.
+ * See <a href="http://www.theora.org/doc/Theora.pdf">the Theora
+ *  specification</a>, Section 4.4, for more details.
+ *
+ * Frame rate, in frames per second, is stored as a rational fraction, as is
+ *  the pixel aspect ratio.
+ * Note that this refers to the aspect ratio of the individual pixels, not of
+ *  the overall frame itself.
+ * The frame aspect ratio can be computed from pixel aspect ratio using the
+ *  image dimensions.*/
+typedef struct{
+  /**\name Theora version
+   * Bitstream version information.*/
+  /*@{*/
+  unsigned char version_major;
+  unsigned char version_minor;
+  unsigned char version_subminor;
+  /*@}*/
+  /**The encoded frame width.
+   * This must be a multiple of 16, and less than 1048576.*/
+  ogg_uint32_t  frame_width;
+  /**The encoded frame height.
+   * This must be a multiple of 16, and less than 1048576.*/
+  ogg_uint32_t  frame_height;
+  /**The displayed picture width.
+   * This must be no larger than width.*/
+  ogg_uint32_t  pic_width;
+  /**The displayed picture height.
+   * This must be no larger than height.*/
+  ogg_uint32_t  pic_height;
+  /**The X offset of the displayed picture.
+   * This must be no larger than #frame_width-#pic_width or 255, whichever is
+   *  smaller.*/
+  ogg_uint32_t  pic_x;
+  /**The Y offset of the displayed picture.
+   * This must be no larger than #frame_height-#pic_height, and
+   *  #frame_height-#pic_height-#pic_y must be no larger than 255.
+   * This slightly funny restriction is due to the fact that the offset is
+   *  specified from the top of the image for consistency with the standard
+   *  graphics left-handed coordinate system used throughout this API, while
+   *  it is stored in the encoded stream as an offset from the bottom.*/
+  ogg_uint32_t  pic_y;
+  /**\name Frame rate
+   * The frame rate, as a fraction.
+   * If either is 0, the frame rate is undefined.*/
+  /*@{*/
+  ogg_uint32_t  fps_numerator;
+  ogg_uint32_t  fps_denominator;
+  /*@}*/
+  /**\name Aspect ratio
+   * The aspect ratio of the pixels.
+   * If either value is zero, the aspect ratio is undefined.
+   * If not specified by any external means, 1:1 should be assumed.
+   * The aspect ratio of the full picture can be computed as
+   * \code
+   *  aspect_numerator*pic_width/(aspect_denominator*pic_height).
+   * \endcode */
+  /*@{*/
+  ogg_uint32_t  aspect_numerator;
+  ogg_uint32_t  aspect_denominator;
+  /*@}*/
+  /**The color space.*/
+  th_colorspace colorspace;
+  /**The pixel format.*/
+  th_pixel_fmt  pixel_fmt;
+  /**The target bit-rate in bits per second.
+     If initializing an encoder with this struct, set this field to a non-zero
+      value to activate CBR encoding by default.*/
+  int           target_bitrate;
+  /**The target quality level.
+     Valid values range from 0 to 63, inclusive, with higher values giving
+      higher quality.
+     If initializing an encoder with this struct, and #target_bitrate is set
+      to zero, VBR encoding at this quality will be activated by default.*/
+  /*Currently this is set so that a qi of 0 corresponds to distortions of 24
+     times the JND, and each increase by 16 halves that value.
+    This gives us fine discrimination at low qualities, yet effective rate
+     control at high qualities.
+    The qi value 63 is special, however.
+    For this, the highest quality, we use one half of a JND for our threshold.
+    Due to the lower bounds placed on allowable quantizers in Theora, we will
+     not actually be able to achieve quality this good, but this should
+     provide as close to visually lossless quality as Theora is capable of.
+    We could lift the quantizer restrictions without breaking VP3.1
+     compatibility, but this would result in quantized coefficients that are
+     too large for the current bitstream to be able to store.
+    We'd have to redesign the token syntax to store these large coefficients,
+     which would make transcoding complex.*/
+  int           quality;
+  /**The amount to shift to extract the last keyframe number from the granule
+   *  position.
+   * This can be at most 31.
+   * th_info_init() will set this to a default value (currently <tt>6</tt>,
+   *  which is good for streaming applications), but you can set it to 0 to
+   *  make every frame a keyframe.
+   * The maximum distance between key frames is
+   *  <tt>1<<#keyframe_granule_shift</tt>.
+   * The keyframe frequency can be more finely controlled with
+   *  #TH_ENCCTL_SET_KEYFRAME_FREQUENCY_FORCE, which can also be adjusted
+   *  during encoding (for example, to force the next frame to be a keyframe),
+   *  but it cannot be set larger than the amount permitted by this field after
+   *  the headers have been output.*/
+  int           keyframe_granule_shift;
+}th_info;
+
+/**The comment information.
+ *
+ * This structure holds the in-stream metadata corresponding to
+ *  the 'comment' header packet.
+ * The comment header is meant to be used much like someone jotting a quick
+ *  note on the label of a video.
+ * It should be a short, to the point text note that can be more than a couple
+ *  words, but not more than a short paragraph.
+ *
+ * The metadata is stored as a series of (tag, value) pairs, in
+ *  length-encoded string vectors.
+ * The first occurrence of the '=' character delimits the tag and value.
+ * A particular tag may occur more than once, and order is significant.
+ * The character set encoding for the strings is always UTF-8, but the tag
+ *  names are limited to ASCII, and treated as case-insensitive.
+ * See <a href="http://www.theora.org/doc/Theora.pdf">the Theora
+ *  specification</a>, Section 6.3.3 for details.
+ *
+ * In filling in this structure, th_decode_headerin() will null-terminate
+ *  the user_comment strings for safety.
+ * However, the bitstream format itself treats them as 8-bit clean vectors,
+ *  possibly containing null characters, so the length array should be
+ *  treated as their authoritative length.
+ */
+typedef struct th_comment{
+  /**The array of comment string vectors.*/
+  char **user_comments;
+  /**An array of the corresponding length of each vector, in bytes.*/
+  int   *comment_lengths;
+  /**The total number of comment strings.*/
+  int    comments;
+  /**The null-terminated vendor string.
+     This identifies the software used to encode the stream.*/
+  char  *vendor;
+}th_comment;
+
+
+
+/**A single base matrix.*/
+typedef unsigned char th_quant_base[64];
+
+/**A set of \a qi ranges.*/
+typedef struct{
+  /**The number of ranges in the set.*/
+  int                  nranges;
+  /**The size of each of the #nranges ranges.
+     These must sum to 63.*/
+  const int           *sizes;
+  /**#nranges <tt>+1</tt> base matrices.
+     Matrices \a i and <tt>i+1</tt> form the endpoints of range \a i.*/
+  const th_quant_base *base_matrices;
+}th_quant_ranges;
+
+/**A complete set of quantization parameters.
+   The quantizer for each coefficient is calculated as:
+   \code
+    Q=MAX(MIN(qmin[qti][ci!=0],scale[ci!=0][qi]*base[qti][pli][qi][ci]/100),
+     1024).
+   \endcode
+
+   \a qti is the quantization type index: 0 for intra, 1 for inter.
+   <tt>ci!=0</tt> is 0 for the DC coefficient and 1 for AC coefficients.
+   \a qi is the quality index, ranging between 0 (low quality) and 63 (high
+    quality).
+   \a pli is the color plane index: 0 for Y', 1 for Cb, 2 for Cr.
+   \a ci is the DCT coefficient index.
+   Coefficient indices correspond to the normal 2D DCT block
+    ordering--row-major with low frequencies first--\em not zig-zag order.
+
+   Minimum quantizers are constant, and are given by:
+   \code
+   qmin[2][2]={{4,2},{8,4}}.
+   \endcode
+
+   Parameters that can be stored in the bitstream are as follows:
+    - The two scale matrices ac_scale and dc_scale.
+      \code
+      scale[2][64]={dc_scale,ac_scale}.
+      \endcode
+    - The base matrices for each \a qi, \a qti and \a pli (up to 384 in all).
+      In order to avoid storing a full 384 base matrices, only a sparse set of
+       matrices are stored, and the rest are linearly interpolated.
+      This is done as follows.
+      For each \a qti and \a pli, a series of \a n \a qi ranges is defined.
+      The size of each \a qi range can vary arbitrarily, but they must sum to
+       63.
+      Then, <tt>n+1</tt> matrices are specified, one for each endpoint of the
+       ranges.
+      For interpolation purposes, each range's endpoints are the first \a qi
+       value it contains and one past the last \a qi value it contains.
+      Fractional values are rounded to the nearest integer, with ties rounded
+       away from zero.
+
+      Base matrices are stored by reference, so if the same matrices are used
+       multiple times, they will only appear once in the bitstream.
+      The bitstream is also capable of omitting an entire set of ranges and
+       its associated matrices if they are the same as either the previous
+       set (indexed in row-major order) or if the inter set is the same as the
+       intra set.
+
+    - Loop filter limit values.
+      The same limits are used for the loop filter in all color planes, despite
+       potentially differing levels of quantization in each.
+
+   For the current encoder, <tt>scale[ci!=0][qi]</tt> must be no greater
+    than <tt>scale[ci!=0][qi-1]</tt> and <tt>base[qti][pli][qi][ci]</tt> must
+    be no greater than <tt>base[qti][pli][qi-1][ci]</tt>.
+   These two conditions ensure that the actual quantizer for a given \a qti,
+    \a pli, and \a ci does not increase as \a qi increases.
+   This is not required by the decoder.*/
+typedef struct{
+  /**The DC scaling factors.*/
+  ogg_uint16_t    dc_scale[64];
+  /**The AC scaling factors.*/
+  ogg_uint16_t    ac_scale[64];
+  /**The loop filter limit values.*/
+  unsigned char   loop_filter_limits[64];
+  /**The \a qi ranges for each \a ci and \a pli.*/
+  th_quant_ranges qi_ranges[2][3];
+}th_quant_info;
+
+
+
+/**The number of Huffman tables used by Theora.*/
+#define TH_NHUFFMAN_TABLES (80)
+/**The number of DCT token values in each table.*/
+#define TH_NDCT_TOKENS     (32)
+
+/**A Huffman code for a Theora DCT token.
+ * Each set of Huffman codes in a given table must form a complete, prefix-free
+ *  code.
+ * There is no requirement that all the tokens in a table have a valid code,
+ *  but the current encoder is not optimized to take advantage of this.
+ * If each of the five grouops of 16 tables does not contain at least one table
+ *  with a code for every token, then the encoder may fail to encode certain
+ *  frames.
+ * The complete table in the first group of 16 does not have to be in the same
+ *  place as the complete table in the other groups, but the complete tables in
+ *  the remaining four groups must all be in the same place.*/
+typedef struct{
+  /**The bit pattern for the code, with the LSbit of the pattern aligned in
+   *   the LSbit of the word.*/
+  ogg_uint32_t pattern;
+  /**The number of bits in the code.
+   * This must be between 0 and 32, inclusive.*/
+  int          nbits;
+}th_huff_code;
+
+
+
+/**\defgroup basefuncs Functions Shared by Encode and Decode*/
+/*@{*/
+/**\name Basic shared functions
+ * These functions return information about the library itself,
+ * or provide high-level information about codec state
+ * and packet type.
+ *
+ * You must link to \c libtheoradec if you use any of the
+ * functions in this section.*/
+/*@{*/
+/**Retrieves a human-readable string to identify the library vendor and
+ *  version.
+ * \return the version string.*/
+extern const char *th_version_string(void);
+/**Retrieves the library version number.
+ * This is the highest bitstream version that the encoder library will produce,
+ *  or that the decoder library can decode.
+ * This number is composed of a 16-bit major version, 8-bit minor version
+ * and 8 bit sub-version, composed as follows:
+ * \code
+ * (VERSION_MAJOR<<16)+(VERSION_MINOR<<8)+(VERSION_SUBMINOR)
+ * \endcode
+ * \return the version number.*/
+extern ogg_uint32_t th_version_number(void);
+/**Converts a granule position to an absolute frame index, starting at
+ *  <tt>0</tt>.
+ * The granule position is interpreted in the context of a given
+ *  #th_enc_ctx or #th_dec_ctx handle (either will suffice).
+ * \param _encdec  A previously allocated #th_enc_ctx or #th_dec_ctx
+ *                  handle.
+ * \param _granpos The granule position to convert.
+ * \returns The absolute frame index corresponding to \a _granpos.
+ * \retval -1 The given granule position was invalid (i.e. negative).*/
+extern ogg_int64_t th_granule_frame(void *_encdec,ogg_int64_t _granpos);
+/**Converts a granule position to an absolute time in seconds.
+ * The granule position is interpreted in the context of a given
+ *  #th_enc_ctx or #th_dec_ctx handle (either will suffice).
+ * \param _encdec  A previously allocated #th_enc_ctx or #th_dec_ctx
+ *                  handle.
+ * \param _granpos The granule position to convert.
+ * \return The absolute time in seconds corresponding to \a _granpos.
+ *         This is the "end time" for the frame, or the latest time it should
+ *          be displayed.
+ *         It is not the presentation time.
+ * \retval -1 The given granule position was invalid (i.e. negative).*/
+extern double th_granule_time(void *_encdec,ogg_int64_t _granpos);
+/**Determines whether a Theora packet is a header or not.
+ * This function does no verification beyond checking the packet type bit, so
+ *  it should not be used for bitstream identification; use
+ *  th_decode_headerin() for that.
+ * As per the Theora specification, an empty (0-byte) packet is treated as a
+ *  data packet (a delta frame with no coded blocks).
+ * \param _op An <tt>ogg_packet</tt> containing encoded Theora data.
+ * \retval 1 The packet is a header packet
+ * \retval 0 The packet is a video data packet.*/
+extern int th_packet_isheader(ogg_packet *_op);
+/**Determines whether a theora packet is a key frame or not.
+ * This function does no verification beyond checking the packet type and
+ *  key frame bits, so it should not be used for bitstream identification; use
+ *  th_decode_headerin() for that.
+ * As per the Theora specification, an empty (0-byte) packet is treated as a
+ *  delta frame (with no coded blocks).
+ * \param _op An <tt>ogg_packet</tt> containing encoded Theora data.
+ * \retval 1  The packet contains a key frame.
+ * \retval 0  The packet contains a delta frame.
+ * \retval -1 The packet is not a video data packet.*/
+extern int th_packet_iskeyframe(ogg_packet *_op);
+/*@}*/
+
+
+/**\name Functions for manipulating header data
+ * These functions manipulate the #th_info and #th_comment structures
+ * which describe video parameters and key-value metadata, respectively.
+ *
+ * You must link to \c libtheoradec if you use any of the
+ * functions in this section.*/
+/*@{*/
+/**Initializes a th_info structure.
+ * This should be called on a freshly allocated #th_info structure before
+ *  attempting to use it.
+ * \param _info The #th_info struct to initialize.*/
+extern void th_info_init(th_info *_info);
+/**Clears a #th_info structure.
+ * This should be called on a #th_info structure after it is no longer
+ *  needed.
+ * \param _info The #th_info struct to clear.*/
+extern void th_info_clear(th_info *_info);
+
+/**Initialize a #th_comment structure.
+ * This should be called on a freshly allocated #th_comment structure
+ *  before attempting to use it.
+ * \param _tc The #th_comment struct to initialize.*/
+extern void th_comment_init(th_comment *_tc);
+/**Add a comment to an initialized #th_comment structure.
+ * \note Neither th_comment_add() nor th_comment_add_tag() support
+ *  comments containing null values, although the bitstream format does
+ *  support them.
+ * To add such comments you will need to manipulate the #th_comment
+ *  structure directly.
+ * \param _tc      The #th_comment struct to add the comment to.
+ * \param _comment Must be a null-terminated UTF-8 string containing the
+ *                  comment in "TAG=the value" form.*/
+extern void th_comment_add(th_comment *_tc,const char *_comment);
+/**Add a comment to an initialized #th_comment structure.
+ * \note Neither th_comment_add() nor th_comment_add_tag() support
+ *  comments containing null values, although the bitstream format does
+ *  support them.
+ * To add such comments you will need to manipulate the #th_comment
+ *  structure directly.
+ * \param _tc  The #th_comment struct to add the comment to.
+ * \param _tag A null-terminated string containing the tag associated with
+ *              the comment.
+ * \param _val The corresponding value as a null-terminated string.*/
+extern void th_comment_add_tag(th_comment *_tc,const char *_tag,
+ const char *_val);
+/**Look up a comment value by its tag.
+ * \param _tc    An initialized #th_comment structure.
+ * \param _tag   The tag to look up.
+ * \param _count The instance of the tag.
+ *               The same tag can appear multiple times, each with a distinct
+ *                value, so an index is required to retrieve them all.
+ *               The order in which these values appear is significant and
+ *                should be preserved.
+ *               Use th_comment_query_count() to get the legal range for
+ *                the \a _count parameter.
+ * \return A pointer to the queried tag's value.
+ *         This points directly to data in the #th_comment structure.
+ *         It should not be modified or freed by the application, and
+ *          modifications to the structure may invalidate the pointer.
+ * \retval NULL If no matching tag is found.*/
+extern char *th_comment_query(th_comment *_tc,const char *_tag,int _count);
+/**Look up the number of instances of a tag.
+ * Call this first when querying for a specific tag and then iterate over the
+ *  number of instances with separate calls to th_comment_query() to
+ *  retrieve all the values for that tag in order.
+ * \param _tc    An initialized #th_comment structure.
+ * \param _tag   The tag to look up.
+ * \return The number of instances of this particular tag.*/
+extern int th_comment_query_count(th_comment *_tc,const char *_tag);
+/**Clears a #th_comment structure.
+ * This should be called on a #th_comment structure after it is no longer
+ *  needed.
+ * It will free all memory used by the structure members.
+ * \param _tc The #th_comment struct to clear.*/
+extern void th_comment_clear(th_comment *_tc);
+/*@}*/
+/*@}*/
+
+
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif

lib/theoratest/lib/theora/dct.h

@@ -0,0 +1,31 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+  last mod: $Id$
+
+ ********************************************************************/
+
+/*Definitions shared by the forward and inverse DCT transforms.*/
+#if !defined(_dct_H)
+# define _dct_H (1)
+
+/*cos(n*pi/16) (resp. sin(m*pi/16)) scaled by 65536.*/
+#define OC_C1S7 ((ogg_int32_t)64277)
+#define OC_C2S6 ((ogg_int32_t)60547)
+#define OC_C3S5 ((ogg_int32_t)54491)
+#define OC_C4S4 ((ogg_int32_t)46341)
+#define OC_C5S3 ((ogg_int32_t)36410)
+#define OC_C6S2 ((ogg_int32_t)25080)
+#define OC_C7S1 ((ogg_int32_t)12785)
+
+#endif

lib/theoratest/lib/theora/decapiwrapper.c

@@ -0,0 +1,193 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id: decapiwrapper.c 13596 2007-08-23 20:05:38Z tterribe $
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <limits.h>
+#include "apiwrapper.h"
+#include "decint.h"
+#include "theora/theoradec.h"
+
+static void th_dec_api_clear(th_api_wrapper *_api){
+  if(_api->setup)th_setup_free(_api->setup);
+  if(_api->decode)th_decode_free(_api->decode);
+  memset(_api,0,sizeof(*_api));
+}
+
+static void theora_decode_clear(theora_state *_td){
+  if(_td->i!=NULL)theora_info_clear(_td->i);
+  memset(_td,0,sizeof(*_td));
+}
+
+static int theora_decode_control(theora_state *_td,int _req,
+ void *_buf,size_t _buf_sz){
+  return th_decode_ctl(((th_api_wrapper *)_td->i->codec_setup)->decode,
+   _req,_buf,_buf_sz);
+}
+
+static ogg_int64_t theora_decode_granule_frame(theora_state *_td,
+ ogg_int64_t _gp){
+  return th_granule_frame(((th_api_wrapper *)_td->i->codec_setup)->decode,_gp);
+}
+
+static double theora_decode_granule_time(theora_state *_td,ogg_int64_t _gp){
+  return th_granule_time(((th_api_wrapper *)_td->i->codec_setup)->decode,_gp);
+}
+
+static const oc_state_dispatch_vtable OC_DEC_DISPATCH_VTBL={
+  (oc_state_clear_func)theora_decode_clear,
+  (oc_state_control_func)theora_decode_control,
+  (oc_state_granule_frame_func)theora_decode_granule_frame,
+  (oc_state_granule_time_func)theora_decode_granule_time,
+};
+
+static void th_info2theora_info(theora_info *_ci,const th_info *_info){
+  _ci->version_major=_info->version_major;
+  _ci->version_minor=_info->version_minor;
+  _ci->version_subminor=_info->version_subminor;
+  _ci->width=_info->frame_width;
+  _ci->height=_info->frame_height;
+  _ci->frame_width=_info->pic_width;
+  _ci->frame_height=_info->pic_height;
+  _ci->offset_x=_info->pic_x;
+  _ci->offset_y=_info->pic_y;
+  _ci->fps_numerator=_info->fps_numerator;
+  _ci->fps_denominator=_info->fps_denominator;
+  _ci->aspect_numerator=_info->aspect_numerator;
+  _ci->aspect_denominator=_info->aspect_denominator;
+  switch(_info->colorspace){
+    case TH_CS_ITU_REC_470M:_ci->colorspace=OC_CS_ITU_REC_470M;break;
+    case TH_CS_ITU_REC_470BG:_ci->colorspace=OC_CS_ITU_REC_470BG;break;
+    default:_ci->colorspace=OC_CS_UNSPECIFIED;break;
+  }
+  switch(_info->pixel_fmt){
+    case TH_PF_420:_ci->pixelformat=OC_PF_420;break;
+    case TH_PF_422:_ci->pixelformat=OC_PF_422;break;
+    case TH_PF_444:_ci->pixelformat=OC_PF_444;break;
+    default:_ci->pixelformat=OC_PF_RSVD;
+  }
+  _ci->target_bitrate=_info->target_bitrate;
+  _ci->quality=_info->quality;
+  _ci->keyframe_frequency_force=1<<_info->keyframe_granule_shift;
+}
+
+int theora_decode_init(theora_state *_td,theora_info *_ci){
+  th_api_info    *apiinfo;
+  th_api_wrapper *api;
+  th_info         info;
+  api=(th_api_wrapper *)_ci->codec_setup;
+  /*Allocate our own combined API wrapper/theora_info struct.
+    We put them both in one malloc'd block so that when the API wrapper is
+     freed, the info struct goes with it.
+    This avoids having to figure out whether or not we need to free the info
+     struct in either theora_info_clear() or theora_clear().*/
+  apiinfo=(th_api_info *)_ogg_calloc(1,sizeof(*apiinfo));
+  if(apiinfo==NULL)return OC_FAULT;
+  /*Make our own copy of the info struct, since its lifetime should be
+     independent of the one we were passed in.*/
+  *&apiinfo->info=*_ci;
+  /*Convert the info struct now instead of saving the the one we decoded with
+     theora_decode_header(), since the user might have modified values (i.e.,
+     color space, aspect ratio, etc. can be specified from a higher level).
+    The user also might be doing something "clever" with the header packets if
+     they are not using an Ogg encapsulation.*/
+  oc_theora_info2th_info(&info,_ci);
+  /*Don't bother to copy the setup info; th_decode_alloc() makes its own copy
+     of the stuff it needs.*/
+  apiinfo->api.decode=th_decode_alloc(&info,api->setup);
+  if(apiinfo->api.decode==NULL){
+    _ogg_free(apiinfo);
+    return OC_EINVAL;
+  }
+  apiinfo->api.clear=(oc_setup_clear_func)th_dec_api_clear;
+  _td->internal_encode=NULL;
+  /*Provide entry points for ABI compatibility with old decoder shared libs.*/
+  _td->internal_decode=(void *)&OC_DEC_DISPATCH_VTBL;
+  _td->granulepos=0;
+  _td->i=&apiinfo->info;
+  _td->i->codec_setup=&apiinfo->api;
+  return 0;
+}
+
+int theora_decode_header(theora_info *_ci,theora_comment *_cc,ogg_packet *_op){
+  th_api_wrapper *api;
+  th_info         info;
+  int             ret;
+  api=(th_api_wrapper *)_ci->codec_setup;
+  /*Allocate an API wrapper struct on demand, since it will not also include a
+     theora_info struct like the ones that are used in a theora_state struct.*/
+  if(api==NULL){
+    _ci->codec_setup=_ogg_calloc(1,sizeof(*api));
+    if(_ci->codec_setup==NULL)return OC_FAULT;
+    api=(th_api_wrapper *)_ci->codec_setup;
+    api->clear=(oc_setup_clear_func)th_dec_api_clear;
+  }
+  /*Convert from the theora_info struct instead of saving our own th_info
+     struct between calls.
+    The user might be doing something "clever" with the header packets if they
+     are not using an Ogg encapsulation, and we don't want to break this.*/
+  oc_theora_info2th_info(&info,_ci);
+  /*We rely on the fact that theora_comment and th_comment structures are
+     actually identical.
+    Take care not to change this fact unless you change the code here as
+     well!*/
+  ret=th_decode_headerin(&info,(th_comment *)_cc,&api->setup,_op);
+  /*We also rely on the fact that the error return code values are the same,
+    and that the implementations of these two functions return the same set of
+    them.
+   Note that theora_decode_header() really can return OC_NOTFORMAT, even
+    though it is not currently documented to do so.*/
+  if(ret<0)return ret;
+  th_info2theora_info(_ci,&info);
+  return 0;
+}
+
+int theora_decode_packetin(theora_state *_td,ogg_packet *_op){
+  th_api_wrapper *api;
+  ogg_int64_t     gp;
+  int             ret;
+  if(!_td||!_td->i||!_td->i->codec_setup)return OC_FAULT;
+  api=(th_api_wrapper *)_td->i->codec_setup;
+  ret=th_decode_packetin(api->decode,_op,&gp);
+  if(ret<0)return OC_BADPACKET;
+  _td->granulepos=gp;
+  return 0;
+}
+
+int theora_decode_YUVout(theora_state *_td,yuv_buffer *_yuv){
+  th_api_wrapper  *api;
+  th_dec_ctx      *decode;
+  th_ycbcr_buffer  buf;
+  int              ret;
+  if(!_td||!_td->i||!_td->i->codec_setup)return OC_FAULT;
+  api=(th_api_wrapper *)_td->i->codec_setup;
+  decode=(th_dec_ctx *)api->decode;
+  if(!decode)return OC_FAULT;
+  ret=th_decode_ycbcr_out(decode,buf);
+  if(ret>=0){
+    _yuv->y_width=buf[0].width;
+    _yuv->y_height=buf[0].height;
+    _yuv->y_stride=buf[0].stride;
+    _yuv->uv_width=buf[1].width;
+    _yuv->uv_height=buf[1].height;
+    _yuv->uv_stride=buf[1].stride;
+    _yuv->y=buf[0].data;
+    _yuv->u=buf[1].data;
+    _yuv->v=buf[2].data;
+  }
+  return ret;
+}

lib/theoratest/lib/theora/decinfo.c

@@ -0,0 +1,274 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <limits.h>
+#include "decint.h"
+
+/*Only used for fuzzing.*/
+#if defined(HAVE_MEMORY_CONSTRAINT)
+static const int MAX_FUZZING_WIDTH = 16384;
+static const int MAX_FUZZING_HEIGHT = 16384;
+#endif
+
+
+/*Unpacks a series of octets from a given byte array into the pack buffer.
+  No checking is done to ensure the buffer contains enough data.
+  _opb: The pack buffer to read the octets from.
+  _buf: The byte array to store the unpacked bytes in.
+  _len: The number of octets to unpack.*/
+static void oc_unpack_octets(oc_pack_buf *_opb,char *_buf,size_t _len){
+  while(_len-->0){
+    long val;
+    val=oc_pack_read(_opb,8);
+    *_buf++=(char)val;
+  }
+}
+
+/*Unpacks a 32-bit integer encoded by octets in little-endian form.*/
+static long oc_unpack_length(oc_pack_buf *_opb){
+  long ret[4];
+  int  i;
+  for(i=0;i<4;i++)ret[i]=oc_pack_read(_opb,8);
+  return ret[0]|ret[1]<<8|ret[2]<<16|ret[3]<<24;
+}
+
+static int oc_info_unpack(oc_pack_buf *_opb,th_info *_info){
+  long val;
+  /*Check the codec bitstream version.*/
+  val=oc_pack_read(_opb,8);
+  _info->version_major=(unsigned char)val;
+  val=oc_pack_read(_opb,8);
+  _info->version_minor=(unsigned char)val;
+  val=oc_pack_read(_opb,8);
+  _info->version_subminor=(unsigned char)val;
+  /*verify we can parse this bitstream version.
+     We accept earlier minors and all subminors, by spec*/
+  if(_info->version_major>TH_VERSION_MAJOR||
+   (_info->version_major==TH_VERSION_MAJOR&&
+   _info->version_minor>TH_VERSION_MINOR)){
+    return TH_EVERSION;
+  }
+  /*Read the encoded frame description.*/
+  val=oc_pack_read(_opb,16);
+  _info->frame_width=(ogg_uint32_t)val<<4;
+  val=oc_pack_read(_opb,16);
+  _info->frame_height=(ogg_uint32_t)val<<4;
+  val=oc_pack_read(_opb,24);
+  _info->pic_width=(ogg_uint32_t)val;
+  val=oc_pack_read(_opb,24);
+  _info->pic_height=(ogg_uint32_t)val;
+  val=oc_pack_read(_opb,8);
+  _info->pic_x=(ogg_uint32_t)val;
+  val=oc_pack_read(_opb,8);
+  _info->pic_y=(ogg_uint32_t)val;
+  val=oc_pack_read(_opb,32);
+  _info->fps_numerator=(ogg_uint32_t)val;
+  val=oc_pack_read(_opb,32);
+  _info->fps_denominator=(ogg_uint32_t)val;
+  if(_info->frame_width==0||_info->frame_height==0||
+   _info->pic_width+_info->pic_x>_info->frame_width||
+   _info->pic_height+_info->pic_y>_info->frame_height||
+   _info->fps_numerator==0||_info->fps_denominator==0){
+    return TH_EBADHEADER;
+  }
+#if defined(HAVE_MEMORY_CONSTRAINT)
+  if(_info->frame_width>=MAX_FUZZING_WIDTH&&_info->frame_height>=MAX_FUZZING_HEIGHT){
+    return TH_EBADHEADER;
+  }
+#endif
+  /*Note: The sense of pic_y is inverted in what we pass back to the
+     application compared to how it is stored in the bitstream.
+    This is because the bitstream uses a right-handed coordinate system, while
+     applications expect a left-handed one.*/
+  _info->pic_y=_info->frame_height-_info->pic_height-_info->pic_y;
+  val=oc_pack_read(_opb,24);
+  _info->aspect_numerator=(ogg_uint32_t)val;
+  val=oc_pack_read(_opb,24);
+  _info->aspect_denominator=(ogg_uint32_t)val;
+  val=oc_pack_read(_opb,8);
+  _info->colorspace=(th_colorspace)val;
+  val=oc_pack_read(_opb,24);
+  _info->target_bitrate=(int)val;
+  val=oc_pack_read(_opb,6);
+  _info->quality=(int)val;
+  val=oc_pack_read(_opb,5);
+  _info->keyframe_granule_shift=(int)val;
+  val=oc_pack_read(_opb,2);
+  _info->pixel_fmt=(th_pixel_fmt)val;
+  if(_info->pixel_fmt==TH_PF_RSVD)return TH_EBADHEADER;
+  val=oc_pack_read(_opb,3);
+  if(val!=0||oc_pack_bytes_left(_opb)<0)return TH_EBADHEADER;
+  return 0;
+}
+
+static int oc_comment_unpack(oc_pack_buf *_opb,th_comment *_tc){
+  long len;
+  int  i;
+  /*Read the vendor string.*/
+  len=oc_unpack_length(_opb);
+  if(len<0||len>oc_pack_bytes_left(_opb))return TH_EBADHEADER;
+  _tc->vendor=_ogg_malloc((size_t)len+1);
+  if(_tc->vendor==NULL)return TH_EFAULT;
+  oc_unpack_octets(_opb,_tc->vendor,len);
+  _tc->vendor[len]='\0';
+  /*Read the user comments.*/
+  _tc->comments=(int)oc_unpack_length(_opb);
+  len=_tc->comments;
+  if(len<0||len>(LONG_MAX>>2)||len<<2>oc_pack_bytes_left(_opb)){
+    _tc->comments=0;
+    return TH_EBADHEADER;
+  }
+  _tc->comment_lengths=(int *)_ogg_malloc(
+   _tc->comments*sizeof(_tc->comment_lengths[0]));
+  _tc->user_comments=(char **)_ogg_malloc(
+   _tc->comments*sizeof(_tc->user_comments[0]));
+  if(_tc->comment_lengths==NULL||_tc->user_comments==NULL){
+    _tc->comments=0;
+    return TH_EFAULT;
+  }
+  for(i=0;i<_tc->comments;i++){
+    len=oc_unpack_length(_opb);
+    if(len<0||len>oc_pack_bytes_left(_opb)){
+      _tc->comments=i;
+      return TH_EBADHEADER;
+    }
+    _tc->comment_lengths[i]=len;
+    _tc->user_comments[i]=_ogg_malloc((size_t)len+1);
+    if(_tc->user_comments[i]==NULL){
+      _tc->comments=i;
+      return TH_EFAULT;
+    }
+    oc_unpack_octets(_opb,_tc->user_comments[i],len);
+    _tc->user_comments[i][len]='\0';
+  }
+  return oc_pack_bytes_left(_opb)<0?TH_EBADHEADER:0;
+}
+
+static int oc_setup_unpack(oc_pack_buf *_opb,th_setup_info *_setup){
+  int ret;
+  /*Read the quantizer tables.*/
+  ret=oc_quant_params_unpack(_opb,&_setup->qinfo);
+  if(ret<0)return ret;
+  /*Read the Huffman trees.*/
+  return oc_huff_trees_unpack(_opb,_setup->huff_tables);
+}
+
+static void oc_setup_clear(th_setup_info *_setup){
+  oc_quant_params_clear(&_setup->qinfo);
+  oc_huff_trees_clear(_setup->huff_tables);
+}
+
+static int oc_dec_headerin(oc_pack_buf *_opb,th_info *_info,
+ th_comment *_tc,th_setup_info **_setup,ogg_packet *_op){
+  char buffer[6];
+  long val;
+  int  packtype;
+  int  ret;
+  val=oc_pack_read(_opb,8);
+  packtype=(int)val;
+  /*If we're at a data packet...*/
+  if(!(packtype&0x80)){
+    /*Check to make sure we received all three headers...
+      If we haven't seen any valid headers, assume this is not actually
+       Theora.*/
+    if(_info->frame_width<=0)return TH_ENOTFORMAT;
+    /*Follow our documentation, which says we'll return TH_EFAULT if this
+       are NULL (_info was checked by our caller).*/
+    if(_tc==NULL)return TH_EFAULT;
+    /*And if any other headers were missing, declare this packet "out of
+       sequence" instead.*/
+    if(_tc->vendor==NULL)return TH_EBADHEADER;
+    /*Don't check this until it's needed, since we allow passing NULL for the
+       arguments that we're not expecting the next header to fill in yet.*/
+    if(_setup==NULL)return TH_EFAULT;
+    if(*_setup==NULL)return TH_EBADHEADER;
+    /*If we got everything, we're done.*/
+    return 0;
+  }
+  /*Check the codec string.*/
+  oc_unpack_octets(_opb,buffer,6);
+  if(memcmp(buffer,"theora",6)!=0)return TH_ENOTFORMAT;
+  switch(packtype){
+    /*Codec info header.*/
+    case 0x80:{
+      /*This should be the first packet, and we should not already be
+         initialized.*/
+      if(!_op->b_o_s||_info->frame_width>0)return TH_EBADHEADER;
+      ret=oc_info_unpack(_opb,_info);
+      if(ret<0)th_info_clear(_info);
+      else ret=3;
+    }break;
+    /*Comment header.*/
+    case 0x81:{
+      if(_tc==NULL)return TH_EFAULT;
+      /*We shoud have already decoded the info header, and should not yet have
+         decoded the comment header.*/
+      if(_info->frame_width==0||_tc->vendor!=NULL)return TH_EBADHEADER;
+      ret=oc_comment_unpack(_opb,_tc);
+      if(ret<0)th_comment_clear(_tc);
+      else ret=2;
+    }break;
+    /*Codec setup header.*/
+    case 0x82:{
+      oc_setup_info *setup;
+      if(_tc==NULL||_setup==NULL)return TH_EFAULT;
+      /*We should have already decoded the info header and the comment header,
+         and should not yet have decoded the setup header.*/
+      if(_info->frame_width==0||_tc->vendor==NULL||*_setup!=NULL){
+        return TH_EBADHEADER;
+      }
+      setup=(oc_setup_info *)_ogg_calloc(1,sizeof(*setup));
+      if(setup==NULL)return TH_EFAULT;
+      ret=oc_setup_unpack(_opb,setup);
+      if(ret<0){
+        oc_setup_clear(setup);
+        _ogg_free(setup);
+      }
+      else{
+        *_setup=setup;
+        ret=1;
+      }
+    }break;
+    default:{
+      /*We don't know what this header is.*/
+      return TH_EBADHEADER;
+    }break;
+  }
+  return ret;
+}
+
+
+/*Decodes one header packet.
+  This should be called repeatedly with the packets at the beginning of the
+   stream until it returns 0.*/
+int th_decode_headerin(th_info *_info,th_comment *_tc,
+ th_setup_info **_setup,ogg_packet *_op){
+  oc_pack_buf opb;
+  if(_op==NULL)return TH_EBADHEADER;
+  if(_info==NULL)return TH_EFAULT;
+  oc_pack_readinit(&opb,_op->packet,_op->bytes);
+  return oc_dec_headerin(&opb,_info,_tc,_setup,_op);
+}
+
+void th_setup_free(th_setup_info *_setup){
+  if(_setup!=NULL){
+    oc_setup_clear(_setup);
+    _ogg_free(_setup);
+  }
+}

lib/theoratest/lib/theora/decint.h

@@ -0,0 +1,185 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <limits.h>
+#if !defined(_decint_H)
+# define _decint_H (1)
+# include "theora/theoradec.h"
+# include "state.h"
+# include "bitpack.h"
+# include "huffdec.h"
+# include "dequant.h"
+
+typedef struct th_setup_info         oc_setup_info;
+typedef struct oc_dec_opt_vtable     oc_dec_opt_vtable;
+typedef struct oc_dec_pipeline_state oc_dec_pipeline_state;
+typedef struct th_dec_ctx            oc_dec_ctx;
+
+
+
+/*Decoder-specific accelerated functions.*/
+# if defined(OC_C64X_ASM)
+#  include "c64x/c64xdec.h"
+# endif
+
+# if !defined(oc_dec_accel_init)
+#  define oc_dec_accel_init oc_dec_accel_init_c
+# endif
+# if defined(OC_DEC_USE_VTABLE)
+#  if !defined(oc_dec_dc_unpredict_mcu_plane)
+#   define oc_dec_dc_unpredict_mcu_plane(_dec,_pipe,_pli) \
+ ((*(_dec)->opt_vtable.dc_unpredict_mcu_plane)(_dec,_pipe,_pli))
+#  endif
+# else
+#  if !defined(oc_dec_dc_unpredict_mcu_plane)
+#   define oc_dec_dc_unpredict_mcu_plane oc_dec_dc_unpredict_mcu_plane_c
+#  endif
+# endif
+
+
+
+/*Constants for the packet-in state machine specific to the decoder.*/
+
+/*Next packet to read: Data packet.*/
+#define OC_PACKET_DATA (0)
+
+
+
+struct th_setup_info{
+  /*The Huffman codes.*/
+  ogg_int16_t   *huff_tables[TH_NHUFFMAN_TABLES];
+  /*The quantization parameters.*/
+  th_quant_info  qinfo;
+};
+
+
+
+/*Decoder specific functions with accelerated variants.*/
+struct oc_dec_opt_vtable{
+  void (*dc_unpredict_mcu_plane)(oc_dec_ctx *_dec,
+   oc_dec_pipeline_state *_pipe,int _pli);
+};
+
+
+
+struct oc_dec_pipeline_state{
+  /*Decoded DCT coefficients.
+    These are placed here instead of on the stack so that they can persist
+     between blocks, which makes clearing them back to zero much faster when
+     only a few non-zero coefficients were decoded.
+    It requires at least 65 elements because the zig-zag index array uses the
+     65th element as a dumping ground for out-of-range indices to protect us
+     from buffer overflow.
+    We make it fully twice as large so that the second half can serve as the
+     reconstruction buffer, which saves passing another parameter to all the
+     acceleration functios.
+    It also solves problems with 16-byte alignment for NEON on ARM.
+    gcc (as of 4.2.1) only seems to be able to give stack variables 8-byte
+     alignment, and silently produces incorrect results if you ask for 16.
+    Finally, keeping it off the stack means there's less likely to be a data
+     hazard beween the NEON co-processor and the regular ARM core, which avoids
+     unnecessary stalls.*/
+  OC_ALIGN16(ogg_int16_t dct_coeffs[128]);
+  OC_ALIGN16(signed char bounding_values[256]);
+  ptrdiff_t           ti[3][64];
+  ptrdiff_t           ebi[3][64];
+  ptrdiff_t           eob_runs[3][64];
+  const ptrdiff_t    *coded_fragis[3];
+  const ptrdiff_t    *uncoded_fragis[3];
+  ptrdiff_t           ncoded_fragis[3];
+  ptrdiff_t           nuncoded_fragis[3];
+  const ogg_uint16_t *dequant[3][3][2];
+  int                 fragy0[3];
+  int                 fragy_end[3];
+  int                 pred_last[3][4];
+  int                 mcu_nvfrags;
+  int                 loop_filter;
+  int                 pp_level;
+};
+
+
+struct th_dec_ctx{
+  /*Shared encoder/decoder state.*/
+  oc_theora_state        state;
+  /*Whether or not packets are ready to be emitted.
+    This takes on negative values while there are remaining header packets to
+     be emitted, reaches 0 when the codec is ready for input, and goes to 1
+     when a frame has been processed and a data packet is ready.*/
+  int                    packet_state;
+  /*Buffer in which to assemble packets.*/
+  oc_pack_buf            opb;
+  /*Huffman decode trees.*/
+  ogg_int16_t           *huff_tables[TH_NHUFFMAN_TABLES];
+  /*The index of the first token in each plane for each coefficient.*/
+  ptrdiff_t              ti0[3][64];
+  /*The number of outstanding EOB runs at the start of each coefficient in each
+     plane.*/
+  ptrdiff_t              eob_runs[3][64];
+  /*The DCT token lists.*/
+  unsigned char         *dct_tokens;
+  /*The extra bits associated with DCT tokens.*/
+  unsigned char         *extra_bits;
+  /*The number of dct tokens unpacked so far.*/
+  int                    dct_tokens_count;
+  /*The out-of-loop post-processing level.*/
+  int                    pp_level;
+  /*The DC scale used for out-of-loop deblocking.*/
+  int                    pp_dc_scale[64];
+  /*The sharpen modifier used for out-of-loop deringing.*/
+  int                    pp_sharp_mod[64];
+  /*The DC quantization index of each block.*/
+  unsigned char         *dc_qis;
+  /*The variance of each block.*/
+  int                   *variances;
+  /*The storage for the post-processed frame buffer.*/
+  unsigned char         *pp_frame_data;
+  /*Whether or not the post-processsed frame buffer has space for chroma.*/
+  int                    pp_frame_state;
+  /*The buffer used for the post-processed frame.
+    Note that this is _not_ guaranteed to have the same strides and offsets as
+     the reference frame buffers.*/
+  th_ycbcr_buffer        pp_frame_buf;
+  /*The striped decode callback function.*/
+  th_stripe_callback     stripe_cb;
+  oc_dec_pipeline_state  pipe;
+# if defined(OC_DEC_USE_VTABLE)
+  /*Table for decoder acceleration functions.*/
+  oc_dec_opt_vtable      opt_vtable;
+# endif
+# if defined(HAVE_CAIRO)
+  /*Output metrics for debugging.*/
+  int                    telemetry_mbmode;
+  int                    telemetry_mv;
+  int                    telemetry_qi;
+  int                    telemetry_bits;
+  int                    telemetry_frame_bytes;
+  int                    telemetry_coding_bytes;
+  int                    telemetry_mode_bytes;
+  int                    telemetry_mv_bytes;
+  int                    telemetry_qi_bytes;
+  int                    telemetry_dc_bytes;
+  unsigned char         *telemetry_frame_data;
+# endif
+};
+
+/*Default pure-C implementations of decoder-specific accelerated functions.*/
+void oc_dec_accel_init_c(oc_dec_ctx *_dec);
+
+void oc_dec_dc_unpredict_mcu_plane_c(oc_dec_ctx *_dec,
+ oc_dec_pipeline_state *_pipe,int _pli);
+
+#endif

lib/theoratest/lib/theora/dequant.c

@@ -0,0 +1,182 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <ogg/ogg.h>
+#include "dequant.h"
+#include "decint.h"
+
+int oc_quant_params_unpack(oc_pack_buf *_opb,th_quant_info *_qinfo){
+  th_quant_base *base_mats;
+  long           val;
+  int            nbase_mats;
+  int            sizes[64];
+  int            indices[64];
+  int            nbits;
+  int            bmi;
+  int            ci;
+  int            qti;
+  int            pli;
+  int            qri;
+  int            qi;
+  int            i;
+  val=oc_pack_read(_opb,3);
+  nbits=(int)val;
+  for(qi=0;qi<64;qi++){
+    val=oc_pack_read(_opb,nbits);
+    _qinfo->loop_filter_limits[qi]=(unsigned char)val;
+  }
+  val=oc_pack_read(_opb,4);
+  nbits=(int)val+1;
+  for(qi=0;qi<64;qi++){
+    val=oc_pack_read(_opb,nbits);
+    _qinfo->ac_scale[qi]=(ogg_uint16_t)val;
+  }
+  val=oc_pack_read(_opb,4);
+  nbits=(int)val+1;
+  for(qi=0;qi<64;qi++){
+    val=oc_pack_read(_opb,nbits);
+    _qinfo->dc_scale[qi]=(ogg_uint16_t)val;
+  }
+  val=oc_pack_read(_opb,9);
+  nbase_mats=(int)val+1;
+  base_mats=_ogg_malloc(nbase_mats*sizeof(base_mats[0]));
+  if(base_mats==NULL)return TH_EFAULT;
+  for(bmi=0;bmi<nbase_mats;bmi++){
+    for(ci=0;ci<64;ci++){
+      val=oc_pack_read(_opb,8);
+      base_mats[bmi][ci]=(unsigned char)val;
+    }
+  }
+  nbits=oc_ilog(nbase_mats-1);
+  for(i=0;i<6;i++){
+    th_quant_ranges *qranges;
+    th_quant_base   *qrbms;
+    int             *qrsizes;
+    qti=i/3;
+    pli=i%3;
+    qranges=_qinfo->qi_ranges[qti]+pli;
+    if(i>0){
+      val=oc_pack_read1(_opb);
+      if(!val){
+        int qtj;
+        int plj;
+        if(qti>0){
+          val=oc_pack_read1(_opb);
+          if(val){
+            qtj=qti-1;
+            plj=pli;
+          }
+          else{
+            qtj=(i-1)/3;
+            plj=(i-1)%3;
+          }
+        }
+        else{
+          qtj=(i-1)/3;
+          plj=(i-1)%3;
+        }
+        *qranges=*(_qinfo->qi_ranges[qtj]+plj);
+        continue;
+      }
+    }
+    val=oc_pack_read(_opb,nbits);
+    indices[0]=(int)val;
+    for(qi=qri=0;qi<63;){
+      val=oc_pack_read(_opb,oc_ilog(62-qi));
+      sizes[qri]=(int)val+1;
+      qi+=(int)val+1;
+      val=oc_pack_read(_opb,nbits);
+      indices[++qri]=(int)val;
+    }
+    /*Note: The caller is responsible for cleaning up any partially
+       constructed qinfo.*/
+    if(qi>63){
+      _ogg_free(base_mats);
+      return TH_EBADHEADER;
+    }
+    qranges->nranges=qri;
+    qranges->sizes=qrsizes=(int *)_ogg_malloc(qri*sizeof(qrsizes[0]));
+    if(qranges->sizes==NULL){
+      /*Note: The caller is responsible for cleaning up any partially
+         constructed qinfo.*/
+      _ogg_free(base_mats);
+      return TH_EFAULT;
+    }
+    memcpy(qrsizes,sizes,qri*sizeof(qrsizes[0]));
+    qrbms=(th_quant_base *)_ogg_malloc((qri+1)*sizeof(qrbms[0]));
+    if(qrbms==NULL){
+      /*Note: The caller is responsible for cleaning up any partially
+         constructed qinfo.*/
+      _ogg_free(base_mats);
+      return TH_EFAULT;
+    }
+    qranges->base_matrices=(const th_quant_base *)qrbms;
+    do{
+      bmi=indices[qri];
+      /*Note: The caller is responsible for cleaning up any partially
+         constructed qinfo.*/
+      if(bmi>=nbase_mats){
+        _ogg_free(base_mats);
+        return TH_EBADHEADER;
+      }
+      memcpy(qrbms[qri],base_mats[bmi],sizeof(qrbms[qri]));
+    }
+    while(qri-->0);
+  }
+  _ogg_free(base_mats);
+  return 0;
+}
+
+void oc_quant_params_clear(th_quant_info *_qinfo){
+  int i;
+  for(i=6;i-->0;){
+    int qti;
+    int pli;
+    qti=i/3;
+    pli=i%3;
+    /*Clear any duplicate pointer references.*/
+    if(i>0){
+      int qtj;
+      int plj;
+      qtj=(i-1)/3;
+      plj=(i-1)%3;
+      if(_qinfo->qi_ranges[qti][pli].sizes==
+       _qinfo->qi_ranges[qtj][plj].sizes){
+        _qinfo->qi_ranges[qti][pli].sizes=NULL;
+      }
+      if(_qinfo->qi_ranges[qti][pli].base_matrices==
+       _qinfo->qi_ranges[qtj][plj].base_matrices){
+        _qinfo->qi_ranges[qti][pli].base_matrices=NULL;
+      }
+    }
+    if(qti>0){
+      if(_qinfo->qi_ranges[1][pli].sizes==
+       _qinfo->qi_ranges[0][pli].sizes){
+        _qinfo->qi_ranges[1][pli].sizes=NULL;
+      }
+      if(_qinfo->qi_ranges[1][pli].base_matrices==
+       _qinfo->qi_ranges[0][pli].base_matrices){
+        _qinfo->qi_ranges[1][pli].base_matrices=NULL;
+      }
+    }
+    /*Now free all the non-duplicate storage.*/
+    _ogg_free((void *)_qinfo->qi_ranges[qti][pli].sizes);
+    _ogg_free((void *)_qinfo->qi_ranges[qti][pli].base_matrices);
+  }
+}

lib/theoratest/lib/theora/dequant.h

@@ -0,0 +1,27 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#if !defined(_dequant_H)
+# define _dequant_H (1)
+# include "quant.h"
+# include "bitpack.h"
+
+int oc_quant_params_unpack(oc_pack_buf *_opb,
+ th_quant_info *_qinfo);
+void oc_quant_params_clear(th_quant_info *_qinfo);
+
+#endif

lib/theoratest/lib/theora/fragment.c

@@ -0,0 +1,82 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+#include <string.h>
+#include "internal.h"
+
+void oc_frag_copy_c(unsigned char *_dst,const unsigned char *_src,int _ystride){
+  int i;
+  for(i=8;i-->0;){
+    memcpy(_dst,_src,8*sizeof(*_dst));
+    _dst+=_ystride;
+    _src+=_ystride;
+  }
+}
+
+/*Copies the fragments specified by the lists of fragment indices from one
+   frame to another.
+  _dst_frame:     The reference frame to copy to.
+  _src_frame:     The reference frame to copy from.
+  _ystride:       The row stride of the reference frames.
+  _fragis:        A pointer to a list of fragment indices.
+  _nfragis:       The number of fragment indices to copy.
+  _frag_buf_offs: The offsets of fragments in the reference frames.*/
+void oc_frag_copy_list_c(unsigned char *_dst_frame,
+ const unsigned char *_src_frame,int _ystride,
+ const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs){
+  ptrdiff_t fragii;
+  for(fragii=0;fragii<_nfragis;fragii++){
+    ptrdiff_t frag_buf_off;
+    frag_buf_off=_frag_buf_offs[_fragis[fragii]];
+    oc_frag_copy_c(_dst_frame+frag_buf_off,
+     _src_frame+frag_buf_off,_ystride);
+  }
+}
+
+void oc_frag_recon_intra_c(unsigned char *_dst,int _ystride,
+ const ogg_int16_t _residue[64]){
+  int i;
+  for(i=0;i<8;i++){
+    int j;
+    for(j=0;j<8;j++)_dst[j]=OC_CLAMP255(_residue[i*8+j]+128);
+    _dst+=_ystride;
+  }
+}
+
+void oc_frag_recon_inter_c(unsigned char *_dst,
+ const unsigned char *_src,int _ystride,const ogg_int16_t _residue[64]){
+  int i;
+  for(i=0;i<8;i++){
+    int j;
+    for(j=0;j<8;j++)_dst[j]=OC_CLAMP255(_residue[i*8+j]+_src[j]);
+    _dst+=_ystride;
+    _src+=_ystride;
+  }
+}
+
+void oc_frag_recon_inter2_c(unsigned char *_dst,const unsigned char *_src1,
+ const unsigned char *_src2,int _ystride,const ogg_int16_t _residue[64]){
+  int i;
+  for(i=0;i<8;i++){
+    int j;
+    for(j=0;j<8;j++)_dst[j]=OC_CLAMP255(_residue[i*8+j]+(_src1[j]+_src2[j]>>1));
+    _dst+=_ystride;
+    _src1+=_ystride;
+    _src2+=_ystride;
+  }
+}
+
+void oc_restore_fpu_c(void){}

lib/theoratest/lib/theora/huffdec.c

@@ -0,0 +1,512 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <ogg/ogg.h>
+#include "huffdec.h"
+#include "decint.h"
+
+
+
+/*Instead of storing every branching in the tree, subtrees can be collapsed
+   into one node, with a table of size 1<<nbits pointing directly to its
+   descedents nbits levels down.
+  This allows more than one bit to be read at a time, and avoids following all
+   the intermediate branches with next to no increased code complexity once
+   the collapsed tree has been built.
+  We do _not_ require that a subtree be complete to be collapsed, but instead
+   store duplicate pointers in the table, and record the actual depth of the
+   node below its parent.
+  This tells us the number of bits to advance the stream after reaching it.
+
+  This turns out to be equivalent to the method described in \cite{Hash95},
+   without the requirement that codewords be sorted by length.
+  If the codewords were sorted by length (so-called ``canonical-codes''), they
+   could be decoded much faster via either Lindell and Moffat's approach or
+   Hashemian's Condensed Huffman Code approach, the latter of which has an
+   extremely small memory footprint.
+  We can't use Choueka et al.'s finite state machine approach, which is
+   extremely fast, because we can't allow multiple symbols to be output at a
+   time; the codebook can and does change between symbols.
+  It also has very large memory requirements, which impairs cache coherency.
+
+  We store the tree packed in an array of 16-bit integers (words).
+  Each node consists of a single word, followed consecutively by two or more
+   indices of its children.
+  Let n be the value of this first word.
+  This is the number of bits that need to be read to traverse the node, and
+   must be positive.
+  1<<n entries follow in the array, each an index to a child node.
+  If the child is positive, then it is the index of another internal node in
+   the table.
+  If the child is negative or zero, then it is a leaf node.
+  These are stored directly in the child pointer to save space, since they only
+   require a single word.
+  If a leaf node would have been encountered before reading n bits, then it is
+   duplicated the necessary number of times in this table.
+  Leaf nodes pack both a token value and their actual depth in the tree.
+  The token in the leaf node is (-leaf&255).
+  The number of bits that need to be consumed to reach the leaf, starting from
+   the current node, is (-leaf>>8).
+
+  @ARTICLE{Hash95,
+    author="Reza Hashemian",
+    title="Memory Efficient and High-Speed Search {Huffman} Coding",
+    journal="{IEEE} Transactions on Communications",
+    volume=43,
+    number=10,
+    pages="2576--2581",
+    month=Oct,
+    year=1995
+  }*/
+
+
+
+/*The map from external spec-defined tokens to internal tokens.
+  This is constructed so that any extra bits read with the original token value
+   can be masked off the least significant bits of its internal token index.
+  In addition, all of the tokens which require additional extra bits are placed
+   at the start of the list, and grouped by type.
+  OC_DCT_REPEAT_RUN3_TOKEN is placed first, as it is an extra-special case, so
+   giving it index 0 may simplify comparisons on some architectures.
+  These requirements require some substantial reordering.*/
+static const unsigned char OC_DCT_TOKEN_MAP[TH_NDCT_TOKENS]={
+  /*OC_DCT_EOB1_TOKEN (0 extra bits)*/
+  15,
+  /*OC_DCT_EOB2_TOKEN (0 extra bits)*/
+  16,
+  /*OC_DCT_EOB3_TOKEN (0 extra bits)*/
+  17,
+  /*OC_DCT_REPEAT_RUN0_TOKEN (2 extra bits)*/
+  88,
+  /*OC_DCT_REPEAT_RUN1_TOKEN (3 extra bits)*/
+  80,
+  /*OC_DCT_REPEAT_RUN2_TOKEN (4 extra bits)*/
+   1,
+  /*OC_DCT_REPEAT_RUN3_TOKEN (12 extra bits)*/
+   0,
+  /*OC_DCT_SHORT_ZRL_TOKEN (3 extra bits)*/
+  48,
+  /*OC_DCT_ZRL_TOKEN (6 extra bits)*/
+  14,
+  /*OC_ONE_TOKEN (0 extra bits)*/
+  56,
+  /*OC_MINUS_ONE_TOKEN (0 extra bits)*/
+  57,
+  /*OC_TWO_TOKEN (0 extra bits)*/
+  58,
+  /*OC_MINUS_TWO_TOKEN (0 extra bits)*/
+  59,
+  /*OC_DCT_VAL_CAT2 (1 extra bit)*/
+  60,
+  62,
+  64,
+  66,
+  /*OC_DCT_VAL_CAT3 (2 extra bits)*/
+  68,
+  /*OC_DCT_VAL_CAT4 (3 extra bits)*/
+  72,
+  /*OC_DCT_VAL_CAT5 (4 extra bits)*/
+   2,
+  /*OC_DCT_VAL_CAT6 (5 extra bits)*/
+   4,
+  /*OC_DCT_VAL_CAT7 (6 extra bits)*/
+   6,
+  /*OC_DCT_VAL_CAT8 (10 extra bits)*/
+   8,
+  /*OC_DCT_RUN_CAT1A (1 extra bit)*/
+  18,
+  20,
+  22,
+  24,
+  26,
+  /*OC_DCT_RUN_CAT1B (3 extra bits)*/
+  32,
+  /*OC_DCT_RUN_CAT1C (4 extra bits)*/
+  12,
+  /*OC_DCT_RUN_CAT2A (2 extra bits)*/
+  28,
+  /*OC_DCT_RUN_CAT2B (3 extra bits)*/
+  40
+};
+
+/*The log base 2 of number of internal tokens associated with each of the spec
+   tokens (i.e., how many of the extra bits are folded into the token value).
+  Increasing the maximum value beyond 3 will enlarge the amount of stack
+   required for tree construction.*/
+static const unsigned char OC_DCT_TOKEN_MAP_LOG_NENTRIES[TH_NDCT_TOKENS]={
+  0,0,0,2,3,0,0,3,0,0,0,0,0,1,1,1,1,2,3,1,1,1,2,1,1,1,1,1,3,1,2,3
+};
+
+
+/*The size a lookup table is allowed to grow to relative to the number of
+   unique nodes it contains.
+  E.g., if OC_HUFF_SLUSH is 4, then at most 75% of the space in the tree is
+   wasted (1/4 of the space must be used).
+  Larger numbers can decode tokens with fewer read operations, while smaller
+   numbers may save more space.
+  With a sample file:
+  32233473 read calls are required when no tree collapsing is done (100.0%).
+  19269269 read calls are required when OC_HUFF_SLUSH is 1 (59.8%).
+  11144969 read calls are required when OC_HUFF_SLUSH is 2 (34.6%).
+  10538563 read calls are required when OC_HUFF_SLUSH is 4 (32.7%).
+  10192578 read calls are required when OC_HUFF_SLUSH is 8 (31.6%).
+  Since a value of 2 gets us the vast majority of the speed-up with only a
+   small amount of wasted memory, this is what we use.
+  This value must be less than 128, or you could create a tree with more than
+   32767 entries, which would overflow the 16-bit words used to index it.*/
+#define OC_HUFF_SLUSH (2)
+/*The root of the tree is on the fast path, and a larger value here is more
+   beneficial than elsewhere in the tree.
+  7 appears to give the best performance, trading off between increased use of
+   the single-read fast path and cache footprint for the tables, though
+   obviously this will depend on your cache size.
+  Using 7 here, the VP3 tables are about twice as large compared to using 2.*/
+#define OC_ROOT_HUFF_SLUSH (7)
+
+
+
+/*Unpacks a Huffman codebook.
+  _opb:    The buffer to unpack from.
+  _tokens: Stores a list of internal tokens, in the order they were found in
+            the codebook, and the lengths of their corresponding codewords.
+           This is enough to completely define the codebook, while minimizing
+            stack usage and avoiding temporary allocations (for platforms
+            where free() is a no-op).
+  Return: The number of internal tokens in the codebook, or a negative value
+   on error.*/
+int oc_huff_tree_unpack(oc_pack_buf *_opb,unsigned char _tokens[256][2]){
+  ogg_uint32_t code;
+  int          len;
+  int          ntokens;
+  int          nleaves;
+  code=0;
+  len=ntokens=nleaves=0;
+  for(;;){
+    long bits;
+    bits=oc_pack_read1(_opb);
+    /*Only process nodes so long as there's more bits in the buffer.*/
+    if(oc_pack_bytes_left(_opb)<0)return TH_EBADHEADER;
+    /*Read an internal node:*/
+    if(!bits){
+      len++;
+      /*Don't allow codewords longer than 32 bits.*/
+      if(len>32)return TH_EBADHEADER;
+    }
+    /*Read a leaf node:*/
+    else{
+      ogg_uint32_t code_bit;
+      int          neb;
+      int          nentries;
+      int          token;
+      /*Don't allow more than 32 spec-tokens per codebook.*/
+      if(++nleaves>32)return TH_EBADHEADER;
+      bits=oc_pack_read(_opb,OC_NDCT_TOKEN_BITS);
+      neb=OC_DCT_TOKEN_MAP_LOG_NENTRIES[bits];
+      token=OC_DCT_TOKEN_MAP[bits];
+      nentries=1<<neb;
+      while(nentries-->0){
+        _tokens[ntokens][0]=(unsigned char)token++;
+        _tokens[ntokens][1]=(unsigned char)(len+neb);
+        ntokens++;
+      }
+      code_bit=0x80000000U>>len-1;
+      while(len>0&&(code&code_bit)){
+        code^=code_bit;
+        code_bit<<=1;
+        len--;
+      }
+      if(len<=0)break;
+      code|=code_bit;
+    }
+  }
+  return ntokens;
+}
+
+/*Count how many tokens would be required to fill a subtree at depth _depth.
+  _tokens: A list of internal tokens, in the order they are found in the
+            codebook, and the lengths of their corresponding codewords.
+  _depth:  The depth of the desired node in the corresponding tree structure.
+  Return: The number of tokens that belong to that subtree.*/
+static int oc_huff_subtree_tokens(unsigned char _tokens[][2],int _depth){
+  ogg_uint32_t code;
+  int          ti;
+  code=0;
+  ti=0;
+  do{
+    if(_tokens[ti][1]-_depth<32)code+=0x80000000U>>_tokens[ti++][1]-_depth;
+    else{
+      /*Because of the expanded internal tokens, we can have codewords as long
+         as 35 bits.
+        A single recursion here is enough to advance past them.*/
+      code++;
+      ti+=oc_huff_subtree_tokens(_tokens+ti,_depth+31);
+    }
+  }
+  while(code<0x80000000U);
+  return ti;
+}
+
+/*Compute the number of bits to use for a collapsed tree node at the given
+   depth.
+  _tokens:  A list of internal tokens, in the order they are found in the
+             codebook, and the lengths of their corresponding codewords.
+  _ntokens: The number of tokens corresponding to this tree node.
+  _depth:   The depth of this tree node.
+  Return: The number of bits to use for a collapsed tree node rooted here.
+          This is always at least one, even if this was a leaf node.*/
+static int oc_huff_tree_collapse_depth(unsigned char _tokens[][2],
+ int _ntokens,int _depth){
+  int got_leaves;
+  int loccupancy;
+  int occupancy;
+  int slush;
+  int nbits;
+  int best_nbits;
+  slush=_depth>0?OC_HUFF_SLUSH:OC_ROOT_HUFF_SLUSH;
+  /*It's legal to have a tree with just a single node, which requires no bits
+     to decode and always returns the same token.
+    However, no encoder actually does this (yet).
+    To avoid a special case in oc_huff_token_decode(), we force the number of
+     lookahead bits to be at least one.
+    This will produce a tree that looks ahead one bit and then advances the
+     stream zero bits.*/
+  nbits=1;
+  occupancy=2;
+  got_leaves=1;
+  do{
+    int ti;
+    if(got_leaves)best_nbits=nbits;
+    nbits++;
+    got_leaves=0;
+    loccupancy=occupancy;
+    for(occupancy=ti=0;ti<_ntokens;occupancy++){
+      if(_tokens[ti][1]<_depth+nbits)ti++;
+      else if(_tokens[ti][1]==_depth+nbits){
+        got_leaves=1;
+        ti++;
+      }
+      else ti+=oc_huff_subtree_tokens(_tokens+ti,_depth+nbits);
+    }
+  }
+  while(occupancy>loccupancy&&occupancy*slush>=1<<nbits);
+  return best_nbits;
+}
+
+/*Determines the size in words of a Huffman tree node that represents a
+   subtree of depth _nbits.
+  _nbits: The depth of the subtree.
+          This must be greater than zero.
+  Return: The number of words required to store the node.*/
+static size_t oc_huff_node_size(int _nbits){
+  return 1+(1<<_nbits);
+}
+
+/*Produces a collapsed-tree representation of the given token list.
+  _tree: The storage for the collapsed Huffman tree.
+         This may be NULL to compute the required storage size instead of
+          constructing the tree.
+  _tokens:  A list of internal tokens, in the order they are found in the
+             codebook, and the lengths of their corresponding codewords.
+  _ntokens: The number of tokens corresponding to this tree node.
+  Return: The number of words required to store the tree.*/
+static size_t oc_huff_tree_collapse(ogg_int16_t *_tree,
+ unsigned char _tokens[][2],int _ntokens){
+  ogg_int16_t   node[34];
+  unsigned char depth[34];
+  unsigned char last[34];
+  size_t        ntree;
+  int           ti;
+  int           l;
+  depth[0]=0;
+  last[0]=(unsigned char)(_ntokens-1);
+  ntree=0;
+  ti=0;
+  l=0;
+  do{
+    int nbits;
+    nbits=oc_huff_tree_collapse_depth(_tokens+ti,last[l]+1-ti,depth[l]);
+    node[l]=(ogg_int16_t)ntree;
+    ntree+=oc_huff_node_size(nbits);
+    if(_tree!=NULL)_tree[node[l]++]=(ogg_int16_t)nbits;
+    do{
+      while(ti<=last[l]&&_tokens[ti][1]<=depth[l]+nbits){
+        if(_tree!=NULL){
+          ogg_int16_t leaf;
+          int         nentries;
+          nentries=1<<depth[l]+nbits-_tokens[ti][1];
+          leaf=(ogg_int16_t)-(_tokens[ti][1]-depth[l]<<8|_tokens[ti][0]);
+          while(nentries-->0)_tree[node[l]++]=leaf;
+        }
+        ti++;
+      }
+      if(ti<=last[l]){
+        /*We need to recurse*/
+        depth[l+1]=(unsigned char)(depth[l]+nbits);
+        if(_tree!=NULL)_tree[node[l]++]=(ogg_int16_t)ntree;
+        l++;
+        last[l]=
+         (unsigned char)(ti+oc_huff_subtree_tokens(_tokens+ti,depth[l])-1);
+        break;
+      }
+      /*Pop back up a level of recursion.*/
+      else if(l-->0)nbits=depth[l+1]-depth[l];
+    }
+    while(l>=0);
+  }
+  while(l>=0);
+  return ntree;
+}
+
+/*Unpacks a set of Huffman trees, and reduces them to a collapsed
+   representation.
+  _opb:   The buffer to unpack the trees from.
+  _nodes: The table to fill with the Huffman trees.
+  Return: 0 on success, or a negative value on error.
+          The caller is responsible for cleaning up any partially initialized
+           _nodes on failure.*/
+int oc_huff_trees_unpack(oc_pack_buf *_opb,
+ ogg_int16_t *_nodes[TH_NHUFFMAN_TABLES]){
+  int i;
+  for(i=0;i<TH_NHUFFMAN_TABLES;i++){
+    unsigned char  tokens[256][2];
+    int            ntokens;
+    ogg_int16_t   *tree;
+    size_t         size;
+    /*Unpack the full tree into a temporary buffer.*/
+    ntokens=oc_huff_tree_unpack(_opb,tokens);
+    if(ntokens<0)return ntokens;
+    /*Figure out how big the collapsed tree will be and allocate space for it.*/
+    size=oc_huff_tree_collapse(NULL,tokens,ntokens);
+    /*This should never happen; if it does it means you set OC_HUFF_SLUSH or
+       OC_ROOT_HUFF_SLUSH too large.*/
+    if(size>32767)return TH_EIMPL;
+    tree=(ogg_int16_t *)_ogg_malloc(size*sizeof(*tree));
+    if(tree==NULL)return TH_EFAULT;
+    /*Construct the collapsed the tree.*/
+    oc_huff_tree_collapse(tree,tokens,ntokens);
+    _nodes[i]=tree;
+  }
+  return 0;
+}
+
+/*Determines the size in words of a Huffman subtree.
+  _tree: The complete Huffman tree.
+  _node: The index of the root of the desired subtree.
+  Return: The number of words required to store the tree.*/
+static size_t oc_huff_tree_size(const ogg_int16_t *_tree,int _node){
+  size_t size;
+  int    nchildren;
+  int    n;
+  int    i;
+  n=_tree[_node];
+  size=oc_huff_node_size(n);
+  nchildren=1<<n;
+  i=0;
+  do{
+    int child;
+    child=_tree[_node+i+1];
+    if(child<=0)i+=1<<n-(-child>>8);
+    else{
+      size+=oc_huff_tree_size(_tree,child);
+      i++;
+    }
+  }
+  while(i<nchildren);
+  return size;
+}
+
+/*Makes a copy of the given set of Huffman trees.
+  _dst: The array to store the copy in.
+  _src: The array of trees to copy.*/
+int oc_huff_trees_copy(ogg_int16_t *_dst[TH_NHUFFMAN_TABLES],
+ const ogg_int16_t *const _src[TH_NHUFFMAN_TABLES]){
+  int i;
+  for(i=0;i<TH_NHUFFMAN_TABLES;i++){
+    size_t size;
+    size=oc_huff_tree_size(_src[i],0);
+    _dst[i]=(ogg_int16_t *)_ogg_malloc(size*sizeof(*_dst[i]));
+    if(_dst[i]==NULL){
+      while(i-->0)_ogg_free(_dst[i]);
+      return TH_EFAULT;
+    }
+    memcpy(_dst[i],_src[i],size*sizeof(*_dst[i]));
+  }
+  return 0;
+}
+
+/*Frees the memory used by a set of Huffman trees.
+  _nodes: The array of trees to free.*/
+void oc_huff_trees_clear(ogg_int16_t *_nodes[TH_NHUFFMAN_TABLES]){
+  int i;
+  for(i=0;i<TH_NHUFFMAN_TABLES;i++)_ogg_free(_nodes[i]);
+}
+
+
+/*Unpacks a single token using the given Huffman tree.
+  _opb:  The buffer to unpack the token from.
+  _node: The tree to unpack the token with.
+  Return: The token value.*/
+int oc_huff_token_decode_c(oc_pack_buf *_opb,const ogg_int16_t *_tree){
+  const unsigned char *ptr;
+  const unsigned char *stop;
+  oc_pb_window         window;
+  int                  available;
+  long                 bits;
+  int                  node;
+  int                  n;
+  ptr=_opb->ptr;
+  window=_opb->window;
+  stop=_opb->stop;
+  available=_opb->bits;
+  node=0;
+  for(;;){
+    n=_tree[node];
+    if(n>available){
+      unsigned shift;
+      shift=OC_PB_WINDOW_SIZE-available;
+      do{
+        /*We don't bother setting eof because we won't check for it after we've
+           started decoding DCT tokens.*/
+        if(ptr>=stop){
+          shift=(unsigned)-OC_LOTS_OF_BITS;
+          break;
+        }
+        shift-=8;
+        window|=(oc_pb_window)*ptr++<<shift;
+      }
+      while(shift>=8);
+      /*Note: We never request more than 24 bits, so there's no need to fill in
+         the last partial byte here.*/
+      available=OC_PB_WINDOW_SIZE-shift;
+    }
+    bits=window>>OC_PB_WINDOW_SIZE-n;
+    node=_tree[node+1+bits];
+    if(node<=0)break;
+    window<<=n;
+    available-=n;
+  }
+  node=-node;
+  n=node>>8;
+  window<<=n;
+  available-=n;
+  _opb->ptr=ptr;
+  _opb->window=window;
+  _opb->bits=available;
+  return node&255;
+}

lib/theoratest/lib/theora/huffdec.h

@@ -0,0 +1,32 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#if !defined(_huffdec_H)
+# define _huffdec_H (1)
+# include "huffman.h"
+# include "bitpack.h"
+
+
+
+int oc_huff_trees_unpack(oc_pack_buf *_opb,
+ ogg_int16_t *_nodes[TH_NHUFFMAN_TABLES]);
+int oc_huff_trees_copy(ogg_int16_t *_dst[TH_NHUFFMAN_TABLES],
+ const ogg_int16_t *const _src[TH_NHUFFMAN_TABLES]);
+void oc_huff_trees_clear(ogg_int16_t *_nodes[TH_NHUFFMAN_TABLES]);
+int oc_huff_token_decode_c(oc_pack_buf *_opb,const ogg_int16_t *_node);
+
+#endif

lib/theoratest/lib/theora/huffman.h

@@ -0,0 +1,70 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#if !defined(_huffman_H)
+# define _huffman_H (1)
+# include "theora/codec.h"
+# include "ocintrin.h"
+
+/*The range of valid quantized DCT coefficient values.
+  VP3 used 511 in the encoder, but the bitstream is capable of 580.*/
+#define OC_DCT_VAL_RANGE         (580)
+
+#define OC_NDCT_TOKEN_BITS       (5)
+
+#define OC_DCT_EOB1_TOKEN        (0)
+#define OC_DCT_EOB2_TOKEN        (1)
+#define OC_DCT_EOB3_TOKEN        (2)
+#define OC_DCT_REPEAT_RUN0_TOKEN (3)
+#define OC_DCT_REPEAT_RUN1_TOKEN (4)
+#define OC_DCT_REPEAT_RUN2_TOKEN (5)
+#define OC_DCT_REPEAT_RUN3_TOKEN (6)
+
+#define OC_DCT_SHORT_ZRL_TOKEN   (7)
+#define OC_DCT_ZRL_TOKEN         (8)
+
+#define OC_ONE_TOKEN             (9)
+#define OC_MINUS_ONE_TOKEN       (10)
+#define OC_TWO_TOKEN             (11)
+#define OC_MINUS_TWO_TOKEN       (12)
+
+#define OC_DCT_VAL_CAT2          (13)
+#define OC_DCT_VAL_CAT3          (17)
+#define OC_DCT_VAL_CAT4          (18)
+#define OC_DCT_VAL_CAT5          (19)
+#define OC_DCT_VAL_CAT6          (20)
+#define OC_DCT_VAL_CAT7          (21)
+#define OC_DCT_VAL_CAT8          (22)
+
+#define OC_DCT_RUN_CAT1A         (23)
+#define OC_DCT_RUN_CAT1B         (28)
+#define OC_DCT_RUN_CAT1C         (29)
+#define OC_DCT_RUN_CAT2A         (30)
+#define OC_DCT_RUN_CAT2B         (31)
+
+#define OC_NDCT_EOB_TOKEN_MAX    (7)
+#define OC_NDCT_ZRL_TOKEN_MAX    (9)
+#define OC_NDCT_VAL_MAX          (23)
+#define OC_NDCT_VAL_CAT1_MAX     (13)
+#define OC_NDCT_VAL_CAT2_MAX     (17)
+#define OC_NDCT_VAL_CAT2_SIZE    (OC_NDCT_VAL_CAT2_MAX-OC_DCT_VAL_CAT2)
+#define OC_NDCT_RUN_MAX          (32)
+#define OC_NDCT_RUN_CAT1A_MAX    (28)
+
+extern const unsigned char OC_DCT_TOKEN_EXTRA_BITS[TH_NDCT_TOKENS];
+
+#endif

lib/theoratest/lib/theora/idct.c

@@ -0,0 +1,330 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <string.h>
+#include "internal.h"
+#include "dct.h"
+
+/*Performs an inverse 8 point Type-II DCT transform.
+  The output is scaled by a factor of 2 relative to the orthonormal version of
+   the transform.
+  _y: The buffer to store the result in.
+      Data will be placed in every 8th entry (e.g., in a column of an 8x8
+       block).
+  _x: The input coefficients.
+      The first 8 entries are used (e.g., from a row of an 8x8 block).*/
+static void idct8(ogg_int16_t *_y,const ogg_int16_t _x[8]){
+  ogg_int32_t t[8];
+  ogg_int32_t r;
+  /*Stage 1:*/
+  /*0-1 butterfly.*/
+  t[0]=OC_C4S4*(ogg_int16_t)(_x[0]+_x[4])>>16;
+  t[1]=OC_C4S4*(ogg_int16_t)(_x[0]-_x[4])>>16;
+  /*2-3 rotation by 6pi/16.*/
+  t[2]=(OC_C6S2*_x[2]>>16)-(OC_C2S6*_x[6]>>16);
+  t[3]=(OC_C2S6*_x[2]>>16)+(OC_C6S2*_x[6]>>16);
+  /*4-7 rotation by 7pi/16.*/
+  t[4]=(OC_C7S1*_x[1]>>16)-(OC_C1S7*_x[7]>>16);
+  /*5-6 rotation by 3pi/16.*/
+  t[5]=(OC_C3S5*_x[5]>>16)-(OC_C5S3*_x[3]>>16);
+  t[6]=(OC_C5S3*_x[5]>>16)+(OC_C3S5*_x[3]>>16);
+  t[7]=(OC_C1S7*_x[1]>>16)+(OC_C7S1*_x[7]>>16);
+  /*Stage 2:*/
+  /*4-5 butterfly.*/
+  r=t[4]+t[5];
+  t[5]=OC_C4S4*(ogg_int16_t)(t[4]-t[5])>>16;
+  t[4]=r;
+  /*7-6 butterfly.*/
+  r=t[7]+t[6];
+  t[6]=OC_C4S4*(ogg_int16_t)(t[7]-t[6])>>16;
+  t[7]=r;
+  /*Stage 3:*/
+  /*0-3 butterfly.*/
+  r=t[0]+t[3];
+  t[3]=t[0]-t[3];
+  t[0]=r;
+  /*1-2 butterfly.*/
+  r=t[1]+t[2];
+  t[2]=t[1]-t[2];
+  t[1]=r;
+  /*6-5 butterfly.*/
+  r=t[6]+t[5];
+  t[5]=t[6]-t[5];
+  t[6]=r;
+  /*Stage 4:*/
+  /*0-7 butterfly.*/
+  _y[0<<3]=(ogg_int16_t)(t[0]+t[7]);
+  /*1-6 butterfly.*/
+  _y[1<<3]=(ogg_int16_t)(t[1]+t[6]);
+  /*2-5 butterfly.*/
+  _y[2<<3]=(ogg_int16_t)(t[2]+t[5]);
+  /*3-4 butterfly.*/
+  _y[3<<3]=(ogg_int16_t)(t[3]+t[4]);
+  _y[4<<3]=(ogg_int16_t)(t[3]-t[4]);
+  _y[5<<3]=(ogg_int16_t)(t[2]-t[5]);
+  _y[6<<3]=(ogg_int16_t)(t[1]-t[6]);
+  _y[7<<3]=(ogg_int16_t)(t[0]-t[7]);
+}
+
+/*Performs an inverse 8 point Type-II DCT transform.
+  The output is scaled by a factor of 2 relative to the orthonormal version of
+   the transform.
+  _y: The buffer to store the result in.
+      Data will be placed in every 8th entry (e.g., in a column of an 8x8
+       block).
+  _x: The input coefficients.
+      Only the first 4 entries are used.
+      The other 4 are assumed to be 0.*/
+static void idct8_4(ogg_int16_t *_y,const ogg_int16_t _x[8]){
+  ogg_int32_t t[8];
+  ogg_int32_t r;
+  /*Stage 1:*/
+  t[0]=OC_C4S4*_x[0]>>16;
+  t[2]=OC_C6S2*_x[2]>>16;
+  t[3]=OC_C2S6*_x[2]>>16;
+  t[4]=OC_C7S1*_x[1]>>16;
+  t[5]=-(OC_C5S3*_x[3]>>16);
+  t[6]=OC_C3S5*_x[3]>>16;
+  t[7]=OC_C1S7*_x[1]>>16;
+  /*Stage 2:*/
+  r=t[4]+t[5];
+  t[5]=OC_C4S4*(ogg_int16_t)(t[4]-t[5])>>16;
+  t[4]=r;
+  r=t[7]+t[6];
+  t[6]=OC_C4S4*(ogg_int16_t)(t[7]-t[6])>>16;
+  t[7]=r;
+  /*Stage 3:*/
+  t[1]=t[0]+t[2];
+  t[2]=t[0]-t[2];
+  r=t[0]+t[3];
+  t[3]=t[0]-t[3];
+  t[0]=r;
+  r=t[6]+t[5];
+  t[5]=t[6]-t[5];
+  t[6]=r;
+  /*Stage 4:*/
+  _y[0<<3]=(ogg_int16_t)(t[0]+t[7]);
+  _y[1<<3]=(ogg_int16_t)(t[1]+t[6]);
+  _y[2<<3]=(ogg_int16_t)(t[2]+t[5]);
+  _y[3<<3]=(ogg_int16_t)(t[3]+t[4]);
+  _y[4<<3]=(ogg_int16_t)(t[3]-t[4]);
+  _y[5<<3]=(ogg_int16_t)(t[2]-t[5]);
+  _y[6<<3]=(ogg_int16_t)(t[1]-t[6]);
+  _y[7<<3]=(ogg_int16_t)(t[0]-t[7]);
+}
+
+/*Performs an inverse 8 point Type-II DCT transform.
+  The output is scaled by a factor of 2 relative to the orthonormal version of
+   the transform.
+  _y: The buffer to store the result in.
+      Data will be placed in every 8th entry (e.g., in a column of an 8x8
+       block).
+  _x: The input coefficients.
+      Only the first 3 entries are used.
+      The other 5 are assumed to be 0.*/
+static void idct8_3(ogg_int16_t *_y,const ogg_int16_t _x[8]){
+  ogg_int32_t t[8];
+  ogg_int32_t r;
+  /*Stage 1:*/
+  t[0]=OC_C4S4*_x[0]>>16;
+  t[2]=OC_C6S2*_x[2]>>16;
+  t[3]=OC_C2S6*_x[2]>>16;
+  t[4]=OC_C7S1*_x[1]>>16;
+  t[7]=OC_C1S7*_x[1]>>16;
+  /*Stage 2:*/
+  t[5]=OC_C4S4*t[4]>>16;
+  t[6]=OC_C4S4*t[7]>>16;
+  /*Stage 3:*/
+  t[1]=t[0]+t[2];
+  t[2]=t[0]-t[2];
+  r=t[0]+t[3];
+  t[3]=t[0]-t[3];
+  t[0]=r;
+  r=t[6]+t[5];
+  t[5]=t[6]-t[5];
+  t[6]=r;
+  /*Stage 4:*/
+  _y[0<<3]=(ogg_int16_t)(t[0]+t[7]);
+  _y[1<<3]=(ogg_int16_t)(t[1]+t[6]);
+  _y[2<<3]=(ogg_int16_t)(t[2]+t[5]);
+  _y[3<<3]=(ogg_int16_t)(t[3]+t[4]);
+  _y[4<<3]=(ogg_int16_t)(t[3]-t[4]);
+  _y[5<<3]=(ogg_int16_t)(t[2]-t[5]);
+  _y[6<<3]=(ogg_int16_t)(t[1]-t[6]);
+  _y[7<<3]=(ogg_int16_t)(t[0]-t[7]);
+}
+
+/*Performs an inverse 8 point Type-II DCT transform.
+  The output is scaled by a factor of 2 relative to the orthonormal version of
+   the transform.
+  _y: The buffer to store the result in.
+      Data will be placed in every 8th entry (e.g., in a column of an 8x8
+       block).
+  _x: The input coefficients.
+      Only the first 2 entries are used.
+      The other 6 are assumed to be 0.*/
+static void idct8_2(ogg_int16_t *_y,const ogg_int16_t _x[8]){
+  ogg_int32_t t[8];
+  ogg_int32_t r;
+  /*Stage 1:*/
+  t[0]=OC_C4S4*_x[0]>>16;
+  t[4]=OC_C7S1*_x[1]>>16;
+  t[7]=OC_C1S7*_x[1]>>16;
+  /*Stage 2:*/
+  t[5]=OC_C4S4*t[4]>>16;
+  t[6]=OC_C4S4*t[7]>>16;
+  /*Stage 3:*/
+  r=t[6]+t[5];
+  t[5]=t[6]-t[5];
+  t[6]=r;
+  /*Stage 4:*/
+  _y[0<<3]=(ogg_int16_t)(t[0]+t[7]);
+  _y[1<<3]=(ogg_int16_t)(t[0]+t[6]);
+  _y[2<<3]=(ogg_int16_t)(t[0]+t[5]);
+  _y[3<<3]=(ogg_int16_t)(t[0]+t[4]);
+  _y[4<<3]=(ogg_int16_t)(t[0]-t[4]);
+  _y[5<<3]=(ogg_int16_t)(t[0]-t[5]);
+  _y[6<<3]=(ogg_int16_t)(t[0]-t[6]);
+  _y[7<<3]=(ogg_int16_t)(t[0]-t[7]);
+}
+
+/*Performs an inverse 8 point Type-II DCT transform.
+  The output is scaled by a factor of 2 relative to the orthonormal version of
+   the transform.
+  _y: The buffer to store the result in.
+      Data will be placed in every 8th entry (e.g., in a column of an 8x8
+       block).
+  _x: The input coefficients.
+      Only the first entry is used.
+      The other 7 are assumed to be 0.*/
+static void idct8_1(ogg_int16_t *_y,const ogg_int16_t _x[1]){
+  _y[0<<3]=_y[1<<3]=_y[2<<3]=_y[3<<3]=
+   _y[4<<3]=_y[5<<3]=_y[6<<3]=_y[7<<3]=(ogg_int16_t)(OC_C4S4*_x[0]>>16);
+}
+
+/*Performs an inverse 8x8 Type-II DCT transform.
+  The input is assumed to be scaled by a factor of 4 relative to orthonormal
+   version of the transform.
+  All coefficients but the first 3 in zig-zag scan order are assumed to be 0:
+   x  x  0  0  0  0  0  0
+   x  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+  _y: The buffer to store the result in.
+      This may be the same as _x.
+  _x: The input coefficients.*/
+static void oc_idct8x8_3(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  ogg_int16_t w[64];
+  int         i;
+  /*Transform rows of x into columns of w.*/
+  idct8_2(w,_x);
+  idct8_1(w+1,_x+8);
+  /*Transform rows of w into columns of y.*/
+  for(i=0;i<8;i++)idct8_2(_y+i,w+i*8);
+  /*Adjust for the scale factor.*/
+  for(i=0;i<64;i++)_y[i]=(ogg_int16_t)(_y[i]+8>>4);
+  /*Clear input data for next block.*/
+  _x[0]=_x[1]=_x[8]=0;
+}
+
+/*Performs an inverse 8x8 Type-II DCT transform.
+  The input is assumed to be scaled by a factor of 4 relative to orthonormal
+   version of the transform.
+  All coefficients but the first 10 in zig-zag scan order are assumed to be 0:
+   x  x  x  x  0  0  0  0
+   x  x  x  0  0  0  0  0
+   x  x  0  0  0  0  0  0
+   x  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+   0  0  0  0  0  0  0  0
+  _y: The buffer to store the result in.
+      This may be the same as _x.
+  _x: The input coefficients.*/
+static void oc_idct8x8_10(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  ogg_int16_t w[64];
+  int         i;
+  /*Transform rows of x into columns of w.*/
+  idct8_4(w,_x);
+  idct8_3(w+1,_x+8);
+  idct8_2(w+2,_x+16);
+  idct8_1(w+3,_x+24);
+  /*Transform rows of w into columns of y.*/
+  for(i=0;i<8;i++)idct8_4(_y+i,w+i*8);
+  /*Adjust for the scale factor.*/
+  for(i=0;i<64;i++)_y[i]=(ogg_int16_t)(_y[i]+8>>4);
+  /*Clear input data for next block.*/
+  _x[0]=_x[1]=_x[2]=_x[3]=_x[8]=_x[9]=_x[10]=_x[16]=_x[17]=_x[24]=0;
+}
+
+/*Performs an inverse 8x8 Type-II DCT transform.
+  The input is assumed to be scaled by a factor of 4 relative to orthonormal
+   version of the transform.
+  _y: The buffer to store the result in.
+      This may be the same as _x.
+  _x: The input coefficients.*/
+static void oc_idct8x8_slow(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  ogg_int16_t w[64];
+  int         i;
+  /*Transform rows of x into columns of w.*/
+  for(i=0;i<8;i++)idct8(w+i,_x+i*8);
+  /*Transform rows of w into columns of y.*/
+  for(i=0;i<8;i++)idct8(_y+i,w+i*8);
+  /*Adjust for the scale factor.*/
+  for(i=0;i<64;i++)_y[i]=(ogg_int16_t)(_y[i]+8>>4);
+  /*Clear input data for next block.*/
+  for(i=0;i<64;i++)_x[i]=0;
+}
+
+/*Performs an inverse 8x8 Type-II DCT transform.
+  The input is assumed to be scaled by a factor of 4 relative to orthonormal
+   version of the transform.*/
+void oc_idct8x8_c(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi){
+  /*_last_zzi is subtly different from an actual count of the number of
+     coefficients we decoded for this block.
+    It contains the value of zzi BEFORE the final token in the block was
+     decoded.
+    In most cases this is an EOB token (the continuation of an EOB run from a
+     previous block counts), and so this is the same as the coefficient count.
+    However, in the case that the last token was NOT an EOB token, but filled
+     the block up with exactly 64 coefficients, _last_zzi will be less than 64.
+    Provided the last token was not a pure zero run, the minimum value it can
+     be is 46, and so that doesn't affect any of the cases in this routine.
+    However, if the last token WAS a pure zero run of length 63, then _last_zzi
+     will be 1 while the number of coefficients decoded is 64.
+    Thus, we will trigger the following special case, where the real
+     coefficient count would not.
+    Note also that a zero run of length 64 will give _last_zzi a value of 0,
+     but we still process the DC coefficient, which might have a non-zero value
+     due to DC prediction.
+    Although convoluted, this is arguably the correct behavior: it allows us to
+     use a smaller transform when the block ends with a long zero run instead
+     of a normal EOB token.
+    It could be smarter... multiple separate zero runs at the end of a block
+     will fool it, but an encoder that generates these really deserves what it
+     gets.
+    Needless to say we inherited this approach from VP3.*/
+  /*Then perform the iDCT.*/
+  if(_last_zzi<=3)oc_idct8x8_3(_y,_x);
+  else if(_last_zzi<=10)oc_idct8x8_10(_y,_x);
+  else oc_idct8x8_slow(_y,_x);
+}

lib/theoratest/lib/theora/internal.c

@@ -0,0 +1,210 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <limits.h>
+#include <string.h>
+#include "internal.h"
+
+
+
+/*A map from the index in the zig zag scan to the coefficient number in a
+   block.
+  All zig zag indices beyond 63 are sent to coefficient 64, so that zero runs
+   past the end of a block in bogus streams get mapped to a known location.*/
+const unsigned char OC_FZIG_ZAG[128]={
+   0, 1, 8,16, 9, 2, 3,10,
+  17,24,32,25,18,11, 4, 5,
+  12,19,26,33,40,48,41,34,
+  27,20,13, 6, 7,14,21,28,
+  35,42,49,56,57,50,43,36,
+  29,22,15,23,30,37,44,51,
+  58,59,52,45,38,31,39,46,
+  53,60,61,54,47,55,62,63,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64
+};
+
+/*A map from the coefficient number in a block to its index in the zig zag
+   scan.*/
+const unsigned char OC_IZIG_ZAG[64]={
+   0, 1, 5, 6,14,15,27,28,
+   2, 4, 7,13,16,26,29,42,
+   3, 8,12,17,25,30,41,43,
+   9,11,18,24,31,40,44,53,
+  10,19,23,32,39,45,52,54,
+  20,22,33,38,46,51,55,60,
+  21,34,37,47,50,56,59,61,
+  35,36,48,49,57,58,62,63
+};
+
+/*A map from physical macro block ordering to bitstream macro block
+   ordering within a super block.*/
+const unsigned char OC_MB_MAP[2][2]={{0,3},{1,2}};
+
+/*A list of the indices in the oc_mb.map array that can be valid for each of
+   the various chroma decimation types.*/
+const unsigned char OC_MB_MAP_IDXS[TH_PF_NFORMATS][12]={
+  {0,1,2,3,4,8},
+  {0,1,2,3,4,5,8,9},
+  {0,1,2,3,4,6,8,10},
+  {0,1,2,3,4,5,6,7,8,9,10,11}
+};
+
+/*The number of indices in the oc_mb.map array that can be valid for each of
+   the various chroma decimation types.*/
+const unsigned char OC_MB_MAP_NIDXS[TH_PF_NFORMATS]={6,8,8,12};
+
+/*The number of extra bits that are coded with each of the DCT tokens.
+  Each DCT token has some fixed number of additional bits (possibly 0) stored
+   after the token itself, containing, for example, coefficient magnitude,
+   sign bits, etc.*/
+const unsigned char OC_DCT_TOKEN_EXTRA_BITS[TH_NDCT_TOKENS]={
+  0,0,0,2,3,4,12,3,6,
+  0,0,0,0,
+  1,1,1,1,2,3,4,5,6,10,
+  1,1,1,1,1,3,4,
+  2,3
+};
+
+
+
+int oc_ilog(unsigned _v){
+  int ret;
+  for(ret=0;_v;ret++)_v>>=1;
+  return ret;
+}
+
+
+
+void *oc_aligned_malloc(size_t _sz,size_t _align){
+  unsigned char *p;
+  if(_align-1>UCHAR_MAX||(_align&_align-1)||_sz>~(size_t)0-_align)return NULL;
+  p=(unsigned char *)_ogg_malloc(_sz+_align);
+  if(p!=NULL){
+    int offs;
+    offs=((p-(unsigned char *)0)-1&_align-1);
+    p[offs]=offs;
+    p+=offs+1;
+  }
+  return p;
+}
+
+void oc_aligned_free(void *_ptr){
+  unsigned char *p;
+  p=(unsigned char *)_ptr;
+  if(p!=NULL){
+    int offs;
+    offs=*--p;
+    _ogg_free(p-offs);
+  }
+}
+
+
+void **oc_malloc_2d(size_t _height,size_t _width,size_t _sz){
+  size_t  rowsz;
+  size_t  colsz;
+  size_t  datsz;
+  char   *ret;
+  colsz=_height*sizeof(void *);
+  rowsz=_sz*_width;
+  datsz=rowsz*_height;
+  /*Alloc array and row pointers.*/
+  ret=(char *)_ogg_malloc(datsz+colsz);
+  /*Initialize the array.*/
+  if(ret!=NULL){
+    size_t   i;
+    void   **p;
+    char    *datptr;
+    p=(void **)ret;
+    i=_height;
+    for(datptr=ret+colsz;i-->0;p++,datptr+=rowsz)*p=(void *)datptr;
+  }
+  return (void **)ret;
+}
+
+void **oc_calloc_2d(size_t _height,size_t _width,size_t _sz){
+  size_t  colsz;
+  size_t  rowsz;
+  size_t  datsz;
+  char   *ret;
+  colsz=_height*sizeof(void *);
+  rowsz=_sz*_width;
+  datsz=rowsz*_height;
+  /*Alloc array and row pointers.*/
+  ret=(char *)_ogg_calloc(datsz+colsz,1);
+  /*Initialize the array.*/
+  if(ret!=NULL){
+    size_t   i;
+    void   **p;
+    char    *datptr;
+    p=(void **)ret;
+    i=_height;
+    for(datptr=ret+colsz;i-->0;p++,datptr+=rowsz)*p=(void *)datptr;
+  }
+  return (void **)ret;
+}
+
+void oc_free_2d(void *_ptr){
+  _ogg_free(_ptr);
+}
+
+/*Fills in a Y'CbCr buffer with a pointer to the image data in the first
+   buffer, but with the opposite vertical orientation.
+  _dst: The destination buffer.
+        This can be the same as _src.
+  _src: The source buffer.*/
+void oc_ycbcr_buffer_flip(th_ycbcr_buffer _dst,
+ const th_ycbcr_buffer _src){
+  int pli;
+  for(pli=0;pli<3;pli++){
+    _dst[pli].width=_src[pli].width;
+    _dst[pli].height=_src[pli].height;
+    _dst[pli].stride=-_src[pli].stride;
+    _dst[pli].data=_src[pli].data
+     +(1-_dst[pli].height)*(ptrdiff_t)_dst[pli].stride;
+  }
+}
+
+const char *th_version_string(void){
+  return OC_VENDOR_STRING;
+}
+
+ogg_uint32_t th_version_number(void){
+  return (TH_VERSION_MAJOR<<16)+(TH_VERSION_MINOR<<8)+TH_VERSION_SUB;
+}
+
+/*Determines the packet type.
+  Note that this correctly interprets a 0-byte packet as a video data packet.
+  Return: 1 for a header packet, 0 for a data packet.*/
+int th_packet_isheader(ogg_packet *_op){
+  return _op->bytes>0?_op->packet[0]>>7:0;
+}
+
+/*Determines the frame type of a video data packet.
+  Note that this correctly interprets a 0-byte packet as a delta frame.
+  Return: 1 for a key frame, 0 for a delta frame, and -1 for a header
+           packet.*/
+int th_packet_iskeyframe(ogg_packet *_op){
+  return _op->bytes<=0?0:_op->packet[0]&0x80?-1:!(_op->packet[0]&0x40);
+}

lib/theoratest/lib/theora/internal.h

@@ -0,0 +1,116 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+#if !defined(_internal_H)
+# define _internal_H (1)
+# include <stdlib.h>
+# include <limits.h>
+# if defined(HAVE_CONFIG_H)
+#  include "config.h"
+# endif
+# include "theora/codec.h"
+# include "theora/theora.h"
+# include "ocintrin.h"
+
+# if !defined(__GNUC_PREREQ)
+#  if defined(__GNUC__)&&defined(__GNUC_MINOR__)
+#   define __GNUC_PREREQ(_maj,_min) \
+ ((__GNUC__<<16)+__GNUC_MINOR__>=((_maj)<<16)+(_min))
+#  else
+#   define __GNUC_PREREQ(_maj,_min) 0
+#  endif
+# endif
+
+# if defined(_MSC_VER)
+/*Disable missing EMMS warnings.*/
+#  pragma warning(disable:4799)
+/*Thank you Microsoft, I know the order of operations.*/
+#  pragma warning(disable:4554)
+# endif
+/*You, too, gcc.*/
+# if __GNUC_PREREQ(4,2)
+#  pragma GCC diagnostic ignored "-Wparentheses"
+# endif
+
+/*Some assembly constructs require aligned operands.
+  The following macros are _only_ intended for structure member declarations.
+  Although they will sometimes work on stack variables, gcc will often silently
+   ignore them.
+  A separate set of macros could be made for manual stack alignment, but we
+   don't actually require it anywhere.*/
+# if defined(OC_X86_ASM)||defined(OC_ARM_ASM)
+#  if defined(__GNUC__)
+#   define OC_ALIGN8(expr) expr __attribute__((aligned(8)))
+#   define OC_ALIGN16(expr) expr __attribute__((aligned(16)))
+#  elif defined(_MSC_VER)
+#   define OC_ALIGN8(expr) __declspec (align(8)) expr
+#   define OC_ALIGN16(expr) __declspec (align(16)) expr
+#  else
+#   error "Alignment macros required for this platform."
+#  endif
+# endif
+# if !defined(OC_ALIGN8)
+#  define OC_ALIGN8(expr) expr
+# endif
+# if !defined(OC_ALIGN16)
+#  define OC_ALIGN16(expr) expr
+# endif
+
+
+
+/*This library's version.*/
+# define OC_VENDOR_STRING "Xiph.Org libtheora 1.2.0alpha 20100924 (Ptalarbvorm)"
+
+/*Theora bitstream version.*/
+# define TH_VERSION_MAJOR (3)
+# define TH_VERSION_MINOR (2)
+# define TH_VERSION_SUB   (1)
+# define TH_VERSION_CHECK(_info,_maj,_min,_sub) \
+ ((_info)->version_major>(_maj)||(_info)->version_major==(_maj)&& \
+ ((_info)->version_minor>(_min)||(_info)->version_minor==(_min)&& \
+ (_info)->version_subminor>=(_sub)))
+
+
+
+/*A map from the index in the zig zag scan to the coefficient number in a
+   block.*/
+extern const unsigned char OC_FZIG_ZAG[128];
+/*A map from the coefficient number in a block to its index in the zig zag
+   scan.*/
+extern const unsigned char OC_IZIG_ZAG[64];
+/*A map from physical macro block ordering to bitstream macro block
+   ordering within a super block.*/
+extern const unsigned char OC_MB_MAP[2][2];
+/*A list of the indices in the oc_mb_map array that can be valid for each of
+   the various chroma decimation types.*/
+extern const unsigned char OC_MB_MAP_IDXS[TH_PF_NFORMATS][12];
+/*The number of indices in the oc_mb_map array that can be valid for each of
+   the various chroma decimation types.*/
+extern const unsigned char OC_MB_MAP_NIDXS[TH_PF_NFORMATS];
+
+
+
+int oc_ilog(unsigned _v);
+void *oc_aligned_malloc(size_t _sz,size_t _align);
+void oc_aligned_free(void *_ptr);
+void **oc_malloc_2d(size_t _height,size_t _width,size_t _sz);
+void **oc_calloc_2d(size_t _height,size_t _width,size_t _sz);
+void oc_free_2d(void *_ptr);
+
+void oc_ycbcr_buffer_flip(th_ycbcr_buffer _dst,
+ const th_ycbcr_buffer _src);
+
+#endif

lib/theoratest/lib/theora/ocintrin.h

@@ -0,0 +1,128 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+/*Some common macros for potential platform-specific optimization.*/
+#include <math.h>
+#if !defined(_ocintrin_H)
+# define _ocintrin_H (1)
+
+/*Some specific platforms may have optimized intrinsic or inline assembly
+   versions of these functions which can substantially improve performance.
+  We define macros for them to allow easy incorporation of these non-ANSI
+   features.*/
+
+/*Note that we do not provide a macro for abs(), because it is provided as a
+   library function, which we assume is translated into an intrinsic to avoid
+   the function call overhead and then implemented in the smartest way for the
+   target platform.
+  With modern gcc (4.x), this is true: it uses cmov instructions if the
+   architecture supports it and branchless bit-twiddling if it does not (the
+   speed difference between the two approaches is not measurable).
+  Interestingly, the bit-twiddling method was patented in 2000 (US 6,073,150)
+   by Sun Microsystems, despite prior art dating back to at least 1996:
+   http://web.archive.org/web/19961201174141/www.x86.org/ftp/articles/pentopt/PENTOPT.TXT
+  On gcc 3.x, however, our assumption is not true, as abs() is translated to a
+   conditional jump, which is horrible on deeply piplined architectures (e.g.,
+   all consumer architectures for the past decade or more).
+  Also be warned that -C*abs(x) where C is a constant is mis-optimized as
+   abs(C*x) on every gcc release before 4.2.3.
+  See bug http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34130 */
+
+/*Modern gcc (4.x) can compile the naive versions of min and max with cmov if
+   given an appropriate architecture, but the branchless bit-twiddling versions
+   are just as fast, and do not require any special target architecture.
+  Earlier gcc versions (3.x) compiled both code to the same assembly
+   instructions, because of the way they represented ((_b)>(_a)) internally.*/
+#define OC_MAXI(_a,_b)      ((_a)-((_a)-(_b)&-((_b)>(_a))))
+#define OC_MINI(_a,_b)      ((_a)+((_b)-(_a)&-((_b)<(_a))))
+/*Clamps an integer into the given range.
+  If _a>_c, then the lower bound _a is respected over the upper bound _c (this
+   behavior is required to meet our documented API behavior).
+  _a: The lower bound.
+  _b: The value to clamp.
+  _c: The upper boud.*/
+#define OC_CLAMPI(_a,_b,_c) (OC_MAXI(_a,OC_MINI(_b,_c)))
+#define OC_CLAMP255(_x)     ((unsigned char)((((_x)<0)-1)&((_x)|-((_x)>255))))
+/*This has a chance of compiling branchless, and is just as fast as the
+   bit-twiddling method, which is slightly less portable, since it relies on a
+   sign-extended rightshift, which is not guaranteed by ANSI (but present on
+   every relevant platform).*/
+#define OC_SIGNI(_a)        (((_a)>0)-((_a)<0))
+/*Slightly more portable than relying on a sign-extended right-shift (which is
+   not guaranteed by ANSI), and just as fast, since gcc (3.x and 4.x both)
+   compile it into the right-shift anyway.*/
+#define OC_SIGNMASK(_a)     (-((_a)<0))
+/*Divides an integer by a power of two, truncating towards 0.
+  _dividend: The integer to divide.
+  _shift:    The non-negative power of two to divide by.
+  _rmask:    (1<<_shift)-1*/
+#define OC_DIV_POW2(_dividend,_shift,_rmask)\
+  ((_dividend)+(OC_SIGNMASK(_dividend)&(_rmask))>>(_shift))
+/*Divides _x by 65536, truncating towards 0.*/
+#define OC_DIV2_16(_x) OC_DIV_POW2(_x,16,0xFFFF)
+/*Divides _x by 2, truncating towards 0.*/
+#define OC_DIV2(_x) OC_DIV_POW2(_x,1,0x1)
+/*Divides _x by 8, truncating towards 0.*/
+#define OC_DIV8(_x) OC_DIV_POW2(_x,3,0x7)
+/*Divides _x by 16, truncating towards 0.*/
+#define OC_DIV16(_x) OC_DIV_POW2(_x,4,0xF)
+/*Right shifts _dividend by _shift, adding _rval, and subtracting one for
+   negative dividends first.
+  When _rval is (1<<_shift-1), this is equivalent to division with rounding
+   ties away from zero.*/
+#define OC_DIV_ROUND_POW2(_dividend,_shift,_rval)\
+  ((_dividend)+OC_SIGNMASK(_dividend)+(_rval)>>(_shift))
+/*Divides a _x by 2, rounding towards even numbers.*/
+#define OC_DIV2_RE(_x) ((_x)+((_x)>>1&1)>>1)
+/*Divides a _x by (1<<(_shift)), rounding towards even numbers.*/
+#define OC_DIV_POW2_RE(_x,_shift) \
+  ((_x)+((_x)>>(_shift)&1)+((1<<(_shift))-1>>1)>>(_shift))
+/*Swaps two integers _a and _b if _a>_b.*/
+#define OC_SORT2I(_a,_b) \
+  do{ \
+    int t__; \
+    t__=((_a)^(_b))&-((_b)<(_a)); \
+    (_a)^=t__; \
+    (_b)^=t__; \
+  } \
+  while(0)
+
+/*Accesses one of four (signed) bytes given an index.
+  This can be used to avoid small lookup tables.*/
+#define OC_BYTE_TABLE32(_a,_b,_c,_d,_i) \
+  ((signed char) \
+   (((_a)&0xFF|((_b)&0xFF)<<8|((_c)&0xFF)<<16|((_d)&0xFF)<<24)>>(_i)*8))
+/*Accesses one of eight (unsigned) nibbles given an index.
+  This can be used to avoid small lookup tables.*/
+#define OC_UNIBBLE_TABLE32(_a,_b,_c,_d,_e,_f,_g,_h,_i) \
+  ((((_a)&0xF|((_b)&0xF)<<4|((_c)&0xF)<<8|((_d)&0xF)<<12| \
+   ((_e)&0xF)<<16|((_f)&0xF)<<20|((_g)&0xF)<<24|((_h)&0xF)<<28)>>(_i)*4)&0xF)
+
+
+
+/*All of these macros should expect floats as arguments.*/
+#define OC_MAXF(_a,_b)      ((_a)<(_b)?(_b):(_a))
+#define OC_MINF(_a,_b)      ((_a)>(_b)?(_b):(_a))
+#define OC_CLAMPF(_a,_b,_c) (OC_MINF(_a,OC_MAXF(_b,_c)))
+#define OC_FABSF(_f)        ((float)fabs(_f))
+#define OC_SQRTF(_f)        ((float)sqrt(_f))
+#define OC_POWF(_b,_e)      ((float)pow(_b,_e))
+#define OC_LOGF(_f)         ((float)log(_f))
+#define OC_IFLOORF(_f)      ((int)floor(_f))
+#define OC_ICEILF(_f)       ((int)ceil(_f))
+
+#endif

lib/theoratest/lib/theora/quant.c

@@ -0,0 +1,127 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <ogg/ogg.h>
+#include "quant.h"
+#include "decint.h"
+
+/*The maximum output of the DCT with +/- 255 inputs is +/- 8157.
+  These minimum quantizers ensure the result after quantization (and after
+   prediction for DC) will be no more than +/- 510.
+  The tokenization system can handle values up to +/- 580, so there is no need
+   to do any coefficient clamping.
+  I would rather have allowed smaller quantizers and had to clamp, but these
+   minimums were required when constructing the original VP3 matrices and have
+   been formalized in the spec.*/
+static const unsigned OC_DC_QUANT_MIN[2]={4<<2,8<<2};
+static const unsigned OC_AC_QUANT_MIN[2]={2<<2,4<<2};
+
+/*Initializes the dequantization tables from a set of quantizer info.
+  Currently the dequantizer (and elsewhere enquantizer) tables are expected to
+   be initialized as pointing to the storage reserved for them in the
+   oc_theora_state (resp. oc_enc_ctx) structure.
+  If some tables are duplicates of others, the pointers will be adjusted to
+   point to a single copy of the tables, but the storage for them will not be
+   freed.
+  If you're concerned about the memory footprint, the obvious thing to do is
+   to move the storage out of its fixed place in the structures and allocate
+   it on demand.
+  However, a much, much better option is to only store the quantization
+   matrices being used for the current frame, and to recalculate these as the
+   qi values change between frames (this is what VP3 did).*/
+void oc_dequant_tables_init(ogg_uint16_t *_dequant[64][3][2],
+ int _pp_dc_scale[64],const th_quant_info *_qinfo){
+  /*Coding mode: intra or inter.*/
+  int          qti;
+  /*Y', C_b, C_r*/
+  int          pli;
+  for(qti=0;qti<2;qti++)for(pli=0;pli<3;pli++){
+    /*Quality index.*/
+    int qi;
+    /*Range iterator.*/
+    int qri;
+    for(qi=0,qri=0;qri<=_qinfo->qi_ranges[qti][pli].nranges;qri++){
+      th_quant_base base;
+      ogg_uint32_t  q;
+      int           qi_start;
+      int           qi_end;
+      memcpy(base,_qinfo->qi_ranges[qti][pli].base_matrices[qri],
+       sizeof(base));
+      qi_start=qi;
+      if(qri==_qinfo->qi_ranges[qti][pli].nranges)qi_end=qi+1;
+      else qi_end=qi+_qinfo->qi_ranges[qti][pli].sizes[qri];
+      /*Iterate over quality indicies in this range.*/
+      for(;;){
+        ogg_uint32_t qfac;
+        int          zzi;
+        int          ci;
+        /*In the original VP3.2 code, the rounding offset and the size of the
+           dead zone around 0 were controlled by a "sharpness" parameter.
+          The size of our dead zone is now controlled by the per-coefficient
+           quality thresholds returned by our HVS module.
+          We round down from a more accurate value when the quality of the
+           reconstruction does not fall below our threshold and it saves bits.
+          Hence, all of that VP3.2 code is gone from here, and the remaining
+           floating point code has been implemented as equivalent integer code
+           with exact precision.*/
+        qfac=(ogg_uint32_t)_qinfo->dc_scale[qi]*base[0];
+        /*For postprocessing, not dequantization.*/
+        if(_pp_dc_scale!=NULL)_pp_dc_scale[qi]=(int)(qfac/160);
+        /*Scale DC the coefficient from the proper table.*/
+        q=(qfac/100)<<2;
+        q=OC_CLAMPI(OC_DC_QUANT_MIN[qti],q,OC_QUANT_MAX);
+        _dequant[qi][pli][qti][0]=(ogg_uint16_t)q;
+        /*Now scale AC coefficients from the proper table.*/
+        for(zzi=1;zzi<64;zzi++){
+          q=((ogg_uint32_t)_qinfo->ac_scale[qi]*base[OC_FZIG_ZAG[zzi]]/100)<<2;
+          q=OC_CLAMPI(OC_AC_QUANT_MIN[qti],q,OC_QUANT_MAX);
+          _dequant[qi][pli][qti][zzi]=(ogg_uint16_t)q;
+        }
+        /*If this is a duplicate of a previous matrix, use that instead.
+          This simple check helps us improve cache coherency later.*/
+        {
+          int dupe;
+          int qtj;
+          int plj;
+          dupe=0;
+          for(qtj=0;qtj<=qti;qtj++){
+            for(plj=0;plj<(qtj<qti?3:pli);plj++){
+              if(!memcmp(_dequant[qi][pli][qti],_dequant[qi][plj][qtj],
+               sizeof(oc_quant_table))){
+                dupe=1;
+                break;
+              }
+            }
+            if(dupe)break;
+          }
+          if(dupe)_dequant[qi][pli][qti]=_dequant[qi][plj][qtj];
+        }
+        if(++qi>=qi_end)break;
+        /*Interpolate the next base matrix.*/
+        for(ci=0;ci<64;ci++){
+          base[ci]=(unsigned char)(
+           (2*((qi_end-qi)*_qinfo->qi_ranges[qti][pli].base_matrices[qri][ci]+
+           (qi-qi_start)*_qinfo->qi_ranges[qti][pli].base_matrices[qri+1][ci])
+           +_qinfo->qi_ranges[qti][pli].sizes[qri])/
+           (2*_qinfo->qi_ranges[qti][pli].sizes[qri]));
+        }
+      }
+    }
+  }
+}

lib/theoratest/lib/theora/quant.h

@@ -0,0 +1,33 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#if !defined(_quant_H)
+# define _quant_H (1)
+# include "theora/codec.h"
+# include "ocintrin.h"
+
+typedef ogg_uint16_t   oc_quant_table[64];
+
+
+/*Maximum scaled quantizer value.*/
+#define OC_QUANT_MAX          (1024<<2)
+
+
+void oc_dequant_tables_init(ogg_uint16_t *_dequant[64][3][2],
+ int _pp_dc_scale[64],const th_quant_info *_qinfo);
+
+#endif

lib/theoratest/lib/theora/state.h

@@ -0,0 +1,555 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id: internal.h 17337 2010-07-19 16:08:54Z tterribe $
+
+ ********************************************************************/
+#if !defined(_state_H)
+# define _state_H (1)
+# include "internal.h"
+# include "huffman.h"
+# include "quant.h"
+
+
+
+/*A single quadrant of the map from a super block to fragment numbers.*/
+typedef ptrdiff_t       oc_sb_map_quad[4];
+/*A map from a super block to fragment numbers.*/
+typedef oc_sb_map_quad  oc_sb_map[4];
+/*A single plane of the map from a macro block to fragment numbers.*/
+typedef ptrdiff_t       oc_mb_map_plane[4];
+/*A map from a macro block to fragment numbers.*/
+typedef oc_mb_map_plane oc_mb_map[3];
+/*A motion vector.*/
+typedef ogg_int16_t     oc_mv;
+
+typedef struct oc_sb_flags              oc_sb_flags;
+typedef struct oc_border_info           oc_border_info;
+typedef struct oc_fragment              oc_fragment;
+typedef struct oc_fragment_plane        oc_fragment_plane;
+typedef struct oc_base_opt_vtable       oc_base_opt_vtable;
+typedef struct oc_base_opt_data         oc_base_opt_data;
+typedef struct oc_state_dispatch_vtable oc_state_dispatch_vtable;
+typedef struct oc_theora_state          oc_theora_state;
+
+
+
+/*Shared accelerated functions.*/
+# if defined(OC_X86_ASM)
+#  if defined(_MSC_VER)
+#   include "x86_vc/x86int.h"
+#  else
+#   include "x86/x86int.h"
+#  endif
+# endif
+# if defined(OC_ARM_ASM)
+#  include "arm-intrinsics/armint.h"
+# endif
+# if defined(OC_C64X_ASM)
+#  include "c64x/c64xint.h"
+# endif
+# if defined(OC_WASM_SIMD128)
+#  include "wasm-simd/wav128int.h"
+# endif
+
+# if !defined(oc_state_accel_init)
+#  define oc_state_accel_init oc_state_accel_init_c
+# endif
+# if defined(OC_STATE_USE_VTABLE)
+#  if !defined(oc_frag_copy)
+#   define oc_frag_copy(_state,_dst,_src,_ystride) \
+  ((*(_state)->opt_vtable.frag_copy)(_dst,_src,_ystride))
+#  endif
+#  if !defined(oc_frag_copy_list)
+#   define oc_frag_copy_list(_state,_dst_frame,_src_frame,_ystride, \
+ _fragis,_nfragis,_frag_buf_offs) \
+ ((*(_state)->opt_vtable.frag_copy_list)(_dst_frame,_src_frame,_ystride, \
+  _fragis,_nfragis,_frag_buf_offs))
+#  endif
+#  if !defined(oc_frag_recon_intra)
+#   define oc_frag_recon_intra(_state,_dst,_dst_ystride,_residue) \
+  ((*(_state)->opt_vtable.frag_recon_intra)(_dst,_dst_ystride,_residue))
+#  endif
+#  if !defined(oc_frag_recon_inter)
+#   define oc_frag_recon_inter(_state,_dst,_src,_ystride,_residue) \
+  ((*(_state)->opt_vtable.frag_recon_inter)(_dst,_src,_ystride,_residue))
+#  endif
+#  if !defined(oc_frag_recon_inter2)
+#   define oc_frag_recon_inter2(_state,_dst,_src1,_src2,_ystride,_residue) \
+  ((*(_state)->opt_vtable.frag_recon_inter2)(_dst, \
+   _src1,_src2,_ystride,_residue))
+#  endif
+# if !defined(oc_idct8x8)
+#   define oc_idct8x8(_state,_y,_x,_last_zzi) \
+  ((*(_state)->opt_vtable.idct8x8)(_y,_x,_last_zzi))
+#  endif
+#  if !defined(oc_state_frag_recon)
+#   define oc_state_frag_recon(_state,_fragi, \
+ _pli,_dct_coeffs,_last_zzi,_dc_quant) \
+  ((*(_state)->opt_vtable.state_frag_recon)(_state,_fragi, \
+   _pli,_dct_coeffs,_last_zzi,_dc_quant))
+#  endif
+#  if !defined(oc_loop_filter_init)
+#   define oc_loop_filter_init(_state,_bv,_flimit) \
+  ((*(_state)->opt_vtable.loop_filter_init)(_bv,_flimit))
+#  endif
+#  if !defined(oc_state_loop_filter_frag_rows)
+#   define oc_state_loop_filter_frag_rows(_state, \
+ _bv,_refi,_pli,_fragy0,_fragy_end) \
+  ((*(_state)->opt_vtable.state_loop_filter_frag_rows)(_state, \
+   _bv,_refi,_pli,_fragy0,_fragy_end))
+#  endif
+#  if !defined(oc_restore_fpu)
+#   define oc_restore_fpu(_state) \
+  ((*(_state)->opt_vtable.restore_fpu)())
+#  endif
+# else
+#  if !defined(oc_frag_copy)
+#   define oc_frag_copy(_state,_dst,_src,_ystride) \
+  oc_frag_copy_c(_dst,_src,_ystride)
+#  endif
+#  if !defined(oc_frag_copy_list)
+#   define oc_frag_copy_list(_state,_dst_frame,_src_frame,_ystride, \
+ _fragis,_nfragis,_frag_buf_offs) \
+  oc_frag_copy_list_c(_dst_frame,_src_frame,_ystride, \
+  _fragis,_nfragis,_frag_buf_offs)
+#  endif
+#  if !defined(oc_frag_recon_intra)
+#   define oc_frag_recon_intra(_state,_dst,_dst_ystride,_residue) \
+  oc_frag_recon_intra_c(_dst,_dst_ystride,_residue)
+#  endif
+#  if !defined(oc_frag_recon_inter)
+#   define oc_frag_recon_inter(_state,_dst,_src,_ystride,_residue) \
+  oc_frag_recon_inter_c(_dst,_src,_ystride,_residue)
+#  endif
+#  if !defined(oc_frag_recon_inter2)
+#   define oc_frag_recon_inter2(_state,_dst,_src1,_src2,_ystride,_residue) \
+  oc_frag_recon_inter2_c(_dst,_src1,_src2,_ystride,_residue)
+#  endif
+#  if !defined(oc_idct8x8)
+#   define oc_idct8x8(_state,_y,_x,_last_zzi) oc_idct8x8_c(_y,_x,_last_zzi)
+#  endif
+#  if !defined(oc_state_frag_recon)
+#   define oc_state_frag_recon oc_state_frag_recon_c
+#  endif
+#  if !defined(oc_loop_filter_init)
+#   define oc_loop_filter_init(_state,_bv,_flimit) \
+  oc_loop_filter_init_c(_bv,_flimit)
+#  endif
+#  if !defined(oc_state_loop_filter_frag_rows)
+#   define oc_state_loop_filter_frag_rows oc_state_loop_filter_frag_rows_c
+#  endif
+#  if !defined(oc_restore_fpu)
+#   define oc_restore_fpu(_state) do{}while(0)
+#  endif
+# endif
+
+
+
+/*A keyframe.*/
+# define OC_INTRA_FRAME (0)
+/*A predicted frame.*/
+# define OC_INTER_FRAME (1)
+/*A frame of unknown type (frame type decision has not yet been made).*/
+# define OC_UNKWN_FRAME (-1)
+
+/*The amount of padding to add to the reconstructed frame buffers on all
+   sides.
+  This is used to allow unrestricted motion vectors without special casing.
+  This must be a multiple of 2.*/
+# define OC_UMV_PADDING (16)
+
+/*Frame classification indices.*/
+/*The previous golden frame.*/
+# define OC_FRAME_GOLD      (0)
+/*The previous frame.*/
+# define OC_FRAME_PREV      (1)
+/*The current frame.*/
+# define OC_FRAME_SELF      (2)
+/*Used to mark uncoded fragments (for DC prediction).*/
+# define OC_FRAME_NONE      (3)
+
+/*The input or output buffer.*/
+# define OC_FRAME_IO        (3)
+/*Uncompressed prev golden frame.*/
+# define OC_FRAME_GOLD_ORIG (4)
+/*Uncompressed previous frame. */
+# define OC_FRAME_PREV_ORIG (5)
+
+/*Macroblock modes.*/
+/*Macro block is invalid: It is never coded.*/
+# define OC_MODE_INVALID        (-1)
+/*Encoded difference from the same macro block in the previous frame.*/
+# define OC_MODE_INTER_NOMV     (0)
+/*Encoded with no motion compensated prediction.*/
+# define OC_MODE_INTRA          (1)
+/*Encoded difference from the previous frame offset by the given motion
+   vector.*/
+# define OC_MODE_INTER_MV       (2)
+/*Encoded difference from the previous frame offset by the last coded motion
+   vector.*/
+# define OC_MODE_INTER_MV_LAST  (3)
+/*Encoded difference from the previous frame offset by the second to last
+   coded motion vector.*/
+# define OC_MODE_INTER_MV_LAST2 (4)
+/*Encoded difference from the same macro block in the previous golden
+   frame.*/
+# define OC_MODE_GOLDEN_NOMV    (5)
+/*Encoded difference from the previous golden frame offset by the given motion
+   vector.*/
+# define OC_MODE_GOLDEN_MV      (6)
+/*Encoded difference from the previous frame offset by the individual motion
+   vectors given for each block.*/
+# define OC_MODE_INTER_MV_FOUR  (7)
+/*The number of (coded) modes.*/
+# define OC_NMODES              (8)
+
+/*Determines the reference frame used for a given MB mode.*/
+# define OC_FRAME_FOR_MODE(_x) \
+ OC_UNIBBLE_TABLE32(OC_FRAME_PREV,OC_FRAME_SELF,OC_FRAME_PREV,OC_FRAME_PREV, \
+  OC_FRAME_PREV,OC_FRAME_GOLD,OC_FRAME_GOLD,OC_FRAME_PREV,(_x))
+
+/*Constants for the packet state machine common between encoder and decoder.*/
+
+/*Next packet to emit/read: Codec info header.*/
+# define OC_PACKET_INFO_HDR    (-3)
+/*Next packet to emit/read: Comment header.*/
+# define OC_PACKET_COMMENT_HDR (-2)
+/*Next packet to emit/read: Codec setup header.*/
+# define OC_PACKET_SETUP_HDR   (-1)
+/*No more packets to emit/read.*/
+# define OC_PACKET_DONE        (INT_MAX)
+
+
+
+#define OC_MV(_x,_y)         ((oc_mv)((_x)&0xFF|(_y)<<8))
+#define OC_MV_X(_mv)         ((signed char)(_mv))
+#define OC_MV_Y(_mv)         ((_mv)>>8)
+#define OC_MV_ADD(_mv1,_mv2) \
+  OC_MV(OC_MV_X(_mv1)+OC_MV_X(_mv2), \
+   OC_MV_Y(_mv1)+OC_MV_Y(_mv2))
+#define OC_MV_SUB(_mv1,_mv2) \
+  OC_MV(OC_MV_X(_mv1)-OC_MV_X(_mv2), \
+   OC_MV_Y(_mv1)-OC_MV_Y(_mv2))
+
+
+
+/*Super blocks are 32x32 segments of pixels in a single color plane indexed
+   in image order.
+  Internally, super blocks are broken up into four quadrants, each of which
+   contains a 2x2 pattern of blocks, each of which is an 8x8 block of pixels.
+  Quadrants, and the blocks within them, are indexed in a special order called
+   a "Hilbert curve" within the super block.
+
+  In order to differentiate between the Hilbert-curve indexing strategy and
+   the regular image order indexing strategy, blocks indexed in image order
+   are called "fragments".
+  Fragments are indexed in image order, left to right, then bottom to top,
+   from Y' plane to Cb plane to Cr plane.
+
+  The co-located fragments in all image planes corresponding to the location
+   of a single quadrant of a luma plane super block form a macro block.
+  Thus there is only a single set of macro blocks for all planes, each of which
+   contains between 6 and 12 fragments, depending on the pixel format.
+  Therefore macro block information is kept in a separate set of arrays from
+   super blocks to avoid unused space in the other planes.
+  The lists are indexed in super block order.
+  That is, the macro block corresponding to the macro block mbi in (luma plane)
+   super block sbi is at index (sbi<<2|mbi).
+  Thus the number of macro blocks in each dimension is always twice the number
+   of super blocks, even when only an odd number fall inside the coded frame.
+  These "extra" macro blocks are just an artifact of our internal data layout,
+   and not part of the coded stream; they are flagged with a negative MB mode.*/
+
+
+
+/*Super block information.*/
+struct oc_sb_flags{
+  unsigned char coded_fully:1;
+  unsigned char coded_partially:1;
+  unsigned char quad_valid:4;
+};
+
+
+
+/*Information about a fragment which intersects the border of the displayable
+   region.
+  This marks which pixels belong to the displayable region.*/
+struct oc_border_info{
+  /*A bit mask marking which pixels are in the displayable region.
+    Pixel (x,y) corresponds to bit (y<<3|x).*/
+  ogg_int64_t mask;
+  /*The number of pixels in the displayable region.
+    This is always positive, and always less than 64.*/
+  int         npixels;
+};
+
+
+
+/*Fragment information.*/
+struct oc_fragment{
+  /*A flag indicating whether or not this fragment is coded.*/
+  unsigned   coded:1;
+  /*A flag indicating that this entire fragment lies outside the displayable
+     region of the frame.
+    Note the contrast with an invalid macro block, which is outside the coded
+     frame, not just the displayable one.
+    There are no fragments outside the coded frame by construction.*/
+  unsigned   invalid:1;
+  /*The index of the quality index used for this fragment's AC coefficients.*/
+  unsigned   qii:4;
+  /*The index of the reference frame this fragment is predicted from.*/
+  unsigned   refi:2;
+  /*The mode of the macroblock this fragment belongs to.*/
+  unsigned   mb_mode:3;
+  /*The index of the associated border information for fragments which lie
+     partially outside the displayable region.
+    For fragments completely inside or outside this region, this is -1.
+    Note that the C standard requires an explicit signed keyword for bitfield
+     types, since some compilers may treat them as unsigned without it.*/
+  signed int borderi:5;
+  /*The prediction-corrected DC component.
+    Note that the C standard requires an explicit signed keyword for bitfield
+     types, since some compilers may treat them as unsigned without it.*/
+  signed int dc:16;
+};
+
+
+
+/*A description of each fragment plane.*/
+struct oc_fragment_plane{
+  /*The number of fragments in the horizontal direction.*/
+  int       nhfrags;
+  /*The number of fragments in the vertical direction.*/
+  int       nvfrags;
+  /*The offset of the first fragment in the plane.*/
+  ptrdiff_t froffset;
+  /*The total number of fragments in the plane.*/
+  ptrdiff_t nfrags;
+  /*The number of super blocks in the horizontal direction.*/
+  unsigned  nhsbs;
+  /*The number of super blocks in the vertical direction.*/
+  unsigned  nvsbs;
+  /*The offset of the first super block in the plane.*/
+  unsigned  sboffset;
+  /*The total number of super blocks in the plane.*/
+  unsigned  nsbs;
+};
+
+
+typedef void (*oc_state_loop_filter_frag_rows_func)(
+ const oc_theora_state *_state,signed char _bv[256],int _refi,int _pli,
+ int _fragy0,int _fragy_end);
+
+/*The shared (encoder and decoder) functions that have accelerated variants.*/
+struct oc_base_opt_vtable{
+  void (*frag_copy)(unsigned char *_dst,
+   const unsigned char *_src,int _ystride);
+  void (*frag_copy_list)(unsigned char *_dst_frame,
+   const unsigned char *_src_frame,int _ystride,
+   const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs);
+  void (*frag_recon_intra)(unsigned char *_dst,int _ystride,
+   const ogg_int16_t _residue[64]);
+  void (*frag_recon_inter)(unsigned char *_dst,
+   const unsigned char *_src,int _ystride,const ogg_int16_t _residue[64]);
+  void (*frag_recon_inter2)(unsigned char *_dst,const unsigned char *_src1,
+   const unsigned char *_src2,int _ystride,const ogg_int16_t _residue[64]);
+  void (*idct8x8)(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi);
+  void (*state_frag_recon)(const oc_theora_state *_state,ptrdiff_t _fragi,
+   int _pli,ogg_int16_t _dct_coeffs[128],int _last_zzi,ogg_uint16_t _dc_quant);
+  void (*loop_filter_init)(signed char _bv[256],int _flimit);
+  oc_state_loop_filter_frag_rows_func state_loop_filter_frag_rows;
+  void (*restore_fpu)(void);
+};
+
+/*The shared (encoder and decoder) tables that vary according to which variants
+   of the above functions are used.*/
+struct oc_base_opt_data{
+  const unsigned char *dct_fzig_zag;
+};
+
+
+/*State information common to both the encoder and decoder.*/
+struct oc_theora_state{
+  /*The stream information.*/
+  th_info             info;
+# if defined(OC_STATE_USE_VTABLE)
+  /*Table for shared accelerated functions.*/
+  oc_base_opt_vtable  opt_vtable;
+# endif
+  /*Table for shared data used by accelerated functions.*/
+  oc_base_opt_data    opt_data;
+  /*CPU flags to detect the presence of extended instruction sets.*/
+  ogg_uint32_t        cpu_flags;
+  /*The fragment plane descriptions.*/
+  oc_fragment_plane   fplanes[3];
+  /*The list of fragments, indexed in image order.*/
+  oc_fragment        *frags;
+  /*The the offset into the reference frame buffer to the upper-left pixel of
+     each fragment.*/
+  ptrdiff_t          *frag_buf_offs;
+  /*The motion vector for each fragment.*/
+  oc_mv              *frag_mvs;
+  /*The total number of fragments in a single frame.*/
+  ptrdiff_t           nfrags;
+  /*The list of super block maps, indexed in image order.*/
+  oc_sb_map          *sb_maps;
+  /*The list of super block flags, indexed in image order.*/
+  oc_sb_flags        *sb_flags;
+  /*The total number of super blocks in a single frame.*/
+  unsigned            nsbs;
+  /*The fragments from each color plane that belong to each macro block.
+    Fragments are stored in image order (left to right then top to bottom).
+    When chroma components are decimated, the extra fragments have an index of
+     -1.*/
+  oc_mb_map          *mb_maps;
+  /*The list of macro block modes.
+    A negative number indicates the macro block lies entirely outside the
+     coded frame.*/
+  signed char        *mb_modes;
+  /*The number of macro blocks in the X direction.*/
+  unsigned            nhmbs;
+  /*The number of macro blocks in the Y direction.*/
+  unsigned            nvmbs;
+  /*The total number of macro blocks.*/
+  size_t              nmbs;
+  /*The list of coded fragments, in coded order.
+    Uncoded fragments are stored in reverse order from the end of the list.*/
+  ptrdiff_t          *coded_fragis;
+  /*The number of coded fragments in each plane.*/
+  ptrdiff_t           ncoded_fragis[3];
+  /*The total number of coded fragments.*/
+  ptrdiff_t           ntotal_coded_fragis;
+  /*The actual buffers used for the reference frames.*/
+  th_ycbcr_buffer     ref_frame_bufs[6];
+  /*The index of the buffers being used for each OC_FRAME_* reference frame.*/
+  int                 ref_frame_idx[6];
+  /*The storage for the reference frame buffers.
+    This is just ref_frame_bufs[ref_frame_idx[i]][0].data, but is cached here
+     for faster look-up.*/
+  unsigned char      *ref_frame_data[6];
+  /*The handle used to allocate the reference frame buffers.*/
+  unsigned char      *ref_frame_handle;
+  /*The strides for each plane in the reference frames.*/
+  int                 ref_ystride[3];
+  /*The number of unique border patterns.*/
+  int                 nborders;
+  /*The unique border patterns for all border fragments.
+    The borderi field of fragments which straddle the border indexes this
+     list.*/
+  oc_border_info      borders[16];
+  /*The frame number of the last keyframe.*/
+  ogg_int64_t         keyframe_num;
+  /*The frame number of the current frame.*/
+  ogg_int64_t         curframe_num;
+  /*The granpos of the current frame.*/
+  ogg_int64_t         granpos;
+  /*The type of the current frame.*/
+  signed char         frame_type;
+  /*The bias to add to the frame count when computing granule positions.*/
+  unsigned char       granpos_bias;
+  /*The number of quality indices used in the current frame.*/
+  unsigned char       nqis;
+  /*The quality indices of the current frame.*/
+  unsigned char       qis[3];
+  /*The dequantization tables, stored in zig-zag order, and indexed by
+     qi, pli, qti, and zzi.*/
+  ogg_uint16_t       *dequant_tables[64][3][2];
+  OC_ALIGN16(oc_quant_table      dequant_table_data[64][3][2]);
+  /*Loop filter strength parameters.*/
+  unsigned char       loop_filter_limits[64];
+};
+
+
+
+/*The function type used to fill in the chroma plane motion vectors for a
+   macro block when 4 different motion vectors are specified in the luma
+   plane.
+  _cbmvs: The chroma block-level motion vectors to fill in.
+  _lmbmv: The luma macro-block level motion vector to fill in for use in
+           prediction.
+  _lbmvs: The luma block-level motion vectors.*/
+typedef void (*oc_set_chroma_mvs_func)(oc_mv _cbmvs[4],const oc_mv _lbmvs[4]);
+
+
+
+/*A table of functions used to fill in the Cb,Cr plane motion vectors for a
+   macro block when 4 different motion vectors are specified in the luma
+   plane.*/
+extern const oc_set_chroma_mvs_func OC_SET_CHROMA_MVS_TABLE[TH_PF_NFORMATS];
+
+
+
+int oc_state_init(oc_theora_state *_state,const th_info *_info,int _nrefs);
+void oc_state_clear(oc_theora_state *_state);
+void oc_state_accel_init_c(oc_theora_state *_state);
+void oc_state_borders_fill_rows(oc_theora_state *_state,int _refi,int _pli,
+ int _y0,int _yend);
+void oc_state_borders_fill_caps(oc_theora_state *_state,int _refi,int _pli);
+void oc_state_borders_fill(oc_theora_state *_state,int _refi);
+void oc_state_fill_buffer_ptrs(oc_theora_state *_state,int _buf_idx,
+ th_ycbcr_buffer _img);
+int oc_state_mbi_for_pos(oc_theora_state *_state,int _mbx,int _mby);
+int oc_state_get_mv_offsets(const oc_theora_state *_state,int _offsets[2],
+ int _pli,oc_mv _mv);
+
+void oc_loop_filter_init_c(signed char _bv[256],int _flimit);
+void oc_state_loop_filter(oc_theora_state *_state,int _frame);
+# if defined(OC_DUMP_IMAGES)
+int oc_state_dump_frame(const oc_theora_state *_state,int _frame,
+ const char *_suf);
+# endif
+
+/*Default pure-C implementations of shared accelerated functions.*/
+void oc_frag_copy_c(unsigned char *_dst,
+ const unsigned char *_src,int _src_ystride);
+void oc_frag_copy_list_c(unsigned char *_dst_frame,
+ const unsigned char *_src_frame,int _ystride,
+ const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs);
+void oc_frag_recon_intra_c(unsigned char *_dst,int _dst_ystride,
+ const ogg_int16_t _residue[64]);
+void oc_frag_recon_inter_c(unsigned char *_dst,
+ const unsigned char *_src,int _ystride,const ogg_int16_t _residue[64]);
+void oc_frag_recon_inter2_c(unsigned char *_dst,const unsigned char *_src1,
+ const unsigned char *_src2,int _ystride,const ogg_int16_t _residue[64]);
+void oc_idct8x8_c(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi);
+void oc_state_frag_recon_c(const oc_theora_state *_state,ptrdiff_t _fragi,
+ int _pli,ogg_int16_t _dct_coeffs[128],int _last_zzi,ogg_uint16_t _dc_quant);
+void oc_state_loop_filter_frag_rows_c(const oc_theora_state *_state,
+ signed char _bv[256],int _refi,int _pli,int _fragy0,int _fragy_end);
+void oc_restore_fpu_c(void);
+
+/*We need a way to call a few encoder functions without introducing a link-time
+   dependency into the decoder, while still allowing the old alpha API which
+   does not distinguish between encoder and decoder objects to be used.
+  We do this by placing a function table at the start of the encoder object
+   which can dispatch into the encoder library.
+  We do a similar thing for the decoder in case we ever decide to split off a
+   common base library.*/
+typedef void (*oc_state_clear_func)(theora_state *_th);
+typedef int (*oc_state_control_func)(theora_state *th,int _req,
+ void *_buf,size_t _buf_sz);
+typedef ogg_int64_t (*oc_state_granule_frame_func)(theora_state *_th,
+ ogg_int64_t _granulepos);
+typedef double (*oc_state_granule_time_func)(theora_state *_th,
+ ogg_int64_t _granulepos);
+
+
+struct oc_state_dispatch_vtable{
+  oc_state_clear_func         clear;
+  oc_state_control_func       control;
+  oc_state_granule_frame_func granule_frame;
+  oc_state_granule_time_func  granule_time;
+};
+
+#endif

lib/theoratest/lib/theora/theora.h

@@ -0,0 +1,786 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+  last mod: $Id: theora.h,v 1.17 2003/12/06 18:06:19 arc Exp $
+
+ ********************************************************************/
+
+#ifndef _O_THEORA_H_
+#define _O_THEORA_H_
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif /* __cplusplus */
+
+#include <stddef.h>	/* for size_t */
+
+#include <ogg/ogg.h>
+
+/** \file
+ * The libtheora pre-1.0 legacy C API.
+ *
+ * \ingroup oldfuncs
+ *
+ * \section intro Introduction
+ *
+ * This is the documentation for the libtheora legacy C API, declared in
+ * the theora.h header, which describes the old interface used before
+ * the 1.0 release. This API was widely deployed for several years and
+ * remains supported, but for new code we recommend the cleaner API
+ * declared in theoradec.h and theoraenc.h.
+ *
+ * libtheora is the reference implementation for
+ * <a href="http://www.theora.org/">Theora</a>, a free video codec.
+ * Theora is derived from On2's VP3 codec with improved integration with
+ * Ogg multimedia formats by <a href="http://www.xiph.org/">Xiph.Org</a>.
+ *
+ * \section overview Overview
+ *
+ * This library will both decode and encode theora packets to/from raw YUV
+ * frames.  In either case, the packets will most likely either come from or
+ * need to be embedded in an Ogg stream.  Use
+ * <a href="http://xiph.org/ogg/">libogg</a> or
+ * <a href="http://www.annodex.net/software/liboggz/index.html">liboggz</a>
+ * to extract/package these packets.
+ *
+ * \section decoding Decoding Process
+ *
+ * Decoding can be separated into the following steps:
+ * -# initialise theora_info and theora_comment structures using
+ *    theora_info_init() and theora_comment_init():
+ \verbatim
+ theora_info     info;
+ theora_comment  comment;
+
+ theora_info_init(&info);
+ theora_comment_init(&comment);
+ \endverbatim
+ * -# retrieve header packets from Ogg stream (there should be 3) and decode
+ *    into theora_info and theora_comment structures using
+ *    theora_decode_header().  See \ref identification for more information on
+ *    identifying which packets are theora packets.
+ \verbatim
+ int i;
+ for (i = 0; i < 3; i++)
+ {
+   (get a theora packet "op" from the Ogg stream)
+   theora_decode_header(&info, &comment, op);
+ }
+ \endverbatim
+ * -# initialise the decoder based on the information retrieved into the
+ *    theora_info struct by theora_decode_header().  You will need a
+ *    theora_state struct.
+ \verbatim
+ theora_state state;
+
+ theora_decode_init(&state, &info);
+ \endverbatim
+ * -# pass in packets and retrieve decoded frames!  See the yuv_buffer
+ *    documentation for information on how to retrieve raw YUV data.
+ \verbatim
+ yuf_buffer buffer;
+ while (last packet was not e_o_s) {
+   (get a theora packet "op" from the Ogg stream)
+   theora_decode_packetin(&state, op);
+   theora_decode_YUVout(&state, &buffer);
+ }
+ \endverbatim
+ *
+ *
+ * \subsection identification Identifying Theora Packets
+ *
+ * All streams inside an Ogg file have a unique serial_no attached to the
+ * stream.  Typically, you will want to
+ *  - retrieve the serial_no for each b_o_s (beginning of stream) page
+ *    encountered within the Ogg file;
+ *  - test the first (only) packet on that page to determine if it is a theora
+ *    packet;
+ *  - once you have found a theora b_o_s page then use the retrieved serial_no
+ *    to identify future packets belonging to the same theora stream.
+ *
+ * Note that you \e cannot use theora_packet_isheader() to determine if a
+ * packet is a theora packet or not, as this function does not perform any
+ * checking beyond whether a header bit is present.  Instead, use the
+ * theora_decode_header() function and check the return value; or examine the
+ * header bytes at the beginning of the Ogg page.
+ */
+
+
+/** \defgroup oldfuncs Legacy pre-1.0 C API */
+/*  @{ */
+
+/**
+ * A YUV buffer for passing uncompressed frames to and from the codec.
+ * This holds a Y'CbCr frame in planar format. The CbCr planes can be
+ * subsampled and have their own separate dimensions and row stride
+ * offsets. Note that the strides may be negative in some
+ * configurations. For theora the width and height of the largest plane
+ * must be a multiple of 16. The actual meaningful picture size and
+ * offset are stored in the theora_info structure; frames returned by
+ * the decoder may need to be cropped for display.
+ *
+ * All samples are 8 bits. Within each plane samples are ordered by
+ * row from the top of the frame to the bottom. Within each row samples
+ * are ordered from left to right.
+ *
+ * During decode, the yuv_buffer struct is allocated by the user, but all
+ * fields (including luma and chroma pointers) are filled by the library.
+ * These pointers address library-internal memory and their contents should
+ * not be modified.
+ *
+ * Conversely, during encode the user allocates the struct and fills out all
+ * fields.  The user also manages the data addressed by the luma and chroma
+ * pointers.  See the encoder_example.c and dump_video.c example files in
+ * theora/examples/ for more information.
+ */
+typedef struct {
+    int   y_width;      /**< Width of the Y' luminance plane */
+    int   y_height;     /**< Height of the luminance plane */
+    int   y_stride;     /**< Offset in bytes between successive rows */
+
+    int   uv_width;     /**< Width of the Cb and Cr chroma planes */
+    int   uv_height;    /**< Height of the chroma planes */
+    int   uv_stride;    /**< Offset between successive chroma rows */
+    unsigned char *y;   /**< Pointer to start of luminance data */
+    unsigned char *u;   /**< Pointer to start of Cb data */
+    unsigned char *v;   /**< Pointer to start of Cr data */
+
+} yuv_buffer;
+
+/**
+ * A Colorspace.
+ */
+typedef enum {
+  OC_CS_UNSPECIFIED,    /**< The colorspace is unknown or unspecified */
+  OC_CS_ITU_REC_470M,   /**< This is the best option for 'NTSC' content */
+  OC_CS_ITU_REC_470BG,  /**< This is the best option for 'PAL' content */
+  OC_CS_NSPACES         /**< This marks the end of the defined colorspaces */
+} theora_colorspace;
+
+/**
+ * A Chroma subsampling
+ *
+ * These enumerate the available chroma subsampling options supported
+ * by the theora format. See Section 4.4 of the specification for
+ * exact definitions.
+ */
+typedef enum {
+  OC_PF_420,    /**< Chroma subsampling by 2 in each direction (4:2:0) */
+  OC_PF_RSVD,   /**< Reserved value */
+  OC_PF_422,    /**< Horizonatal chroma subsampling by 2 (4:2:2) */
+  OC_PF_444     /**< No chroma subsampling at all (4:4:4) */
+} theora_pixelformat;
+
+/**
+ * Theora bitstream info.
+ * Contains the basic playback parameters for a stream,
+ * corresponding to the initial 'info' header packet.
+ *
+ * Encoded theora frames must be a multiple of 16 in width and height.
+ * To handle other frame sizes, a crop rectangle is specified in
+ * frame_height and frame_width, offset_x and * offset_y. The offset
+ * and size should still be a multiple of 2 to avoid chroma sampling
+ * shifts. Offset values in this structure are measured from the
+ * upper left of the image.
+ *
+ * Frame rate, in frames per second, is stored as a rational
+ * fraction. Aspect ratio is also stored as a rational fraction, and
+ * refers to the aspect ratio of the frame pixels, not of the
+ * overall frame itself.
+ *
+ * See <a href="http://svn.xiph.org/trunk/theora/examples/encoder_example.c">
+ * examples/encoder_example.c</a> for usage examples of the
+ * other parameters and good default settings for the encoder parameters.
+ */
+typedef struct {
+  ogg_uint32_t  width;		/**< encoded frame width  */
+  ogg_uint32_t  height;		/**< encoded frame height */
+  ogg_uint32_t  frame_width;	/**< display frame width  */
+  ogg_uint32_t  frame_height;	/**< display frame height */
+  ogg_uint32_t  offset_x;	/**< horizontal offset of the displayed frame */
+  ogg_uint32_t  offset_y;	/**< vertical offset of the displayed frame */
+  ogg_uint32_t  fps_numerator;	    /**< frame rate numerator **/
+  ogg_uint32_t  fps_denominator;    /**< frame rate denominator **/
+  ogg_uint32_t  aspect_numerator;   /**< pixel aspect ratio numerator */
+  ogg_uint32_t  aspect_denominator; /**< pixel aspect ratio denominator */
+  theora_colorspace colorspace;	    /**< colorspace */
+  int           target_bitrate;	    /**< nominal bitrate in bits per second */
+  int           quality;  /**< Nominal quality setting, 0-63 */
+  int           quick_p;  /**< Quick encode/decode */
+
+  /* decode only */
+  unsigned char version_major;
+  unsigned char version_minor;
+  unsigned char version_subminor;
+
+  void *codec_setup;
+
+  /* encode only */
+  int           dropframes_p;
+  int           keyframe_auto_p;
+  ogg_uint32_t  keyframe_frequency;
+  ogg_uint32_t  keyframe_frequency_force;  /* also used for decode init to
+                                              get granpos shift correct */
+  ogg_uint32_t  keyframe_data_target_bitrate;
+  ogg_int32_t   keyframe_auto_threshold;
+  ogg_uint32_t  keyframe_mindistance;
+  ogg_int32_t   noise_sensitivity;
+  ogg_int32_t   sharpness;
+
+  theora_pixelformat pixelformat;	/**< chroma subsampling mode to expect */
+
+} theora_info;
+
+/** Codec internal state and context.
+ */
+typedef struct{
+  theora_info *i;
+  ogg_int64_t granulepos;
+
+  void *internal_encode;
+  void *internal_decode;
+
+} theora_state;
+
+/**
+ * Comment header metadata.
+ *
+ * This structure holds the in-stream metadata corresponding to
+ * the 'comment' header packet.
+ *
+ * Meta data is stored as a series of (tag, value) pairs, in
+ * length-encoded string vectors. The first occurence of the
+ * '=' character delimits the tag and value. A particular tag
+ * may occur more than once. The character set encoding for
+ * the strings is always UTF-8, but the tag names are limited
+ * to case-insensitive ASCII. See the spec for details.
+ *
+ * In filling in this structure, theora_decode_header() will
+ * null-terminate the user_comment strings for safety. However,
+ * the bitstream format itself treats them as 8-bit clean,
+ * and so the length array should be treated as authoritative
+ * for their length.
+ */
+typedef struct theora_comment{
+  char **user_comments;         /**< An array of comment string vectors */
+  int   *comment_lengths;       /**< An array of corresponding string vector lengths in bytes */
+  int    comments;              /**< The total number of comment string vectors */
+  char  *vendor;                /**< The vendor string identifying the encoder, null terminated */
+
+} theora_comment;
+
+
+/**\name theora_control() codes */
+/* \anchor decctlcodes_old
+ * These are the available request codes for theora_control()
+ * when called with a decoder instance.
+ * By convention decoder control codes are odd, to distinguish
+ * them from \ref encctlcodes_old "encoder control codes" which
+ * are even.
+ *
+ * Note that since the 1.0 release, both the legacy and the final
+ * implementation accept all the same control codes, but only the
+ * final API declares the newer codes.
+ *
+ * Keep any experimental or vendor-specific values above \c 0x8000.*/
+
+/*@{*/
+
+/**Get the maximum post-processing level.
+ * The decoder supports a post-processing filter that can improve
+ * the appearance of the decoded images. This returns the highest
+ * level setting for this post-processor, corresponding to maximum
+ * improvement and computational expense.
+ */
+#define TH_DECCTL_GET_PPLEVEL_MAX (1)
+
+/**Set the post-processing level.
+ * Sets the level of post-processing to use when decoding the
+ * compressed stream. This must be a value between zero (off)
+ * and the maximum returned by TH_DECCTL_GET_PPLEVEL_MAX.
+ */
+#define TH_DECCTL_SET_PPLEVEL (3)
+
+/**Sets the maximum distance between key frames.
+ * This can be changed during an encode, but will be bounded by
+ *  <tt>1<<th_info#keyframe_granule_shift</tt>.
+ * If it is set before encoding begins, th_info#keyframe_granule_shift will
+ *  be enlarged appropriately.
+ *
+ * \param[in]  buf <tt>ogg_uint32_t</tt>: The maximum distance between key
+ *                   frames.
+ * \param[out] buf <tt>ogg_uint32_t</tt>: The actual maximum distance set.
+ * \retval OC_FAULT  \a theora_state or \a buf is <tt>NULL</tt>.
+ * \retval OC_EINVAL \a buf_sz is not <tt>sizeof(ogg_uint32_t)</tt>.
+ * \retval OC_IMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_SET_KEYFRAME_FREQUENCY_FORCE (4)
+
+/**Set the granule position.
+ * Call this after a seek, to update the internal granulepos
+ * in the decoder, to insure that subsequent frames are marked
+ * properly. If you track timestamps yourself and do not use
+ * the granule postion returned by the decoder, then you do
+ * not need to use this control.
+ */
+#define TH_DECCTL_SET_GRANPOS (5)
+
+/**\anchor encctlcodes_old */
+
+/**Sets the quantization parameters to use.
+ * The parameters are copied, not stored by reference, so they can be freed
+ *  after this call.
+ * <tt>NULL</tt> may be specified to revert to the default parameters.
+ *
+ * \param[in] buf #th_quant_info
+ * \retval OC_FAULT  \a theora_state is <tt>NULL</tt>.
+ * \retval OC_EINVAL Encoding has already begun, the quantization parameters
+ *                    are not acceptable to this version of the encoder,
+ *                    \a buf is <tt>NULL</tt> and \a buf_sz is not zero,
+ *                    or \a buf is non-<tt>NULL</tt> and \a buf_sz is
+ *                    not <tt>sizeof(#th_quant_info)</tt>.
+ * \retval OC_IMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_SET_QUANT_PARAMS (2)
+
+/**Disables any encoder features that would prevent lossless transcoding back
+ *  to VP3.
+ * This primarily means disabling block-level QI values and not using 4MV mode
+ *  when any of the luma blocks in a macro block are not coded.
+ * It also includes using the VP3 quantization tables and Huffman codes; if you
+ *  set them explicitly after calling this function, the resulting stream will
+ *  not be VP3-compatible.
+ * If you enable VP3-compatibility when encoding 4:2:2 or 4:4:4 source
+ *  material, or when using a picture region smaller than the full frame (e.g.
+ *  a non-multiple-of-16 width or height), then non-VP3 bitstream features will
+ *  still be disabled, but the stream will still not be VP3-compatible, as VP3
+ *  was not capable of encoding such formats.
+ * If you call this after encoding has already begun, then the quantization
+ *  tables and codebooks cannot be changed, but the frame-level features will
+ *  be enabled or disabled as requested.
+ *
+ * \param[in]  buf <tt>int</tt>: a non-zero value to enable VP3 compatibility,
+ *                   or 0 to disable it (the default).
+ * \param[out] buf <tt>int</tt>: 1 if all bitstream features required for
+ *                   VP3-compatibility could be set, and 0 otherwise.
+ *                  The latter will be returned if the pixel format is not
+ *                   4:2:0, the picture region is smaller than the full frame,
+ *                   or if encoding has begun, preventing the quantization
+ *                   tables and codebooks from being set.
+ * \retval OC_FAULT  \a theora_state or \a buf is <tt>NULL</tt>.
+ * \retval OC_EINVAL \a buf_sz is not <tt>sizeof(int)</tt>.
+ * \retval OC_IMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_SET_VP3_COMPATIBLE (10)
+
+/**Gets the maximum speed level.
+ * Higher speed levels favor quicker encoding over better quality per bit.
+ * Depending on the encoding mode, and the internal algorithms used, quality
+ *  may actually improve, but in this case bitrate will also likely increase.
+ * In any case, overall rate/distortion performance will probably decrease.
+ * The maximum value, and the meaning of each value, may change depending on
+ *  the current encoding mode (VBR vs. CQI, etc.).
+ *
+ * \param[out] buf int: The maximum encoding speed level.
+ * \retval OC_FAULT  \a theora_state or \a buf is <tt>NULL</tt>.
+ * \retval OC_EINVAL \a buf_sz is not <tt>sizeof(int)</tt>.
+ * \retval OC_IMPL   Not supported by this implementation in the current
+ *                    encoding mode.*/
+#define TH_ENCCTL_GET_SPLEVEL_MAX (12)
+
+/**Sets the speed level.
+ * By default a speed value of 1 is used.
+ *
+ * \param[in] buf int: The new encoding speed level.
+ *                      0 is slowest, larger values use less CPU.
+ * \retval OC_FAULT  \a theora_state or \a buf is <tt>NULL</tt>.
+ * \retval OC_EINVAL \a buf_sz is not <tt>sizeof(int)</tt>, or the
+ *                    encoding speed level is out of bounds.
+ *                   The maximum encoding speed level may be
+ *                    implementation- and encoding mode-specific, and can be
+ *                    obtained via #TH_ENCCTL_GET_SPLEVEL_MAX.
+ * \retval OC_IMPL   Not supported by this implementation in the current
+ *                    encoding mode.*/
+#define TH_ENCCTL_SET_SPLEVEL (14)
+
+/*@}*/
+
+#define OC_FAULT       -1       /**< General failure */
+#define OC_EINVAL      -10      /**< Library encountered invalid internal data */
+#define OC_DISABLED    -11      /**< Requested action is disabled */
+#define OC_BADHEADER   -20      /**< Header packet was corrupt/invalid */
+#define OC_NOTFORMAT   -21      /**< Packet is not a theora packet */
+#define OC_VERSION     -22      /**< Bitstream version is not handled */
+#define OC_IMPL        -23      /**< Feature or action not implemented */
+#define OC_BADPACKET   -24      /**< Packet is corrupt */
+#define OC_NEWPACKET   -25      /**< Packet is an (ignorable) unhandled extension */
+#define OC_DUPFRAME    1        /**< Packet is a dropped frame */
+
+/**
+ * Retrieve a human-readable string to identify the encoder vendor and version.
+ * \returns A version string.
+ */
+extern const char *theora_version_string(void);
+
+/**
+ * Retrieve a 32-bit version number.
+ * This number is composed of a 16-bit major version, 8-bit minor version
+ * and 8 bit sub-version, composed as follows:
+<pre>
+   (VERSION_MAJOR<<16) + (VERSION_MINOR<<8) + (VERSION_SUB)
+</pre>
+* \returns The version number.
+*/
+extern ogg_uint32_t theora_version_number(void);
+
+/**
+ * Initialize the theora encoder.
+ * \param th The theora_state handle to initialize for encoding.
+ * \param ti A theora_info struct filled with the desired encoding parameters.
+ * \retval 0 Success
+ */
+extern int theora_encode_init(theora_state *th, theora_info *ti);
+
+/**
+ * Submit a YUV buffer to the theora encoder.
+ * \param t A theora_state handle previously initialized for encoding.
+ * \param yuv A buffer of YUV data to encode.  Note that both the yuv_buffer
+ *            struct and the luma/chroma buffers within should be allocated by
+ *            the user.
+ * \retval OC_EINVAL Encoder is not ready, or is finished.
+ * \retval -1 The size of the given frame differs from those previously input
+ * \retval 0 Success
+ */
+extern int theora_encode_YUVin(theora_state *t, yuv_buffer *yuv);
+
+/**
+ * Request the next packet of encoded video.
+ * The encoded data is placed in a user-provided ogg_packet structure.
+ * \param t A theora_state handle previously initialized for encoding.
+ * \param last_p whether this is the last packet the encoder should produce.
+ * \param op An ogg_packet structure to fill. libtheora will set all
+ *           elements of this structure, including a pointer to encoded
+ *           data. The memory for the encoded data is owned by libtheora.
+ * \retval 0 No internal storage exists OR no packet is ready
+ * \retval -1 The encoding process has completed
+ * \retval 1 Success
+ */
+extern int theora_encode_packetout( theora_state *t, int last_p,
+                                    ogg_packet *op);
+
+/**
+ * Request a packet containing the initial header.
+ * A pointer to the header data is placed in a user-provided ogg_packet
+ * structure.
+ * \param t A theora_state handle previously initialized for encoding.
+ * \param op An ogg_packet structure to fill. libtheora will set all
+ *           elements of this structure, including a pointer to the header
+ *           data. The memory for the header data is owned by libtheora.
+ * \retval 0 Success
+ */
+extern int theora_encode_header(theora_state *t, ogg_packet *op);
+
+/**
+ * Request a comment header packet from provided metadata.
+ * A pointer to the comment data is placed in a user-provided ogg_packet
+ * structure.
+ * \param tc A theora_comment structure filled with the desired metadata
+ * \param op An ogg_packet structure to fill. libtheora will set all
+ *           elements of this structure, including a pointer to the encoded
+ *           comment data. The memory for the comment data is owned by
+ *           the application, and must be freed by it using _ogg_free().
+ *           On some systems (such as Windows when using dynamic linking), this
+ *           may mean the free is executed in a different module from the
+ *           malloc, which will crash; there is no way to free this memory on
+ *           such systems.
+ * \retval 0 Success
+ */
+extern int theora_encode_comment(theora_comment *tc, ogg_packet *op);
+
+/**
+ * Request a packet containing the codebook tables for the stream.
+ * A pointer to the codebook data is placed in a user-provided ogg_packet
+ * structure.
+ * \param t A theora_state handle previously initialized for encoding.
+ * \param op An ogg_packet structure to fill. libtheora will set all
+ *           elements of this structure, including a pointer to the codebook
+ *           data. The memory for the header data is owned by libtheora.
+ * \retval 0 Success
+ */
+extern int theora_encode_tables(theora_state *t, ogg_packet *op);
+
+/**
+ * Decode an Ogg packet, with the expectation that the packet contains
+ * an initial header, comment data or codebook tables.
+ *
+ * \param ci A theora_info structure to fill. This must have been previously
+ *           initialized with theora_info_init(). If \a op contains an initial
+ *           header, theora_decode_header() will fill \a ci with the
+ *           parsed header values. If \a op contains codebook tables,
+ *           theora_decode_header() will parse these and attach an internal
+ *           representation to \a ci->codec_setup.
+ * \param cc A theora_comment structure to fill. If \a op contains comment
+ *           data, theora_decode_header() will fill \a cc with the parsed
+ *           comments.
+ * \param op An ogg_packet structure which you expect contains an initial
+ *           header, comment data or codebook tables.
+ *
+ * \retval OC_BADHEADER \a op is NULL; OR the first byte of \a op->packet
+ *                      has the signature of an initial packet, but op is
+ *                      not a b_o_s packet; OR this packet has the signature
+ *                      of an initial header packet, but an initial header
+ *                      packet has already been seen; OR this packet has the
+ *                      signature of a comment packet, but the initial header
+ *                      has not yet been seen; OR this packet has the signature
+ *                      of a comment packet, but contains invalid data; OR
+ *                      this packet has the signature of codebook tables,
+ *                      but the initial header or comments have not yet
+ *                      been seen; OR this packet has the signature of codebook
+ *                      tables, but contains invalid data;
+ *                      OR the stream being decoded has a compatible version
+ *                      but this packet does not have the signature of a
+ *                      theora initial header, comments, or codebook packet
+ * \retval OC_VERSION   The packet data of \a op is an initial header with
+ *                      a version which is incompatible with this version of
+ *                      libtheora.
+ * \retval OC_NEWPACKET the stream being decoded has an incompatible (future)
+ *                      version and contains an unknown signature.
+ * \retval 0            Success
+ *
+ * \note The normal usage is that theora_decode_header() be called on the
+ *       first three packets of a theora logical bitstream in succession.
+ */
+extern int theora_decode_header(theora_info *ci, theora_comment *cc,
+                                ogg_packet *op);
+
+/**
+ * Initialize a theora_state handle for decoding.
+ * \param th The theora_state handle to initialize.
+ * \param c  A theora_info struct filled with the desired decoding parameters.
+ *           This is of course usually obtained from a previous call to
+ *           theora_decode_header().
+ * \retval 0 Success
+ */
+extern int theora_decode_init(theora_state *th, theora_info *c);
+
+/**
+ * Input a packet containing encoded data into the theora decoder.
+ * \param th A theora_state handle previously initialized for decoding.
+ * \param op An ogg_packet containing encoded theora data.
+ * \retval 0 Success
+ * \retval OC_BADPACKET \a op does not contain encoded video data
+ */
+extern int theora_decode_packetin(theora_state *th,ogg_packet *op);
+
+/**
+ * Output the next available frame of decoded YUV data.
+ * \param th A theora_state handle previously initialized for decoding.
+ * \param yuv A yuv_buffer in which libtheora should place the decoded data.
+ *            Note that the buffer struct itself is allocated by the user, but
+ *            that the luma and chroma pointers will be filled in by the
+ *            library.  Also note that these luma and chroma regions should be
+ *            considered read-only by the user.
+ * \retval 0 Success
+ */
+extern int theora_decode_YUVout(theora_state *th,yuv_buffer *yuv);
+
+/**
+ * Report whether a theora packet is a header or not
+ * This function does no verification beyond checking the header
+ * flag bit so it should not be used for bitstream identification;
+ * use theora_decode_header() for that.
+ *
+ * \param op An ogg_packet containing encoded theora data.
+ * \retval 1 The packet is a header packet
+ * \retval 0 The packet is not a header packet (and so contains frame data)
+ *
+ * Thus function was added in the 1.0alpha4 release.
+ */
+extern int theora_packet_isheader(ogg_packet *op);
+
+/**
+ * Report whether a theora packet is a keyframe or not
+ *
+ * \param op An ogg_packet containing encoded theora data.
+ * \retval 1 The packet contains a keyframe image
+ * \retval 0 The packet is contains an interframe delta
+ * \retval -1 The packet is not an image data packet at all
+ *
+ * Thus function was added in the 1.0alpha4 release.
+ */
+extern int theora_packet_iskeyframe(ogg_packet *op);
+
+/**
+ * Report the granulepos shift radix
+ *
+ * When embedded in Ogg, Theora uses a two-part granulepos,
+ * splitting the 64-bit field into two pieces. The more-significant
+ * section represents the frame count at the last keyframe,
+ * and the less-significant section represents the count of
+ * frames since the last keyframe. In this way the overall
+ * field is still non-decreasing with time, but usefully encodes
+ * a pointer to the last keyframe, which is necessary for
+ * correctly restarting decode after a seek.
+ *
+ * This function reports the number of bits used to represent
+ * the distance to the last keyframe, and thus how the granulepos
+ * field must be shifted or masked to obtain the two parts.
+ *
+ * Since libtheora returns compressed data in an ogg_packet
+ * structure, this may be generally useful even if the Theora
+ * packets are not being used in an Ogg container.
+ *
+ * \param ti A previously initialized theora_info struct
+ * \returns The bit shift dividing the two granulepos fields
+ *
+ * This function was added in the 1.0alpha5 release.
+ */
+int theora_granule_shift(theora_info *ti);
+
+/**
+ * Convert a granulepos to an absolute frame index, starting at 0.
+ * The granulepos is interpreted in the context of a given theora_state handle.
+ *
+ * Note that while the granulepos encodes the frame count (i.e. starting
+ * from 1) this call returns the frame index, starting from zero. Thus
+ * One can calculate the presentation time by multiplying the index by
+ * the rate.
+ *
+ * \param th A previously initialized theora_state handle (encode or decode)
+ * \param granulepos The granulepos to convert.
+ * \returns The frame index corresponding to \a granulepos.
+ * \retval -1 The given granulepos is undefined (i.e. negative)
+ *
+ * Thus function was added in the 1.0alpha4 release.
+ */
+extern ogg_int64_t theora_granule_frame(theora_state *th,ogg_int64_t granulepos);
+
+/**
+ * Convert a granulepos to absolute time in seconds. The granulepos is
+ * interpreted in the context of a given theora_state handle, and gives
+ * the end time of a frame's presentation as used in Ogg mux ordering.
+ *
+ * \param th A previously initialized theora_state handle (encode or decode)
+ * \param granulepos The granulepos to convert.
+ * \returns The absolute time in seconds corresponding to \a granulepos.
+ *          This is the "end time" for the frame, or the latest time it should
+ *           be displayed.
+ *          It is not the presentation time.
+ * \retval -1. The given granulepos is undefined (i.e. negative).
+ */
+extern double theora_granule_time(theora_state *th,ogg_int64_t granulepos);
+
+/**
+ * Initialize a theora_info structure. All values within the given theora_info
+ * structure are initialized, and space is allocated within libtheora for
+ * internal codec setup data.
+ * \param c A theora_info struct to initialize.
+ */
+extern void theora_info_init(theora_info *c);
+
+/**
+ * Clear a theora_info structure. All values within the given theora_info
+ * structure are cleared, and associated internal codec setup data is freed.
+ * \param c A theora_info struct to initialize.
+ */
+extern void theora_info_clear(theora_info *c);
+
+/**
+ * Free all internal data associated with a theora_state handle.
+ * \param t A theora_state handle.
+ */
+extern void theora_clear(theora_state *t);
+
+/**
+ * Initialize an allocated theora_comment structure
+ * \param tc An allocated theora_comment structure
+ **/
+extern void theora_comment_init(theora_comment *tc);
+
+/**
+ * Add a comment to an initialized theora_comment structure
+ * \param tc A previously initialized theora comment structure
+ * \param comment A null-terminated string encoding the comment in the form
+ *                "TAG=the value"
+ *
+ * Neither theora_comment_add() nor theora_comment_add_tag() support
+ * comments containing null values, although the bitstream format
+ * supports this. To add such comments you will need to manipulate
+ * the theora_comment structure directly.
+ **/
+
+extern void theora_comment_add(theora_comment *tc, char *comment);
+
+/**
+ * Add a comment to an initialized theora_comment structure.
+ * \param tc A previously initialized theora comment structure
+ * \param tag A null-terminated string containing the tag
+ *            associated with the comment.
+ * \param value The corresponding value as a null-terminated string
+ *
+ * Neither theora_comment_add() nor theora_comment_add_tag() support
+ * comments containing null values, although the bitstream format
+ * supports this. To add such comments you will need to manipulate
+ * the theora_comment structure directly.
+ **/
+extern void theora_comment_add_tag(theora_comment *tc,
+                                       char *tag, char *value);
+
+/**
+ * Look up a comment value by tag.
+ * \param tc Tn initialized theora_comment structure
+ * \param tag The tag to look up
+ * \param count The instance of the tag. The same tag can appear multiple
+ *              times, each with a distinct and ordered value, so an index
+ *              is required to retrieve them all.
+ * \returns A pointer to the queried tag's value
+ * \retval NULL No matching tag is found
+ *
+ * \note Use theora_comment_query_count() to get the legal range for the
+ * count parameter.
+ **/
+
+extern char *theora_comment_query(theora_comment *tc, char *tag, int count);
+
+/** Look up the number of instances of a tag.
+ *  \param tc An initialized theora_comment structure
+ *  \param tag The tag to look up
+ *  \returns The number on instances of a particular tag.
+ *
+ *  Call this first when querying for a specific tag and then interate
+ *  over the number of instances with separate calls to
+ *  theora_comment_query() to retrieve all instances in order.
+ **/
+extern int   theora_comment_query_count(theora_comment *tc, char *tag);
+
+/**
+ * Clear an allocated theora_comment struct so that it can be freed.
+ * \param tc An allocated theora_comment structure.
+ **/
+extern void  theora_comment_clear(theora_comment *tc);
+
+/**Encoder control function.
+ * This is used to provide advanced control the encoding process.
+ * \param th     A #theora_state handle.
+ * \param req    The control code to process.
+ *                See \ref encctlcodes_old "the list of available
+ *			control codes" for details.
+ * \param buf    The parameters for this control code.
+ * \param buf_sz The size of the parameter buffer.*/
+extern int theora_control(theora_state *th,int req,void *buf,size_t buf_sz);
+
+/* @} */ /* end oldfuncs doxygen group */
+
+#ifdef __cplusplus
+}
+#endif /* __cplusplus */
+
+#endif /* _O_THEORA_H_ */

lib/theoratest/lib/theora/theoradec.h

@@ -0,0 +1,333 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+  last mod: $Id: theora.h,v 1.8 2004/03/15 22:17:32 derf Exp $
+
+ ********************************************************************/
+
+/**\file
+ * The <tt>libtheoradec</tt> C decoding API.*/
+
+#if !defined(_O_THEORA_THEORADEC_H_)
+# define _O_THEORA_THEORADEC_H_ (1)
+# include <stddef.h>
+# include <ogg/ogg.h>
+# include "codec.h"
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+
+
+/**\name th_decode_ctl() codes
+ * \anchor decctlcodes
+ * These are the available request codes for th_decode_ctl().
+ * By convention, these are odd, to distinguish them from the
+ *  \ref encctlcodes "encoder control codes".
+ * Keep any experimental or vendor-specific values above \c 0x8000.*/
+/*@{*/
+/**Gets the maximum post-processing level.
+ * The decoder supports a post-processing filter that can improve
+ * the appearance of the decoded images. This returns the highest
+ * level setting for this post-processor, corresponding to maximum
+ * improvement and computational expense.
+ *
+ * \param[out] _buf int: The maximum post-processing level.
+ * \retval TH_EFAULT  \a _dec_ctx or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL  \a _buf_sz is not <tt>sizeof(int)</tt>.
+ * \retval TH_EIMPL   Not supported by this implementation.*/
+#define TH_DECCTL_GET_PPLEVEL_MAX (1)
+/**Sets the post-processing level.
+ * By default, post-processing is disabled.
+ *
+ * Sets the level of post-processing to use when decoding the
+ * compressed stream. This must be a value between zero (off)
+ * and the maximum returned by TH_DECCTL_GET_PPLEVEL_MAX.
+ *
+ * \param[in] _buf int: The new post-processing level.
+ *                      0 to disable; larger values use more CPU.
+ * \retval TH_EFAULT  \a _dec_ctx or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL  \a _buf_sz is not <tt>sizeof(int)</tt>, or the
+ *                     post-processing level is out of bounds.
+ *                    The maximum post-processing level may be
+ *                     implementation-specific, and can be obtained via
+ *                     #TH_DECCTL_GET_PPLEVEL_MAX.
+ * \retval TH_EIMPL   Not supported by this implementation.*/
+#define TH_DECCTL_SET_PPLEVEL (3)
+/**Sets the granule position.
+ * Call this after a seek, before decoding the first frame, to ensure that the
+ *  proper granule position is returned for all subsequent frames.
+ * If you track timestamps yourself and do not use the granule position
+ *  returned by the decoder, then you need not call this function.
+ *
+ * \param[in] _buf <tt>ogg_int64_t</tt>: The granule position of the next
+ *                  frame.
+ * \retval TH_EFAULT  \a _dec_ctx or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL  \a _buf_sz is not <tt>sizeof(ogg_int64_t)</tt>, or the
+ *                     granule position is negative.*/
+#define TH_DECCTL_SET_GRANPOS (5)
+/**Sets the striped decode callback function.
+ * If set, this function will be called as each piece of a frame is fully
+ *  decoded in th_decode_packetin().
+ * You can pass in a #th_stripe_callback with
+ *  th_stripe_callback#stripe_decoded set to <tt>NULL</tt> to disable the
+ *  callbacks at any point.
+ * Enabling striped decode does not prevent you from calling
+ *  th_decode_ycbcr_out() after the frame is fully decoded.
+ *
+ * \param[in]  _buf #th_stripe_callback: The callback parameters.
+ * \retval TH_EFAULT  \a _dec_ctx or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL  \a _buf_sz is not
+ *                     <tt>sizeof(th_stripe_callback)</tt>.*/
+#define TH_DECCTL_SET_STRIPE_CB (7)
+
+/**Sets the macroblock display mode. Set to 0 to disable displaying
+ * macroblocks.*/
+#define TH_DECCTL_SET_TELEMETRY_MBMODE (9)
+/**Sets the motion vector display mode. Set to 0 to disable displaying motion
+ * vectors.*/
+#define TH_DECCTL_SET_TELEMETRY_MV (11)
+/**Sets the adaptive quantization display mode. Set to 0 to disable displaying
+ * adaptive quantization. */
+#define TH_DECCTL_SET_TELEMETRY_QI (13)
+/**Sets the bitstream breakdown visualization mode. Set to 0 to disable
+ * displaying bitstream breakdown.*/
+#define TH_DECCTL_SET_TELEMETRY_BITS (15)
+/*@}*/
+
+
+
+/**A callback function for striped decode.
+ * This is a function pointer to an application-provided function that will be
+ *  called each time a section of the image is fully decoded in
+ *  th_decode_packetin().
+ * This allows the application to process the section immediately, while it is
+ *  still in cache.
+ * Note that the frame is decoded bottom to top, so \a _yfrag0 will steadily
+ *  decrease with each call until it reaches 0, at which point the full frame
+ *  is decoded.
+ * The number of fragment rows made available in each call depends on the pixel
+ *  format and the number of post-processing filters enabled, and may not even
+ *  be constant for the entire frame.
+ * If a non-<tt>NULL</tt> \a _granpos pointer is passed to
+ *  th_decode_packetin(), the granule position for the frame will be stored
+ *  in it before the first callback is made.
+ * If an entire frame is dropped (a 0-byte packet), then no callbacks will be
+ *  made at all for that frame.
+ * \param _ctx       An application-provided context pointer.
+ * \param _buf       The image buffer for the decoded frame.
+ * \param _yfrag0    The Y coordinate of the first row of 8x8 fragments
+ *                    decoded.
+ *                   Multiply this by 8 to obtain the pixel row number in the
+ *                    luma plane.
+ *                   If the chroma planes are subsampled in the Y direction,
+ *                    this will always be divisible by two.
+ * \param _yfrag_end The Y coordinate of the first row of 8x8 fragments past
+ *                    the newly decoded section.
+ *                   If the chroma planes are subsampled in the Y direction,
+ *                    this will always be divisible by two.
+ *                   I.e., this section contains fragment rows
+ *                    <tt>\a _yfrag0 ...\a _yfrag_end -1</tt>.*/
+typedef void (*th_stripe_decoded_func)(void *_ctx,th_ycbcr_buffer _buf,
+ int _yfrag0,int _yfrag_end);
+
+/**The striped decode callback data to pass to #TH_DECCTL_SET_STRIPE_CB.*/
+typedef struct{
+  /**An application-provided context pointer.
+   * This will be passed back verbatim to the application.*/
+  void                   *ctx;
+  /**The callback function pointer.*/
+  th_stripe_decoded_func  stripe_decoded;
+}th_stripe_callback;
+
+
+
+/**\name Decoder state
+   The following data structures are opaque, and their contents are not
+    publicly defined by this API.
+   Referring to their internals directly is unsupported, and may break without
+    warning.*/
+/*@{*/
+/**The decoder context.*/
+typedef struct th_dec_ctx    th_dec_ctx;
+/**Setup information.
+   This contains auxiliary information (Huffman tables and quantization
+    parameters) decoded from the setup header by th_decode_headerin() to be
+    passed to th_decode_alloc().
+   It can be re-used to initialize any number of decoders, and can be freed
+    via th_setup_free() at any time.*/
+typedef struct th_setup_info th_setup_info;
+/*@}*/
+
+
+
+/**\defgroup decfuncs Functions for Decoding*/
+/*@{*/
+/**\name Functions for decoding
+ * You must link to <tt>libtheoradec</tt> if you use any of the
+ * functions in this section.
+ *
+ * The functions are listed in the order they are used in a typical decode.
+ * The basic steps are:
+ * - Parse the header packets by repeatedly calling th_decode_headerin().
+ * - Allocate a #th_dec_ctx handle with th_decode_alloc().
+ * - Call th_setup_free() to free any memory used for codec setup
+ *    information.
+ * - Perform any additional decoder configuration with th_decode_ctl().
+ * - For each video data packet:
+ *   - Submit the packet to the decoder via th_decode_packetin().
+ *   - Retrieve the uncompressed video data via th_decode_ycbcr_out().
+ * - Call th_decode_free() to release all decoder memory.*/
+/*@{*/
+/**Decodes the header packets of a Theora stream.
+ * This should be called on the initial packets of the stream, in succession,
+ *  until it returns <tt>0</tt>, indicating that all headers have been
+ *  processed, or an error is encountered.
+ * At least three header packets are required, and additional optional header
+ *  packets may follow.
+ * This can be used on the first packet of any logical stream to determine if
+ *  that stream is a Theora stream.
+ * \param _info  A #th_info structure to fill in.
+ *               This must have been previously initialized with
+ *                th_info_init().
+ *               The application may immediately begin using the contents of
+ *                this structure after the first header is decoded, though it
+ *                must continue to be passed in on all subsequent calls.
+ * \param _tc    A #th_comment structure to fill in.
+ *               The application may immediately begin using the contents of
+ *                this structure after the second header is decoded, though it
+ *                must continue to be passed in on all subsequent calls.
+ * \param _setup Returns a pointer to additional, private setup information
+ *                needed by the decoder.
+ *               The contents of this pointer must be initialized to
+ *                <tt>NULL</tt> on the first call, and the returned value must
+ *                continue to be passed in on all subsequent calls.
+ * \param _op    An <tt>ogg_packet</tt> structure which contains one of the
+ *                initial packets of an Ogg logical stream.
+ * \return A positive value indicates that a Theora header was successfully
+ *          processed.
+ * \retval 0             The first video data packet was encountered after all
+ *                        required header packets were parsed.
+ *                       The packet just passed in on this call should be saved
+ *                        and fed to th_decode_packetin() to begin decoding
+ *                        video data.
+ * \retval TH_EFAULT     One of \a _info, \a _tc, or \a _setup was
+ *                        <tt>NULL</tt>.
+ * \retval TH_EBADHEADER \a _op was <tt>NULL</tt>, the packet was not the next
+ *                        header packet in the expected sequence, or the format
+ *                        of the header data was invalid.
+ * \retval TH_EVERSION   The packet data was a Theora info header, but for a
+ *                        bitstream version not decodable with this version of
+ *                        <tt>libtheoradec</tt>.
+ * \retval TH_ENOTFORMAT The packet was not a Theora header.
+ */
+extern int th_decode_headerin(th_info *_info,th_comment *_tc,
+ th_setup_info **_setup,ogg_packet *_op);
+/**Allocates a decoder instance.
+ *
+ * <b>Security Warning:</b> The Theora format supports very large frame sizes,
+ *  potentially even larger than the address space of a 32-bit machine, and
+ *  creating a decoder context allocates the space for several frames of data.
+ * If the allocation fails here, your program will crash, possibly at some
+ *  future point because the OS kernel returned a valid memory range and will
+ *  only fail when it tries to map the pages in it the first time they are
+ *  used.
+ * Even if it succeeds, you may experience a denial of service if the frame
+ *  size is large enough to cause excessive paging.
+ * If you are integrating libtheora in a larger application where such things
+ *  are undesirable, it is highly recommended that you check the frame size in
+ *  \a _info before calling this function and refuse to decode streams where it
+ *  is larger than some reasonable maximum.
+ * libtheora will not check this for you, because there may be machines that
+ *  can handle such streams and applications that wish to.
+ * \param _info  A #th_info struct filled via th_decode_headerin().
+ * \param _setup A #th_setup_info handle returned via
+ *                th_decode_headerin().
+ * \return The initialized #th_dec_ctx handle.
+ * \retval NULL If the decoding parameters were invalid.*/
+extern th_dec_ctx *th_decode_alloc(const th_info *_info,
+ const th_setup_info *_setup);
+/**Releases all storage used for the decoder setup information.
+ * This should be called after you no longer want to create any decoders for
+ *  a stream whose headers you have parsed with th_decode_headerin().
+ * \param _setup The setup information to free.
+ *               This can safely be <tt>NULL</tt>.*/
+extern void th_setup_free(th_setup_info *_setup);
+/**Decoder control function.
+ * This is used to provide advanced control of the decoding process.
+ * \param _dec    A #th_dec_ctx handle.
+ * \param _req    The control code to process.
+ *                See \ref decctlcodes "the list of available control codes"
+ *                 for details.
+ * \param _buf    The parameters for this control code.
+ * \param _buf_sz The size of the parameter buffer.
+ * \return Possible return values depend on the control code used.
+ *          See \ref decctlcodes "the list of control codes" for
+ *          specific values. Generally 0 indicates success.*/
+extern int th_decode_ctl(th_dec_ctx *_dec,int _req,void *_buf,
+ size_t _buf_sz);
+/**Submits a packet containing encoded video data to the decoder.
+ * \param _dec     A #th_dec_ctx handle.
+ * \param _op      An <tt>ogg_packet</tt> containing encoded video data.
+ * \param _granpos Returns the granule position of the decoded packet.
+ *                 If non-<tt>NULL</tt>, the granule position for this specific
+ *                  packet is stored in this location.
+ *                 This is computed incrementally from previously decoded
+ *                  packets.
+ *                 After a seek, the correct granule position must be set via
+ *                  #TH_DECCTL_SET_GRANPOS for this to work properly.
+ * \retval 0             Success.
+ *                       A new decoded frame can be retrieved by calling
+ *                        th_decode_ycbcr_out().
+ * \retval TH_DUPFRAME   The packet represented a dropped frame (either a
+ *                        0-byte frame or an INTER frame with no coded blocks).
+ *                       The player can skip the call to th_decode_ycbcr_out(),
+ *                        as the contents of the decoded frame buffer have not
+ *                        changed.
+ * \retval TH_EFAULT     \a _dec or \a _op was <tt>NULL</tt>.
+ * \retval TH_EBADPACKET \a _op does not contain encoded video data.
+ * \retval TH_EIMPL      The video data uses bitstream features which this
+ *                        library does not support.*/
+extern int th_decode_packetin(th_dec_ctx *_dec,const ogg_packet *_op,
+ ogg_int64_t *_granpos);
+/**Outputs the next available frame of decoded Y'CbCr data.
+ * If a striped decode callback has been set with #TH_DECCTL_SET_STRIPE_CB,
+ *  then the application does not need to call this function.
+ * \param _dec   A #th_dec_ctx handle.
+ * \param _ycbcr A video buffer structure to fill in.
+ *               <tt>libtheoradec</tt> will fill in all the members of this
+ *                structure, including the pointers to the uncompressed video
+ *                data.
+ *               The memory for this video data is owned by
+ *                <tt>libtheoradec</tt>.
+ *               It may be freed or overwritten without notification when
+ *                subsequent frames are decoded.
+ * \retval 0 Success
+ * \retval TH_EFAULT     \a _dec or \a _ycbcr was <tt>NULL</tt>.
+ */
+extern int th_decode_ycbcr_out(th_dec_ctx *_dec,
+ th_ycbcr_buffer _ycbcr);
+/**Frees an allocated decoder instance.
+ * \param _dec A #th_dec_ctx handle.*/
+extern void th_decode_free(th_dec_ctx *_dec);
+/*@}*/
+/*@}*/
+
+
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif

lib/theoratest/lib/theora/theoraenc.h

@@ -0,0 +1,548 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+  last mod: $Id: theora.h,v 1.8 2004/03/15 22:17:32 derf Exp $
+
+ ********************************************************************/
+
+/**\file
+ * The <tt>libtheoraenc</tt> C encoding API.*/
+
+#if !defined(_O_THEORA_THEORAENC_H_)
+# define _O_THEORA_THEORAENC_H_ (1)
+# include <stddef.h>
+# include <ogg/ogg.h>
+# include "codec.h"
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+
+
+/**\name th_encode_ctl() codes
+ * \anchor encctlcodes
+ * These are the available request codes for th_encode_ctl().
+ * By convention, these are even, to distinguish them from the
+ *  \ref decctlcodes "decoder control codes".
+ * Keep any experimental or vendor-specific values above \c 0x8000.*/
+/*@{*/
+/**Sets the Huffman tables to use.
+ * The tables are copied, not stored by reference, so they can be freed after
+ *  this call.
+ * <tt>NULL</tt> may be specified to revert to the default tables.
+ *
+ * \param[in] _buf <tt>#th_huff_code[#TH_NHUFFMAN_TABLES][#TH_NDCT_TOKENS]</tt>
+ * \retval TH_EFAULT \a _enc is <tt>NULL</tt>.
+ * \retval TH_EINVAL Encoding has already begun or one or more of the given
+ *                     tables is not full or prefix-free, \a _buf is
+ *                     <tt>NULL</tt> and \a _buf_sz is not zero, or \a _buf is
+ *                     non-<tt>NULL</tt> and \a _buf_sz is not
+ *                     <tt>sizeof(#th_huff_code)*#TH_NHUFFMAN_TABLES*#TH_NDCT_TOKENS</tt>.
+ * \retval TH_EIMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_SET_HUFFMAN_CODES (0)
+/**Sets the quantization parameters to use.
+ * The parameters are copied, not stored by reference, so they can be freed
+ *  after this call.
+ * <tt>NULL</tt> may be specified to revert to the default parameters.
+ *
+ * \param[in] _buf #th_quant_info
+ * \retval TH_EFAULT \a _enc is <tt>NULL</tt>.
+ * \retval TH_EINVAL Encoding has already begun, \a _buf is
+ *                    <tt>NULL</tt> and \a _buf_sz is not zero,
+ *                    or \a _buf is non-<tt>NULL</tt> and
+ *                    \a _buf_sz is not <tt>sizeof(#th_quant_info)</tt>.
+ * \retval TH_EIMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_SET_QUANT_PARAMS (2)
+/**Sets the maximum distance between key frames.
+ * This can be changed during an encode, but will be bounded by
+ *  <tt>1<<th_info#keyframe_granule_shift</tt>.
+ * If it is set before encoding begins, th_info#keyframe_granule_shift will
+ *  be enlarged appropriately.
+ *
+ * \param[in]  _buf <tt>ogg_uint32_t</tt>: The maximum distance between key
+ *                   frames.
+ * \param[out] _buf <tt>ogg_uint32_t</tt>: The actual maximum distance set.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(ogg_uint32_t)</tt>.
+ * \retval TH_EIMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_SET_KEYFRAME_FREQUENCY_FORCE (4)
+/**Disables any encoder features that would prevent lossless transcoding back
+ *  to VP3.
+ * This primarily means disabling block-adaptive quantization and always coding
+ *  all four luma blocks in a macro block when 4MV is used.
+ * It also includes using the VP3 quantization tables and Huffman codes; if you
+ *  set them explicitly after calling this function, the resulting stream will
+ *  not be VP3-compatible.
+ * If you enable VP3-compatibility when encoding 4:2:2 or 4:4:4 source
+ *  material, or when using a picture region smaller than the full frame (e.g.
+ *  a non-multiple-of-16 width or height), then non-VP3 bitstream features will
+ *  still be disabled, but the stream will still not be VP3-compatible, as VP3
+ *  was not capable of encoding such formats.
+ * If you call this after encoding has already begun, then the quantization
+ *  tables and codebooks cannot be changed, but the frame-level features will
+ *  be enabled or disabled as requested.
+ *
+ * \param[in]  _buf <tt>int</tt>: a non-zero value to enable VP3 compatibility,
+ *                   or 0 to disable it (the default).
+ * \param[out] _buf <tt>int</tt>: 1 if all bitstream features required for
+ *                   VP3-compatibility could be set, and 0 otherwise.
+ *                  The latter will be returned if the pixel format is not
+ *                   4:2:0, the picture region is smaller than the full frame,
+ *                   or if encoding has begun, preventing the quantization
+ *                   tables and codebooks from being set.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(int)</tt>.
+ * \retval TH_EIMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_SET_VP3_COMPATIBLE (10)
+/**Gets the maximum speed level.
+ * Higher speed levels favor quicker encoding over better quality per bit.
+ * Depending on the encoding mode, and the internal algorithms used, quality
+ *  may actually improve, but in this case bitrate will also likely increase.
+ * In any case, overall rate/distortion performance will probably decrease.
+ * The maximum value, and the meaning of each value, may change depending on
+ *  the current encoding mode (VBR vs. constant quality, etc.).
+ *
+ * \param[out] _buf <tt>int</tt>: The maximum encoding speed level.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(int)</tt>.
+ * \retval TH_EIMPL   Not supported by this implementation in the current
+ *                    encoding mode.*/
+#define TH_ENCCTL_GET_SPLEVEL_MAX (12)
+/**Sets the speed level.
+ * The current speed level may be retrieved using #TH_ENCCTL_GET_SPLEVEL.
+ *
+ * \param[in] _buf <tt>int</tt>: The new encoding speed level.
+ *                 0 is slowest, larger values use less CPU.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(int)</tt>, or the
+ *                    encoding speed level is out of bounds.
+ *                   The maximum encoding speed level may be
+ *                    implementation- and encoding mode-specific, and can be
+ *                    obtained via #TH_ENCCTL_GET_SPLEVEL_MAX.
+ * \retval TH_EIMPL   Not supported by this implementation in the current
+ *                    encoding mode.*/
+#define TH_ENCCTL_SET_SPLEVEL (14)
+/**Gets the current speed level.
+ * The default speed level may vary according to encoder implementation, but if
+ *  this control code is not supported (it returns #TH_EIMPL), the default may
+ *  be assumed to be the slowest available speed (0).
+ * The maximum encoding speed level may be implementation- and encoding
+ *  mode-specific, and can be obtained via #TH_ENCCTL_GET_SPLEVEL_MAX.
+ *
+ * \param[out] _buf <tt>int</tt>: The current encoding speed level.
+ *                  0 is slowest, larger values use less CPU.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(int)</tt>.
+ * \retval TH_EIMPL   Not supported by this implementation in the current
+ *                    encoding mode.*/
+#define TH_ENCCTL_GET_SPLEVEL (16)
+/**Sets the number of duplicates of the next frame to produce.
+ * Although libtheora can encode duplicate frames very cheaply, it costs some
+ *  amount of CPU to detect them, and a run of duplicates cannot span a
+ *  keyframe boundary.
+ * This control code tells the encoder to produce the specified number of extra
+ *  duplicates of the next frame.
+ * This allows the encoder to make smarter keyframe placement decisions and
+ *  rate control decisions, and reduces CPU usage as well, when compared to
+ *  just submitting the same frame for encoding multiple times.
+ * This setting only applies to the next frame submitted for encoding.
+ * You MUST call th_encode_packetout() repeatedly until it returns 0, or the
+ *  extra duplicate frames will be lost.
+ *
+ * \param[in] _buf <tt>int</tt>: The number of duplicates to produce.
+ *                 If this is negative or zero, no duplicates will be produced.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(int)</tt>, or the
+ *                    number of duplicates is greater than or equal to the
+ *                    maximum keyframe interval.
+ *                   In the latter case, NO duplicate frames will be produced.
+ *                   You must ensure that the maximum keyframe interval is set
+ *                    larger than the maximum number of duplicates you will
+ *                    ever wish to insert prior to encoding.
+ * \retval TH_EIMPL   Not supported by this implementation in the current
+ *                    encoding mode.*/
+#define TH_ENCCTL_SET_DUP_COUNT (18)
+/**Modifies the default bitrate management behavior.
+ * Use to allow or disallow frame dropping, and to enable or disable capping
+ *  bit reservoir overflows and underflows.
+ * See \ref encctlcodes "the list of available flags".
+ * The flags are set by default to
+ *  <tt>#TH_RATECTL_DROP_FRAMES|#TH_RATECTL_CAP_OVERFLOW</tt>.
+ *
+ * \param[in] _buf <tt>int</tt>: Any combination of
+ *                  \ref ratectlflags "the available flags":
+ *                 - #TH_RATECTL_DROP_FRAMES: Enable frame dropping.
+ *                 - #TH_RATECTL_CAP_OVERFLOW: Don't bank excess bits for later
+ *                    use.
+ *                 - #TH_RATECTL_CAP_UNDERFLOW: Don't try to make up shortfalls
+ *                    later.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(int)</tt> or rate control
+ *                    is not enabled.
+ * \retval TH_EIMPL   Not supported by this implementation in the current
+ *                    encoding mode.*/
+#define TH_ENCCTL_SET_RATE_FLAGS (20)
+/**Sets the size of the bitrate management bit reservoir as a function
+ *  of number of frames.
+ * The reservoir size affects how quickly bitrate management reacts to
+ *  instantaneous changes in the video complexity.
+ * Larger reservoirs react more slowly, and provide better overall quality, but
+ *  require more buffering by a client, adding more latency to live streams.
+ * By default, libtheora sets the reservoir to the maximum distance between
+ *  keyframes, subject to a minimum and maximum limit.
+ * This call may be used to increase or decrease the reservoir, increasing or
+ *  decreasing the allowed temporary variance in bitrate.
+ * An implementation may impose some limits on the size of a reservoir it can
+ *  handle, in which case the actual reservoir size may not be exactly what was
+ *  requested.
+ * The actual value set will be returned.
+ *
+ * \param[in]  _buf <tt>int</tt>: Requested size of the reservoir measured in
+ *                   frames.
+ * \param[out] _buf <tt>int</tt>: The actual size of the reservoir set.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(int)</tt>, or rate control
+ *                    is not enabled.  The buffer has an implementation
+ *                    defined minimum and maximum size and the value in _buf
+ *                    will be adjusted to match the actual value set.
+ * \retval TH_EIMPL   Not supported by this implementation in the current
+ *                    encoding mode.*/
+#define TH_ENCCTL_SET_RATE_BUFFER (22)
+/**Enable pass 1 of two-pass encoding mode and retrieve the first pass metrics.
+ * Pass 1 mode must be enabled before the first frame is encoded, and a target
+ *  bitrate must have already been specified to the encoder.
+ * Although this does not have to be the exact rate that will be used in the
+ *  second pass, closer values may produce better results.
+ * The first call returns the size of the two-pass header data, along with some
+ *  placeholder content, and sets the encoder into pass 1 mode implicitly.
+ * This call sets the encoder to pass 1 mode implicitly.
+ * Then, a subsequent call must be made after each call to
+ *  th_encode_ycbcr_in() to retrieve the metrics for that frame.
+ * An additional, final call must be made to retrieve the summary data,
+ *  containing such information as the total number of frames, etc.
+ * This must be stored in place of the placeholder data that was returned
+ *  in the first call, before the frame metrics data.
+ * All of this data must be presented back to the encoder during pass 2 using
+ *  #TH_ENCCTL_2PASS_IN.
+ *
+ * \param[out] <tt>char *</tt>_buf: Returns a pointer to internal storage
+ *              containing the two pass metrics data.
+ *             This storage is only valid until the next call, or until the
+ *              encoder context is freed, and must be copied by the
+ *              application.
+ * \retval >=0       The number of bytes of metric data available in the
+ *                    returned buffer.
+ * \retval TH_EFAULT \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL \a _buf_sz is not <tt>sizeof(char *)</tt>, no target
+ *                    bitrate has been set, or the first call was made after
+ *                    the first frame was submitted for encoding.
+ * \retval TH_EIMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_2PASS_OUT (24)
+/**Submits two-pass encoding metric data collected the first encoding pass to
+ *  the second pass.
+ * The first call must be made before the first frame is encoded, and a target
+ *  bitrate must have already been specified to the encoder.
+ * It sets the encoder to pass 2 mode implicitly; this cannot be disabled.
+ * The encoder may require reading data from some or all of the frames in
+ *  advance, depending on, e.g., the reservoir size used in the second pass.
+ * You must call this function repeatedly before each frame to provide data
+ *  until either a) it fails to consume all of the data presented or b) all of
+ *  the pass 1 data has been consumed.
+ * In the first case, you must save the remaining data to be presented after
+ *  the next frame.
+ * You can call this function with a NULL argument to get an upper bound on
+ *  the number of bytes that will be required before the next frame.
+ *
+ * When pass 2 is first enabled, the default bit reservoir is set to the entire
+ *  file; this gives maximum flexibility but can lead to very high peak rates.
+ * You can subsequently set it to another value with #TH_ENCCTL_SET_RATE_BUFFER
+ *  (e.g., to set it to the keyframe interval for non-live streaming), however,
+ *  you may then need to provide more data before the next frame.
+ *
+ * \param[in] _buf <tt>char[]</tt>: A buffer containing the data returned by
+ *                  #TH_ENCCTL_2PASS_OUT in pass 1.
+ *                 You may pass <tt>NULL</tt> for \a _buf to return an upper
+ *                  bound on the number of additional bytes needed before the
+ *                  next frame.
+ *                 The summary data returned at the end of pass 1 must be at
+ *                  the head of the buffer on the first call with a
+ *                  non-<tt>NULL</tt> \a _buf, and the placeholder data
+ *                  returned at the start of pass 1 should be omitted.
+ *                 After each call you should advance this buffer by the number
+ *                  of bytes consumed.
+ * \retval >0            The number of bytes of metric data required/consumed.
+ * \retval 0             No more data is required before the next frame.
+ * \retval TH_EFAULT     \a _enc is <tt>NULL</tt>.
+ * \retval TH_EINVAL     No target bitrate has been set, or the first call was
+ *                        made after the first frame was submitted for
+ *                        encoding.
+ * \retval TH_ENOTFORMAT The data did not appear to be pass 1 from a compatible
+ *                        implementation of this library.
+ * \retval TH_EBADHEADER The data was invalid; this may be returned when
+ *                        attempting to read an aborted pass 1 file that still
+ *                        has the placeholder data in place of the summary
+ *                        data.
+ * \retval TH_EIMPL       Not supported by this implementation.*/
+#define TH_ENCCTL_2PASS_IN (26)
+/**Sets the current encoding quality.
+ * This is only valid so long as no bitrate has been specified, either through
+ *  the #th_info struct used to initialize the encoder or through
+ *  #TH_ENCCTL_SET_BITRATE (this restriction may be relaxed in a future
+ *  version).
+ * If it is set before the headers are emitted, the target quality encoded in
+ *  them will be updated.
+ *
+ * \param[in] _buf <tt>int</tt>: The new target quality, in the range 0...63,
+ *                  inclusive.
+ * \retval 0             Success.
+ * \retval TH_EFAULT     \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL     A target bitrate has already been specified, or the
+ *                        quality index was not in the range 0...63.
+ * \retval TH_EIMPL       Not supported by this implementation.*/
+#define TH_ENCCTL_SET_QUALITY (28)
+/**Sets the current encoding bitrate.
+ * Once a bitrate is set, the encoder must use a rate-controlled mode for all
+ *  future frames (this restriction may be relaxed in a future version).
+ * If it is set before the headers are emitted, the target bitrate encoded in
+ *  them will be updated.
+ * Due to the buffer delay, the exact bitrate of each section of the encode is
+ *  not guaranteed.
+ * The encoder may have already used more bits than allowed for the frames it
+ *  has encoded, expecting to make them up in future frames, or it may have
+ *  used fewer, holding the excess in reserve.
+ * The exact transition between the two bitrates is not well-defined by this
+ *  API, but may be affected by flags set with #TH_ENCCTL_SET_RATE_FLAGS.
+ * After a number of frames equal to the buffer delay, one may expect further
+ *  output to average at the target bitrate.
+ *
+ * \param[in] _buf <tt>long</tt>: The new target bitrate, in bits per second.
+ * \retval 0             Success.
+ * \retval TH_EFAULT     \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL     The target bitrate was not positive.
+ *                       A future version of this library may allow passing 0
+ *                        to disabled rate-controlled mode and return to a
+ *                        quality-based mode, in which case this function will
+ *                        not return an error for that value.
+ * \retval TH_EIMPL      Not supported by this implementation.*/
+#define TH_ENCCTL_SET_BITRATE (30)
+/**Sets the configuration to be compatible with that from the given setup
+ *  header.
+ * This sets the Huffman codebooks and quantization parameters to match those
+ *  found in the given setup header.
+ * This guarantees that packets encoded by this encoder will be decodable using
+ *  a decoder configured with the passed-in setup header.
+ * It does <em>not</em> guarantee that th_encode_flushheader() will produce a
+ *  bit-identical setup header, only that they will be compatible.
+ * If you need a bit-identical setup header, then use the one you passed into
+ *  this command, and not the one returned by th_encode_flushheader().
+ *
+ * This also does <em>not</em> enable or disable VP3 compatibility; that is not
+ *  signaled in the setup header (or anywhere else in the encoded stream), and
+ *  is controlled independently by the #TH_ENCCTL_SET_VP3_COMPATIBLE function.
+ * If you wish to enable VP3 compatibility mode <em>and</em> want the codebooks
+ *  and quantization parameters to match the given setup header, you should
+ *  enable VP3 compatibility before invoking this command, otherwise the
+ *  codebooks and quantization parameters will be reset to the VP3 defaults.
+ *
+ * The current encoder does not support Huffman codebooks which do not contain
+ *  codewords for all 32 tokens.
+ * Such codebooks are legal, according to the specification, but cannot be
+ *  configured with this function.
+ *
+ * \param[in] _buf <tt>unsigned char[]</tt>: The encoded setup header to copy
+ *                                            the configuration from.
+ *                                           This should be the original,
+ *                                            undecoded setup header packet,
+ *                                            and <em>not</em> a #th_setup_info
+ *                                            structure filled in by
+ *                                            th_decode_headerin().
+ * \retval TH_EFAULT     \a _enc or \a _buf is <tt>NULL</tt>.
+ * \retval TH_EINVAL     Encoding has already begun, so the codebooks and
+ *                        quantization parameters cannot be changed, or the
+ *                        data in the setup header was not supported by this
+ *                        encoder.
+ * \retval TH_EBADHEADER \a _buf did not contain a valid setup header packet.
+ * \retval TH_ENOTFORMAT \a _buf did not contain a Theora header at all.
+ * \retval TH_EIMPL   Not supported by this implementation.*/
+#define TH_ENCCTL_SET_COMPAT_CONFIG (32)
+
+/*@}*/
+
+
+/**\name TH_ENCCTL_SET_RATE_FLAGS flags
+ * \anchor ratectlflags
+ * These are the flags available for use with #TH_ENCCTL_SET_RATE_FLAGS.*/
+/*@{*/
+/**Drop frames to keep within bitrate buffer constraints.
+ * This can have a severe impact on quality, but is the only way to ensure that
+ *  bitrate targets are met at low rates during sudden bursts of activity.
+ * It is enabled by default.*/
+#define TH_RATECTL_DROP_FRAMES   (0x1)
+/**Ignore bitrate buffer overflows.
+ * If the encoder uses so few bits that the reservoir of available bits
+ *  overflows, ignore the excess.
+ * The encoder will not try to use these extra bits in future frames.
+ * At high rates this may cause the result to be undersized, but allows a
+ *  client to play the stream using a finite buffer; it should normally be
+ *  enabled, which is the default.*/
+#define TH_RATECTL_CAP_OVERFLOW  (0x2)
+/**Ignore bitrate buffer underflows.
+ * If the encoder uses so many bits that the reservoir of available bits
+ *  underflows, ignore the deficit.
+ * The encoder will not try to make up these extra bits in future frames.
+ * At low rates this may cause the result to be oversized; it should normally
+ *  be disabled, which is the default.*/
+#define TH_RATECTL_CAP_UNDERFLOW (0x4)
+/*@}*/
+
+
+
+/**The quantization parameters used by VP3.*/
+extern const th_quant_info TH_VP31_QUANT_INFO;
+
+/**The Huffman tables used by VP3.*/
+extern const th_huff_code
+ TH_VP31_HUFF_CODES[TH_NHUFFMAN_TABLES][TH_NDCT_TOKENS];
+
+
+
+/**\name Encoder state
+   The following data structure is opaque, and its contents are not publicly
+    defined by this API.
+   Referring to its internals directly is unsupported, and may break without
+    warning.*/
+/*@{*/
+/**The encoder context.*/
+typedef struct th_enc_ctx    th_enc_ctx;
+/*@}*/
+
+
+
+/**\defgroup encfuncs Functions for Encoding*/
+/*@{*/
+/**\name Functions for encoding
+ * You must link to <tt>libtheoraenc</tt> and <tt>libtheoradec</tt>
+ *  if you use any of the functions in this section.
+ *
+ * The functions are listed in the order they are used in a typical encode.
+ * The basic steps are:
+ * - Fill in a #th_info structure with details on the format of the video you
+ *    wish to encode.
+ * - Allocate a #th_enc_ctx handle with th_encode_alloc().
+ * - Perform any additional encoder configuration required with
+ *    th_encode_ctl().
+ * - Repeatedly call th_encode_flushheader() to retrieve all the header
+ *    packets.
+ * - For each uncompressed frame:
+ *   - Submit the uncompressed frame via th_encode_ycbcr_in()
+ *   - Repeatedly call th_encode_packetout() to retrieve any video
+ *      data packets that are ready.
+ * - Call th_encode_free() to release all encoder memory.*/
+/*@{*/
+/**Allocates an encoder instance.
+ * \param _info A #th_info struct filled with the desired encoding parameters.
+ * \return The initialized #th_enc_ctx handle.
+ * \retval NULL If the encoding parameters were invalid.*/
+extern th_enc_ctx *th_encode_alloc(const th_info *_info);
+/**Encoder control function.
+ * This is used to provide advanced control the encoding process.
+ * \param _enc    A #th_enc_ctx handle.
+ * \param _req    The control code to process.
+ *                See \ref encctlcodes "the list of available control codes"
+ *                 for details.
+ * \param _buf    The parameters for this control code.
+ * \param _buf_sz The size of the parameter buffer.
+ * \return Possible return values depend on the control code used.
+ *          See \ref encctlcodes "the list of control codes" for
+ *          specific values. Generally 0 indicates success.*/
+extern int th_encode_ctl(th_enc_ctx *_enc,int _req,void *_buf,size_t _buf_sz);
+/**Outputs the next header packet.
+ * This should be called repeatedly after encoder initialization until it
+ *  returns 0 in order to get all of the header packets, in order, before
+ *  encoding actual video data.
+ * \param _enc      A #th_enc_ctx handle.
+ * \param _comments The metadata to place in the comment header, when it is
+ *                   encoded.
+ * \param _op       An <tt>ogg_packet</tt> structure to fill.
+ *                  All of the elements of this structure will be set,
+ *                   including a pointer to the header data.
+ *                  The memory for the header data is owned by
+ *                   <tt>libtheoraenc</tt>, and may be invalidated when the
+ *                   next encoder function is called.
+ * \return A positive value indicates that a header packet was successfully
+ *          produced.
+ * \retval 0         No packet was produced, and no more header packets remain.
+ * \retval TH_EFAULT \a _enc, \a _comments, or \a _op was <tt>NULL</tt>.*/
+extern int th_encode_flushheader(th_enc_ctx *_enc,
+ th_comment *_comments,ogg_packet *_op);
+/**Submits an uncompressed frame to the encoder.
+ * \param _enc   A #th_enc_ctx handle.
+ * \param _ycbcr A buffer of Y'CbCr data to encode.
+ *               If the width and height of the buffer matches the frame size
+ *                the encoder was initialized with, the encoder will only
+ *                reference the portion inside the picture region.
+ *               Any data outside this region will be ignored, and need not map
+ *                to a valid address.
+ *               Alternatively, you can pass a buffer equal to the size of the
+ *                picture region, if this is less than the full frame size.
+ *               When using subsampled chroma planes, odd picture sizes or odd
+ *                picture offsets may require an unexpected chroma plane size,
+ *                and their use is generally discouraged, as they will not be
+ *                well-supported by players and other media frameworks.
+ *               See Section 4.4 of
+ *                <a href="http://www.theora.org/doc/Theora.pdf">the Theora
+ *                specification</a> for details if you wish to use them anyway.
+ * \retval 0         Success.
+ * \retval TH_EFAULT \a _enc or \a _ycbcr is <tt>NULL</tt>.
+ * \retval TH_EINVAL The buffer size matches neither the frame size nor the
+ *                    picture size the encoder was initialized with, or
+ *                    encoding has already completed.*/
+extern int th_encode_ycbcr_in(th_enc_ctx *_enc,th_ycbcr_buffer _ycbcr);
+/**Retrieves encoded video data packets.
+ * This should be called repeatedly after each frame is submitted to flush any
+ *  encoded packets, until it returns 0.
+ * The encoder will not buffer these packets as subsequent frames are
+ *  compressed, so a failure to do so will result in lost video data.
+ * \note Currently the encoder operates in a one-frame-in, one-packet-out
+ *        manner.
+ *       However, this may be changed in the future.
+ * \param _enc  A #th_enc_ctx handle.
+ * \param _last Set this flag to a non-zero value if no more uncompressed
+ *               frames will be submitted.
+ *              This ensures that a proper EOS flag is set on the last packet.
+ * \param _op   An <tt>ogg_packet</tt> structure to fill.
+ *              All of the elements of this structure will be set, including a
+ *               pointer to the video data.
+ *              The memory for the video data is owned by
+ *               <tt>libtheoraenc</tt>, and may be invalidated when the next
+ *               encoder function is called.
+ * \return A positive value indicates that a video data packet was successfully
+ *          produced.
+ * \retval 0         No packet was produced, and no more encoded video data
+ *                    remains.
+ * \retval TH_EFAULT \a _enc or \a _op was <tt>NULL</tt>.*/
+extern int th_encode_packetout(th_enc_ctx *_enc,int _last,ogg_packet *_op);
+/**Frees an allocated encoder instance.
+ * \param _enc A #th_enc_ctx handle.*/
+extern void th_encode_free(th_enc_ctx *_enc);
+/*@}*/
+/*@}*/
+
+
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif

lib/theoratest/lib/theora/tinfo.c

@@ -0,0 +1,131 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <ctype.h>
+#include <string.h>
+#include "internal.h"
+
+
+
+/*This is more or less the same as strncasecmp, but that doesn't exist
+   everywhere, and this is a fairly trivial function, so we include it.
+  Note: We take advantage of the fact that we know _n is less than or equal to
+   the length of at least one of the strings.*/
+static int oc_tagcompare(const char *_s1,const char *_s2,int _n){
+  int c;
+  for(c=0;c<_n;c++){
+    if(toupper(_s1[c])!=toupper(_s2[c]))return !0;
+  }
+  return _s1[c]!='=';
+}
+
+
+
+void th_info_init(th_info *_info){
+  memset(_info,0,sizeof(*_info));
+  _info->version_major=TH_VERSION_MAJOR;
+  _info->version_minor=TH_VERSION_MINOR;
+  _info->version_subminor=TH_VERSION_SUB;
+  _info->keyframe_granule_shift=6;
+}
+
+void th_info_clear(th_info *_info){
+  memset(_info,0,sizeof(*_info));
+}
+
+
+
+void th_comment_init(th_comment *_tc){
+  memset(_tc,0,sizeof(*_tc));
+}
+
+void th_comment_add(th_comment *_tc,const char *_comment){
+  char **user_comments;
+  int   *comment_lengths;
+  int    comment_len;
+  user_comments=_ogg_realloc(_tc->user_comments,
+   (_tc->comments+2)*sizeof(*_tc->user_comments));
+  if(user_comments==NULL)return;
+  _tc->user_comments=user_comments;
+  comment_lengths=_ogg_realloc(_tc->comment_lengths,
+   (_tc->comments+2)*sizeof(*_tc->comment_lengths));
+  if(comment_lengths==NULL)return;
+  _tc->comment_lengths=comment_lengths;
+  comment_len=strlen(_comment);
+  comment_lengths[_tc->comments]=comment_len;
+  user_comments[_tc->comments]=_ogg_malloc(comment_len+1);
+  if(user_comments[_tc->comments]==NULL)return;
+  memcpy(_tc->user_comments[_tc->comments],_comment,comment_len+1);
+  _tc->comments++;
+  _tc->user_comments[_tc->comments]=NULL;
+}
+
+void th_comment_add_tag(th_comment *_tc,const char *_tag,const char *_val){
+  char *comment;
+  int   tag_len;
+  int   val_len;
+  tag_len=strlen(_tag);
+  val_len=strlen(_val);
+  /*+2 for '=' and '\0'.*/
+  comment=_ogg_malloc(tag_len+val_len+2);
+  if(comment==NULL)return;
+  memcpy(comment,_tag,tag_len);
+  comment[tag_len]='=';
+  memcpy(comment+tag_len+1,_val,val_len+1);
+  th_comment_add(_tc,comment);
+  _ogg_free(comment);
+}
+
+char *th_comment_query(th_comment *_tc,const char *_tag,int _count){
+  long i;
+  int  found;
+  int  tag_len;
+  tag_len=strlen(_tag);
+  found=0;
+  for(i=0;i<_tc->comments;i++){
+    if(!oc_tagcompare(_tc->user_comments[i],_tag,tag_len)){
+      /*We return a pointer to the data, not a copy.*/
+      if(_count==found++)return _tc->user_comments[i]+tag_len+1;
+    }
+  }
+  /*Didn't find anything.*/
+  return NULL;
+}
+
+int th_comment_query_count(th_comment *_tc,const char *_tag){
+  long i;
+  int  tag_len;
+  int  count;
+  tag_len=strlen(_tag);
+  count=0;
+  for(i=0;i<_tc->comments;i++){
+    if(!oc_tagcompare(_tc->user_comments[i],_tag,tag_len))count++;
+  }
+  return count;
+}
+
+void th_comment_clear(th_comment *_tc){
+  if(_tc!=NULL){
+    long i;
+    for(i=0;i<_tc->comments;i++)_ogg_free(_tc->user_comments[i]);
+    _ogg_free(_tc->user_comments);
+    _ogg_free(_tc->comment_lengths);
+    _ogg_free(_tc->vendor);
+    memset(_tc,0,sizeof(*_tc));
+  }
+}

lib/theoratest/lib/theora/x86/mmxfrag.c

@@ -0,0 +1,368 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+/*MMX acceleration of fragment reconstruction for motion compensation.
+  Originally written by Rudolf Marek.
+  Additional optimization by Nils Pipenbrinck.
+  Note: Loops are unrolled for best performance.
+  The iteration each instruction belongs to is marked in the comments as #i.*/
+#include <stddef.h>
+#include "x86int.h"
+
+#if defined(OC_X86_ASM)
+
+/*Copies an 8x8 block of pixels from _src to _dst, assuming _ystride bytes
+   between rows.*/
+# define OC_FRAG_COPY_MMX(_dst,_src,_ystride) \
+  do{ \
+    const unsigned char *src; \
+    unsigned char       *dst; \
+    ptrdiff_t            ystride3; \
+    src=(_src); \
+    dst=(_dst); \
+    __asm__ __volatile__( \
+      /*src+0*ystride*/ \
+      "movq (%[src]),%%mm0\n\t" \
+      /*src+1*ystride*/ \
+      "movq (%[src],%[ystride]),%%mm1\n\t" \
+      /*ystride3=ystride*3*/ \
+      "lea (%[ystride],%[ystride],2),%[ystride3]\n\t" \
+      /*src+2*ystride*/ \
+      "movq (%[src],%[ystride],2),%%mm2\n\t" \
+      /*src+3*ystride*/ \
+      "movq (%[src],%[ystride3]),%%mm3\n\t" \
+      /*dst+0*ystride*/ \
+      "movq %%mm0,(%[dst])\n\t" \
+      /*dst+1*ystride*/ \
+      "movq %%mm1,(%[dst],%[ystride])\n\t" \
+      /*Pointer to next 4.*/ \
+      "lea (%[src],%[ystride],4),%[src]\n\t" \
+      /*dst+2*ystride*/ \
+      "movq %%mm2,(%[dst],%[ystride],2)\n\t" \
+      /*dst+3*ystride*/ \
+      "movq %%mm3,(%[dst],%[ystride3])\n\t" \
+      /*Pointer to next 4.*/ \
+      "lea (%[dst],%[ystride],4),%[dst]\n\t" \
+      /*src+0*ystride*/ \
+      "movq (%[src]),%%mm0\n\t" \
+      /*src+1*ystride*/ \
+      "movq (%[src],%[ystride]),%%mm1\n\t" \
+      /*src+2*ystride*/ \
+      "movq (%[src],%[ystride],2),%%mm2\n\t" \
+      /*src+3*ystride*/ \
+      "movq (%[src],%[ystride3]),%%mm3\n\t" \
+      /*dst+0*ystride*/ \
+      "movq %%mm0,(%[dst])\n\t" \
+      /*dst+1*ystride*/ \
+      "movq %%mm1,(%[dst],%[ystride])\n\t" \
+      /*dst+2*ystride*/ \
+      "movq %%mm2,(%[dst],%[ystride],2)\n\t" \
+      /*dst+3*ystride*/ \
+      "movq %%mm3,(%[dst],%[ystride3])\n\t" \
+      :[dst]"+r"(dst),[src]"+r"(src),[ystride3]"=&r"(ystride3) \
+      :[ystride]"r"((ptrdiff_t)(_ystride)) \
+      :"memory" \
+    ); \
+  } \
+  while(0)
+
+/*Copies an 8x8 block of pixels from _src to _dst, assuming _ystride bytes
+   between rows.*/
+void oc_frag_copy_mmx(unsigned char *_dst,
+ const unsigned char *_src,int _ystride){
+  OC_FRAG_COPY_MMX(_dst,_src,_ystride);
+}
+
+/*Copies the fragments specified by the lists of fragment indices from one
+   frame to another.
+  _dst_frame:     The reference frame to copy to.
+  _src_frame:     The reference frame to copy from.
+  _ystride:       The row stride of the reference frames.
+  _fragis:        A pointer to a list of fragment indices.
+  _nfragis:       The number of fragment indices to copy.
+  _frag_buf_offs: The offsets of fragments in the reference frames.*/
+void oc_frag_copy_list_mmx(unsigned char *_dst_frame,
+ const unsigned char *_src_frame,int _ystride,
+ const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs){
+  ptrdiff_t fragii;
+  for(fragii=0;fragii<_nfragis;fragii++){
+    ptrdiff_t frag_buf_off;
+    frag_buf_off=_frag_buf_offs[_fragis[fragii]];
+    OC_FRAG_COPY_MMX(_dst_frame+frag_buf_off,
+     _src_frame+frag_buf_off,_ystride);
+  }
+}
+
+
+void oc_frag_recon_intra_mmx(unsigned char *_dst,int _ystride,
+ const ogg_int16_t *_residue){
+  __asm__ __volatile__(
+    /*Set mm0 to 0xFFFFFFFFFFFFFFFF.*/
+    "pcmpeqw %%mm0,%%mm0\n\t"
+    /*#0 Load low residue.*/
+    "movq 0*8(%[residue]),%%mm1\n\t"
+    /*#0 Load high residue.*/
+    "movq 1*8(%[residue]),%%mm2\n\t"
+    /*Set mm0 to 0x8000800080008000.*/
+    "psllw $15,%%mm0\n\t"
+    /*#1 Load low residue.*/
+    "movq 2*8(%[residue]),%%mm3\n\t"
+    /*#1 Load high residue.*/
+    "movq 3*8(%[residue]),%%mm4\n\t"
+    /*Set mm0 to 0x0080008000800080.*/
+    "psrlw $8,%%mm0\n\t"
+    /*#2 Load low residue.*/
+    "movq 4*8(%[residue]),%%mm5\n\t"
+    /*#2 Load high residue.*/
+    "movq 5*8(%[residue]),%%mm6\n\t"
+    /*#0 Bias low  residue.*/
+    "paddsw %%mm0,%%mm1\n\t"
+    /*#0 Bias high residue.*/
+    "paddsw %%mm0,%%mm2\n\t"
+    /*#0 Pack to byte.*/
+    "packuswb %%mm2,%%mm1\n\t"
+    /*#1 Bias low  residue.*/
+    "paddsw %%mm0,%%mm3\n\t"
+    /*#1 Bias high residue.*/
+    "paddsw %%mm0,%%mm4\n\t"
+    /*#1 Pack to byte.*/
+    "packuswb %%mm4,%%mm3\n\t"
+    /*#2 Bias low  residue.*/
+    "paddsw %%mm0,%%mm5\n\t"
+    /*#2 Bias high residue.*/
+    "paddsw %%mm0,%%mm6\n\t"
+    /*#2 Pack to byte.*/
+    "packuswb %%mm6,%%mm5\n\t"
+    /*#0 Write row.*/
+    "movq %%mm1,(%[dst])\n\t"
+    /*#1 Write row.*/
+    "movq %%mm3,(%[dst],%[ystride])\n\t"
+    /*#2 Write row.*/
+    "movq %%mm5,(%[dst],%[ystride],2)\n\t"
+    /*#3 Load low residue.*/
+    "movq 6*8(%[residue]),%%mm1\n\t"
+    /*#3 Load high residue.*/
+    "movq 7*8(%[residue]),%%mm2\n\t"
+    /*#4 Load high residue.*/
+    "movq 8*8(%[residue]),%%mm3\n\t"
+    /*#4 Load high residue.*/
+    "movq 9*8(%[residue]),%%mm4\n\t"
+    /*#5 Load high residue.*/
+    "movq 10*8(%[residue]),%%mm5\n\t"
+    /*#5 Load high residue.*/
+    "movq 11*8(%[residue]),%%mm6\n\t"
+    /*#3 Bias low  residue.*/
+    "paddsw %%mm0,%%mm1\n\t"
+    /*#3 Bias high residue.*/
+    "paddsw %%mm0,%%mm2\n\t"
+    /*#3 Pack to byte.*/
+    "packuswb %%mm2,%%mm1\n\t"
+    /*#4 Bias low  residue.*/
+    "paddsw %%mm0,%%mm3\n\t"
+    /*#4 Bias high residue.*/
+    "paddsw %%mm0,%%mm4\n\t"
+    /*#4 Pack to byte.*/
+    "packuswb %%mm4,%%mm3\n\t"
+    /*#5 Bias low  residue.*/
+    "paddsw %%mm0,%%mm5\n\t"
+    /*#5 Bias high residue.*/
+    "paddsw %%mm0,%%mm6\n\t"
+    /*#5 Pack to byte.*/
+    "packuswb %%mm6,%%mm5\n\t"
+    /*#3 Write row.*/
+    "movq %%mm1,(%[dst],%[ystride3])\n\t"
+    /*#4 Write row.*/
+    "movq %%mm3,(%[dst4])\n\t"
+    /*#5 Write row.*/
+    "movq %%mm5,(%[dst4],%[ystride])\n\t"
+    /*#6 Load low residue.*/
+    "movq 12*8(%[residue]),%%mm1\n\t"
+    /*#6 Load high residue.*/
+    "movq 13*8(%[residue]),%%mm2\n\t"
+    /*#7 Load low residue.*/
+    "movq 14*8(%[residue]),%%mm3\n\t"
+    /*#7 Load high residue.*/
+    "movq 15*8(%[residue]),%%mm4\n\t"
+    /*#6 Bias low  residue.*/
+    "paddsw %%mm0,%%mm1\n\t"
+    /*#6 Bias high residue.*/
+    "paddsw %%mm0,%%mm2\n\t"
+    /*#6 Pack to byte.*/
+    "packuswb %%mm2,%%mm1\n\t"
+    /*#7 Bias low  residue.*/
+    "paddsw %%mm0,%%mm3\n\t"
+    /*#7 Bias high residue.*/
+    "paddsw %%mm0,%%mm4\n\t"
+    /*#7 Pack to byte.*/
+    "packuswb %%mm4,%%mm3\n\t"
+    /*#6 Write row.*/
+    "movq %%mm1,(%[dst4],%[ystride],2)\n\t"
+    /*#7 Write row.*/
+    "movq %%mm3,(%[dst4],%[ystride3])\n\t"
+    :
+    :[residue]"r"(_residue),
+     [dst]"r"(_dst),
+     [dst4]"r"(_dst+(_ystride<<2)),
+     [ystride]"r"((ptrdiff_t)_ystride),
+     [ystride3]"r"((ptrdiff_t)_ystride*3)
+    :"memory"
+  );
+}
+
+void oc_frag_recon_inter_mmx(unsigned char *_dst,const unsigned char *_src,
+ int _ystride,const ogg_int16_t *_residue){
+  int i;
+  /*Zero mm0.*/
+  __asm__ __volatile__("pxor %%mm0,%%mm0\n\t"::);
+  for(i=4;i-->0;){
+    __asm__ __volatile__(
+      /*#0 Load source.*/
+      "movq (%[src]),%%mm3\n\t"
+      /*#1 Load source.*/
+      "movq (%[src],%[ystride]),%%mm7\n\t"
+      /*#0 Get copy of src.*/
+      "movq %%mm3,%%mm4\n\t"
+      /*#0 Expand high source.*/
+      "punpckhbw %%mm0,%%mm4\n\t"
+      /*#0 Expand low  source.*/
+      "punpcklbw %%mm0,%%mm3\n\t"
+      /*#0 Add residue high.*/
+      "paddsw 8(%[residue]),%%mm4\n\t"
+      /*#1 Get copy of src.*/
+      "movq %%mm7,%%mm2\n\t"
+      /*#0 Add residue low.*/
+      "paddsw (%[residue]), %%mm3\n\t"
+      /*#1 Expand high source.*/
+      "punpckhbw %%mm0,%%mm2\n\t"
+      /*#0 Pack final row pixels.*/
+      "packuswb %%mm4,%%mm3\n\t"
+      /*#1 Expand low  source.*/
+      "punpcklbw %%mm0,%%mm7\n\t"
+      /*#1 Add residue low.*/
+      "paddsw 16(%[residue]),%%mm7\n\t"
+      /*#1 Add residue high.*/
+      "paddsw 24(%[residue]),%%mm2\n\t"
+      /*Advance residue.*/
+      "lea 32(%[residue]),%[residue]\n\t"
+      /*#1 Pack final row pixels.*/
+      "packuswb %%mm2,%%mm7\n\t"
+      /*Advance src.*/
+      "lea (%[src],%[ystride],2),%[src]\n\t"
+      /*#0 Write row.*/
+      "movq %%mm3,(%[dst])\n\t"
+      /*#1 Write row.*/
+      "movq %%mm7,(%[dst],%[ystride])\n\t"
+      /*Advance dst.*/
+      "lea (%[dst],%[ystride],2),%[dst]\n\t"
+      :[residue]"+r"(_residue),[dst]"+r"(_dst),[src]"+r"(_src)
+      :[ystride]"r"((ptrdiff_t)_ystride)
+      :"memory"
+    );
+  }
+}
+
+void oc_frag_recon_inter2_mmx(unsigned char *_dst,const unsigned char *_src1,
+ const unsigned char *_src2,int _ystride,const ogg_int16_t *_residue){
+  int i;
+  /*Zero mm7.*/
+  __asm__ __volatile__("pxor %%mm7,%%mm7\n\t"::);
+  for(i=4;i-->0;){
+    __asm__ __volatile__(
+      /*#0 Load src1.*/
+      "movq (%[src1]),%%mm0\n\t"
+      /*#0 Load src2.*/
+      "movq (%[src2]),%%mm2\n\t"
+      /*#0 Copy src1.*/
+      "movq %%mm0,%%mm1\n\t"
+      /*#0 Copy src2.*/
+      "movq %%mm2,%%mm3\n\t"
+      /*#1 Load src1.*/
+      "movq (%[src1],%[ystride]),%%mm4\n\t"
+      /*#0 Unpack lower src1.*/
+      "punpcklbw %%mm7,%%mm0\n\t"
+      /*#1 Load src2.*/
+      "movq (%[src2],%[ystride]),%%mm5\n\t"
+      /*#0 Unpack higher src1.*/
+      "punpckhbw %%mm7,%%mm1\n\t"
+      /*#0 Unpack lower src2.*/
+      "punpcklbw %%mm7,%%mm2\n\t"
+      /*#0 Unpack higher src2.*/
+      "punpckhbw %%mm7,%%mm3\n\t"
+      /*Advance src1 ptr.*/
+      "lea (%[src1],%[ystride],2),%[src1]\n\t"
+      /*Advance src2 ptr.*/
+      "lea (%[src2],%[ystride],2),%[src2]\n\t"
+      /*#0 Lower src1+src2.*/
+      "paddsw %%mm2,%%mm0\n\t"
+      /*#0 Higher src1+src2.*/
+      "paddsw %%mm3,%%mm1\n\t"
+      /*#1 Copy src1.*/
+      "movq %%mm4,%%mm2\n\t"
+      /*#0 Build lo average.*/
+      "psraw $1,%%mm0\n\t"
+      /*#1 Copy src2.*/
+      "movq %%mm5,%%mm3\n\t"
+      /*#1 Unpack lower src1.*/
+      "punpcklbw %%mm7,%%mm4\n\t"
+      /*#0 Build hi average.*/
+      "psraw $1,%%mm1\n\t"
+      /*#1 Unpack higher src1.*/
+      "punpckhbw %%mm7,%%mm2\n\t"
+      /*#0 low+=residue.*/
+      "paddsw (%[residue]),%%mm0\n\t"
+      /*#1 Unpack lower src2.*/
+      "punpcklbw %%mm7,%%mm5\n\t"
+      /*#0 high+=residue.*/
+      "paddsw 8(%[residue]),%%mm1\n\t"
+      /*#1 Unpack higher src2.*/
+      "punpckhbw %%mm7,%%mm3\n\t"
+      /*#1 Lower src1+src2.*/
+      "paddsw %%mm4,%%mm5\n\t"
+      /*#0 Pack and saturate.*/
+      "packuswb %%mm1,%%mm0\n\t"
+      /*#1 Higher src1+src2.*/
+      "paddsw %%mm2,%%mm3\n\t"
+      /*#0 Write row.*/
+      "movq %%mm0,(%[dst])\n\t"
+      /*#1 Build lo average.*/
+      "psraw $1,%%mm5\n\t"
+      /*#1 Build hi average.*/
+      "psraw $1,%%mm3\n\t"
+      /*#1 low+=residue.*/
+      "paddsw 16(%[residue]),%%mm5\n\t"
+      /*#1 high+=residue.*/
+      "paddsw 24(%[residue]),%%mm3\n\t"
+      /*#1 Pack and saturate.*/
+      "packuswb  %%mm3,%%mm5\n\t"
+      /*#1 Write row ptr.*/
+      "movq %%mm5,(%[dst],%[ystride])\n\t"
+      /*Advance residue ptr.*/
+      "add $32,%[residue]\n\t"
+      /*Advance dest ptr.*/
+      "lea (%[dst],%[ystride],2),%[dst]\n\t"
+     :[dst]"+r"(_dst),[residue]"+r"(_residue),
+      [src1]"+r"(_src1),[src2]"+r"(_src2)
+     :[ystride]"r"((ptrdiff_t)_ystride)
+     :"memory"
+    );
+  }
+}
+
+void oc_restore_fpu_mmx(void){
+  __asm__ __volatile__("emms\n\t");
+}
+#endif

lib/theoratest/lib/theora/x86/mmxidct.c

@@ -0,0 +1,558 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+/*MMX acceleration of Theora's iDCT.
+  Originally written by Rudolf Marek, based on code from On2's VP3.*/
+#include "x86int.h"
+#include "../dct.h"
+
+#if defined(OC_X86_ASM)
+
+/*These are offsets into the table of constants below.*/
+/*7 rows of cosines, in order: pi/16 * (1 ... 7).*/
+#define OC_COSINE_OFFSET (0)
+/*A row of 8's.*/
+#define OC_EIGHT_OFFSET  (56)
+
+
+
+/*38 cycles*/
+#define OC_IDCT_BEGIN(_y,_x) \
+  "#OC_IDCT_BEGIN\n\t" \
+  "movq "OC_I(3,_x)",%%mm2\n\t" \
+  "movq "OC_MEM_OFFS(0x30,c)",%%mm6\n\t" \
+  "movq %%mm2,%%mm4\n\t" \
+  "movq "OC_J(5,_x)",%%mm7\n\t" \
+  "pmulhw %%mm6,%%mm4\n\t" \
+  "movq "OC_MEM_OFFS(0x50,c)",%%mm1\n\t" \
+  "pmulhw %%mm7,%%mm6\n\t" \
+  "movq %%mm1,%%mm5\n\t" \
+  "pmulhw %%mm2,%%mm1\n\t" \
+  "movq "OC_I(1,_x)",%%mm3\n\t" \
+  "pmulhw %%mm7,%%mm5\n\t" \
+  "movq "OC_MEM_OFFS(0x10,c)",%%mm0\n\t" \
+  "paddw %%mm2,%%mm4\n\t" \
+  "paddw %%mm7,%%mm6\n\t" \
+  "paddw %%mm1,%%mm2\n\t" \
+  "movq "OC_J(7,_x)",%%mm1\n\t" \
+  "paddw %%mm5,%%mm7\n\t" \
+  "movq %%mm0,%%mm5\n\t" \
+  "pmulhw %%mm3,%%mm0\n\t" \
+  "paddw %%mm7,%%mm4\n\t" \
+  "pmulhw %%mm1,%%mm5\n\t" \
+  "movq "OC_MEM_OFFS(0x70,c)",%%mm7\n\t" \
+  "psubw %%mm2,%%mm6\n\t" \
+  "paddw %%mm3,%%mm0\n\t" \
+  "pmulhw %%mm7,%%mm3\n\t" \
+  "movq "OC_I(2,_x)",%%mm2\n\t" \
+  "pmulhw %%mm1,%%mm7\n\t" \
+  "paddw %%mm1,%%mm5\n\t" \
+  "movq %%mm2,%%mm1\n\t" \
+  "pmulhw "OC_MEM_OFFS(0x20,c)",%%mm2\n\t" \
+  "psubw %%mm5,%%mm3\n\t" \
+  "movq "OC_J(6,_x)",%%mm5\n\t" \
+  "paddw %%mm7,%%mm0\n\t" \
+  "movq %%mm5,%%mm7\n\t" \
+  "psubw %%mm4,%%mm0\n\t" \
+  "pmulhw "OC_MEM_OFFS(0x20,c)",%%mm5\n\t" \
+  "paddw %%mm1,%%mm2\n\t" \
+  "pmulhw "OC_MEM_OFFS(0x60,c)",%%mm1\n\t" \
+  "paddw %%mm4,%%mm4\n\t" \
+  "paddw %%mm0,%%mm4\n\t" \
+  "psubw %%mm6,%%mm3\n\t" \
+  "paddw %%mm7,%%mm5\n\t" \
+  "paddw %%mm6,%%mm6\n\t" \
+  "pmulhw "OC_MEM_OFFS(0x60,c)",%%mm7\n\t" \
+  "paddw %%mm3,%%mm6\n\t" \
+  "movq %%mm4,"OC_I(1,_y)"\n\t" \
+  "psubw %%mm5,%%mm1\n\t" \
+  "movq "OC_MEM_OFFS(0x40,c)",%%mm4\n\t" \
+  "movq %%mm3,%%mm5\n\t" \
+  "pmulhw %%mm4,%%mm3\n\t" \
+  "paddw %%mm2,%%mm7\n\t" \
+  "movq %%mm6,"OC_I(2,_y)"\n\t" \
+  "movq %%mm0,%%mm2\n\t" \
+  "movq "OC_I(0,_x)",%%mm6\n\t" \
+  "pmulhw %%mm4,%%mm0\n\t" \
+  "paddw %%mm3,%%mm5\n\t" \
+  "movq "OC_J(4,_x)",%%mm3\n\t" \
+  "psubw %%mm1,%%mm5\n\t" \
+  "paddw %%mm0,%%mm2\n\t" \
+  "psubw %%mm3,%%mm6\n\t" \
+  "movq %%mm6,%%mm0\n\t" \
+  "pmulhw %%mm4,%%mm6\n\t" \
+  "paddw %%mm3,%%mm3\n\t" \
+  "paddw %%mm1,%%mm1\n\t" \
+  "paddw %%mm0,%%mm3\n\t" \
+  "paddw %%mm5,%%mm1\n\t" \
+  "pmulhw %%mm3,%%mm4\n\t" \
+  "paddw %%mm0,%%mm6\n\t" \
+  "psubw %%mm2,%%mm6\n\t" \
+  "paddw %%mm2,%%mm2\n\t" \
+  "movq "OC_I(1,_y)",%%mm0\n\t" \
+  "paddw %%mm6,%%mm2\n\t" \
+  "paddw %%mm3,%%mm4\n\t" \
+  "psubw %%mm1,%%mm2\n\t" \
+  "#end OC_IDCT_BEGIN\n\t" \
+
+/*38+8=46 cycles.*/
+#define OC_ROW_IDCT(_y,_x) \
+  "#OC_ROW_IDCT\n" \
+  OC_IDCT_BEGIN(_y,_x) \
+  /*r3=D'*/ \
+  "movq "OC_I(2,_y)",%%mm3\n\t" \
+  /*r4=E'=E-G*/ \
+  "psubw %%mm7,%%mm4\n\t" \
+  /*r1=H'+H'*/ \
+  "paddw %%mm1,%%mm1\n\t" \
+  /*r7=G+G*/ \
+  "paddw %%mm7,%%mm7\n\t" \
+  /*r1=R1=A''+H'*/ \
+  "paddw %%mm2,%%mm1\n\t" \
+  /*r7=G'=E+G*/ \
+  "paddw %%mm4,%%mm7\n\t" \
+  /*r4=R4=E'-D'*/ \
+  "psubw %%mm3,%%mm4\n\t" \
+  "paddw %%mm3,%%mm3\n\t" \
+  /*r6=R6=F'-B''*/ \
+  "psubw %%mm5,%%mm6\n\t" \
+  "paddw %%mm5,%%mm5\n\t" \
+  /*r3=R3=E'+D'*/ \
+  "paddw %%mm4,%%mm3\n\t" \
+  /*r5=R5=F'+B''*/ \
+  "paddw %%mm6,%%mm5\n\t" \
+  /*r7=R7=G'-C'*/ \
+  "psubw %%mm0,%%mm7\n\t" \
+  "paddw %%mm0,%%mm0\n\t" \
+  /*Save R1.*/ \
+  "movq %%mm1,"OC_I(1,_y)"\n\t" \
+  /*r0=R0=G.+C.*/ \
+  "paddw %%mm7,%%mm0\n\t" \
+  "#end OC_ROW_IDCT\n\t" \
+
+/*The following macro does two 4x4 transposes in place.
+  At entry, we assume:
+    r0 = a3 a2 a1 a0
+  I(1) = b3 b2 b1 b0
+    r2 = c3 c2 c1 c0
+    r3 = d3 d2 d1 d0
+
+    r4 = e3 e2 e1 e0
+    r5 = f3 f2 f1 f0
+    r6 = g3 g2 g1 g0
+    r7 = h3 h2 h1 h0
+
+  At exit, we have:
+  I(0) = d0 c0 b0 a0
+  I(1) = d1 c1 b1 a1
+  I(2) = d2 c2 b2 a2
+  I(3) = d3 c3 b3 a3
+
+  J(4) = h0 g0 f0 e0
+  J(5) = h1 g1 f1 e1
+  J(6) = h2 g2 f2 e2
+  J(7) = h3 g3 f3 e3
+
+  I(0) I(1) I(2) I(3) is the transpose of r0 I(1) r2 r3.
+  J(4) J(5) J(6) J(7) is the transpose of r4  r5  r6 r7.
+
+  Since r1 is free at entry, we calculate the Js first.*/
+/*19 cycles.*/
+#define OC_TRANSPOSE(_y) \
+  "#OC_TRANSPOSE\n\t" \
+  "movq %%mm4,%%mm1\n\t" \
+  "punpcklwd %%mm5,%%mm4\n\t" \
+  "movq %%mm0,"OC_I(0,_y)"\n\t" \
+  "punpckhwd %%mm5,%%mm1\n\t" \
+  "movq %%mm6,%%mm0\n\t" \
+  "punpcklwd %%mm7,%%mm6\n\t" \
+  "movq %%mm4,%%mm5\n\t" \
+  "punpckldq %%mm6,%%mm4\n\t" \
+  "punpckhdq %%mm6,%%mm5\n\t" \
+  "movq %%mm1,%%mm6\n\t" \
+  "movq %%mm4,"OC_J(4,_y)"\n\t" \
+  "punpckhwd %%mm7,%%mm0\n\t" \
+  "movq %%mm5,"OC_J(5,_y)"\n\t" \
+  "punpckhdq %%mm0,%%mm6\n\t" \
+  "movq "OC_I(0,_y)",%%mm4\n\t" \
+  "punpckldq %%mm0,%%mm1\n\t" \
+  "movq "OC_I(1,_y)",%%mm5\n\t" \
+  "movq %%mm4,%%mm0\n\t" \
+  "movq %%mm6,"OC_J(7,_y)"\n\t" \
+  "punpcklwd %%mm5,%%mm0\n\t" \
+  "movq %%mm1,"OC_J(6,_y)"\n\t" \
+  "punpckhwd %%mm5,%%mm4\n\t" \
+  "movq %%mm2,%%mm5\n\t" \
+  "punpcklwd %%mm3,%%mm2\n\t" \
+  "movq %%mm0,%%mm1\n\t" \
+  "punpckldq %%mm2,%%mm0\n\t" \
+  "punpckhdq %%mm2,%%mm1\n\t" \
+  "movq %%mm4,%%mm2\n\t" \
+  "movq %%mm0,"OC_I(0,_y)"\n\t" \
+  "punpckhwd %%mm3,%%mm5\n\t" \
+  "movq %%mm1,"OC_I(1,_y)"\n\t" \
+  "punpckhdq %%mm5,%%mm4\n\t" \
+  "punpckldq %%mm5,%%mm2\n\t" \
+  "movq %%mm4,"OC_I(3,_y)"\n\t" \
+  "movq %%mm2,"OC_I(2,_y)"\n\t" \
+  "#end OC_TRANSPOSE\n\t" \
+
+/*38+19=57 cycles.*/
+#define OC_COLUMN_IDCT(_y) \
+  "#OC_COLUMN_IDCT\n" \
+  OC_IDCT_BEGIN(_y,_y) \
+  "paddw "OC_MEM_OFFS(0x00,c)",%%mm2\n\t" \
+  /*r1=H'+H'*/ \
+  "paddw %%mm1,%%mm1\n\t" \
+  /*r1=R1=A''+H'*/ \
+  "paddw %%mm2,%%mm1\n\t" \
+  /*r2=NR2*/ \
+  "psraw $4,%%mm2\n\t" \
+  /*r4=E'=E-G*/ \
+  "psubw %%mm7,%%mm4\n\t" \
+  /*r1=NR1*/ \
+  "psraw $4,%%mm1\n\t" \
+  /*r3=D'*/ \
+  "movq "OC_I(2,_y)",%%mm3\n\t" \
+  /*r7=G+G*/ \
+  "paddw %%mm7,%%mm7\n\t" \
+  /*Store NR2 at I(2).*/ \
+  "movq %%mm2,"OC_I(2,_y)"\n\t" \
+  /*r7=G'=E+G*/ \
+  "paddw %%mm4,%%mm7\n\t" \
+  /*Store NR1 at I(1).*/ \
+  "movq %%mm1,"OC_I(1,_y)"\n\t" \
+  /*r4=R4=E'-D'*/ \
+  "psubw %%mm3,%%mm4\n\t" \
+  "paddw "OC_MEM_OFFS(0x00,c)",%%mm4\n\t" \
+  /*r3=D'+D'*/ \
+  "paddw %%mm3,%%mm3\n\t" \
+  /*r3=R3=E'+D'*/ \
+  "paddw %%mm4,%%mm3\n\t" \
+  /*r4=NR4*/ \
+  "psraw $4,%%mm4\n\t" \
+  /*r6=R6=F'-B''*/ \
+  "psubw %%mm5,%%mm6\n\t" \
+  /*r3=NR3*/ \
+  "psraw $4,%%mm3\n\t" \
+  "paddw "OC_MEM_OFFS(0x00,c)",%%mm6\n\t" \
+  /*r5=B''+B''*/ \
+  "paddw %%mm5,%%mm5\n\t" \
+  /*r5=R5=F'+B''*/ \
+  "paddw %%mm6,%%mm5\n\t" \
+  /*r6=NR6*/ \
+  "psraw $4,%%mm6\n\t" \
+  /*Store NR4 at J(4).*/ \
+  "movq %%mm4,"OC_J(4,_y)"\n\t" \
+  /*r5=NR5*/ \
+  "psraw $4,%%mm5\n\t" \
+  /*Store NR3 at I(3).*/ \
+  "movq %%mm3,"OC_I(3,_y)"\n\t" \
+  /*r7=R7=G'-C'*/ \
+  "psubw %%mm0,%%mm7\n\t" \
+  "paddw "OC_MEM_OFFS(0x00,c)",%%mm7\n\t" \
+  /*r0=C'+C'*/ \
+  "paddw %%mm0,%%mm0\n\t" \
+  /*r0=R0=G'+C'*/ \
+  "paddw %%mm7,%%mm0\n\t" \
+  /*r7=NR7*/ \
+  "psraw $4,%%mm7\n\t" \
+  /*Store NR6 at J(6).*/ \
+  "movq %%mm6,"OC_J(6,_y)"\n\t" \
+  /*r0=NR0*/ \
+  "psraw $4,%%mm0\n\t" \
+  /*Store NR5 at J(5).*/ \
+  "movq %%mm5,"OC_J(5,_y)"\n\t" \
+  /*Store NR7 at J(7).*/ \
+  "movq %%mm7,"OC_J(7,_y)"\n\t" \
+  /*Store NR0 at I(0).*/ \
+  "movq %%mm0,"OC_I(0,_y)"\n\t" \
+  "#end OC_COLUMN_IDCT\n\t" \
+
+static void oc_idct8x8_slow_mmx(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  int i;
+  /*This routine accepts an 8x8 matrix, but in partially transposed form.
+    Every 4x4 block is transposed.*/
+  __asm__ __volatile__(
+#define OC_I(_k,_y)   OC_MEM_OFFS((_k)*16,_y)
+#define OC_J(_k,_y)   OC_MEM_OFFS(((_k)-4)*16+8,_y)
+    OC_ROW_IDCT(y,x)
+    OC_TRANSPOSE(y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y)   OC_MEM_OFFS((_k)*16+64,_y)
+#define OC_J(_k,_y)   OC_MEM_OFFS(((_k)-4)*16+72,_y)
+    OC_ROW_IDCT(y,x)
+    OC_TRANSPOSE(y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y)   OC_MEM_OFFS((_k)*16,_y)
+#define OC_J(_k,_y)   OC_I(_k,_y)
+    OC_COLUMN_IDCT(y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y)   OC_MEM_OFFS((_k)*16+8,_y)
+#define OC_J(_k,_y)   OC_I(_k,_y)
+    OC_COLUMN_IDCT(y)
+#undef  OC_I
+#undef  OC_J
+    :[y]"=m"OC_ARRAY_OPERAND(ogg_int16_t,_y,64)
+    :[x]"m"OC_CONST_ARRAY_OPERAND(ogg_int16_t,_x,64),
+     [c]"m"OC_CONST_ARRAY_OPERAND(ogg_int16_t,OC_IDCT_CONSTS,128)
+  );
+  __asm__ __volatile__("pxor %%mm0,%%mm0\n\t"::);
+  for(i=0;i<4;i++){
+    __asm__ __volatile__(
+      "movq %%mm0,"OC_MEM_OFFS(0x00,x)"\n\t"
+      "movq %%mm0,"OC_MEM_OFFS(0x08,x)"\n\t"
+      "movq %%mm0,"OC_MEM_OFFS(0x10,x)"\n\t"
+      "movq %%mm0,"OC_MEM_OFFS(0x18,x)"\n\t"
+      :[x]"=m"OC_ARRAY_OPERAND(ogg_int16_t,_x+16*i,16)
+    );
+  }
+}
+
+/*25 cycles.*/
+#define OC_IDCT_BEGIN_10(_y,_x) \
+ "#OC_IDCT_BEGIN_10\n\t" \
+ "movq "OC_I(3,_x)",%%mm2\n\t" \
+ "nop\n\t" \
+ "movq "OC_MEM_OFFS(0x30,c)",%%mm6\n\t" \
+ "movq %%mm2,%%mm4\n\t" \
+ "movq "OC_MEM_OFFS(0x50,c)",%%mm1\n\t" \
+ "pmulhw %%mm6,%%mm4\n\t" \
+ "movq "OC_I(1,_x)",%%mm3\n\t" \
+ "pmulhw %%mm2,%%mm1\n\t" \
+ "movq "OC_MEM_OFFS(0x10,c)",%%mm0\n\t" \
+ "paddw %%mm2,%%mm4\n\t" \
+ "pxor %%mm6,%%mm6\n\t" \
+ "paddw %%mm1,%%mm2\n\t" \
+ "movq "OC_I(2,_x)",%%mm5\n\t" \
+ "pmulhw %%mm3,%%mm0\n\t" \
+ "movq %%mm5,%%mm1\n\t" \
+ "paddw %%mm3,%%mm0\n\t" \
+ "pmulhw "OC_MEM_OFFS(0x70,c)",%%mm3\n\t" \
+ "psubw %%mm2,%%mm6\n\t" \
+ "pmulhw "OC_MEM_OFFS(0x20,c)",%%mm5\n\t" \
+ "psubw %%mm4,%%mm0\n\t" \
+ "movq "OC_I(2,_x)",%%mm7\n\t" \
+ "paddw %%mm4,%%mm4\n\t" \
+ "paddw %%mm5,%%mm7\n\t" \
+ "paddw %%mm0,%%mm4\n\t" \
+ "pmulhw "OC_MEM_OFFS(0x60,c)",%%mm1\n\t" \
+ "psubw %%mm6,%%mm3\n\t" \
+ "movq %%mm4,"OC_I(1,_y)"\n\t" \
+ "paddw %%mm6,%%mm6\n\t" \
+ "movq "OC_MEM_OFFS(0x40,c)",%%mm4\n\t" \
+ "paddw %%mm3,%%mm6\n\t" \
+ "movq %%mm3,%%mm5\n\t" \
+ "pmulhw %%mm4,%%mm3\n\t" \
+ "movq %%mm6,"OC_I(2,_y)"\n\t" \
+ "movq %%mm0,%%mm2\n\t" \
+ "movq "OC_I(0,_x)",%%mm6\n\t" \
+ "pmulhw %%mm4,%%mm0\n\t" \
+ "paddw %%mm3,%%mm5\n\t" \
+ "paddw %%mm0,%%mm2\n\t" \
+ "psubw %%mm1,%%mm5\n\t" \
+ "pmulhw %%mm4,%%mm6\n\t" \
+ "paddw "OC_I(0,_x)",%%mm6\n\t" \
+ "paddw %%mm1,%%mm1\n\t" \
+ "movq %%mm6,%%mm4\n\t" \
+ "paddw %%mm5,%%mm1\n\t" \
+ "psubw %%mm2,%%mm6\n\t" \
+ "paddw %%mm2,%%mm2\n\t" \
+ "movq "OC_I(1,_y)",%%mm0\n\t" \
+ "paddw %%mm6,%%mm2\n\t" \
+ "psubw %%mm1,%%mm2\n\t" \
+ "nop\n\t" \
+ "#end OC_IDCT_BEGIN_10\n\t" \
+
+/*25+8=33 cycles.*/
+#define OC_ROW_IDCT_10(_y,_x) \
+ "#OC_ROW_IDCT_10\n\t" \
+ OC_IDCT_BEGIN_10(_y,_x) \
+ /*r3=D'*/ \
+ "movq "OC_I(2,_y)",%%mm3\n\t" \
+ /*r4=E'=E-G*/ \
+ "psubw %%mm7,%%mm4\n\t" \
+ /*r1=H'+H'*/ \
+ "paddw %%mm1,%%mm1\n\t" \
+ /*r7=G+G*/ \
+ "paddw %%mm7,%%mm7\n\t" \
+ /*r1=R1=A''+H'*/ \
+ "paddw %%mm2,%%mm1\n\t" \
+ /*r7=G'=E+G*/ \
+ "paddw %%mm4,%%mm7\n\t" \
+ /*r4=R4=E'-D'*/ \
+ "psubw %%mm3,%%mm4\n\t" \
+ "paddw %%mm3,%%mm3\n\t" \
+ /*r6=R6=F'-B''*/ \
+ "psubw %%mm5,%%mm6\n\t" \
+ "paddw %%mm5,%%mm5\n\t" \
+ /*r3=R3=E'+D'*/ \
+ "paddw %%mm4,%%mm3\n\t" \
+ /*r5=R5=F'+B''*/ \
+ "paddw %%mm6,%%mm5\n\t" \
+ /*r7=R7=G'-C'*/ \
+ "psubw %%mm0,%%mm7\n\t" \
+ "paddw %%mm0,%%mm0\n\t" \
+ /*Save R1.*/ \
+ "movq %%mm1,"OC_I(1,_y)"\n\t" \
+ /*r0=R0=G'+C'*/ \
+ "paddw %%mm7,%%mm0\n\t" \
+ "#end OC_ROW_IDCT_10\n\t" \
+
+/*25+19=44 cycles'*/
+#define OC_COLUMN_IDCT_10(_y) \
+ "#OC_COLUMN_IDCT_10\n\t" \
+ OC_IDCT_BEGIN_10(_y,_y) \
+ "paddw "OC_MEM_OFFS(0x00,c)",%%mm2\n\t" \
+ /*r1=H'+H'*/ \
+ "paddw %%mm1,%%mm1\n\t" \
+ /*r1=R1=A''+H'*/ \
+ "paddw %%mm2,%%mm1\n\t" \
+ /*r2=NR2*/ \
+ "psraw $4,%%mm2\n\t" \
+ /*r4=E'=E-G*/ \
+ "psubw %%mm7,%%mm4\n\t" \
+ /*r1=NR1*/ \
+ "psraw $4,%%mm1\n\t" \
+ /*r3=D'*/ \
+ "movq "OC_I(2,_y)",%%mm3\n\t" \
+ /*r7=G+G*/ \
+ "paddw %%mm7,%%mm7\n\t" \
+ /*Store NR2 at I(2).*/ \
+ "movq %%mm2,"OC_I(2,_y)"\n\t" \
+ /*r7=G'=E+G*/ \
+ "paddw %%mm4,%%mm7\n\t" \
+ /*Store NR1 at I(1).*/ \
+ "movq %%mm1,"OC_I(1,_y)"\n\t" \
+ /*r4=R4=E'-D'*/ \
+ "psubw %%mm3,%%mm4\n\t" \
+ "paddw "OC_MEM_OFFS(0x00,c)",%%mm4\n\t" \
+ /*r3=D'+D'*/ \
+ "paddw %%mm3,%%mm3\n\t" \
+ /*r3=R3=E'+D'*/ \
+ "paddw %%mm4,%%mm3\n\t" \
+ /*r4=NR4*/ \
+ "psraw $4,%%mm4\n\t" \
+ /*r6=R6=F'-B''*/ \
+ "psubw %%mm5,%%mm6\n\t" \
+ /*r3=NR3*/ \
+ "psraw $4,%%mm3\n\t" \
+ "paddw "OC_MEM_OFFS(0x00,c)",%%mm6\n\t" \
+ /*r5=B''+B''*/ \
+ "paddw %%mm5,%%mm5\n\t" \
+ /*r5=R5=F'+B''*/ \
+ "paddw %%mm6,%%mm5\n\t" \
+ /*r6=NR6*/ \
+ "psraw $4,%%mm6\n\t" \
+ /*Store NR4 at J(4).*/ \
+ "movq %%mm4,"OC_J(4,_y)"\n\t" \
+ /*r5=NR5*/ \
+ "psraw $4,%%mm5\n\t" \
+ /*Store NR3 at I(3).*/ \
+ "movq %%mm3,"OC_I(3,_y)"\n\t" \
+ /*r7=R7=G'-C'*/ \
+ "psubw %%mm0,%%mm7\n\t" \
+ "paddw "OC_MEM_OFFS(0x00,c)",%%mm7\n\t" \
+ /*r0=C'+C'*/ \
+ "paddw %%mm0,%%mm0\n\t" \
+ /*r0=R0=G'+C'*/ \
+ "paddw %%mm7,%%mm0\n\t" \
+ /*r7=NR7*/ \
+ "psraw $4,%%mm7\n\t" \
+ /*Store NR6 at J(6).*/ \
+ "movq %%mm6,"OC_J(6,_y)"\n\t" \
+ /*r0=NR0*/ \
+ "psraw $4,%%mm0\n\t" \
+ /*Store NR5 at J(5).*/ \
+ "movq %%mm5,"OC_J(5,_y)"\n\t" \
+ /*Store NR7 at J(7).*/ \
+ "movq %%mm7,"OC_J(7,_y)"\n\t" \
+ /*Store NR0 at I(0).*/ \
+ "movq %%mm0,"OC_I(0,_y)"\n\t" \
+ "#end OC_COLUMN_IDCT_10\n\t" \
+
+static void oc_idct8x8_10_mmx(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  __asm__ __volatile__(
+#define OC_I(_k,_y) OC_MEM_OFFS((_k)*16,_y)
+#define OC_J(_k,_y) OC_MEM_OFFS(((_k)-4)*16+8,_y)
+    /*Done with dequant, descramble, and partial transpose.
+      Now do the iDCT itself.*/
+    OC_ROW_IDCT_10(y,x)
+    OC_TRANSPOSE(y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y) OC_MEM_OFFS((_k)*16,_y)
+#define OC_J(_k,_y) OC_I(_k,_y)
+    OC_COLUMN_IDCT_10(y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y) OC_MEM_OFFS((_k)*16+8,_y)
+#define OC_J(_k,_y) OC_I(_k,_y)
+    OC_COLUMN_IDCT_10(y)
+#undef  OC_I
+#undef  OC_J
+    :[y]"=m"OC_ARRAY_OPERAND(ogg_int16_t,_y,64)
+    :[x]"m"OC_CONST_ARRAY_OPERAND(ogg_int16_t,_x,64),
+     [c]"m"OC_CONST_ARRAY_OPERAND(ogg_int16_t,OC_IDCT_CONSTS,128)
+  );
+  __asm__ __volatile__(
+    "pxor %%mm0,%%mm0\n\t"
+    "movq %%mm0,"OC_MEM_OFFS(0x00,x)"\n\t"
+    "movq %%mm0,"OC_MEM_OFFS(0x10,x)"\n\t"
+    "movq %%mm0,"OC_MEM_OFFS(0x20,x)"\n\t"
+    "movq %%mm0,"OC_MEM_OFFS(0x30,x)"\n\t"
+    :[x]"+m"OC_ARRAY_OPERAND(ogg_int16_t,_x,28)
+  );
+}
+
+/*Performs an inverse 8x8 Type-II DCT transform.
+  The input is assumed to be scaled by a factor of 4 relative to orthonormal
+   version of the transform.*/
+void oc_idct8x8_mmx(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi){
+  /*_last_zzi is subtly different from an actual count of the number of
+     coefficients we decoded for this block.
+    It contains the value of zzi BEFORE the final token in the block was
+     decoded.
+    In most cases this is an EOB token (the continuation of an EOB run from a
+     previous block counts), and so this is the same as the coefficient count.
+    However, in the case that the last token was NOT an EOB token, but filled
+     the block up with exactly 64 coefficients, _last_zzi will be less than 64.
+    Provided the last token was not a pure zero run, the minimum value it can
+     be is 46, and so that doesn't affect any of the cases in this routine.
+    However, if the last token WAS a pure zero run of length 63, then _last_zzi
+     will be 1 while the number of coefficients decoded is 64.
+    Thus, we will trigger the following special case, where the real
+     coefficient count would not.
+    Note also that a zero run of length 64 will give _last_zzi a value of 0,
+     but we still process the DC coefficient, which might have a non-zero value
+     due to DC prediction.
+    Although convoluted, this is arguably the correct behavior: it allows us to
+     use a smaller transform when the block ends with a long zero run instead
+     of a normal EOB token.
+    It could be smarter... multiple separate zero runs at the end of a block
+     will fool it, but an encoder that generates these really deserves what it
+     gets.
+    Needless to say we inherited this approach from VP3.*/
+  /*Then perform the iDCT.*/
+  if(_last_zzi<=10)oc_idct8x8_10_mmx(_y,_x);
+  else oc_idct8x8_slow_mmx(_y,_x);
+}
+
+#endif

lib/theoratest/lib/theora/x86/mmxloop.h

@@ -0,0 +1,318 @@
+#if !defined(_x86_mmxloop_H)
+# define _x86_mmxloop_H (1)
+# include <stddef.h>
+# include "x86int.h"
+
+#if defined(OC_X86_ASM)
+
+/*On entry, mm0={a0,...,a7}, mm1={b0,...,b7}, mm2={c0,...,c7}, mm3={d0,...d7}.
+  On exit, mm1={b0+lflim(R_0,L),...,b7+lflim(R_7,L)} and
+   mm2={c0-lflim(R_0,L),...,c7-lflim(R_7,L)}; mm0 and mm3 are clobbered.*/
+#define OC_LOOP_FILTER8_MMX \
+  "#OC_LOOP_FILTER8_MMX\n\t" \
+  /*mm7=0*/ \
+  "pxor %%mm7,%%mm7\n\t" \
+  /*mm6:mm0={a0,...,a7}*/ \
+  "movq %%mm0,%%mm6\n\t" \
+  "punpcklbw %%mm7,%%mm0\n\t" \
+  "punpckhbw %%mm7,%%mm6\n\t" \
+  /*mm3:mm5={d0,...,d7}*/ \
+  "movq %%mm3,%%mm5\n\t" \
+  "punpcklbw %%mm7,%%mm3\n\t" \
+  "punpckhbw %%mm7,%%mm5\n\t" \
+  /*mm6:mm0={a0-d0,...,a7-d7}*/ \
+  "psubw %%mm3,%%mm0\n\t" \
+  "psubw %%mm5,%%mm6\n\t" \
+  /*mm3:mm1={b0,...,b7}*/ \
+  "movq %%mm1,%%mm3\n\t" \
+  "punpcklbw %%mm7,%%mm1\n\t" \
+  "movq %%mm2,%%mm4\n\t" \
+  "punpckhbw %%mm7,%%mm3\n\t" \
+  /*mm5:mm4={c0,...,c7}*/ \
+  "movq %%mm2,%%mm5\n\t" \
+  "punpcklbw %%mm7,%%mm4\n\t" \
+  "punpckhbw %%mm7,%%mm5\n\t" \
+  /*mm7={3}x4 \
+    mm5:mm4={c0-b0,...,c7-b7}*/ \
+  "pcmpeqw %%mm7,%%mm7\n\t" \
+  "psubw %%mm1,%%mm4\n\t" \
+  "psrlw $14,%%mm7\n\t" \
+  "psubw %%mm3,%%mm5\n\t" \
+  /*Scale by 3.*/ \
+  "pmullw %%mm7,%%mm4\n\t" \
+  "pmullw %%mm7,%%mm5\n\t" \
+  /*mm7={4}x4 \
+    mm5:mm4=f={a0-d0+3*(c0-b0),...,a7-d7+3*(c7-b7)}*/ \
+  "psrlw $1,%%mm7\n\t" \
+  "paddw %%mm0,%%mm4\n\t" \
+  "psllw $2,%%mm7\n\t" \
+  "movq (%[ll]),%%mm0\n\t" \
+  "paddw %%mm6,%%mm5\n\t" \
+  /*R_i has the range [-127,128], so we compute -R_i instead. \
+    mm4=-R_i=-(f+4>>3)=0xFF^(f-4>>3)*/ \
+  "psubw %%mm7,%%mm4\n\t" \
+  "psubw %%mm7,%%mm5\n\t" \
+  "psraw $3,%%mm4\n\t" \
+  "psraw $3,%%mm5\n\t" \
+  "pcmpeqb %%mm7,%%mm7\n\t" \
+  "packsswb %%mm5,%%mm4\n\t" \
+  "pxor %%mm6,%%mm6\n\t" \
+  "pxor %%mm7,%%mm4\n\t" \
+  "packuswb %%mm3,%%mm1\n\t" \
+  /*Now compute lflim of -mm4 cf. Section 7.10 of the sepc.*/ \
+  /*There's no unsigned byte+signed byte with unsigned saturation op code, so \
+     we have to split things by sign (the other option is to work in 16 bits, \
+     but working in 8 bits gives much better parallelism). \
+    We compute abs(R_i), but save a mask of which terms were negative in mm6. \
+    Then we compute mm4=abs(lflim(R_i,L))=min(abs(R_i),max(2*L-abs(R_i),0)). \
+    Finally, we split mm4 into positive and negative pieces using the mask in \
+     mm6, and add and subtract them as appropriate.*/ \
+  /*mm4=abs(-R_i)*/ \
+  /*mm7=255-2*L*/ \
+  "pcmpgtb %%mm4,%%mm6\n\t" \
+  "psubb %%mm0,%%mm7\n\t" \
+  "pxor %%mm6,%%mm4\n\t" \
+  "psubb %%mm0,%%mm7\n\t" \
+  "psubb %%mm6,%%mm4\n\t" \
+  /*mm7=255-max(2*L-abs(R_i),0)*/ \
+  "paddusb %%mm4,%%mm7\n\t" \
+  /*mm4=min(abs(R_i),max(2*L-abs(R_i),0))*/ \
+  "paddusb %%mm7,%%mm4\n\t" \
+  "psubusb %%mm7,%%mm4\n\t" \
+  /*Now split mm4 by the original sign of -R_i.*/ \
+  "movq %%mm4,%%mm5\n\t" \
+  "pand %%mm6,%%mm4\n\t" \
+  "pandn %%mm5,%%mm6\n\t" \
+  /*mm1={b0+lflim(R_0,L),...,b7+lflim(R_7,L)}*/ \
+  /*mm2={c0-lflim(R_0,L),...,c7-lflim(R_7,L)}*/ \
+  "paddusb %%mm4,%%mm1\n\t" \
+  "psubusb %%mm4,%%mm2\n\t" \
+  "psubusb %%mm6,%%mm1\n\t" \
+  "paddusb %%mm6,%%mm2\n\t" \
+
+/*On entry, mm0={a0,...,a7}, mm1={b0,...,b7}, mm2={c0,...,c7}, mm3={d0,...d7}.
+  On exit, mm1={b0+lflim(R_0,L),...,b7+lflim(R_7,L)} and
+   mm2={c0-lflim(R_0,L),...,c7-lflim(R_7,L)}.
+  All other MMX registers are clobbered.*/
+#define OC_LOOP_FILTER8_MMXEXT \
+  "#OC_LOOP_FILTER8_MMXEXT\n\t" \
+  /*R_i=(a_i-3*b_i+3*c_i-d_i+4>>3) has the range [-127,128], so we compute \
+     -R_i=(-a_i+3*b_i-3*c_i+d_i+3>>3) instead.*/ \
+  /*This first part is based on the transformation \
+      f = -(3*(c-b)+a-d+4>>3) \
+        = -(3*(c+255-b)+(a+255-d)+4-1020>>3) \
+        = -(3*(c+~b)+(a+~d)-1016>>3) \
+        = 127-(3*(c+~b)+(a+~d)>>3) \
+        = 128+~(3*(c+~b)+(a+~d)>>3) (mod 256). \
+    Although pavgb(a,b) = (a+b+1>>1) (biased up), we rely heavily on the \
+     fact that ~pavgb(~a,~b) = (a+b>>1) (biased down). \
+    Using this, the last expression above can be computed in 8 bits of working \
+     precision via: \
+      u = ~pavgb(~b,c); \
+      v = pavgb(b,~c); \
+      This mask is 0 or 0xFF, and controls whether t is biased up or down: \
+      m = u-v; \
+      t = m^pavgb(m^~a,m^d); \
+      f = 128+pavgb(pavgb(t,u),v); \
+    This required some careful analysis to ensure that carries are propagated \
+     correctly in all cases, but has been checked exhaustively.*/ \
+  /*input (a, b, c, d, ., ., ., .)*/ \
+  /*ff=0xFF; \
+    u=b; \
+    v=c; \
+    ll=255-2*L;*/ \
+  "pcmpeqb %%mm7,%%mm7\n\t" \
+  "movq %%mm1,%%mm4\n\t" \
+  "movq %%mm2,%%mm5\n\t" \
+  "movq (%[ll]),%%mm6\n\t" \
+  /*allocated u, v, ll, ff: (a, b, c, d, u, v, ll, ff)*/ \
+  /*u^=ff; \
+    v^=ff;*/ \
+  "pxor %%mm7,%%mm4\n\t" \
+  "pxor %%mm7,%%mm5\n\t" \
+  /*allocated ll: (a, b, c, d, u, v, ll, ff)*/ \
+  /*u=pavgb(u,c); \
+    v=pavgb(v,b);*/ \
+  "pavgb %%mm2,%%mm4\n\t" \
+  "pavgb %%mm1,%%mm5\n\t" \
+  /*u^=ff; \
+    a^=ff;*/ \
+  "pxor %%mm7,%%mm4\n\t" \
+  "pxor %%mm7,%%mm0\n\t" \
+  /*m=u-v;*/ \
+  "psubb %%mm5,%%mm4\n\t" \
+  /*freed u, allocated m: (a, b, c, d, m, v, ll, ff)*/ \
+  /*a^=m; \
+    d^=m;*/ \
+  "pxor %%mm4,%%mm0\n\t" \
+  "pxor %%mm4,%%mm3\n\t" \
+  /*t=pavgb(a,d);*/ \
+  "pavgb %%mm3,%%mm0\n\t" \
+  "psllw $7,%%mm7\n\t" \
+  /*freed a, d, ff, allocated t, of: (t, b, c, ., m, v, ll, of)*/ \
+  /*t^=m; \
+    u=m+v;*/ \
+  "pxor %%mm4,%%mm0\n\t" \
+  "paddb %%mm5,%%mm4\n\t" \
+  /*freed t, m, allocated f, u: (f, b, c, ., u, v, ll, of)*/ \
+  /*f=pavgb(f,u); \
+    of=128;*/ \
+  "pavgb %%mm4,%%mm0\n\t" \
+  "packsswb %%mm7,%%mm7\n\t" \
+  /*freed u, ff, allocated ll: (f, b, c, ., ll, v, ll, of)*/ \
+  /*f=pavgb(f,v);*/ \
+  "pavgb %%mm5,%%mm0\n\t" \
+  "movq %%mm7,%%mm3\n\t" \
+  "movq %%mm6,%%mm4\n\t" \
+  /*freed v, allocated of: (f, b, c, of, ll, ., ll, of)*/ \
+  /*Now compute lflim of R_i=-(128+mm0) cf. Section 7.10 of the sepc.*/ \
+  /*There's no unsigned byte+signed byte with unsigned saturation op code, so \
+     we have to split things by sign (the other option is to work in 16 bits, \
+     but staying in 8 bits gives much better parallelism).*/ \
+  /*Instead of adding the offset of 128 in mm3, we use it to split mm0. \
+    This is the same number of instructions as computing a mask and splitting \
+     after the lflim computation, but has shorter dependency chains.*/ \
+  /*mm0=R_i<0?-R_i:0 (denoted abs(R_i<0))\
+    mm3=R_i>0?R_i:0* (denoted abs(R_i>0))*/ \
+  "psubusb %%mm0,%%mm3\n\t" \
+  "psubusb %%mm7,%%mm0\n\t" \
+  /*mm6=255-max(2*L-abs(R_i<0),0) \
+    mm4=255-max(2*L-abs(R_i>0),0)*/ \
+  "paddusb %%mm3,%%mm4\n\t" \
+  "paddusb %%mm0,%%mm6\n\t" \
+  /*mm0=min(abs(R_i<0),max(2*L-abs(R_i<0),0)) \
+    mm3=min(abs(R_i>0),max(2*L-abs(R_i>0),0))*/ \
+  "paddusb %%mm4,%%mm3\n\t" \
+  "paddusb %%mm6,%%mm0\n\t" \
+  "psubusb %%mm4,%%mm3\n\t" \
+  "psubusb %%mm6,%%mm0\n\t" \
+  /*mm1={b0+lflim(R_0,L),...,b7+lflim(R_7,L)}*/ \
+  /*mm2={c0-lflim(R_0,L),...,c7-lflim(R_7,L)}*/ \
+  "paddusb %%mm3,%%mm1\n\t" \
+  "psubusb %%mm3,%%mm2\n\t" \
+  "psubusb %%mm0,%%mm1\n\t" \
+  "paddusb %%mm0,%%mm2\n\t" \
+
+#define OC_LOOP_FILTER_V(_filter,_pix,_ystride,_ll) \
+  do{ \
+    ptrdiff_t ystride3__; \
+    __asm__ __volatile__( \
+      /*mm0={a0,...,a7}*/ \
+      "movq (%[pix]),%%mm0\n\t" \
+      /*ystride3=_ystride*3*/ \
+      "lea (%[ystride],%[ystride],2),%[ystride3]\n\t" \
+      /*mm3={d0,...,d7}*/ \
+      "movq (%[pix],%[ystride3]),%%mm3\n\t" \
+      /*mm1={b0,...,b7}*/ \
+      "movq (%[pix],%[ystride]),%%mm1\n\t" \
+      /*mm2={c0,...,c7}*/ \
+      "movq (%[pix],%[ystride],2),%%mm2\n\t" \
+      _filter \
+      /*Write it back out.*/ \
+      "movq %%mm1,(%[pix],%[ystride])\n\t" \
+      "movq %%mm2,(%[pix],%[ystride],2)\n\t" \
+      :[ystride3]"=&r"(ystride3__) \
+      :[pix]"r"(_pix-_ystride*2),[ystride]"r"((ptrdiff_t)(_ystride)), \
+       [ll]"r"(_ll) \
+      :"memory" \
+    ); \
+  } \
+  while(0)
+
+#define OC_LOOP_FILTER_H(_filter,_pix,_ystride,_ll) \
+  do{ \
+    unsigned char *pix__; \
+    ptrdiff_t      ystride3__; \
+    ptrdiff_t      d__; \
+    pix__=(_pix)-2; \
+    __asm__ __volatile__( \
+      /*x x x x d0 c0 b0 a0*/ \
+      "movd (%[pix]),%%mm0\n\t" \
+      /*x x x x d1 c1 b1 a1*/ \
+      "movd (%[pix],%[ystride]),%%mm1\n\t" \
+      /*ystride3=_ystride*3*/ \
+      "lea (%[ystride],%[ystride],2),%[ystride3]\n\t" \
+      /*x x x x d2 c2 b2 a2*/ \
+      "movd (%[pix],%[ystride],2),%%mm2\n\t" \
+      /*x x x x d3 c3 b3 a3*/ \
+      "lea (%[pix],%[ystride],4),%[d]\n\t" \
+      "movd (%[pix],%[ystride3]),%%mm3\n\t" \
+      /*x x x x d4 c4 b4 a4*/ \
+      "movd (%[d]),%%mm4\n\t" \
+      /*x x x x d5 c5 b5 a5*/ \
+      "movd (%[d],%[ystride]),%%mm5\n\t" \
+      /*x x x x d6 c6 b6 a6*/ \
+      "movd (%[d],%[ystride],2),%%mm6\n\t" \
+      /*x x x x d7 c7 b7 a7*/ \
+      "movd (%[d],%[ystride3]),%%mm7\n\t" \
+      /*mm0=d1 d0 c1 c0 b1 b0 a1 a0*/ \
+      "punpcklbw %%mm1,%%mm0\n\t" \
+      /*mm2=d3 d2 c3 c2 b3 b2 a3 a2*/ \
+      "punpcklbw %%mm3,%%mm2\n\t" \
+      /*mm3=d1 d0 c1 c0 b1 b0 a1 a0*/ \
+      "movq %%mm0,%%mm3\n\t" \
+      /*mm0=b3 b2 b1 b0 a3 a2 a1 a0*/ \
+      "punpcklwd %%mm2,%%mm0\n\t" \
+      /*mm3=d3 d2 d1 d0 c3 c2 c1 c0*/ \
+      "punpckhwd %%mm2,%%mm3\n\t" \
+      /*mm1=b3 b2 b1 b0 a3 a2 a1 a0*/ \
+      "movq %%mm0,%%mm1\n\t" \
+      /*mm4=d5 d4 c5 c4 b5 b4 a5 a4*/ \
+      "punpcklbw %%mm5,%%mm4\n\t" \
+      /*mm6=d7 d6 c7 c6 b7 b6 a7 a6*/ \
+      "punpcklbw %%mm7,%%mm6\n\t" \
+      /*mm5=d5 d4 c5 c4 b5 b4 a5 a4*/ \
+      "movq %%mm4,%%mm5\n\t" \
+      /*mm4=b7 b6 b5 b4 a7 a6 a5 a4*/ \
+      "punpcklwd %%mm6,%%mm4\n\t" \
+      /*mm5=d7 d6 d5 d4 c7 c6 c5 c4*/ \
+      "punpckhwd %%mm6,%%mm5\n\t" \
+      /*mm2=d3 d2 d1 d0 c3 c2 c1 c0*/ \
+      "movq %%mm3,%%mm2\n\t" \
+      /*mm0=a7 a6 a5 a4 a3 a2 a1 a0*/ \
+      "punpckldq %%mm4,%%mm0\n\t" \
+      /*mm1=b7 b6 b5 b4 b3 b2 b1 b0*/ \
+      "punpckhdq %%mm4,%%mm1\n\t" \
+      /*mm2=c7 c6 c5 c4 c3 c2 c1 c0*/ \
+      "punpckldq %%mm5,%%mm2\n\t" \
+      /*mm3=d7 d6 d5 d4 d3 d2 d1 d0*/ \
+      "punpckhdq %%mm5,%%mm3\n\t" \
+      _filter \
+      /*mm2={b0+R_0'',...,b7+R_7''}*/ \
+      "movq %%mm1,%%mm0\n\t" \
+      /*mm1={b0+R_0'',c0-R_0'',...,b3+R_3'',c3-R_3''}*/ \
+      "punpcklbw %%mm2,%%mm1\n\t" \
+      /*mm2={b4+R_4'',c4-R_4'',...,b7+R_7'',c7-R_7''}*/ \
+      "punpckhbw %%mm2,%%mm0\n\t" \
+      /*[d]=c1 b1 c0 b0*/ \
+      "movd %%mm1,%[d]\n\t" \
+      "movw %w[d],1(%[pix])\n\t" \
+      "psrlq $32,%%mm1\n\t" \
+      "shr $16,%[d]\n\t" \
+      "movw %w[d],1(%[pix],%[ystride])\n\t" \
+      /*[d]=c3 b3 c2 b2*/ \
+      "movd %%mm1,%[d]\n\t" \
+      "movw %w[d],1(%[pix],%[ystride],2)\n\t" \
+      "shr $16,%[d]\n\t" \
+      "movw %w[d],1(%[pix],%[ystride3])\n\t" \
+      "lea (%[pix],%[ystride],4),%[pix]\n\t" \
+      /*[d]=c5 b5 c4 b4*/ \
+      "movd %%mm0,%[d]\n\t" \
+      "movw %w[d],1(%[pix])\n\t" \
+      "psrlq $32,%%mm0\n\t" \
+      "shr $16,%[d]\n\t" \
+      "movw %w[d],1(%[pix],%[ystride])\n\t" \
+      /*[d]=c7 b7 c6 b6*/ \
+      "movd %%mm0,%[d]\n\t" \
+      "movw %w[d],1(%[pix],%[ystride],2)\n\t" \
+      "shr $16,%[d]\n\t" \
+      "movw %w[d],1(%[pix],%[ystride3])\n\t" \
+      :[pix]"+r"(pix__),[ystride3]"=&r"(ystride3__),[d]"=&r"(d__) \
+      :[ystride]"r"((ptrdiff_t)(_ystride)),[ll]"r"(_ll) \
+      :"memory" \
+    ); \
+  } \
+  while(0)
+
+# endif
+#endif

lib/theoratest/lib/theora/x86/mmxstate.c

@@ -0,0 +1,226 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+/*MMX acceleration of complete fragment reconstruction algorithm.
+  Originally written by Rudolf Marek.*/
+#include <string.h>
+#include "x86int.h"
+#include "mmxloop.h"
+
+#if defined(OC_X86_ASM)
+
+void oc_state_frag_recon_mmx(const oc_theora_state *_state,ptrdiff_t _fragi,
+ int _pli,ogg_int16_t _dct_coeffs[128],int _last_zzi,ogg_uint16_t _dc_quant){
+  unsigned char *dst;
+  ptrdiff_t      frag_buf_off;
+  int            ystride;
+  int            refi;
+  /*Apply the inverse transform.*/
+  /*Special case only having a DC component.*/
+  if(_last_zzi<2){
+    /*Note that this value must be unsigned, to keep the __asm__ block from
+       sign-extending it when it puts it in a register.*/
+    ogg_uint16_t p;
+    int          i;
+    /*We round this dequant product (and not any of the others) because there's
+       no iDCT rounding.*/
+    p=(ogg_int16_t)(_dct_coeffs[0]*(ogg_int32_t)_dc_quant+15>>5);
+    /*Fill _dct_coeffs with p.*/
+    __asm__ __volatile__(
+      /*mm0=0000 0000 0000 AAAA*/
+      "movd %[p],%%mm0\n\t"
+      /*mm0=0000 0000 AAAA AAAA*/
+      "punpcklwd %%mm0,%%mm0\n\t"
+      /*mm0=AAAA AAAA AAAA AAAA*/
+      "punpckldq %%mm0,%%mm0\n\t"
+      :
+      :[p]"r"((unsigned)p)
+    );
+    for(i=0;i<4;i++){
+      __asm__ __volatile__(
+        "movq %%mm0,"OC_MEM_OFFS(0x00,y)"\n\t"
+        "movq %%mm0,"OC_MEM_OFFS(0x08,y)"\n\t"
+        "movq %%mm0,"OC_MEM_OFFS(0x10,y)"\n\t"
+        "movq %%mm0,"OC_MEM_OFFS(0x18,y)"\n\t"
+        :[y]"=m"OC_ARRAY_OPERAND(ogg_int16_t,_dct_coeffs+64+16*i,16)
+      );
+    }
+  }
+  else{
+    /*Dequantize the DC coefficient.*/
+    _dct_coeffs[0]=(ogg_int16_t)(_dct_coeffs[0]*(int)_dc_quant);
+    oc_idct8x8(_state,_dct_coeffs+64,_dct_coeffs,_last_zzi);
+  }
+  /*Fill in the target buffer.*/
+  frag_buf_off=_state->frag_buf_offs[_fragi];
+  refi=_state->frags[_fragi].refi;
+  ystride=_state->ref_ystride[_pli];
+  dst=_state->ref_frame_data[OC_FRAME_SELF]+frag_buf_off;
+  if(refi==OC_FRAME_SELF)oc_frag_recon_intra_mmx(dst,ystride,_dct_coeffs+64);
+  else{
+    const unsigned char *ref;
+    int                  mvoffsets[2];
+    ref=_state->ref_frame_data[refi]+frag_buf_off;
+    if(oc_state_get_mv_offsets(_state,mvoffsets,_pli,
+     _state->frag_mvs[_fragi])>1){
+      oc_frag_recon_inter2_mmx(dst,ref+mvoffsets[0],ref+mvoffsets[1],ystride,
+       _dct_coeffs+64);
+    }
+    else oc_frag_recon_inter_mmx(dst,ref+mvoffsets[0],ystride,_dct_coeffs+64);
+  }
+}
+
+/*We copy these entire function to inline the actual MMX routines so that we
+   use only a single indirect call.*/
+
+void oc_loop_filter_init_mmx(signed char _bv[256],int _flimit){
+  memset(_bv,_flimit,8);
+}
+
+/*Apply the loop filter to a given set of fragment rows in the given plane.
+  The filter may be run on the bottom edge, affecting pixels in the next row of
+   fragments, so this row also needs to be available.
+  _bv:        The bounding values array.
+  _refi:      The index of the frame buffer to filter.
+  _pli:       The color plane to filter.
+  _fragy0:    The Y coordinate of the first fragment row to filter.
+  _fragy_end: The Y coordinate of the fragment row to stop filtering at.*/
+void oc_state_loop_filter_frag_rows_mmx(const oc_theora_state *_state,
+ signed char _bv[256],int _refi,int _pli,int _fragy0,int _fragy_end){
+  OC_ALIGN8(unsigned char   ll[8]);
+  const oc_fragment_plane *fplane;
+  const oc_fragment       *frags;
+  const ptrdiff_t         *frag_buf_offs;
+  unsigned char           *ref_frame_data;
+  ptrdiff_t                fragi_top;
+  ptrdiff_t                fragi_bot;
+  ptrdiff_t                fragi0;
+  ptrdiff_t                fragi0_end;
+  int                      ystride;
+  int                      nhfrags;
+  memset(ll,_state->loop_filter_limits[_state->qis[0]],sizeof(ll));
+  fplane=_state->fplanes+_pli;
+  nhfrags=fplane->nhfrags;
+  fragi_top=fplane->froffset;
+  fragi_bot=fragi_top+fplane->nfrags;
+  fragi0=fragi_top+_fragy0*(ptrdiff_t)nhfrags;
+  fragi0_end=fragi0+(_fragy_end-_fragy0)*(ptrdiff_t)nhfrags;
+  ystride=_state->ref_ystride[_pli];
+  frags=_state->frags;
+  frag_buf_offs=_state->frag_buf_offs;
+  ref_frame_data=_state->ref_frame_data[_refi];
+  /*The following loops are constructed somewhat non-intuitively on purpose.
+    The main idea is: if a block boundary has at least one coded fragment on
+     it, the filter is applied to it.
+    However, the order that the filters are applied in matters, and VP3 chose
+     the somewhat strange ordering used below.*/
+  while(fragi0<fragi0_end){
+    ptrdiff_t fragi;
+    ptrdiff_t fragi_end;
+    fragi=fragi0;
+    fragi_end=fragi+nhfrags;
+    while(fragi<fragi_end){
+      if(frags[fragi].coded){
+        unsigned char *ref;
+        ref=ref_frame_data+frag_buf_offs[fragi];
+        if(fragi>fragi0){
+          OC_LOOP_FILTER_H(OC_LOOP_FILTER8_MMX,ref,ystride,ll);
+        }
+        if(fragi0>fragi_top){
+          OC_LOOP_FILTER_V(OC_LOOP_FILTER8_MMX,ref,ystride,ll);
+        }
+        if(fragi+1<fragi_end&&!frags[fragi+1].coded){
+          OC_LOOP_FILTER_H(OC_LOOP_FILTER8_MMX,ref+8,ystride,ll);
+        }
+        if(fragi+nhfrags<fragi_bot&&!frags[fragi+nhfrags].coded){
+          OC_LOOP_FILTER_V(OC_LOOP_FILTER8_MMX,ref+(ystride<<3),ystride,ll);
+        }
+      }
+      fragi++;
+    }
+    fragi0+=nhfrags;
+  }
+}
+
+void oc_loop_filter_init_mmxext(signed char _bv[256],int _flimit){
+  memset(_bv,~(_flimit<<1),8);
+}
+
+/*Apply the loop filter to a given set of fragment rows in the given plane.
+  The filter may be run on the bottom edge, affecting pixels in the next row of
+   fragments, so this row also needs to be available.
+  _bv:        The bounding values array.
+  _refi:      The index of the frame buffer to filter.
+  _pli:       The color plane to filter.
+  _fragy0:    The Y coordinate of the first fragment row to filter.
+  _fragy_end: The Y coordinate of the fragment row to stop filtering at.*/
+void oc_state_loop_filter_frag_rows_mmxext(const oc_theora_state *_state,
+ signed char _bv[256],int _refi,int _pli,int _fragy0,int _fragy_end){
+  const oc_fragment_plane *fplane;
+  const oc_fragment       *frags;
+  const ptrdiff_t         *frag_buf_offs;
+  unsigned char           *ref_frame_data;
+  ptrdiff_t                fragi_top;
+  ptrdiff_t                fragi_bot;
+  ptrdiff_t                fragi0;
+  ptrdiff_t                fragi0_end;
+  int                      ystride;
+  int                      nhfrags;
+  fplane=_state->fplanes+_pli;
+  nhfrags=fplane->nhfrags;
+  fragi_top=fplane->froffset;
+  fragi_bot=fragi_top+fplane->nfrags;
+  fragi0=fragi_top+_fragy0*(ptrdiff_t)nhfrags;
+  fragi0_end=fragi_top+_fragy_end*(ptrdiff_t)nhfrags;
+  ystride=_state->ref_ystride[_pli];
+  frags=_state->frags;
+  frag_buf_offs=_state->frag_buf_offs;
+  ref_frame_data=_state->ref_frame_data[_refi];
+  /*The following loops are constructed somewhat non-intuitively on purpose.
+    The main idea is: if a block boundary has at least one coded fragment on
+     it, the filter is applied to it.
+    However, the order that the filters are applied in matters, and VP3 chose
+     the somewhat strange ordering used below.*/
+  while(fragi0<fragi0_end){
+    ptrdiff_t fragi;
+    ptrdiff_t fragi_end;
+    fragi=fragi0;
+    fragi_end=fragi+nhfrags;
+    while(fragi<fragi_end){
+      if(frags[fragi].coded){
+        unsigned char *ref;
+        ref=ref_frame_data+frag_buf_offs[fragi];
+        if(fragi>fragi0){
+          OC_LOOP_FILTER_H(OC_LOOP_FILTER8_MMXEXT,ref,ystride,_bv);
+        }
+        if(fragi0>fragi_top){
+          OC_LOOP_FILTER_V(OC_LOOP_FILTER8_MMXEXT,ref,ystride,_bv);
+        }
+        if(fragi+1<fragi_end&&!frags[fragi+1].coded){
+          OC_LOOP_FILTER_H(OC_LOOP_FILTER8_MMXEXT,ref+8,ystride,_bv);
+        }
+        if(fragi+nhfrags<fragi_bot&&!frags[fragi+nhfrags].coded){
+          OC_LOOP_FILTER_V(OC_LOOP_FILTER8_MMXEXT,ref+(ystride<<3),ystride,_bv);
+        }
+      }
+      fragi++;
+    }
+    fragi0+=nhfrags;
+  }
+}
+
+#endif

lib/theoratest/lib/theora/x86/sse2idct.c

@@ -0,0 +1,456 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id: mmxidct.c 16503 2009-08-22 18:14:02Z giles $
+
+ ********************************************************************/
+
+/*SSE2 acceleration of Theora's iDCT.*/
+#include "x86int.h"
+#include "sse2trans.h"
+#include "../dct.h"
+
+#if defined(OC_X86_ASM)
+
+/*A table of constants used by the MMX routines.*/
+const unsigned short __attribute__((aligned(16),used)) OC_IDCT_CONSTS[64]={
+        8,      8,      8,      8,      8,      8,      8,      8,
+  OC_C1S7,OC_C1S7,OC_C1S7,OC_C1S7,OC_C1S7,OC_C1S7,OC_C1S7,OC_C1S7,
+  OC_C2S6,OC_C2S6,OC_C2S6,OC_C2S6,OC_C2S6,OC_C2S6,OC_C2S6,OC_C2S6,
+  OC_C3S5,OC_C3S5,OC_C3S5,OC_C3S5,OC_C3S5,OC_C3S5,OC_C3S5,OC_C3S5,
+  OC_C4S4,OC_C4S4,OC_C4S4,OC_C4S4,OC_C4S4,OC_C4S4,OC_C4S4,OC_C4S4,
+  OC_C5S3,OC_C5S3,OC_C5S3,OC_C5S3,OC_C5S3,OC_C5S3,OC_C5S3,OC_C5S3,
+  OC_C6S2,OC_C6S2,OC_C6S2,OC_C6S2,OC_C6S2,OC_C6S2,OC_C6S2,OC_C6S2,
+  OC_C7S1,OC_C7S1,OC_C7S1,OC_C7S1,OC_C7S1,OC_C7S1,OC_C7S1,OC_C7S1
+};
+
+
+/*Performs the first three stages of the iDCT.
+  xmm2, xmm6, xmm3, and xmm5 must contain the corresponding rows of the input
+   (accessed in that order).
+  The remaining rows must be in _x at their corresponding locations.
+  On output, xmm7 down to xmm4 contain rows 0 through 3, and xmm0 up to xmm3
+   contain rows 4 through 7.*/
+#define OC_IDCT_8x8_ABC(_x) \
+  "#OC_IDCT_8x8_ABC\n\t" \
+  /*Stage 1:*/ \
+  /*2-3 rotation by 6pi/16. \
+    xmm4=xmm7=C6, xmm0=xmm1=C2, xmm2=X2, xmm6=X6.*/ \
+  "movdqa "OC_MEM_OFFS(0x20,c)",%%xmm1\n\t" \
+  "movdqa "OC_MEM_OFFS(0x60,c)",%%xmm4\n\t" \
+  "movdqa %%xmm1,%%xmm0\n\t" \
+  "pmulhw %%xmm2,%%xmm1\n\t" \
+  "movdqa %%xmm4,%%xmm7\n\t" \
+  "pmulhw %%xmm6,%%xmm0\n\t" \
+  "pmulhw %%xmm2,%%xmm7\n\t" \
+  "pmulhw %%xmm6,%%xmm4\n\t" \
+  "paddw %%xmm6,%%xmm0\n\t" \
+  "movdqa "OC_MEM_OFFS(0x30,c)",%%xmm6\n\t" \
+  "paddw %%xmm1,%%xmm2\n\t" \
+  "psubw %%xmm0,%%xmm7\n\t" \
+  "movdqa %%xmm7,"OC_MEM_OFFS(0x00,buf)"\n\t" \
+  "paddw %%xmm4,%%xmm2\n\t" \
+  "movdqa "OC_MEM_OFFS(0x50,c)",%%xmm4\n\t" \
+  "movdqa %%xmm2,"OC_MEM_OFFS(0x10,buf)"\n\t" \
+  /*5-6 rotation by 3pi/16. \
+    xmm4=xmm2=C5, xmm1=xmm6=C3, xmm3=X3, xmm5=X5.*/ \
+  "movdqa %%xmm4,%%xmm2\n\t" \
+  "movdqa %%xmm6,%%xmm1\n\t" \
+  "pmulhw %%xmm3,%%xmm4\n\t" \
+  "pmulhw %%xmm5,%%xmm1\n\t" \
+  "pmulhw %%xmm3,%%xmm6\n\t" \
+  "pmulhw %%xmm5,%%xmm2\n\t" \
+  "paddw %%xmm3,%%xmm4\n\t" \
+  "paddw %%xmm5,%%xmm3\n\t" \
+  "paddw %%xmm6,%%xmm3\n\t" \
+  "movdqa "OC_MEM_OFFS(0x70,_x)",%%xmm6\n\t" \
+  "paddw %%xmm5,%%xmm1\n\t" \
+  "movdqa "OC_MEM_OFFS(0x10,_x)",%%xmm5\n\t" \
+  "paddw %%xmm3,%%xmm2\n\t" \
+  "movdqa "OC_MEM_OFFS(0x70,c)",%%xmm3\n\t" \
+  "psubw %%xmm4,%%xmm1\n\t" \
+  "movdqa "OC_MEM_OFFS(0x10,c)",%%xmm4\n\t" \
+  /*4-7 rotation by 7pi/16. \
+    xmm4=xmm7=C1, xmm3=xmm0=C7, xmm5=X1, xmm6=X7.*/ \
+  "movdqa %%xmm3,%%xmm0\n\t" \
+  "movdqa %%xmm4,%%xmm7\n\t" \
+  "pmulhw %%xmm5,%%xmm3\n\t" \
+  "pmulhw %%xmm5,%%xmm7\n\t" \
+  "pmulhw %%xmm6,%%xmm4\n\t" \
+  "pmulhw %%xmm6,%%xmm0\n\t" \
+  "paddw %%xmm6,%%xmm4\n\t" \
+  "movdqa "OC_MEM_OFFS(0x40,_x)",%%xmm6\n\t" \
+  "paddw %%xmm5,%%xmm7\n\t" \
+  "psubw %%xmm4,%%xmm3\n\t" \
+  "movdqa "OC_MEM_OFFS(0x40,c)",%%xmm4\n\t" \
+  "paddw %%xmm7,%%xmm0\n\t" \
+  "movdqa "OC_MEM_OFFS(0x00,_x)",%%xmm7\n\t" \
+  /*0-1 butterfly. \
+    xmm4=xmm5=C4, xmm7=X0, xmm6=X4.*/ \
+  "paddw %%xmm7,%%xmm6\n\t" \
+  "movdqa %%xmm4,%%xmm5\n\t" \
+  "pmulhw %%xmm6,%%xmm4\n\t" \
+  "paddw %%xmm7,%%xmm7\n\t" \
+  "psubw %%xmm6,%%xmm7\n\t" \
+  "paddw %%xmm6,%%xmm4\n\t" \
+  /*Stage 2:*/ \
+  /*4-5 butterfly: xmm3=t[4], xmm1=t[5] \
+    7-6 butterfly: xmm2=t[6], xmm0=t[7]*/ \
+  "movdqa %%xmm3,%%xmm6\n\t" \
+  "paddw %%xmm1,%%xmm3\n\t" \
+  "psubw %%xmm1,%%xmm6\n\t" \
+  "movdqa %%xmm5,%%xmm1\n\t" \
+  "pmulhw %%xmm7,%%xmm5\n\t" \
+  "paddw %%xmm7,%%xmm5\n\t" \
+  "movdqa %%xmm0,%%xmm7\n\t" \
+  "paddw %%xmm2,%%xmm0\n\t" \
+  "psubw %%xmm2,%%xmm7\n\t" \
+  "movdqa %%xmm1,%%xmm2\n\t" \
+  "pmulhw %%xmm6,%%xmm1\n\t" \
+  "pmulhw %%xmm7,%%xmm2\n\t" \
+  "paddw %%xmm6,%%xmm1\n\t" \
+  "movdqa "OC_MEM_OFFS(0x00,buf)",%%xmm6\n\t" \
+  "paddw %%xmm7,%%xmm2\n\t" \
+  "movdqa "OC_MEM_OFFS(0x10,buf)",%%xmm7\n\t" \
+  /*Stage 3: \
+    6-5 butterfly: xmm1=t[5], xmm2=t[6] -> xmm1=t[6]+t[5], xmm2=t[6]-t[5] \
+    0-3 butterfly: xmm4=t[0], xmm7=t[3] -> xmm7=t[0]+t[3], xmm4=t[0]-t[3] \
+    1-2 butterfly: xmm5=t[1], xmm6=t[2] -> xmm6=t[1]+t[2], xmm5=t[1]-t[2]*/ \
+  "paddw %%xmm2,%%xmm1\n\t" \
+  "paddw %%xmm5,%%xmm6\n\t" \
+  "paddw %%xmm4,%%xmm7\n\t" \
+  "paddw %%xmm2,%%xmm2\n\t" \
+  "paddw %%xmm4,%%xmm4\n\t" \
+  "paddw %%xmm5,%%xmm5\n\t" \
+  "psubw %%xmm1,%%xmm2\n\t" \
+  "psubw %%xmm7,%%xmm4\n\t" \
+  "psubw %%xmm6,%%xmm5\n\t" \
+
+/*Performs the last stage of the iDCT.
+  On input, xmm7 down to xmm4 contain rows 0 through 3, and xmm0 up to xmm3
+   contain rows 4 through 7.
+  On output, xmm0 through xmm7 contain the corresponding rows.*/
+#define OC_IDCT_8x8_D \
+  "#OC_IDCT_8x8_D\n\t" \
+  /*Stage 4: \
+    0-7 butterfly: xmm7=t[0], xmm0=t[7] -> xmm0=t[0]+t[7], xmm7=t[0]-t[7] \
+    1-6 butterfly: xmm6=t[1], xmm1=t[6] -> xmm1=t[1]+t[6], xmm6=t[1]-t[6] \
+    2-5 butterfly: xmm5=t[2], xmm2=t[5] -> xmm2=t[2]+t[5], xmm5=t[2]-t[5] \
+    3-4 butterfly: xmm4=t[3], xmm3=t[4] -> xmm3=t[3]+t[4], xmm4=t[3]-t[4]*/ \
+  "psubw %%xmm0,%%xmm7\n\t" \
+  "psubw %%xmm1,%%xmm6\n\t" \
+  "psubw %%xmm2,%%xmm5\n\t" \
+  "psubw %%xmm3,%%xmm4\n\t" \
+  "paddw %%xmm0,%%xmm0\n\t" \
+  "paddw %%xmm1,%%xmm1\n\t" \
+  "paddw %%xmm2,%%xmm2\n\t" \
+  "paddw %%xmm3,%%xmm3\n\t" \
+  "paddw %%xmm7,%%xmm0\n\t" \
+  "paddw %%xmm6,%%xmm1\n\t" \
+  "paddw %%xmm5,%%xmm2\n\t" \
+  "paddw %%xmm4,%%xmm3\n\t" \
+
+/*Performs the last stage of the iDCT.
+  On input, xmm7 down to xmm4 contain rows 0 through 3, and xmm0 up to xmm3
+   contain rows 4 through 7.
+  On output, xmm0 through xmm7 contain the corresponding rows.*/
+#define OC_IDCT_8x8_D_STORE \
+  "#OC_IDCT_8x8_D_STORE\n\t" \
+  /*Stage 4: \
+    0-7 butterfly: xmm7=t[0], xmm0=t[7] -> xmm0=t[0]+t[7], xmm7=t[0]-t[7] \
+    1-6 butterfly: xmm6=t[1], xmm1=t[6] -> xmm1=t[1]+t[6], xmm6=t[1]-t[6] \
+    2-5 butterfly: xmm5=t[2], xmm2=t[5] -> xmm2=t[2]+t[5], xmm5=t[2]-t[5] \
+    3-4 butterfly: xmm4=t[3], xmm3=t[4] -> xmm3=t[3]+t[4], xmm4=t[3]-t[4]*/ \
+  "psubw %%xmm3,%%xmm4\n\t" \
+  "movdqa %%xmm4,"OC_MEM_OFFS(0x40,y)"\n\t" \
+  "movdqa "OC_MEM_OFFS(0x00,c)",%%xmm4\n\t" \
+  "psubw %%xmm0,%%xmm7\n\t" \
+  "psubw %%xmm1,%%xmm6\n\t" \
+  "psubw %%xmm2,%%xmm5\n\t" \
+  "paddw %%xmm4,%%xmm7\n\t" \
+  "paddw %%xmm4,%%xmm6\n\t" \
+  "paddw %%xmm4,%%xmm5\n\t" \
+  "paddw "OC_MEM_OFFS(0x40,y)",%%xmm4\n\t" \
+  "paddw %%xmm0,%%xmm0\n\t" \
+  "paddw %%xmm1,%%xmm1\n\t" \
+  "paddw %%xmm2,%%xmm2\n\t" \
+  "paddw %%xmm3,%%xmm3\n\t" \
+  "paddw %%xmm7,%%xmm0\n\t" \
+  "paddw %%xmm6,%%xmm1\n\t" \
+  "psraw $4,%%xmm0\n\t" \
+  "paddw %%xmm5,%%xmm2\n\t" \
+  "movdqa %%xmm0,"OC_MEM_OFFS(0x00,y)"\n\t" \
+  "psraw $4,%%xmm1\n\t" \
+  "paddw %%xmm4,%%xmm3\n\t" \
+  "movdqa %%xmm1,"OC_MEM_OFFS(0x10,y)"\n\t" \
+  "psraw $4,%%xmm2\n\t" \
+  "movdqa %%xmm2,"OC_MEM_OFFS(0x20,y)"\n\t" \
+  "psraw $4,%%xmm3\n\t" \
+  "movdqa %%xmm3,"OC_MEM_OFFS(0x30,y)"\n\t" \
+  "psraw $4,%%xmm4\n\t" \
+  "movdqa %%xmm4,"OC_MEM_OFFS(0x40,y)"\n\t" \
+  "psraw $4,%%xmm5\n\t" \
+  "movdqa %%xmm5,"OC_MEM_OFFS(0x50,y)"\n\t" \
+  "psraw $4,%%xmm6\n\t" \
+  "movdqa %%xmm6,"OC_MEM_OFFS(0x60,y)"\n\t" \
+  "psraw $4,%%xmm7\n\t" \
+  "movdqa %%xmm7,"OC_MEM_OFFS(0x70,y)"\n\t" \
+
+static void oc_idct8x8_slow_sse2(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  OC_ALIGN16(ogg_int16_t buf[16]);
+  int i;
+  /*This routine accepts an 8x8 matrix pre-transposed.*/
+  __asm__ __volatile__(
+    /*Load rows 2, 3, 5, and 6 for the first stage of the iDCT.*/
+    "movdqa "OC_MEM_OFFS(0x20,x)",%%xmm2\n\t"
+    "movdqa "OC_MEM_OFFS(0x60,x)",%%xmm6\n\t"
+    "movdqa "OC_MEM_OFFS(0x30,x)",%%xmm3\n\t"
+    "movdqa "OC_MEM_OFFS(0x50,x)",%%xmm5\n\t"
+    OC_IDCT_8x8_ABC(x)
+    OC_IDCT_8x8_D
+    OC_TRANSPOSE_8x8
+    /*Clear out rows 0, 1, 4, and 7 for the first stage of the iDCT.*/
+    "movdqa %%xmm7,"OC_MEM_OFFS(0x70,y)"\n\t"
+    "movdqa %%xmm4,"OC_MEM_OFFS(0x40,y)"\n\t"
+    "movdqa %%xmm1,"OC_MEM_OFFS(0x10,y)"\n\t"
+    "movdqa %%xmm0,"OC_MEM_OFFS(0x00,y)"\n\t"
+    OC_IDCT_8x8_ABC(y)
+    OC_IDCT_8x8_D_STORE
+    :[buf]"=m"(OC_ARRAY_OPERAND(ogg_int16_t,buf,16)),
+     [y]"=m"(OC_ARRAY_OPERAND(ogg_int16_t,_y,64))
+    :[x]"m"(OC_CONST_ARRAY_OPERAND(ogg_int16_t,_x,64)),
+     [c]"m"(OC_CONST_ARRAY_OPERAND(ogg_int16_t,OC_IDCT_CONSTS,128))
+  );
+  __asm__ __volatile__("pxor %%xmm0,%%xmm0\n\t"::);
+  /*Clear input data for next block (decoder only).*/
+  for(i=0;i<2;i++){
+    __asm__ __volatile__(
+      "movdqa %%xmm0,"OC_MEM_OFFS(0x00,x)"\n\t"
+      "movdqa %%xmm0,"OC_MEM_OFFS(0x10,x)"\n\t"
+      "movdqa %%xmm0,"OC_MEM_OFFS(0x20,x)"\n\t"
+      "movdqa %%xmm0,"OC_MEM_OFFS(0x30,x)"\n\t"
+      :[x]"=m"(OC_ARRAY_OPERAND(ogg_int16_t,_x+i*32,32))
+    );
+  }
+}
+
+/*For the first step of the 10-coefficient version of the 8x8 iDCT, we only
+   need to work with four columns at a time.
+  Doing this in MMX is faster on processors with a 64-bit data path.*/
+#define OC_IDCT_8x8_10_MMX \
+  "#OC_IDCT_8x8_10_MMX\n\t" \
+  /*Stage 1:*/ \
+  /*2-3 rotation by 6pi/16. \
+    mm7=C6, mm6=C2, mm2=X2, X6=0.*/ \
+  "movq "OC_MEM_OFFS(0x60,c)",%%mm7\n\t" \
+  "movq "OC_MEM_OFFS(0x20,c)",%%mm6\n\t" \
+  "pmulhw %%mm2,%%mm6\n\t" \
+  "pmulhw %%mm2,%%mm7\n\t" \
+  "movq "OC_MEM_OFFS(0x50,c)",%%mm5\n\t" \
+  "paddw %%mm6,%%mm2\n\t" \
+  "movq %%mm2,"OC_MEM_OFFS(0x10,buf)"\n\t" \
+  "movq "OC_MEM_OFFS(0x30,c)",%%mm2\n\t" \
+  "movq %%mm7,"OC_MEM_OFFS(0x00,buf)"\n\t" \
+  /*5-6 rotation by 3pi/16. \
+    mm5=C5, mm2=C3, mm3=X3, X5=0.*/ \
+  "pmulhw %%mm3,%%mm5\n\t" \
+  "pmulhw %%mm3,%%mm2\n\t" \
+  "movq "OC_MEM_OFFS(0x10,c)",%%mm7\n\t" \
+  "paddw %%mm3,%%mm5\n\t" \
+  "paddw %%mm3,%%mm2\n\t" \
+  "movq "OC_MEM_OFFS(0x70,c)",%%mm3\n\t" \
+  /*4-7 rotation by 7pi/16. \
+    mm7=C1, mm3=C7, mm1=X1, X7=0.*/ \
+  "pmulhw %%mm1,%%mm3\n\t" \
+  "pmulhw %%mm1,%%mm7\n\t" \
+  "movq "OC_MEM_OFFS(0x40,c)",%%mm4\n\t" \
+  "movq %%mm3,%%mm6\n\t" \
+  "paddw %%mm1,%%mm7\n\t" \
+  /*0-1 butterfly. \
+    mm4=C4, mm0=X0, X4=0.*/ \
+  /*Stage 2:*/ \
+  /*4-5 butterfly: mm3=t[4], mm5=t[5] \
+    7-6 butterfly: mm2=t[6], mm7=t[7]*/ \
+  "psubw %%mm5,%%mm3\n\t" \
+  "paddw %%mm5,%%mm6\n\t" \
+  "movq %%mm4,%%mm1\n\t" \
+  "pmulhw %%mm0,%%mm4\n\t" \
+  "paddw %%mm0,%%mm4\n\t" \
+  "movq %%mm7,%%mm0\n\t" \
+  "movq %%mm4,%%mm5\n\t" \
+  "paddw %%mm2,%%mm0\n\t" \
+  "psubw %%mm2,%%mm7\n\t" \
+  "movq %%mm1,%%mm2\n\t" \
+  "pmulhw %%mm6,%%mm1\n\t" \
+  "pmulhw %%mm7,%%mm2\n\t" \
+  "paddw %%mm6,%%mm1\n\t" \
+  "movq "OC_MEM_OFFS(0x00,buf)",%%mm6\n\t" \
+  "paddw %%mm7,%%mm2\n\t" \
+  "movq "OC_MEM_OFFS(0x10,buf)",%%mm7\n\t" \
+  /*Stage 3: \
+    6-5 butterfly: mm1=t[5], mm2=t[6] -> mm1=t[6]+t[5], mm2=t[6]-t[5] \
+    0-3 butterfly: mm4=t[0], mm7=t[3] -> mm7=t[0]+t[3], mm4=t[0]-t[3] \
+    1-2 butterfly: mm5=t[1], mm6=t[2] -> mm6=t[1]+t[2], mm5=t[1]-t[2]*/ \
+  "paddw %%mm2,%%mm1\n\t" \
+  "paddw %%mm5,%%mm6\n\t" \
+  "paddw %%mm4,%%mm7\n\t" \
+  "paddw %%mm2,%%mm2\n\t" \
+  "paddw %%mm4,%%mm4\n\t" \
+  "paddw %%mm5,%%mm5\n\t" \
+  "psubw %%mm1,%%mm2\n\t" \
+  "psubw %%mm7,%%mm4\n\t" \
+  "psubw %%mm6,%%mm5\n\t" \
+  /*Stage 4: \
+    0-7 butterfly: mm7=t[0], mm0=t[7] -> mm0=t[0]+t[7], mm7=t[0]-t[7] \
+    1-6 butterfly: mm6=t[1], mm1=t[6] -> mm1=t[1]+t[6], mm6=t[1]-t[6] \
+    2-5 butterfly: mm5=t[2], mm2=t[5] -> mm2=t[2]+t[5], mm5=t[2]-t[5] \
+    3-4 butterfly: mm4=t[3], mm3=t[4] -> mm3=t[3]+t[4], mm4=t[3]-t[4]*/ \
+  "psubw %%mm0,%%mm7\n\t" \
+  "psubw %%mm1,%%mm6\n\t" \
+  "psubw %%mm2,%%mm5\n\t" \
+  "psubw %%mm3,%%mm4\n\t" \
+  "paddw %%mm0,%%mm0\n\t" \
+  "paddw %%mm1,%%mm1\n\t" \
+  "paddw %%mm2,%%mm2\n\t" \
+  "paddw %%mm3,%%mm3\n\t" \
+  "paddw %%mm7,%%mm0\n\t" \
+  "paddw %%mm6,%%mm1\n\t" \
+  "paddw %%mm5,%%mm2\n\t" \
+  "paddw %%mm4,%%mm3\n\t" \
+
+#define OC_IDCT_8x8_10_ABC \
+  "#OC_IDCT_8x8_10_ABC\n\t" \
+  /*Stage 1:*/ \
+  /*2-3 rotation by 6pi/16. \
+    xmm7=C6, xmm6=C2, xmm2=X2, X6=0.*/ \
+  "movdqa "OC_MEM_OFFS(0x60,c)",%%xmm7\n\t" \
+  "movdqa "OC_MEM_OFFS(0x20,c)",%%xmm6\n\t" \
+  "pmulhw %%xmm2,%%xmm6\n\t" \
+  "pmulhw %%xmm2,%%xmm7\n\t" \
+  "movdqa "OC_MEM_OFFS(0x50,c)",%%xmm5\n\t" \
+  "paddw %%xmm6,%%xmm2\n\t" \
+  "movdqa %%xmm2,"OC_MEM_OFFS(0x10,buf)"\n\t" \
+  "movdqa "OC_MEM_OFFS(0x30,c)",%%xmm2\n\t" \
+  "movdqa %%xmm7,"OC_MEM_OFFS(0x00,buf)"\n\t" \
+  /*5-6 rotation by 3pi/16. \
+    xmm5=C5, xmm2=C3, xmm3=X3, X5=0.*/ \
+  "pmulhw %%xmm3,%%xmm5\n\t" \
+  "pmulhw %%xmm3,%%xmm2\n\t" \
+  "movdqa "OC_MEM_OFFS(0x10,c)",%%xmm7\n\t" \
+  "paddw %%xmm3,%%xmm5\n\t" \
+  "paddw %%xmm3,%%xmm2\n\t" \
+  "movdqa "OC_MEM_OFFS(0x70,c)",%%xmm3\n\t" \
+  /*4-7 rotation by 7pi/16. \
+    xmm7=C1, xmm3=C7, xmm1=X1, X7=0.*/ \
+  "pmulhw %%xmm1,%%xmm3\n\t" \
+  "pmulhw %%xmm1,%%xmm7\n\t" \
+  "movdqa "OC_MEM_OFFS(0x40,c)",%%xmm4\n\t" \
+  "movdqa %%xmm3,%%xmm6\n\t" \
+  "paddw %%xmm1,%%xmm7\n\t" \
+  /*0-1 butterfly. \
+    xmm4=C4, xmm0=X0, X4=0.*/ \
+  /*Stage 2:*/ \
+  /*4-5 butterfly: xmm3=t[4], xmm5=t[5] \
+    7-6 butterfly: xmm2=t[6], xmm7=t[7]*/ \
+  "psubw %%xmm5,%%xmm3\n\t" \
+  "paddw %%xmm5,%%xmm6\n\t" \
+  "movdqa %%xmm4,%%xmm1\n\t" \
+  "pmulhw %%xmm0,%%xmm4\n\t" \
+  "paddw %%xmm0,%%xmm4\n\t" \
+  "movdqa %%xmm7,%%xmm0\n\t" \
+  "movdqa %%xmm4,%%xmm5\n\t" \
+  "paddw %%xmm2,%%xmm0\n\t" \
+  "psubw %%xmm2,%%xmm7\n\t" \
+  "movdqa %%xmm1,%%xmm2\n\t" \
+  "pmulhw %%xmm6,%%xmm1\n\t" \
+  "pmulhw %%xmm7,%%xmm2\n\t" \
+  "paddw %%xmm6,%%xmm1\n\t" \
+  "movdqa "OC_MEM_OFFS(0x00,buf)",%%xmm6\n\t" \
+  "paddw %%xmm7,%%xmm2\n\t" \
+  "movdqa "OC_MEM_OFFS(0x10,buf)",%%xmm7\n\t" \
+  /*Stage 3: \
+    6-5 butterfly: xmm1=t[5], xmm2=t[6] -> xmm1=t[6]+t[5], xmm2=t[6]-t[5] \
+    0-3 butterfly: xmm4=t[0], xmm7=t[3] -> xmm7=t[0]+t[3], xmm4=t[0]-t[3] \
+    1-2 butterfly: xmm5=t[1], xmm6=t[2] -> xmm6=t[1]+t[2], xmm5=t[1]-t[2]*/ \
+  "paddw %%xmm2,%%xmm1\n\t" \
+  "paddw %%xmm5,%%xmm6\n\t" \
+  "paddw %%xmm4,%%xmm7\n\t" \
+  "paddw %%xmm2,%%xmm2\n\t" \
+  "paddw %%xmm4,%%xmm4\n\t" \
+  "paddw %%xmm5,%%xmm5\n\t" \
+  "psubw %%xmm1,%%xmm2\n\t" \
+  "psubw %%xmm7,%%xmm4\n\t" \
+  "psubw %%xmm6,%%xmm5\n\t" \
+
+static void oc_idct8x8_10_sse2(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  OC_ALIGN16(ogg_int16_t buf[16]);
+  /*This routine accepts an 8x8 matrix pre-transposed.*/
+  __asm__ __volatile__(
+    "movq "OC_MEM_OFFS(0x20,x)",%%mm2\n\t"
+    "movq "OC_MEM_OFFS(0x30,x)",%%mm3\n\t"
+    "movq "OC_MEM_OFFS(0x10,x)",%%mm1\n\t"
+    "movq "OC_MEM_OFFS(0x00,x)",%%mm0\n\t"
+    OC_IDCT_8x8_10_MMX
+    OC_TRANSPOSE_8x4_MMX2SSE
+    OC_IDCT_8x8_10_ABC
+    OC_IDCT_8x8_D_STORE
+    :[buf]"=m"(OC_ARRAY_OPERAND(short,buf,16)),
+     [y]"=m"(OC_ARRAY_OPERAND(ogg_int16_t,_y,64))
+    :[x]"m"OC_CONST_ARRAY_OPERAND(ogg_int16_t,_x,64),
+     [c]"m"(OC_CONST_ARRAY_OPERAND(ogg_int16_t,OC_IDCT_CONSTS,128))
+  );
+  /*Clear input data for next block (decoder only).*/
+  __asm__ __volatile__(
+    "pxor %%mm0,%%mm0\n\t"
+    "movq %%mm0,"OC_MEM_OFFS(0x00,x)"\n\t"
+    "movq %%mm0,"OC_MEM_OFFS(0x10,x)"\n\t"
+    "movq %%mm0,"OC_MEM_OFFS(0x20,x)"\n\t"
+    "movq %%mm0,"OC_MEM_OFFS(0x30,x)"\n\t"
+    :[x]"+m"(OC_ARRAY_OPERAND(ogg_int16_t,_x,28))
+  );
+}
+
+/*Performs an inverse 8x8 Type-II DCT transform.
+  The input is assumed to be scaled by a factor of 4 relative to orthonormal
+   version of the transform.*/
+void oc_idct8x8_sse2(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi){
+  /*_last_zzi is subtly different from an actual count of the number of
+     coefficients we decoded for this block.
+    It contains the value of zzi BEFORE the final token in the block was
+     decoded.
+    In most cases this is an EOB token (the continuation of an EOB run from a
+     previous block counts), and so this is the same as the coefficient count.
+    However, in the case that the last token was NOT an EOB token, but filled
+     the block up with exactly 64 coefficients, _last_zzi will be less than 64.
+    Provided the last token was not a pure zero run, the minimum value it can
+     be is 46, and so that doesn't affect any of the cases in this routine.
+    However, if the last token WAS a pure zero run of length 63, then _last_zzi
+     will be 1 while the number of coefficients decoded is 64.
+    Thus, we will trigger the following special case, where the real
+     coefficient count would not.
+    Note also that a zero run of length 64 will give _last_zzi a value of 0,
+     but we still process the DC coefficient, which might have a non-zero value
+     due to DC prediction.
+    Although convoluted, this is arguably the correct behavior: it allows us to
+     use a smaller transform when the block ends with a long zero run instead
+     of a normal EOB token.
+    It could be smarter... multiple separate zero runs at the end of a block
+     will fool it, but an encoder that generates these really deserves what it
+     gets.
+    Needless to say we inherited this approach from VP3.*/
+  /*Then perform the iDCT.*/
+  if(_last_zzi<=10)oc_idct8x8_10_sse2(_y,_x);
+  else oc_idct8x8_slow_sse2(_y,_x);
+}
+
+#endif

lib/theoratest/lib/theora/x86/sse2trans.h

@@ -0,0 +1,242 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id: sse2trans.h 15675 2009-02-06 09:43:27Z tterribe $
+
+ ********************************************************************/
+
+#if !defined(_x86_sse2trans_H)
+# define _x86_sse2trans_H (1)
+# include "x86int.h"
+
+# if defined(OC_X86_64_ASM)
+/*On x86-64 we can transpose in-place without spilling registers.
+  By clever choices of the order to apply the butterflies and the order of
+   their outputs, we can take the rows in order and output the columns in order
+   without any extra operations and using just one temporary register.*/
+#  define OC_TRANSPOSE_8x8 \
+ "#OC_TRANSPOSE_8x8\n\t" \
+ "movdqa %%xmm4,%%xmm8\n\t" \
+ /*xmm4 = f3 e3 f2 e2 f1 e1 f0 e0*/ \
+ "punpcklwd %%xmm5,%%xmm4\n\t" \
+ /*xmm8 = f7 e7 f6 e6 f5 e5 f4 e4*/ \
+ "punpckhwd %%xmm5,%%xmm8\n\t" \
+ /*xmm5 is free.*/ \
+ "movdqa %%xmm0,%%xmm5\n\t" \
+ /*xmm0 = b3 a3 b2 a2 b1 a1 b0 a0*/ \
+ "punpcklwd %%xmm1,%%xmm0\n\t" \
+ /*xmm5 = b7 a7 b6 a6 b5 a5 b4 a4*/ \
+ "punpckhwd %%xmm1,%%xmm5\n\t" \
+ /*xmm1 is free.*/ \
+ "movdqa %%xmm6,%%xmm1\n\t" \
+ /*xmm6 = h3 g3 h2 g2 h1 g1 h0 g0*/ \
+ "punpcklwd %%xmm7,%%xmm6\n\t" \
+ /*xmm1 = h7 g7 h6 g6 h5 g5 h4 g4*/ \
+ "punpckhwd %%xmm7,%%xmm1\n\t" \
+ /*xmm7 is free.*/ \
+ "movdqa %%xmm2,%%xmm7\n\t" \
+ /*xmm2 = d7 c7 d6 c6 d5 c5 d4 c4*/ \
+ "punpckhwd %%xmm3,%%xmm2\n\t" \
+ /*xmm7 = d3 c3 d2 c2 d1 c1 d0 c0*/ \
+ "punpcklwd %%xmm3,%%xmm7\n\t" \
+ /*xmm3 is free.*/ \
+ "movdqa %%xmm0,%%xmm3\n\t" \
+ /*xmm0 = d1 c1 b1 a1 d0 c0 b0 a0*/ \
+ "punpckldq %%xmm7,%%xmm0\n\t" \
+ /*xmm3 = d3 c3 b3 a3 d2 c2 b2 a2*/ \
+ "punpckhdq %%xmm7,%%xmm3\n\t" \
+ /*xmm7 is free.*/ \
+ "movdqa %%xmm5,%%xmm7\n\t" \
+ /*xmm5 = d5 c5 b5 a5 d4 c4 b4 a4*/ \
+ "punpckldq %%xmm2,%%xmm5\n\t" \
+ /*xmm7 = d7 c7 b7 a7 d6 c6 b6 a6*/ \
+ "punpckhdq %%xmm2,%%xmm7\n\t" \
+ /*xmm2 is free.*/ \
+ "movdqa %%xmm4,%%xmm2\n\t" \
+ /*xmm4 = h3 g3 f3 e3 h2 g2 f2 e2*/ \
+ "punpckhdq %%xmm6,%%xmm4\n\t" \
+ /*xmm2 = h1 g1 f1 e1 h0 g0 f0 e0*/ \
+ "punpckldq %%xmm6,%%xmm2\n\t" \
+ /*xmm6 is free.*/ \
+ "movdqa %%xmm8,%%xmm6\n\t" \
+ /*xmm6 = h5 g5 f5 e5 h4 g4 f4 e4*/ \
+ "punpckldq %%xmm1,%%xmm6\n\t" \
+ /*xmm8 = h7 g7 f7 e7 h6 g6 f6 e6*/ \
+ "punpckhdq %%xmm1,%%xmm8\n\t" \
+ /*xmm1 is free.*/ \
+ "movdqa %%xmm0,%%xmm1\n\t" \
+ /*xmm0 = h0 g0 f0 e0 d0 c0 b0 a0*/ \
+ "punpcklqdq %%xmm2,%%xmm0\n\t" \
+ /*xmm1 = h1 g1 f1 e1 d1 c1 b1 a1*/ \
+ "punpckhqdq %%xmm2,%%xmm1\n\t" \
+ /*xmm2 is free.*/ \
+ "movdqa %%xmm3,%%xmm2\n\t" \
+ /*xmm3 = h3 g3 f3 e3 d3 c3 b3 a3*/ \
+ "punpckhqdq %%xmm4,%%xmm3\n\t" \
+ /*xmm2 = h2 g2 f2 e2 d2 c2 b2 a2*/ \
+ "punpcklqdq %%xmm4,%%xmm2\n\t" \
+ /*xmm4 is free.*/ \
+ "movdqa %%xmm5,%%xmm4\n\t" \
+ /*xmm5 = h5 g5 f5 e5 d5 c5 b5 a5*/ \
+ "punpckhqdq %%xmm6,%%xmm5\n\t" \
+ /*xmm4 = h4 g4 f4 e4 d4 c4 b4 a4*/ \
+ "punpcklqdq %%xmm6,%%xmm4\n\t" \
+ /*xmm6 is free.*/ \
+ "movdqa %%xmm7,%%xmm6\n\t" \
+ /*xmm7 = h7 g7 f7 e7 d7 c7 b7 a7*/ \
+ "punpckhqdq %%xmm8,%%xmm7\n\t" \
+ /*xmm6 = h6 g6 f6 e6 d6 c6 b6 a6*/ \
+ "punpcklqdq %%xmm8,%%xmm6\n\t" \
+ /*xmm8 is free.*/ \
+
+# else
+/*Otherwise, we need to spill some values to %[buf] temporarily.
+  Again, the butterflies are carefully arranged to get the columns to come out
+   in order, minimizing register spills and maximizing the delay between a load
+   and when the value loaded is actually used.*/
+#  define OC_TRANSPOSE_8x8 \
+ "#OC_TRANSPOSE_8x8\n\t" \
+ /*buf[0] = a7 a6 a5 a4 a3 a2 a1 a0*/ \
+ "movdqa %%xmm0,"OC_MEM_OFFS(0x00,buf)"\n\t" \
+ /*xmm0 is free.*/ \
+ "movdqa %%xmm2,%%xmm0\n\t" \
+ /*xmm2 = d7 c7 d6 c6 d5 c5 d4 c4*/ \
+ "punpckhwd %%xmm3,%%xmm2\n\t" \
+ /*xmm0 = d3 c3 d2 c2 d1 c1 d0 c0*/ \
+ "punpcklwd %%xmm3,%%xmm0\n\t" \
+ /*xmm3 = a7 a6 a5 a4 a3 a2 a1 a0*/ \
+ "movdqa "OC_MEM_OFFS(0x00,buf)",%%xmm3\n\t" \
+ /*buf[1] = d7 c7 d6 c6 d5 c5 d4 c4*/ \
+ "movdqa %%xmm2,"OC_MEM_OFFS(0x10,buf)"\n\t" \
+ /*xmm2 is free.*/ \
+ "movdqa %%xmm6,%%xmm2\n\t" \
+ /*xmm6 = h3 g3 h2 g2 h1 g1 h0 g0*/ \
+ "punpcklwd %%xmm7,%%xmm6\n\t" \
+ /*xmm2 = h7 g7 h6 g6 h5 g5 h4 g4*/ \
+ "punpckhwd %%xmm7,%%xmm2\n\t" \
+ /*xmm7 is free.*/ \
+ "movdqa %%xmm4,%%xmm7\n\t" \
+ /*xmm4 = f3 e3 f2 e2 f1 e1 f0 e0*/ \
+ "punpcklwd %%xmm5,%%xmm4\n\t" \
+ /*xmm7 = f7 e7 f6 e6 f5 e5 f4 e4*/ \
+ "punpckhwd %%xmm5,%%xmm7\n\t" \
+ /*xmm5 is free.*/ \
+ "movdqa %%xmm3,%%xmm5\n\t" \
+ /*xmm3 = b3 a3 b2 a2 b1 a1 b0 a0*/ \
+ "punpcklwd %%xmm1,%%xmm3\n\t" \
+ /*xmm5 = b7 a7 b6 a6 b5 a5 b4 a4*/ \
+ "punpckhwd %%xmm1,%%xmm5\n\t" \
+ /*xmm1 is free.*/ \
+ "movdqa %%xmm7,%%xmm1\n\t" \
+ /*xmm7 = h5 g5 f5 e5 h4 g4 f4 e4*/ \
+ "punpckldq %%xmm2,%%xmm7\n\t" \
+ /*xmm1 = h7 g7 f7 e7 h6 g6 f6 e6*/ \
+ "punpckhdq %%xmm2,%%xmm1\n\t" \
+ /*xmm2 = d7 c7 d6 c6 d5 c5 d4 c4*/ \
+ "movdqa "OC_MEM_OFFS(0x10,buf)",%%xmm2\n\t" \
+ /*buf[0] = h7 g7 f7 e7 h6 g6 f6 e6*/ \
+ "movdqa %%xmm1,"OC_MEM_OFFS(0x00,buf)"\n\t" \
+ /*xmm1 is free.*/ \
+ "movdqa %%xmm3,%%xmm1\n\t" \
+ /*xmm3 = d3 c3 b3 a3 d2 c2 b2 a2*/ \
+ "punpckhdq %%xmm0,%%xmm3\n\t" \
+ /*xmm1 = d1 c1 b1 a1 d0 c0 b0 a0*/ \
+ "punpckldq %%xmm0,%%xmm1\n\t" \
+ /*xmm0 is free.*/ \
+ "movdqa %%xmm4,%%xmm0\n\t" \
+ /*xmm4 = h3 g3 f3 e3 h2 g2 f2 e2*/ \
+ "punpckhdq %%xmm6,%%xmm4\n\t" \
+ /*xmm0 = h1 g1 f1 e1 h0 g0 f0 e0*/ \
+ "punpckldq %%xmm6,%%xmm0\n\t" \
+ /*xmm6 is free.*/ \
+ "movdqa %%xmm5,%%xmm6\n\t" \
+ /*xmm5 = d5 c5 b5 a5 d4 c4 b4 a4*/ \
+ "punpckldq %%xmm2,%%xmm5\n\t" \
+ /*xmm6 = d7 c7 b7 a7 d6 c6 b6 a6*/ \
+ "punpckhdq %%xmm2,%%xmm6\n\t" \
+ /*xmm2 is free.*/ \
+ "movdqa %%xmm1,%%xmm2\n\t" \
+ /*xmm1 = h1 g1 f1 e1 d1 c1 b1 a1*/ \
+ "punpckhqdq %%xmm0,%%xmm1\n\t" \
+ /*xmm2 = h0 g0 f0 e0 d0 c0 b0 a0*/ \
+ "punpcklqdq %%xmm0,%%xmm2\n\t" \
+ /*xmm0 = h7 g7 f7 e7 h6 g6 f6 e6*/ \
+ "movdqa "OC_MEM_OFFS(0x00,buf)",%%xmm0\n\t" \
+ /*buf[1] = h0 g0 f0 e0 d0 c0 b0 a0*/ \
+ "movdqa %%xmm2,"OC_MEM_OFFS(0x10,buf)"\n\t" \
+ /*xmm2 is free.*/ \
+ "movdqa %%xmm3,%%xmm2\n\t" \
+ /*xmm3 = h3 g3 f3 e3 d3 c3 b3 a3*/ \
+ "punpckhqdq %%xmm4,%%xmm3\n\t" \
+ /*xmm2 = h2 g2 f2 e2 d2 c2 b2 a2*/ \
+ "punpcklqdq %%xmm4,%%xmm2\n\t" \
+ /*xmm4 is free.*/ \
+ "movdqa %%xmm5,%%xmm4\n\t" \
+ /*xmm5 = h5 g5 f5 e5 d5 c5 b5 a5*/ \
+ "punpckhqdq %%xmm7,%%xmm5\n\t" \
+ /*xmm4 = h4 g4 f4 e4 d4 c4 b4 a4*/ \
+ "punpcklqdq %%xmm7,%%xmm4\n\t" \
+ /*xmm7 is free.*/ \
+ "movdqa %%xmm6,%%xmm7\n\t" \
+ /*xmm6 = h6 g6 f6 e6 d6 c6 b6 a6*/ \
+ "punpcklqdq %%xmm0,%%xmm6\n\t" \
+ /*xmm7 = h7 g7 f7 e7 d7 c7 b7 a7*/ \
+ "punpckhqdq %%xmm0,%%xmm7\n\t" \
+ /*xmm0 = h0 g0 f0 e0 d0 c0 b0 a0*/ \
+ "movdqa "OC_MEM_OFFS(0x10,buf)",%%xmm0\n\t" \
+
+# endif
+
+/*Transpose 4 values in each of 8 MMX registers into 8 values in the first
+   four SSE registers.
+  No need to be clever here; we have plenty of room.*/
+#  define OC_TRANSPOSE_8x4_MMX2SSE \
+ "#OC_TRANSPOSE_8x4_MMX2SSE\n\t" \
+ "movq2dq %%mm0,%%xmm0\n\t" \
+ "movq2dq %%mm1,%%xmm1\n\t" \
+ /*xmmA = b3 a3 b2 a2 b1 a1 b0 a0*/ \
+ "punpcklwd %%xmm1,%%xmm0\n\t" \
+ "movq2dq %%mm2,%%xmm3\n\t" \
+ "movq2dq %%mm3,%%xmm2\n\t" \
+ /*xmmC = d3 c3 d2 c2 d1 c1 d0 c0*/ \
+ "punpcklwd %%xmm2,%%xmm3\n\t" \
+ "movq2dq %%mm4,%%xmm4\n\t" \
+ "movq2dq %%mm5,%%xmm5\n\t" \
+ /*xmmE = f3 e3 f2 e2 f1 e1 f0 e0*/ \
+ "punpcklwd %%xmm5,%%xmm4\n\t" \
+ "movq2dq %%mm6,%%xmm7\n\t" \
+ "movq2dq %%mm7,%%xmm6\n\t" \
+ /*xmmG = h3 g3 h2 g2 h1 g1 h0 g0*/ \
+ "punpcklwd %%xmm6,%%xmm7\n\t" \
+ "movdqa %%xmm0,%%xmm2\n\t" \
+ /*xmm0 = d1 c1 b1 a1 d0 c0 b0 a0*/ \
+ "punpckldq %%xmm3,%%xmm0\n\t" \
+ /*xmm2 = d3 c3 b3 a3 d2 c2 b2 a2*/ \
+ "punpckhdq %%xmm3,%%xmm2\n\t" \
+ "movdqa %%xmm4,%%xmm5\n\t" \
+ /*xmm4 = h1 g1 f1 e1 h0 g0 f0 e0*/ \
+ "punpckldq %%xmm7,%%xmm4\n\t" \
+ /*xmm3 = h3 g3 f3 e3 h2 g2 f2 e2*/ \
+ "punpckhdq %%xmm7,%%xmm5\n\t" \
+ "movdqa %%xmm0,%%xmm1\n\t" \
+ /*xmm0 = h0 g0 f0 e0 d0 c0 b0 a0*/ \
+ "punpcklqdq %%xmm4,%%xmm0\n\t" \
+ /*xmm1 = h1 g1 f1 e1 d1 c1 b1 a1*/ \
+ "punpckhqdq %%xmm4,%%xmm1\n\t" \
+ "movdqa %%xmm2,%%xmm3\n\t" \
+ /*xmm2 = h2 g2 f2 e2 d2 c2 b2 a2*/ \
+ "punpcklqdq %%xmm5,%%xmm2\n\t" \
+ /*xmm3 = h3 g3 f3 e3 d3 c3 b3 a3*/ \
+ "punpckhqdq %%xmm5,%%xmm3\n\t" \
+
+#endif

lib/theoratest/lib/theora/x86/x86cpu.c

@@ -0,0 +1,182 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+ CPU capability detection for x86 processors.
+  Originally written by Rudolf Marek.
+
+ function:
+  last mod: $Id$
+
+ ********************************************************************/
+
+#include "x86cpu.h"
+
+#if !defined(OC_X86_ASM)
+ogg_uint32_t oc_cpu_flags_get(void){
+  return 0;
+}
+#else
+# if defined(__amd64__)||defined(__x86_64__)
+/*On x86-64, gcc seems to be able to figure out how to save %rbx for us when
+   compiling with -fPIC.*/
+#  define cpuid(_op,_eax,_ebx,_ecx,_edx) \
+  __asm__ __volatile__( \
+   "cpuid\n\t" \
+   :[eax]"=a"(_eax),[ebx]"=b"(_ebx),[ecx]"=c"(_ecx),[edx]"=d"(_edx) \
+   :"a"(_op) \
+   :"cc" \
+  )
+# else
+/*On x86-32, not so much.*/
+#  define cpuid(_op,_eax,_ebx,_ecx,_edx) \
+  __asm__ __volatile__( \
+   "xchgl %%ebx,%[ebx]\n\t" \
+   "cpuid\n\t" \
+   "xchgl %%ebx,%[ebx]\n\t" \
+   :[eax]"=a"(_eax),[ebx]"=r"(_ebx),[ecx]"=c"(_ecx),[edx]"=d"(_edx) \
+   :"a"(_op) \
+   :"cc" \
+  )
+# endif
+
+static ogg_uint32_t oc_parse_intel_flags(ogg_uint32_t _edx,ogg_uint32_t _ecx){
+  ogg_uint32_t flags;
+  /*If there isn't even MMX, give up.*/
+  if(!(_edx&0x00800000))return 0;
+  flags=OC_CPU_X86_MMX;
+  if(_edx&0x02000000)flags|=OC_CPU_X86_MMXEXT|OC_CPU_X86_SSE;
+  if(_edx&0x04000000)flags|=OC_CPU_X86_SSE2;
+  if(_ecx&0x00000001)flags|=OC_CPU_X86_PNI;
+  if(_ecx&0x00000100)flags|=OC_CPU_X86_SSSE3;
+  if(_ecx&0x00080000)flags|=OC_CPU_X86_SSE4_1;
+  if(_ecx&0x00100000)flags|=OC_CPU_X86_SSE4_2;
+  return flags;
+}
+
+static ogg_uint32_t oc_parse_amd_flags(ogg_uint32_t _edx,ogg_uint32_t _ecx){
+  ogg_uint32_t flags;
+  /*If there isn't even MMX, give up.*/
+  if(!(_edx&0x00800000))return 0;
+  flags=OC_CPU_X86_MMX;
+  if(_edx&0x00400000)flags|=OC_CPU_X86_MMXEXT;
+  if(_edx&0x80000000)flags|=OC_CPU_X86_3DNOW;
+  if(_edx&0x40000000)flags|=OC_CPU_X86_3DNOWEXT;
+  if(_ecx&0x00000040)flags|=OC_CPU_X86_SSE4A;
+  if(_ecx&0x00000800)flags|=OC_CPU_X86_SSE5;
+  return flags;
+}
+
+ogg_uint32_t oc_cpu_flags_get(void){
+  ogg_uint32_t flags;
+  ogg_uint32_t eax;
+  ogg_uint32_t ebx;
+  ogg_uint32_t ecx;
+  ogg_uint32_t edx;
+# if !defined(__amd64__)&&!defined(__x86_64__)
+  /*Not all x86-32 chips support cpuid, so we have to check.*/
+  __asm__ __volatile__(
+   "pushfl\n\t"
+   "pushfl\n\t"
+   "popl %[a]\n\t"
+   "movl %[a],%[b]\n\t"
+   "xorl $0x200000,%[a]\n\t"
+   "pushl %[a]\n\t"
+   "popfl\n\t"
+   "pushfl\n\t"
+   "popl %[a]\n\t"
+   "popfl\n\t"
+   :[a]"=r"(eax),[b]"=r"(ebx)
+   :
+   :"cc"
+  );
+  /*No cpuid.*/
+  if(eax==ebx)return 0;
+# endif
+  cpuid(0,eax,ebx,ecx,edx);
+  /*         l e t n          I e n i          u n e G*/
+  if(ecx==0x6C65746E&&edx==0x49656E69&&ebx==0x756E6547||
+   /*      6 8 x M          T e n i          u n e G*/
+   ecx==0x3638784D&&edx==0x54656E69&&ebx==0x756E6547){
+    int family;
+    int model;
+    /*Intel, Transmeta (tested with Crusoe TM5800):*/
+    cpuid(1,eax,ebx,ecx,edx);
+    flags=oc_parse_intel_flags(edx,ecx);
+    family=(eax>>8)&0xF;
+    model=(eax>>4)&0xF;
+    /*The SSE unit on the Pentium M and Core Duo is much slower than the MMX
+       unit, so don't use it.*/
+    if(family==6&&(model==9||model==13||model==14)){
+      flags&=~(OC_CPU_X86_SSE2|OC_CPU_X86_PNI);
+    }
+  }
+  /*              D M A c          i t n e          h t u A*/
+  else if(ecx==0x444D4163&&edx==0x69746E65&&ebx==0x68747541||
+   /*      C S N            y b   e          d o e G*/
+   ecx==0x43534e20&&edx==0x79622065&&ebx==0x646f6547){
+    /*AMD, Geode:*/
+    cpuid(0x80000000,eax,ebx,ecx,edx);
+    if(eax<0x80000001)flags=0;
+    else{
+      cpuid(0x80000001,eax,ebx,ecx,edx);
+      flags=oc_parse_amd_flags(edx,ecx);
+    }
+    /*Also check for SSE.*/
+    cpuid(1,eax,ebx,ecx,edx);
+    flags|=oc_parse_intel_flags(edx,ecx);
+  }
+  /*Technically some VIA chips can be configured in the BIOS to return any
+     string here the user wants.
+    There is a special detection method that can be used to identify such
+     processors, but in my opinion, if the user really wants to change it, they
+     deserve what they get.*/
+  /*              s l u a          H r u a          t n e C*/
+  else if(ecx==0x736C7561&&edx==0x48727561&&ebx==0x746E6543){
+    /*VIA:*/
+    /*I only have documentation for the C7 (Esther) and Isaiah (forthcoming)
+       chips (thanks to the engineers from Centaur Technology who provided it).
+      These chips support Intel-like cpuid info.
+      The C3-2 (Nehemiah) cores appear to, as well.*/
+    cpuid(1,eax,ebx,ecx,edx);
+    flags=oc_parse_intel_flags(edx,ecx);
+    if(eax>=0x80000001){
+      /*The (non-Nehemiah) C3 processors support AMD-like cpuid info.
+        We need to check this even if the Intel test succeeds to pick up 3DNow!
+         support on these processors.
+        Unlike actual AMD processors, we cannot _rely_ on this info, since
+         some cores (e.g., the 693 stepping of the Nehemiah) claim to support
+         this function, yet return edx=0, despite the Intel test indicating
+         MMX support.
+        Therefore the features detected here are strictly added to those
+         detected by the Intel test.*/
+      /*TODO: How about earlier chips?*/
+      cpuid(0x80000001,eax,ebx,ecx,edx);
+      /*Note: As of the C7, this function returns Intel-style extended feature
+         flags, not AMD-style.
+        Currently, this only defines bits 11, 20, and 29 (0x20100800), which
+         do not conflict with any of the AMD flags we inspect.
+        For the remaining bits, Intel tells us, "Do not count on their value",
+         but VIA assures us that they will all be zero (at least on the C7 and
+         Isaiah chips).
+        In the (unlikely) event a future processor uses bits 18, 19, 30, or 31
+         (0xC0C00000) for something else, we will have to add code to detect
+         the model to decide when it is appropriate to inspect them.*/
+      flags|=oc_parse_amd_flags(edx,ecx);
+    }
+  }
+  else{
+    /*Implement me.*/
+    flags=0;
+  }
+  return flags;
+}
+#endif

lib/theoratest/lib/theora/x86/x86cpu.h

@@ -0,0 +1,36 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+ function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#if !defined(_x86_x86cpu_H)
+# define _x86_x86cpu_H (1)
+#include "../internal.h"
+
+#define OC_CPU_X86_MMX      (1<<0)
+#define OC_CPU_X86_3DNOW    (1<<1)
+#define OC_CPU_X86_3DNOWEXT (1<<2)
+#define OC_CPU_X86_MMXEXT   (1<<3)
+#define OC_CPU_X86_SSE      (1<<4)
+#define OC_CPU_X86_SSE2     (1<<5)
+#define OC_CPU_X86_PNI      (1<<6)
+#define OC_CPU_X86_SSSE3    (1<<7)
+#define OC_CPU_X86_SSE4_1   (1<<8)
+#define OC_CPU_X86_SSE4_2   (1<<9)
+#define OC_CPU_X86_SSE4A    (1<<10)
+#define OC_CPU_X86_SSE5     (1<<11)
+
+ogg_uint32_t oc_cpu_flags_get(void);
+
+#endif

lib/theoratest/lib/theora/x86/x86int.h

@@ -0,0 +1,122 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#if !defined(_x86_x86int_H)
+# define _x86_x86int_H (1)
+# include "../internal.h"
+
+# if defined(OC_X86_ASM)
+#  define oc_state_accel_init oc_state_accel_init_x86
+#  if defined(OC_X86_64_ASM)
+/*x86-64 guarantees SIMD support up through at least SSE2.
+  If the best routine we have available only needs SSE2 (which at the moment
+   covers all of them), then we can avoid runtime detection and the indirect
+   call.*/
+#   define oc_frag_copy(_state,_dst,_src,_ystride) \
+  oc_frag_copy_mmx(_dst,_src,_ystride)
+#   define oc_frag_copy_list(_state,_dst_frame,_src_frame,_ystride, \
+ _fragis,_nfragis,_frag_buf_offs) \
+  oc_frag_copy_list_mmx(_dst_frame,_src_frame,_ystride, \
+   _fragis,_nfragis,_frag_buf_offs)
+#   define oc_frag_recon_intra(_state,_dst,_ystride,_residue) \
+  oc_frag_recon_intra_mmx(_dst,_ystride,_residue)
+#   define oc_frag_recon_inter(_state,_dst,_src,_ystride,_residue) \
+  oc_frag_recon_inter_mmx(_dst,_src,_ystride,_residue)
+#   define oc_frag_recon_inter2(_state,_dst,_src1,_src2,_ystride,_residue) \
+  oc_frag_recon_inter2_mmx(_dst,_src1,_src2,_ystride,_residue)
+#   define oc_idct8x8(_state,_y,_x,_last_zzi) \
+  oc_idct8x8_sse2(_y,_x,_last_zzi)
+#   define oc_state_frag_recon oc_state_frag_recon_mmx
+#   define oc_loop_filter_init(_state,_bv,_flimit) \
+  oc_loop_filter_init_mmxext(_bv,_flimit)
+#   define oc_state_loop_filter_frag_rows oc_state_loop_filter_frag_rows_mmxext
+#   define oc_restore_fpu(_state) \
+  oc_restore_fpu_mmx()
+#  else
+#   define OC_STATE_USE_VTABLE (1)
+#  endif
+# endif
+
+# include "../state.h"
+# include "x86cpu.h"
+
+/*Converts the expression in the argument to a string.*/
+#define OC_M2STR(_s) #_s
+
+/*Memory operands do not always include an offset.
+  To avoid warnings, we force an offset with %H (which adds 8).*/
+# if __GNUC_PREREQ(4,0)
+#  define OC_MEM_OFFS(_offs,_name) \
+  OC_M2STR(_offs-8+%H[_name])
+# endif
+/*If your gcc version does't support %H, then you get to suffer the warnings.
+  Note that Apple's gas breaks on things like _offs+(%esp): it throws away the
+   whole offset, instead of substituting in 0 for the missing operand to +.*/
+# if !defined(OC_MEM_OFFS)
+#  define OC_MEM_OFFS(_offs,_name) \
+  OC_M2STR(_offs+%[_name])
+# endif
+
+/*Declare an array operand with an exact size.
+  This tells gcc we're going to clobber this memory region, without having to
+   clobber all of "memory" and lets us access local buffers directly using the
+   stack pointer, without allocating a separate register to point to them.*/
+#define OC_ARRAY_OPERAND(_type,_ptr,_size) \
+  (*({ \
+    struct{_type array_value__[(_size)];} *array_addr__=(void *)(_ptr); \
+    array_addr__; \
+  }))
+
+/*Declare an array operand with an exact size.
+  This tells gcc we're going to clobber this memory region, without having to
+   clobber all of "memory" and lets us access local buffers directly using the
+   stack pointer, without allocating a separate register to point to them.*/
+#define OC_CONST_ARRAY_OPERAND(_type,_ptr,_size) \
+  (*({ \
+    const struct{_type array_value__[(_size)];} *array_addr__= \
+     (const void *)(_ptr); \
+    array_addr__; \
+  }))
+
+extern const unsigned short __attribute__((aligned(16))) OC_IDCT_CONSTS[64];
+
+void oc_state_accel_init_x86(oc_theora_state *_state);
+
+void oc_frag_copy_mmx(unsigned char *_dst,
+ const unsigned char *_src,int _ystride);
+void oc_frag_copy_list_mmx(unsigned char *_dst_frame,
+ const unsigned char *_src_frame,int _ystride,
+ const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs);
+void oc_frag_recon_intra_mmx(unsigned char *_dst,int _ystride,
+ const ogg_int16_t *_residue);
+void oc_frag_recon_inter_mmx(unsigned char *_dst,
+ const unsigned char *_src,int _ystride,const ogg_int16_t *_residue);
+void oc_frag_recon_inter2_mmx(unsigned char *_dst,const unsigned char *_src1,
+ const unsigned char *_src2,int _ystride,const ogg_int16_t *_residue);
+void oc_idct8x8_mmx(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi);
+void oc_idct8x8_sse2(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi);
+void oc_state_frag_recon_mmx(const oc_theora_state *_state,ptrdiff_t _fragi,
+ int _pli,ogg_int16_t _dct_coeffs[128],int _last_zzi,ogg_uint16_t _dc_quant);
+void oc_loop_filter_init_mmx(signed char _bv[256],int _flimit);
+void oc_loop_filter_init_mmxext(signed char _bv[256],int _flimit);
+void oc_state_loop_filter_frag_rows_mmx(const oc_theora_state *_state,
+ signed char _bv[256],int _refi,int _pli,int _fragy0,int _fragy_end);
+void oc_state_loop_filter_frag_rows_mmxext(const oc_theora_state *_state,
+ signed char _bv[256],int _refi,int _pli,int _fragy0,int _fragy_end);
+void oc_restore_fpu_mmx(void);
+
+#endif

lib/theoratest/lib/theora/x86/x86state.c

@@ -0,0 +1,97 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include "x86int.h"
+
+#if defined(OC_X86_ASM)
+
+#if defined(OC_STATE_USE_VTABLE)
+/*This table has been modified from OC_FZIG_ZAG by baking a 4x4 transpose into
+   each quadrant of the destination.*/
+static const unsigned char OC_FZIG_ZAG_MMX[128]={
+   0, 8, 1, 2, 9,16,24,17,
+  10, 3,32,11,18,25, 4,12,
+   5,26,19,40,33,34,41,48,
+  27, 6,13,20,28,21,14, 7,
+  56,49,42,35,43,50,57,36,
+  15,22,29,30,23,44,37,58,
+  51,59,38,45,52,31,60,53,
+  46,39,47,54,61,62,55,63,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64
+};
+#endif
+
+/*This table has been modified from OC_FZIG_ZAG by baking an 8x8 transpose into
+   the destination.*/
+static const unsigned char OC_FZIG_ZAG_SSE2[128]={
+   0, 8, 1, 2, 9,16,24,17,
+  10, 3, 4,11,18,25,32,40,
+  33,26,19,12, 5, 6,13,20,
+  27,34,41,48,56,49,42,35,
+  28,21,14, 7,15,22,29,36,
+  43,50,57,58,51,44,37,30,
+  23,31,38,45,52,59,60,53,
+  46,39,47,54,61,62,55,63,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64
+};
+
+void oc_state_accel_init_x86(oc_theora_state *_state){
+  oc_state_accel_init_c(_state);
+  _state->cpu_flags=oc_cpu_flags_get();
+# if defined(OC_STATE_USE_VTABLE)
+  if(_state->cpu_flags&OC_CPU_X86_MMX){
+    _state->opt_vtable.frag_copy=oc_frag_copy_mmx;
+    _state->opt_vtable.frag_copy_list=oc_frag_copy_list_mmx;
+    _state->opt_vtable.frag_recon_intra=oc_frag_recon_intra_mmx;
+    _state->opt_vtable.frag_recon_inter=oc_frag_recon_inter_mmx;
+    _state->opt_vtable.frag_recon_inter2=oc_frag_recon_inter2_mmx;
+    _state->opt_vtable.idct8x8=oc_idct8x8_mmx;
+    _state->opt_vtable.state_frag_recon=oc_state_frag_recon_mmx;
+    _state->opt_vtable.loop_filter_init=oc_loop_filter_init_mmx;
+    _state->opt_vtable.state_loop_filter_frag_rows=
+     oc_state_loop_filter_frag_rows_mmx;
+    _state->opt_vtable.restore_fpu=oc_restore_fpu_mmx;
+    _state->opt_data.dct_fzig_zag=OC_FZIG_ZAG_MMX;
+  }
+  if(_state->cpu_flags&OC_CPU_X86_MMXEXT){
+    _state->opt_vtable.loop_filter_init=oc_loop_filter_init_mmxext;
+    _state->opt_vtable.state_loop_filter_frag_rows=
+     oc_state_loop_filter_frag_rows_mmxext;
+  }
+  if(_state->cpu_flags&OC_CPU_X86_SSE2){
+    _state->opt_vtable.idct8x8=oc_idct8x8_sse2;
+# endif
+    _state->opt_data.dct_fzig_zag=OC_FZIG_ZAG_SSE2;
+# if defined(OC_STATE_USE_VTABLE)
+  }
+# endif
+}
+#endif

lib/theoratest/lib/theora/x86_vc/mmxfrag.c

@@ -0,0 +1,416 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+/*MMX acceleration of fragment reconstruction for motion compensation.
+  Originally written by Rudolf Marek.
+  Additional optimization by Nils Pipenbrinck.
+  Note: Loops are unrolled for best performance.
+  The iteration each instruction belongs to is marked in the comments as #i.*/
+#include <stddef.h>
+#include "x86int.h"
+
+#if defined(OC_X86_ASM)
+
+/*Copies an 8x8 block of pixels from _src to _dst, assuming _ystride bytes
+   between rows.*/
+# define OC_FRAG_COPY_MMX(_dst,_src,_ystride) \
+  do{ \
+    const unsigned char *src; \
+    unsigned char       *dst; \
+    src=(_src); \
+    dst=(_dst); \
+    __asm  mov SRC,src \
+    __asm  mov DST,dst \
+    __asm  mov YSTRIDE,_ystride \
+    /*src+0*ystride*/ \
+    __asm  movq mm0,[SRC] \
+    /*src+1*ystride*/ \
+    __asm  movq mm1,[SRC+YSTRIDE] \
+    /*ystride3=ystride*3*/ \
+    __asm  lea YSTRIDE3,[YSTRIDE+YSTRIDE*2] \
+    /*src+2*ystride*/ \
+    __asm  movq mm2,[SRC+YSTRIDE*2] \
+    /*src+3*ystride*/ \
+    __asm  movq mm3,[SRC+YSTRIDE3] \
+    /*dst+0*ystride*/ \
+    __asm  movq [DST],mm0 \
+    /*dst+1*ystride*/ \
+    __asm  movq [DST+YSTRIDE],mm1 \
+    /*Pointer to next 4.*/ \
+    __asm  lea SRC,[SRC+YSTRIDE*4] \
+    /*dst+2*ystride*/ \
+    __asm  movq [DST+YSTRIDE*2],mm2 \
+    /*dst+3*ystride*/ \
+    __asm  movq [DST+YSTRIDE3],mm3 \
+    /*Pointer to next 4.*/ \
+    __asm  lea DST,[DST+YSTRIDE*4] \
+    /*src+0*ystride*/ \
+    __asm  movq mm0,[SRC] \
+    /*src+1*ystride*/ \
+    __asm  movq mm1,[SRC+YSTRIDE] \
+    /*src+2*ystride*/ \
+    __asm  movq mm2,[SRC+YSTRIDE*2] \
+    /*src+3*ystride*/ \
+    __asm  movq mm3,[SRC+YSTRIDE3] \
+    /*dst+0*ystride*/ \
+    __asm  movq [DST],mm0 \
+    /*dst+1*ystride*/ \
+    __asm  movq [DST+YSTRIDE],mm1 \
+    /*dst+2*ystride*/ \
+    __asm  movq [DST+YSTRIDE*2],mm2 \
+    /*dst+3*ystride*/ \
+    __asm  movq [DST+YSTRIDE3],mm3 \
+  } \
+  while(0)
+
+/*Copies an 8x8 block of pixels from _src to _dst, assuming _ystride bytes
+   between rows.*/
+void oc_frag_copy_mmx(unsigned char *_dst,
+ const unsigned char *_src,int _ystride){
+#define SRC edx
+#define DST eax
+#define YSTRIDE ecx
+#define YSTRIDE3 esi
+  OC_FRAG_COPY_MMX(_dst,_src,_ystride);
+#undef SRC
+#undef DST
+#undef YSTRIDE
+#undef YSTRIDE3
+}
+
+/*Copies the fragments specified by the lists of fragment indices from one
+   frame to another.
+  _dst_frame:     The reference frame to copy to.
+  _src_frame:     The reference frame to copy from.
+  _ystride:       The row stride of the reference frames.
+  _fragis:        A pointer to a list of fragment indices.
+  _nfragis:       The number of fragment indices to copy.
+  _frag_buf_offs: The offsets of fragments in the reference frames.*/
+void oc_frag_copy_list_mmx(unsigned char *_dst_frame,
+ const unsigned char *_src_frame,int _ystride,
+ const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs){
+  ptrdiff_t fragii;
+  for(fragii=0;fragii<_nfragis;fragii++){
+    ptrdiff_t frag_buf_off;
+    frag_buf_off=_frag_buf_offs[_fragis[fragii]];
+#define SRC edx
+#define DST eax
+#define YSTRIDE ecx
+#define YSTRIDE3 edi
+    OC_FRAG_COPY_MMX(_dst_frame+frag_buf_off,
+     _src_frame+frag_buf_off,_ystride);
+#undef SRC
+#undef DST
+#undef YSTRIDE
+#undef YSTRIDE3
+  }
+}
+
+void oc_frag_recon_intra_mmx(unsigned char *_dst,int _ystride,
+ const ogg_int16_t *_residue){
+  __asm{
+#define DST edx
+#define DST4 esi
+#define YSTRIDE eax
+#define YSTRIDE3 edi
+#define RESIDUE ecx
+    mov DST,_dst
+    mov YSTRIDE,_ystride
+    mov RESIDUE,_residue
+    lea DST4,[DST+YSTRIDE*4]
+    lea YSTRIDE3,[YSTRIDE+YSTRIDE*2]
+    /*Set mm0 to 0xFFFFFFFFFFFFFFFF.*/
+    pcmpeqw mm0,mm0
+    /*#0 Load low residue.*/
+    movq mm1,[0*8+RESIDUE]
+    /*#0 Load high residue.*/
+    movq mm2,[1*8+RESIDUE]
+    /*Set mm0 to 0x8000800080008000.*/
+    psllw mm0,15
+    /*#1 Load low residue.*/
+    movq mm3,[2*8+RESIDUE]
+    /*#1 Load high residue.*/
+    movq mm4,[3*8+RESIDUE]
+    /*Set mm0 to 0x0080008000800080.*/
+    psrlw mm0,8
+    /*#2 Load low residue.*/
+    movq mm5,[4*8+RESIDUE]
+    /*#2 Load high residue.*/
+    movq mm6,[5*8+RESIDUE]
+    /*#0 Bias low  residue.*/
+    paddsw mm1,mm0
+    /*#0 Bias high residue.*/
+    paddsw mm2,mm0
+    /*#0 Pack to byte.*/
+    packuswb mm1,mm2
+    /*#1 Bias low  residue.*/
+    paddsw mm3,mm0
+    /*#1 Bias high residue.*/
+    paddsw mm4,mm0
+    /*#1 Pack to byte.*/
+    packuswb mm3,mm4
+    /*#2 Bias low  residue.*/
+    paddsw mm5,mm0
+    /*#2 Bias high residue.*/
+    paddsw mm6,mm0
+    /*#2 Pack to byte.*/
+    packuswb mm5,mm6
+    /*#0 Write row.*/
+    movq [DST],mm1
+    /*#1 Write row.*/
+    movq [DST+YSTRIDE],mm3
+    /*#2 Write row.*/
+    movq [DST+YSTRIDE*2],mm5
+    /*#3 Load low residue.*/
+    movq mm1,[6*8+RESIDUE]
+    /*#3 Load high residue.*/
+    movq mm2,[7*8+RESIDUE]
+    /*#4 Load high residue.*/
+    movq mm3,[8*8+RESIDUE]
+    /*#4 Load high residue.*/
+    movq mm4,[9*8+RESIDUE]
+    /*#5 Load high residue.*/
+    movq mm5,[10*8+RESIDUE]
+    /*#5 Load high residue.*/
+    movq mm6,[11*8+RESIDUE]
+    /*#3 Bias low  residue.*/
+    paddsw mm1,mm0
+    /*#3 Bias high residue.*/
+    paddsw mm2,mm0
+    /*#3 Pack to byte.*/
+    packuswb mm1,mm2
+    /*#4 Bias low  residue.*/
+    paddsw mm3,mm0
+    /*#4 Bias high residue.*/
+    paddsw mm4,mm0
+    /*#4 Pack to byte.*/
+    packuswb mm3,mm4
+    /*#5 Bias low  residue.*/
+    paddsw mm5,mm0
+    /*#5 Bias high residue.*/
+    paddsw mm6,mm0
+    /*#5 Pack to byte.*/
+    packuswb mm5,mm6
+    /*#3 Write row.*/
+    movq [DST+YSTRIDE3],mm1
+    /*#4 Write row.*/
+    movq [DST4],mm3
+    /*#5 Write row.*/
+    movq [DST4+YSTRIDE],mm5
+    /*#6 Load low residue.*/
+    movq mm1,[12*8+RESIDUE]
+    /*#6 Load high residue.*/
+    movq mm2,[13*8+RESIDUE]
+    /*#7 Load low residue.*/
+    movq mm3,[14*8+RESIDUE]
+    /*#7 Load high residue.*/
+    movq mm4,[15*8+RESIDUE]
+    /*#6 Bias low  residue.*/
+    paddsw mm1,mm0
+    /*#6 Bias high residue.*/
+    paddsw mm2,mm0
+    /*#6 Pack to byte.*/
+    packuswb mm1,mm2
+    /*#7 Bias low  residue.*/
+    paddsw mm3,mm0
+    /*#7 Bias high residue.*/
+    paddsw mm4,mm0
+    /*#7 Pack to byte.*/
+    packuswb mm3,mm4
+    /*#6 Write row.*/
+    movq [DST4+YSTRIDE*2],mm1
+    /*#7 Write row.*/
+    movq [DST4+YSTRIDE3],mm3
+#undef DST
+#undef DST4
+#undef YSTRIDE
+#undef YSTRIDE3
+#undef RESIDUE
+  }
+}
+
+void oc_frag_recon_inter_mmx(unsigned char *_dst,const unsigned char *_src,
+ int _ystride,const ogg_int16_t *_residue){
+  int i;
+  /*Zero mm0.*/
+  __asm pxor mm0,mm0;
+  for(i=4;i-->0;){
+    __asm{
+#define DST edx
+#define SRC ecx
+#define YSTRIDE edi
+#define RESIDUE eax
+      mov DST,_dst
+      mov SRC,_src
+      mov YSTRIDE,_ystride
+      mov RESIDUE,_residue
+      /*#0 Load source.*/
+      movq mm3,[SRC]
+      /*#1 Load source.*/
+      movq mm7,[SRC+YSTRIDE]
+      /*#0 Get copy of src.*/
+      movq mm4,mm3
+      /*#0 Expand high source.*/
+      punpckhbw mm4,mm0
+      /*#0 Expand low  source.*/
+      punpcklbw mm3,mm0
+      /*#0 Add residue high.*/
+      paddsw mm4,[8+RESIDUE]
+      /*#1 Get copy of src.*/
+      movq mm2,mm7
+      /*#0 Add residue low.*/
+      paddsw  mm3,[RESIDUE]
+      /*#1 Expand high source.*/
+      punpckhbw mm2,mm0
+      /*#0 Pack final row pixels.*/
+      packuswb mm3,mm4
+      /*#1 Expand low  source.*/
+      punpcklbw mm7,mm0
+      /*#1 Add residue low.*/
+      paddsw mm7,[16+RESIDUE]
+      /*#1 Add residue high.*/
+      paddsw mm2,[24+RESIDUE]
+      /*Advance residue.*/
+      lea RESIDUE,[32+RESIDUE]
+      /*#1 Pack final row pixels.*/
+      packuswb mm7,mm2
+      /*Advance src.*/
+      lea SRC,[SRC+YSTRIDE*2]
+      /*#0 Write row.*/
+      movq [DST],mm3
+      /*#1 Write row.*/
+      movq [DST+YSTRIDE],mm7
+      /*Advance dst.*/
+      lea DST,[DST+YSTRIDE*2]
+      mov _residue,RESIDUE
+      mov _dst,DST
+      mov _src,SRC
+#undef DST
+#undef SRC
+#undef YSTRIDE
+#undef RESIDUE
+    }
+  }
+}
+
+void oc_frag_recon_inter2_mmx(unsigned char *_dst,const unsigned char *_src1,
+ const unsigned char *_src2,int _ystride,const ogg_int16_t *_residue){
+  int i;
+  /*Zero mm7.*/
+  __asm pxor mm7,mm7;
+  for(i=4;i-->0;){
+    __asm{
+#define SRC1 ecx
+#define SRC2 edi
+#define YSTRIDE esi
+#define RESIDUE edx
+#define DST eax
+      mov YSTRIDE,_ystride
+      mov DST,_dst
+      mov RESIDUE,_residue
+      mov SRC1,_src1
+      mov SRC2,_src2
+      /*#0 Load src1.*/
+      movq mm0,[SRC1]
+      /*#0 Load src2.*/
+      movq mm2,[SRC2]
+      /*#0 Copy src1.*/
+      movq mm1,mm0
+      /*#0 Copy src2.*/
+      movq mm3,mm2
+      /*#1 Load src1.*/
+      movq mm4,[SRC1+YSTRIDE]
+      /*#0 Unpack lower src1.*/
+      punpcklbw mm0,mm7
+      /*#1 Load src2.*/
+      movq mm5,[SRC2+YSTRIDE]
+      /*#0 Unpack higher src1.*/
+      punpckhbw mm1,mm7
+      /*#0 Unpack lower src2.*/
+      punpcklbw mm2,mm7
+      /*#0 Unpack higher src2.*/
+      punpckhbw mm3,mm7
+      /*Advance src1 ptr.*/
+      lea SRC1,[SRC1+YSTRIDE*2]
+      /*Advance src2 ptr.*/
+      lea SRC2,[SRC2+YSTRIDE*2]
+      /*#0 Lower src1+src2.*/
+      paddsw mm0,mm2
+      /*#0 Higher src1+src2.*/
+      paddsw mm1,mm3
+      /*#1 Copy src1.*/
+      movq mm2,mm4
+      /*#0 Build lo average.*/
+      psraw mm0,1
+      /*#1 Copy src2.*/
+      movq mm3,mm5
+      /*#1 Unpack lower src1.*/
+      punpcklbw mm4,mm7
+      /*#0 Build hi average.*/
+      psraw mm1,1
+      /*#1 Unpack higher src1.*/
+      punpckhbw mm2,mm7
+      /*#0 low+=residue.*/
+      paddsw mm0,[RESIDUE]
+      /*#1 Unpack lower src2.*/
+      punpcklbw mm5,mm7
+      /*#0 high+=residue.*/
+      paddsw mm1,[8+RESIDUE]
+      /*#1 Unpack higher src2.*/
+      punpckhbw mm3,mm7
+      /*#1 Lower src1+src2.*/
+      paddsw mm5,mm4
+      /*#0 Pack and saturate.*/
+      packuswb mm0,mm1
+      /*#1 Higher src1+src2.*/
+      paddsw mm3,mm2
+      /*#0 Write row.*/
+      movq [DST],mm0
+      /*#1 Build lo average.*/
+      psraw mm5,1
+      /*#1 Build hi average.*/
+      psraw mm3,1
+      /*#1 low+=residue.*/
+      paddsw mm5,[16+RESIDUE]
+      /*#1 high+=residue.*/
+      paddsw mm3,[24+RESIDUE]
+      /*#1 Pack and saturate.*/
+      packuswb  mm5,mm3
+      /*#1 Write row ptr.*/
+      movq [DST+YSTRIDE],mm5
+      /*Advance residue ptr.*/
+      add RESIDUE,32
+      /*Advance dest ptr.*/
+      lea DST,[DST+YSTRIDE*2]
+      mov _dst,DST
+      mov _residue,RESIDUE
+      mov _src1,SRC1
+      mov _src2,SRC2
+#undef SRC1
+#undef SRC2
+#undef YSTRIDE
+#undef RESIDUE
+#undef DST
+    }
+  }
+}
+
+void oc_restore_fpu_mmx(void){
+  __asm emms;
+}
+
+#endif

lib/theoratest/lib/theora/x86_vc/mmxidct.c

@@ -0,0 +1,592 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+/*MMX acceleration of Theora's iDCT.
+  Originally written by Rudolf Marek, based on code from On2's VP3.*/
+#include "x86int.h"
+#include "../dct.h"
+
+#if defined(OC_X86_ASM)
+
+/*These are offsets into the table of constants below.*/
+/*7 rows of cosines, in order: pi/16 * (1 ... 7).*/
+#define OC_COSINE_OFFSET (8)
+/*A row of 8's.*/
+#define OC_EIGHT_OFFSET  (0)
+
+
+
+/*A table of constants used by the MMX routines.*/
+static const OC_ALIGN16(ogg_uint16_t) OC_IDCT_CONSTS[(1+7)*4]={
+      8,    8,    8,    8,
+  (ogg_uint16_t)OC_C1S7,(ogg_uint16_t)OC_C1S7,
+  (ogg_uint16_t)OC_C1S7,(ogg_uint16_t)OC_C1S7,
+  (ogg_uint16_t)OC_C2S6,(ogg_uint16_t)OC_C2S6,
+  (ogg_uint16_t)OC_C2S6,(ogg_uint16_t)OC_C2S6,
+  (ogg_uint16_t)OC_C3S5,(ogg_uint16_t)OC_C3S5,
+  (ogg_uint16_t)OC_C3S5,(ogg_uint16_t)OC_C3S5,
+  (ogg_uint16_t)OC_C4S4,(ogg_uint16_t)OC_C4S4,
+  (ogg_uint16_t)OC_C4S4,(ogg_uint16_t)OC_C4S4,
+  (ogg_uint16_t)OC_C5S3,(ogg_uint16_t)OC_C5S3,
+  (ogg_uint16_t)OC_C5S3,(ogg_uint16_t)OC_C5S3,
+  (ogg_uint16_t)OC_C6S2,(ogg_uint16_t)OC_C6S2,
+  (ogg_uint16_t)OC_C6S2,(ogg_uint16_t)OC_C6S2,
+  (ogg_uint16_t)OC_C7S1,(ogg_uint16_t)OC_C7S1,
+  (ogg_uint16_t)OC_C7S1,(ogg_uint16_t)OC_C7S1
+};
+
+/*38 cycles*/
+#define OC_IDCT_BEGIN(_y,_x) __asm{ \
+  __asm movq mm2,OC_I(3,_x) \
+  __asm movq mm6,OC_C(3) \
+  __asm movq mm4,mm2 \
+  __asm movq mm7,OC_J(5,_x) \
+  __asm pmulhw mm4,mm6 \
+  __asm movq mm1,OC_C(5) \
+  __asm pmulhw mm6,mm7 \
+  __asm movq mm5,mm1 \
+  __asm pmulhw mm1,mm2 \
+  __asm movq mm3,OC_I(1,_x) \
+  __asm pmulhw mm5,mm7 \
+  __asm movq mm0,OC_C(1) \
+  __asm paddw mm4,mm2 \
+  __asm paddw mm6,mm7 \
+  __asm paddw mm2,mm1 \
+  __asm movq mm1,OC_J(7,_x) \
+  __asm paddw mm7,mm5 \
+  __asm movq mm5,mm0 \
+  __asm pmulhw mm0,mm3 \
+  __asm paddw mm4,mm7 \
+  __asm pmulhw mm5,mm1 \
+  __asm movq mm7,OC_C(7) \
+  __asm psubw mm6,mm2 \
+  __asm paddw mm0,mm3 \
+  __asm pmulhw mm3,mm7 \
+  __asm movq mm2,OC_I(2,_x) \
+  __asm pmulhw mm7,mm1 \
+  __asm paddw mm5,mm1 \
+  __asm movq mm1,mm2 \
+  __asm pmulhw mm2,OC_C(2) \
+  __asm psubw mm3,mm5 \
+  __asm movq mm5,OC_J(6,_x) \
+  __asm paddw mm0,mm7 \
+  __asm movq mm7,mm5 \
+  __asm psubw mm0,mm4 \
+  __asm pmulhw mm5,OC_C(2) \
+  __asm paddw mm2,mm1 \
+  __asm pmulhw mm1,OC_C(6) \
+  __asm paddw mm4,mm4 \
+  __asm paddw mm4,mm0 \
+  __asm psubw mm3,mm6 \
+  __asm paddw mm5,mm7 \
+  __asm paddw mm6,mm6 \
+  __asm pmulhw mm7,OC_C(6) \
+  __asm paddw mm6,mm3 \
+  __asm movq OC_I(1,_y),mm4 \
+  __asm psubw mm1,mm5 \
+  __asm movq mm4,OC_C(4) \
+  __asm movq mm5,mm3 \
+  __asm pmulhw mm3,mm4 \
+  __asm paddw mm7,mm2 \
+  __asm movq OC_I(2,_y),mm6 \
+  __asm movq mm2,mm0 \
+  __asm movq mm6,OC_I(0,_x) \
+  __asm pmulhw mm0,mm4 \
+  __asm paddw mm5,mm3 \
+  __asm movq mm3,OC_J(4,_x) \
+  __asm psubw mm5,mm1 \
+  __asm paddw mm2,mm0 \
+  __asm psubw mm6,mm3 \
+  __asm movq mm0,mm6 \
+  __asm pmulhw mm6,mm4 \
+  __asm paddw mm3,mm3 \
+  __asm paddw mm1,mm1 \
+  __asm paddw mm3,mm0 \
+  __asm paddw mm1,mm5 \
+  __asm pmulhw mm4,mm3 \
+  __asm paddw mm6,mm0 \
+  __asm psubw mm6,mm2 \
+  __asm paddw mm2,mm2 \
+  __asm movq mm0,OC_I(1,_y) \
+  __asm paddw mm2,mm6 \
+  __asm paddw mm4,mm3 \
+  __asm psubw mm2,mm1 \
+}
+
+/*38+8=46 cycles.*/
+#define OC_ROW_IDCT(_y,_x) __asm{ \
+  OC_IDCT_BEGIN(_y,_x) \
+  /*r3=D'*/ \
+  __asm  movq mm3,OC_I(2,_y) \
+  /*r4=E'=E-G*/ \
+  __asm  psubw mm4,mm7 \
+  /*r1=H'+H'*/ \
+  __asm  paddw mm1,mm1 \
+  /*r7=G+G*/ \
+  __asm  paddw mm7,mm7 \
+  /*r1=R1=A''+H'*/ \
+  __asm  paddw mm1,mm2 \
+  /*r7=G'=E+G*/ \
+  __asm  paddw mm7,mm4 \
+  /*r4=R4=E'-D'*/ \
+  __asm  psubw mm4,mm3 \
+  __asm  paddw mm3,mm3 \
+  /*r6=R6=F'-B''*/ \
+  __asm  psubw mm6,mm5 \
+  __asm  paddw mm5,mm5 \
+  /*r3=R3=E'+D'*/ \
+  __asm  paddw mm3,mm4 \
+  /*r5=R5=F'+B''*/ \
+  __asm  paddw mm5,mm6 \
+  /*r7=R7=G'-C'*/ \
+  __asm  psubw mm7,mm0 \
+  __asm  paddw mm0,mm0 \
+  /*Save R1.*/ \
+  __asm  movq OC_I(1,_y),mm1 \
+  /*r0=R0=G.+C.*/ \
+  __asm  paddw mm0,mm7 \
+}
+
+/*The following macro does two 4x4 transposes in place.
+  At entry, we assume:
+    r0 = a3 a2 a1 a0
+  I(1) = b3 b2 b1 b0
+    r2 = c3 c2 c1 c0
+    r3 = d3 d2 d1 d0
+
+    r4 = e3 e2 e1 e0
+    r5 = f3 f2 f1 f0
+    r6 = g3 g2 g1 g0
+    r7 = h3 h2 h1 h0
+
+  At exit, we have:
+  I(0) = d0 c0 b0 a0
+  I(1) = d1 c1 b1 a1
+  I(2) = d2 c2 b2 a2
+  I(3) = d3 c3 b3 a3
+
+  J(4) = h0 g0 f0 e0
+  J(5) = h1 g1 f1 e1
+  J(6) = h2 g2 f2 e2
+  J(7) = h3 g3 f3 e3
+
+  I(0) I(1) I(2) I(3) is the transpose of r0 I(1) r2 r3.
+  J(4) J(5) J(6) J(7) is the transpose of r4  r5  r6 r7.
+
+  Since r1 is free at entry, we calculate the Js first.*/
+/*19 cycles.*/
+#define OC_TRANSPOSE(_y) __asm{ \
+  __asm movq mm1,mm4 \
+  __asm punpcklwd mm4,mm5 \
+  __asm movq OC_I(0,_y),mm0 \
+  __asm punpckhwd mm1,mm5 \
+  __asm movq mm0,mm6 \
+  __asm punpcklwd mm6,mm7 \
+  __asm movq mm5,mm4 \
+  __asm punpckldq mm4,mm6 \
+  __asm punpckhdq mm5,mm6 \
+  __asm movq mm6,mm1 \
+  __asm movq OC_J(4,_y),mm4 \
+  __asm punpckhwd mm0,mm7 \
+  __asm movq OC_J(5,_y),mm5 \
+  __asm punpckhdq mm6,mm0 \
+  __asm movq mm4,OC_I(0,_y) \
+  __asm punpckldq mm1,mm0 \
+  __asm movq mm5,OC_I(1,_y) \
+  __asm movq mm0,mm4 \
+  __asm movq OC_J(7,_y),mm6 \
+  __asm punpcklwd mm0,mm5 \
+  __asm movq OC_J(6,_y),mm1 \
+  __asm punpckhwd mm4,mm5 \
+  __asm movq mm5,mm2 \
+  __asm punpcklwd mm2,mm3 \
+  __asm movq mm1,mm0 \
+  __asm punpckldq mm0,mm2 \
+  __asm punpckhdq mm1,mm2 \
+  __asm movq mm2,mm4 \
+  __asm movq OC_I(0,_y),mm0 \
+  __asm punpckhwd mm5,mm3 \
+  __asm movq OC_I(1,_y),mm1 \
+  __asm punpckhdq mm4,mm5 \
+  __asm punpckldq mm2,mm5 \
+  __asm movq OC_I(3,_y),mm4 \
+  __asm movq OC_I(2,_y),mm2 \
+}
+
+/*38+19=57 cycles.*/
+#define OC_COLUMN_IDCT(_y) __asm{ \
+  OC_IDCT_BEGIN(_y,_y) \
+  __asm paddw mm2,OC_8 \
+  /*r1=H'+H'*/ \
+  __asm paddw mm1,mm1 \
+  /*r1=R1=A''+H'*/ \
+  __asm paddw mm1,mm2 \
+  /*r2=NR2*/ \
+  __asm psraw mm2,4 \
+  /*r4=E'=E-G*/ \
+  __asm psubw mm4,mm7 \
+  /*r1=NR1*/ \
+  __asm psraw mm1,4 \
+  /*r3=D'*/ \
+  __asm movq mm3,OC_I(2,_y) \
+  /*r7=G+G*/ \
+  __asm paddw mm7,mm7 \
+  /*Store NR2 at I(2).*/ \
+  __asm movq OC_I(2,_y),mm2 \
+  /*r7=G'=E+G*/ \
+  __asm paddw mm7,mm4 \
+  /*Store NR1 at I(1).*/ \
+  __asm movq OC_I(1,_y),mm1 \
+  /*r4=R4=E'-D'*/ \
+  __asm psubw mm4,mm3 \
+  __asm paddw mm4,OC_8 \
+  /*r3=D'+D'*/ \
+  __asm paddw mm3,mm3 \
+  /*r3=R3=E'+D'*/ \
+  __asm paddw mm3,mm4 \
+  /*r4=NR4*/ \
+  __asm psraw mm4,4 \
+  /*r6=R6=F'-B''*/ \
+  __asm psubw mm6,mm5 \
+  /*r3=NR3*/ \
+  __asm psraw mm3,4 \
+  __asm paddw mm6,OC_8 \
+  /*r5=B''+B''*/ \
+  __asm paddw mm5,mm5 \
+  /*r5=R5=F'+B''*/ \
+  __asm paddw mm5,mm6 \
+  /*r6=NR6*/ \
+  __asm psraw mm6,4 \
+  /*Store NR4 at J(4).*/ \
+  __asm movq OC_J(4,_y),mm4 \
+  /*r5=NR5*/ \
+  __asm psraw mm5,4 \
+  /*Store NR3 at I(3).*/ \
+  __asm movq OC_I(3,_y),mm3 \
+  /*r7=R7=G'-C'*/ \
+  __asm psubw mm7,mm0 \
+  __asm paddw mm7,OC_8 \
+  /*r0=C'+C'*/ \
+  __asm paddw mm0,mm0 \
+  /*r0=R0=G'+C'*/ \
+  __asm paddw mm0,mm7 \
+  /*r7=NR7*/ \
+  __asm psraw mm7,4 \
+  /*Store NR6 at J(6).*/ \
+  __asm movq OC_J(6,_y),mm6 \
+  /*r0=NR0*/ \
+  __asm psraw mm0,4 \
+  /*Store NR5 at J(5).*/ \
+  __asm movq OC_J(5,_y),mm5 \
+  /*Store NR7 at J(7).*/ \
+  __asm movq OC_J(7,_y),mm7 \
+  /*Store NR0 at I(0).*/ \
+  __asm movq OC_I(0,_y),mm0 \
+}
+
+#define OC_MID(_m,_i) [CONSTS+_m+(_i)*8]
+#define OC_C(_i)      OC_MID(OC_COSINE_OFFSET,_i-1)
+#define OC_8          OC_MID(OC_EIGHT_OFFSET,0)
+
+static void oc_idct8x8_slow(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  int i;
+  /*This routine accepts an 8x8 matrix, but in partially transposed form.
+    Every 4x4 block is transposed.*/
+  __asm{
+#define CONSTS eax
+#define Y edx
+#define X ecx
+    mov CONSTS,offset OC_IDCT_CONSTS
+    mov Y,_y
+    mov X,_x
+#define OC_I(_k,_y)   [(_y)+(_k)*16]
+#define OC_J(_k,_y)   [(_y)+((_k)-4)*16+8]
+    OC_ROW_IDCT(Y,X)
+    OC_TRANSPOSE(Y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y)   [(_y)+(_k)*16+64]
+#define OC_J(_k,_y)   [(_y)+((_k)-4)*16+72]
+    OC_ROW_IDCT(Y,X)
+    OC_TRANSPOSE(Y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y)   [(_y)+(_k)*16]
+#define OC_J(_k,_y)   OC_I(_k,_y)
+    OC_COLUMN_IDCT(Y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y)   [(_y)+(_k)*16+8]
+#define OC_J(_k,_y)   OC_I(_k,_y)
+    OC_COLUMN_IDCT(Y)
+#undef  OC_I
+#undef  OC_J
+#undef  CONSTS
+#undef  Y
+#undef  X
+  }
+  __asm pxor mm0,mm0;
+  for(i=0;i<4;i++){
+    ogg_int16_t *x;
+    x=_x+16*i;
+#define X ecx
+    __asm{
+      mov X,x
+      movq [X+0x00],mm0
+      movq [X+0x08],mm0
+      movq [X+0x10],mm0
+      movq [X+0x18],mm0
+    }
+#undef  X
+  }
+}
+
+/*25 cycles.*/
+#define OC_IDCT_BEGIN_10(_y,_x) __asm{ \
+  __asm movq mm2,OC_I(3,_x) \
+  __asm nop \
+  __asm movq mm6,OC_C(3) \
+  __asm movq mm4,mm2 \
+  __asm movq mm1,OC_C(5) \
+  __asm pmulhw mm4,mm6 \
+  __asm movq mm3,OC_I(1,_x) \
+  __asm pmulhw mm1,mm2 \
+  __asm movq mm0,OC_C(1) \
+  __asm paddw mm4,mm2 \
+  __asm pxor mm6,mm6 \
+  __asm paddw mm2,mm1 \
+  __asm movq mm5,OC_I(2,_x) \
+  __asm pmulhw mm0,mm3 \
+  __asm movq mm1,mm5 \
+  __asm paddw mm0,mm3 \
+  __asm pmulhw mm3,OC_C(7) \
+  __asm psubw mm6,mm2 \
+  __asm pmulhw mm5,OC_C(2) \
+  __asm psubw mm0,mm4 \
+  __asm movq mm7,OC_I(2,_x) \
+  __asm paddw mm4,mm4 \
+  __asm paddw mm7,mm5 \
+  __asm paddw mm4,mm0 \
+  __asm pmulhw mm1,OC_C(6) \
+  __asm psubw mm3,mm6 \
+  __asm movq OC_I(1,_y),mm4 \
+  __asm paddw mm6,mm6 \
+  __asm movq mm4,OC_C(4) \
+  __asm paddw mm6,mm3 \
+  __asm movq mm5,mm3 \
+  __asm pmulhw mm3,mm4 \
+  __asm movq OC_I(2,_y),mm6 \
+  __asm movq mm2,mm0 \
+  __asm movq mm6,OC_I(0,_x) \
+  __asm pmulhw mm0,mm4 \
+  __asm paddw mm5,mm3 \
+  __asm paddw mm2,mm0 \
+  __asm psubw mm5,mm1 \
+  __asm pmulhw mm6,mm4 \
+  __asm paddw mm6,OC_I(0,_x) \
+  __asm paddw mm1,mm1 \
+  __asm movq mm4,mm6 \
+  __asm paddw mm1,mm5 \
+  __asm psubw mm6,mm2 \
+  __asm paddw mm2,mm2 \
+  __asm movq mm0,OC_I(1,_y) \
+  __asm paddw mm2,mm6 \
+  __asm psubw mm2,mm1 \
+  __asm nop \
+}
+
+/*25+8=33 cycles.*/
+#define OC_ROW_IDCT_10(_y,_x) __asm{ \
+  OC_IDCT_BEGIN_10(_y,_x) \
+  /*r3=D'*/ \
+   __asm movq mm3,OC_I(2,_y) \
+  /*r4=E'=E-G*/ \
+   __asm psubw mm4,mm7 \
+  /*r1=H'+H'*/ \
+   __asm paddw mm1,mm1 \
+  /*r7=G+G*/ \
+   __asm paddw mm7,mm7 \
+  /*r1=R1=A''+H'*/ \
+   __asm paddw mm1,mm2 \
+  /*r7=G'=E+G*/ \
+   __asm paddw mm7,mm4 \
+  /*r4=R4=E'-D'*/ \
+   __asm psubw mm4,mm3 \
+   __asm paddw mm3,mm3 \
+  /*r6=R6=F'-B''*/ \
+   __asm psubw mm6,mm5 \
+   __asm paddw mm5,mm5 \
+  /*r3=R3=E'+D'*/ \
+   __asm paddw mm3,mm4 \
+  /*r5=R5=F'+B''*/ \
+   __asm paddw mm5,mm6 \
+  /*r7=R7=G'-C'*/ \
+   __asm psubw mm7,mm0 \
+   __asm paddw mm0,mm0 \
+  /*Save R1.*/ \
+   __asm movq OC_I(1,_y),mm1 \
+  /*r0=R0=G'+C'*/ \
+   __asm paddw mm0,mm7 \
+}
+
+/*25+19=44 cycles'*/
+#define OC_COLUMN_IDCT_10(_y) __asm{ \
+  OC_IDCT_BEGIN_10(_y,_y) \
+  __asm paddw mm2,OC_8 \
+  /*r1=H'+H'*/ \
+  __asm paddw mm1,mm1 \
+  /*r1=R1=A''+H'*/ \
+  __asm paddw mm1,mm2 \
+  /*r2=NR2*/ \
+  __asm psraw mm2,4 \
+  /*r4=E'=E-G*/ \
+  __asm psubw mm4,mm7 \
+  /*r1=NR1*/ \
+  __asm psraw mm1,4 \
+  /*r3=D'*/ \
+  __asm movq mm3,OC_I(2,_y) \
+  /*r7=G+G*/ \
+  __asm paddw mm7,mm7 \
+  /*Store NR2 at I(2).*/ \
+  __asm movq OC_I(2,_y),mm2 \
+  /*r7=G'=E+G*/ \
+  __asm paddw mm7,mm4 \
+  /*Store NR1 at I(1).*/ \
+  __asm movq OC_I(1,_y),mm1 \
+  /*r4=R4=E'-D'*/ \
+  __asm psubw mm4,mm3 \
+  __asm paddw mm4,OC_8 \
+  /*r3=D'+D'*/ \
+  __asm paddw mm3,mm3 \
+  /*r3=R3=E'+D'*/ \
+  __asm paddw mm3,mm4 \
+  /*r4=NR4*/ \
+  __asm psraw mm4,4 \
+  /*r6=R6=F'-B''*/ \
+  __asm psubw mm6,mm5 \
+  /*r3=NR3*/ \
+  __asm psraw mm3,4 \
+  __asm paddw mm6,OC_8 \
+  /*r5=B''+B''*/ \
+  __asm paddw mm5,mm5 \
+  /*r5=R5=F'+B''*/ \
+  __asm paddw mm5,mm6 \
+  /*r6=NR6*/ \
+  __asm psraw mm6,4 \
+  /*Store NR4 at J(4).*/ \
+  __asm movq OC_J(4,_y),mm4 \
+  /*r5=NR5*/ \
+  __asm psraw mm5,4 \
+  /*Store NR3 at I(3).*/ \
+  __asm movq OC_I(3,_y),mm3 \
+  /*r7=R7=G'-C'*/ \
+  __asm psubw mm7,mm0 \
+  __asm paddw mm7,OC_8 \
+  /*r0=C'+C'*/ \
+  __asm paddw mm0,mm0 \
+  /*r0=R0=G'+C'*/ \
+  __asm paddw mm0,mm7 \
+  /*r7=NR7*/ \
+  __asm psraw mm7,4 \
+  /*Store NR6 at J(6).*/ \
+  __asm movq OC_J(6,_y),mm6 \
+  /*r0=NR0*/ \
+  __asm psraw mm0,4 \
+  /*Store NR5 at J(5).*/ \
+  __asm movq OC_J(5,_y),mm5 \
+  /*Store NR7 at J(7).*/ \
+  __asm movq OC_J(7,_y),mm7 \
+  /*Store NR0 at I(0).*/ \
+  __asm movq OC_I(0,_y),mm0 \
+}
+
+static void oc_idct8x8_10(ogg_int16_t _y[64],ogg_int16_t _x[64]){
+  __asm{
+#define CONSTS eax
+#define Y edx
+#define X ecx
+    mov CONSTS,offset OC_IDCT_CONSTS
+    mov Y,_y
+    mov X,_x
+#define OC_I(_k,_y) [(_y)+(_k)*16]
+#define OC_J(_k,_y) [(_y)+((_k)-4)*16+8]
+    /*Done with dequant, descramble, and partial transpose.
+      Now do the iDCT itself.*/
+    OC_ROW_IDCT_10(Y,X)
+    OC_TRANSPOSE(Y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y) [(_y)+(_k)*16]
+#define OC_J(_k,_y) OC_I(_k,_y)
+    OC_COLUMN_IDCT_10(Y)
+#undef  OC_I
+#undef  OC_J
+#define OC_I(_k,_y) [(_y)+(_k)*16+8]
+#define OC_J(_k,_y) OC_I(_k,_y)
+    OC_COLUMN_IDCT_10(Y)
+#undef  OC_I
+#undef  OC_J
+#undef  CONSTS
+#undef  Y
+#undef  X
+  }
+#define X ecx
+  __asm{
+    pxor mm0,mm0;
+    mov X,_x
+    movq [X+0x00],mm0
+    movq [X+0x10],mm0
+    movq [X+0x20],mm0
+    movq [X+0x30],mm0
+  }
+#undef  X
+}
+
+/*Performs an inverse 8x8 Type-II DCT transform.
+  The input is assumed to be scaled by a factor of 4 relative to orthonormal
+   version of the transform.*/
+void oc_idct8x8_mmx(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi){
+  /*_last_zzi is subtly different from an actual count of the number of
+     coefficients we decoded for this block.
+    It contains the value of zzi BEFORE the final token in the block was
+     decoded.
+    In most cases this is an EOB token (the continuation of an EOB run from a
+     previous block counts), and so this is the same as the coefficient count.
+    However, in the case that the last token was NOT an EOB token, but filled
+     the block up with exactly 64 coefficients, _last_zzi will be less than 64.
+    Provided the last token was not a pure zero run, the minimum value it can
+     be is 46, and so that doesn't affect any of the cases in this routine.
+    However, if the last token WAS a pure zero run of length 63, then _last_zzi
+     will be 1 while the number of coefficients decoded is 64.
+    Thus, we will trigger the following special case, where the real
+     coefficient count would not.
+    Note also that a zero run of length 64 will give _last_zzi a value of 0,
+     but we still process the DC coefficient, which might have a non-zero value
+     due to DC prediction.
+    Although convoluted, this is arguably the correct behavior: it allows us to
+     use a smaller transform when the block ends with a long zero run instead
+     of a normal EOB token.
+    It could be smarter... multiple separate zero runs at the end of a block
+     will fool it, but an encoder that generates these really deserves what it
+     gets.
+    Needless to say we inherited this approach from VP3.*/
+  /*Perform the iDCT.*/
+  if(_last_zzi<=10)oc_idct8x8_10(_y,_x);
+  else oc_idct8x8_slow(_y,_x);
+}
+
+#endif

lib/theoratest/lib/theora/x86_vc/mmxloop.h

@@ -0,0 +1,219 @@
+#if !defined(_x86_vc_mmxloop_H)
+# define _x86_vc_mmxloop_H (1)
+# include <stddef.h>
+# include "x86int.h"
+
+#if defined(OC_X86_ASM)
+
+/*On entry, mm0={a0,...,a7}, mm1={b0,...,b7}, mm2={c0,...,c7}, mm3={d0,...d7}.
+  On exit, mm1={b0+lflim(R_0,L),...,b7+lflim(R_7,L)} and
+   mm2={c0-lflim(R_0,L),...,c7-lflim(R_7,L)}; mm0 and mm3 are clobbered.*/
+#define OC_LOOP_FILTER8_MMX __asm{ \
+  /*mm7=0*/ \
+  __asm pxor mm7,mm7 \
+  /*mm6:mm0={a0,...,a7}*/ \
+  __asm movq mm6,mm0 \
+  __asm punpcklbw mm0,mm7 \
+  __asm punpckhbw mm6,mm7 \
+  /*mm3:mm5={d0,...,d7}*/ \
+  __asm movq mm5,mm3 \
+  __asm punpcklbw mm3,mm7 \
+  __asm punpckhbw mm5,mm7 \
+  /*mm6:mm0={a0-d0,...,a7-d7}*/ \
+  __asm psubw mm0,mm3 \
+  __asm psubw mm6,mm5 \
+  /*mm3:mm1={b0,...,b7}*/ \
+  __asm movq mm3,mm1 \
+  __asm punpcklbw mm1,mm7 \
+  __asm movq mm4,mm2 \
+  __asm punpckhbw mm3,mm7 \
+  /*mm5:mm4={c0,...,c7}*/ \
+  __asm movq mm5,mm2 \
+  __asm punpcklbw mm4,mm7 \
+  __asm punpckhbw mm5,mm7 \
+  /*mm7={3}x4 \
+    mm5:mm4={c0-b0,...,c7-b7}*/ \
+  __asm pcmpeqw mm7,mm7 \
+  __asm psubw mm4,mm1 \
+  __asm psrlw mm7,14 \
+  __asm psubw mm5,mm3 \
+  /*Scale by 3.*/ \
+  __asm pmullw mm4,mm7 \
+  __asm pmullw mm5,mm7 \
+  /*mm7={4}x4 \
+    mm5:mm4=f={a0-d0+3*(c0-b0),...,a7-d7+3*(c7-b7)}*/ \
+  __asm psrlw mm7,1 \
+  __asm paddw mm4,mm0 \
+  __asm psllw mm7,2 \
+  __asm movq mm0,[LL] \
+  __asm paddw mm5,mm6 \
+  /*R_i has the range [-127,128], so we compute -R_i instead. \
+    mm4=-R_i=-(f+4>>3)=0xFF^(f-4>>3)*/ \
+  __asm psubw mm4,mm7 \
+  __asm psubw mm5,mm7 \
+  __asm psraw mm4,3 \
+  __asm psraw mm5,3 \
+  __asm pcmpeqb mm7,mm7 \
+  __asm packsswb mm4,mm5 \
+  __asm pxor mm6,mm6 \
+  __asm pxor mm4,mm7 \
+  __asm packuswb mm1,mm3 \
+  /*Now compute lflim of -mm4 cf. Section 7.10 of the sepc.*/ \
+  /*There's no unsigned byte+signed byte with unsigned saturation op code, so \
+     we have to split things by sign (the other option is to work in 16 bits, \
+     but working in 8 bits gives much better parallelism). \
+    We compute abs(R_i), but save a mask of which terms were negative in mm6. \
+    Then we compute mm4=abs(lflim(R_i,L))=min(abs(R_i),max(2*L-abs(R_i),0)). \
+    Finally, we split mm4 into positive and negative pieces using the mask in \
+     mm6, and add and subtract them as appropriate.*/ \
+  /*mm4=abs(-R_i)*/ \
+  /*mm7=255-2*L*/ \
+  __asm pcmpgtb mm6,mm4 \
+  __asm psubb mm7,mm0 \
+  __asm pxor mm4,mm6 \
+  __asm psubb mm7,mm0 \
+  __asm psubb mm4,mm6 \
+  /*mm7=255-max(2*L-abs(R_i),0)*/ \
+  __asm paddusb mm7,mm4 \
+  /*mm4=min(abs(R_i),max(2*L-abs(R_i),0))*/ \
+  __asm paddusb mm4,mm7 \
+  __asm psubusb mm4,mm7 \
+  /*Now split mm4 by the original sign of -R_i.*/ \
+  __asm movq mm5,mm4 \
+  __asm pand mm4,mm6 \
+  __asm pandn mm6,mm5 \
+  /*mm1={b0+lflim(R_0,L),...,b7+lflim(R_7,L)}*/ \
+  /*mm2={c0-lflim(R_0,L),...,c7-lflim(R_7,L)}*/ \
+  __asm paddusb mm1,mm4 \
+  __asm psubusb mm2,mm4 \
+  __asm psubusb mm1,mm6 \
+  __asm paddusb mm2,mm6 \
+}
+
+#define OC_LOOP_FILTER_V_MMX(_pix,_ystride,_ll) \
+  do{ \
+    /*Used local variable pix__ in order to fix compilation errors like: \
+       "error C2425: 'SHL' : non-constant expression in 'second operand'".*/ \
+    unsigned char *pix__; \
+    unsigned char *ll__; \
+    ll__=(_ll); \
+    pix__=(_pix); \
+    __asm mov YSTRIDE,_ystride \
+    __asm mov LL,ll__ \
+    __asm mov PIX,pix__ \
+    __asm sub PIX,YSTRIDE \
+    __asm sub PIX,YSTRIDE \
+    /*mm0={a0,...,a7}*/ \
+    __asm movq mm0,[PIX] \
+    /*ystride3=_ystride*3*/ \
+    __asm lea YSTRIDE3,[YSTRIDE+YSTRIDE*2] \
+    /*mm3={d0,...,d7}*/ \
+    __asm movq mm3,[PIX+YSTRIDE3] \
+    /*mm1={b0,...,b7}*/ \
+    __asm movq mm1,[PIX+YSTRIDE] \
+    /*mm2={c0,...,c7}*/ \
+    __asm movq mm2,[PIX+YSTRIDE*2] \
+    OC_LOOP_FILTER8_MMX \
+    /*Write it back out.*/ \
+    __asm movq [PIX+YSTRIDE],mm1 \
+    __asm movq [PIX+YSTRIDE*2],mm2 \
+  } \
+  while(0)
+
+#define OC_LOOP_FILTER_H_MMX(_pix,_ystride,_ll) \
+  do{ \
+    /*Used local variable ll__ in order to fix compilation errors like: \
+       "error C2443: operand size conflict".*/ \
+    unsigned char *ll__; \
+    unsigned char *pix__; \
+    ll__=(_ll); \
+    pix__=(_pix)-2; \
+    __asm mov PIX,pix__ \
+    __asm mov YSTRIDE,_ystride \
+    __asm mov LL,ll__ \
+    /*x x x x d0 c0 b0 a0*/ \
+    __asm movd mm0,[PIX] \
+    /*x x x x d1 c1 b1 a1*/ \
+    __asm movd mm1,[PIX+YSTRIDE] \
+    /*ystride3=_ystride*3*/ \
+    __asm lea YSTRIDE3,[YSTRIDE+YSTRIDE*2] \
+    /*x x x x d2 c2 b2 a2*/ \
+    __asm movd mm2,[PIX+YSTRIDE*2] \
+    /*x x x x d3 c3 b3 a3*/ \
+    __asm lea D,[PIX+YSTRIDE*4] \
+    __asm movd mm3,[PIX+YSTRIDE3] \
+    /*x x x x d4 c4 b4 a4*/ \
+    __asm movd mm4,[D] \
+    /*x x x x d5 c5 b5 a5*/ \
+    __asm movd mm5,[D+YSTRIDE] \
+    /*x x x x d6 c6 b6 a6*/ \
+    __asm movd mm6,[D+YSTRIDE*2] \
+    /*x x x x d7 c7 b7 a7*/ \
+    __asm movd mm7,[D+YSTRIDE3] \
+    /*mm0=d1 d0 c1 c0 b1 b0 a1 a0*/ \
+    __asm punpcklbw mm0,mm1 \
+    /*mm2=d3 d2 c3 c2 b3 b2 a3 a2*/ \
+    __asm punpcklbw mm2,mm3 \
+    /*mm3=d1 d0 c1 c0 b1 b0 a1 a0*/ \
+    __asm movq mm3,mm0 \
+    /*mm0=b3 b2 b1 b0 a3 a2 a1 a0*/ \
+    __asm punpcklwd mm0,mm2 \
+    /*mm3=d3 d2 d1 d0 c3 c2 c1 c0*/ \
+    __asm punpckhwd mm3,mm2 \
+    /*mm1=b3 b2 b1 b0 a3 a2 a1 a0*/ \
+    __asm movq mm1,mm0 \
+    /*mm4=d5 d4 c5 c4 b5 b4 a5 a4*/ \
+    __asm punpcklbw mm4,mm5 \
+    /*mm6=d7 d6 c7 c6 b7 b6 a7 a6*/ \
+    __asm punpcklbw mm6,mm7 \
+    /*mm5=d5 d4 c5 c4 b5 b4 a5 a4*/ \
+    __asm movq mm5,mm4 \
+    /*mm4=b7 b6 b5 b4 a7 a6 a5 a4*/ \
+    __asm punpcklwd mm4,mm6 \
+    /*mm5=d7 d6 d5 d4 c7 c6 c5 c4*/ \
+    __asm punpckhwd mm5,mm6 \
+    /*mm2=d3 d2 d1 d0 c3 c2 c1 c0*/ \
+    __asm movq mm2,mm3 \
+    /*mm0=a7 a6 a5 a4 a3 a2 a1 a0*/ \
+    __asm punpckldq mm0,mm4 \
+    /*mm1=b7 b6 b5 b4 b3 b2 b1 b0*/ \
+    __asm punpckhdq mm1,mm4 \
+    /*mm2=c7 c6 c5 c4 c3 c2 c1 c0*/ \
+    __asm punpckldq mm2,mm5 \
+    /*mm3=d7 d6 d5 d4 d3 d2 d1 d0*/ \
+    __asm punpckhdq mm3,mm5 \
+    OC_LOOP_FILTER8_MMX \
+    /*mm2={b0+R_0'',...,b7+R_7''}*/ \
+    __asm movq mm0,mm1 \
+    /*mm1={b0+R_0'',c0-R_0'',...,b3+R_3'',c3-R_3''}*/ \
+    __asm punpcklbw mm1,mm2 \
+    /*mm2={b4+R_4'',c4-R_4'',...,b7+R_7'',c7-R_7''}*/ \
+    __asm punpckhbw mm0,mm2 \
+    /*[d]=c1 b1 c0 b0*/ \
+    __asm movd D,mm1 \
+    __asm mov [PIX+1],D_WORD \
+    __asm psrlq mm1,32 \
+    __asm shr D,16 \
+    __asm mov [PIX+YSTRIDE+1],D_WORD \
+    /*[d]=c3 b3 c2 b2*/ \
+    __asm movd D,mm1 \
+    __asm mov [PIX+YSTRIDE*2+1],D_WORD \
+    __asm shr D,16 \
+    __asm mov [PIX+YSTRIDE3+1],D_WORD \
+    __asm lea PIX,[PIX+YSTRIDE*4] \
+    /*[d]=c5 b5 c4 b4*/ \
+    __asm movd D,mm0 \
+    __asm mov [PIX+1],D_WORD \
+    __asm psrlq mm0,32 \
+    __asm shr D,16 \
+    __asm mov [PIX+YSTRIDE+1],D_WORD \
+    /*[d]=c7 b7 c6 b6*/ \
+    __asm movd D,mm0 \
+    __asm mov [PIX+YSTRIDE*2+1],D_WORD \
+    __asm shr D,16 \
+    __asm mov [PIX+YSTRIDE3+1],D_WORD \
+  } \
+  while(0)
+
+# endif
+#endif

lib/theoratest/lib/theora/x86_vc/mmxstate.c

@@ -0,0 +1,176 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+/*MMX acceleration of complete fragment reconstruction algorithm.
+  Originally written by Rudolf Marek.*/
+#include <string.h>
+#include "x86int.h"
+#include "mmxloop.h"
+
+#if defined(OC_X86_ASM)
+
+void oc_state_frag_recon_mmx(const oc_theora_state *_state,ptrdiff_t _fragi,
+ int _pli,ogg_int16_t _dct_coeffs[128],int _last_zzi,ogg_uint16_t _dc_quant){
+  unsigned char *dst;
+  ptrdiff_t      frag_buf_off;
+  int            ystride;
+  int            refi;
+  /*Apply the inverse transform.*/
+  /*Special case only having a DC component.*/
+  if(_last_zzi<2){
+    /*Note that this value must be unsigned, to keep the __asm__ block from
+       sign-extending it when it puts it in a register.*/
+    ogg_uint16_t p;
+    /*We round this dequant product (and not any of the others) because there's
+       no iDCT rounding.*/
+    p=(ogg_int16_t)(_dct_coeffs[0]*(ogg_int32_t)_dc_quant+15>>5);
+    /*Fill _dct_coeffs with p.*/
+    __asm{
+#define Y eax
+#define P ecx
+      mov Y,_dct_coeffs
+      movzx P,p
+      lea Y,[Y+128]
+      /*mm0=0000 0000 0000 AAAA*/
+      movd mm0,P
+      /*mm0=0000 0000 AAAA AAAA*/
+      punpcklwd mm0,mm0
+      /*mm0=AAAA AAAA AAAA AAAA*/
+      punpckldq mm0,mm0
+      movq [Y],mm0
+      movq [8+Y],mm0
+      movq [16+Y],mm0
+      movq [24+Y],mm0
+      movq [32+Y],mm0
+      movq [40+Y],mm0
+      movq [48+Y],mm0
+      movq [56+Y],mm0
+      movq [64+Y],mm0
+      movq [72+Y],mm0
+      movq [80+Y],mm0
+      movq [88+Y],mm0
+      movq [96+Y],mm0
+      movq [104+Y],mm0
+      movq [112+Y],mm0
+      movq [120+Y],mm0
+#undef Y
+#undef P
+    }
+  }
+  else{
+    /*Dequantize the DC coefficient.*/
+    _dct_coeffs[0]=(ogg_int16_t)(_dct_coeffs[0]*(int)_dc_quant);
+    oc_idct8x8_mmx(_dct_coeffs+64,_dct_coeffs,_last_zzi);
+  }
+  /*Fill in the target buffer.*/
+  frag_buf_off=_state->frag_buf_offs[_fragi];
+  refi=_state->frags[_fragi].refi;
+  ystride=_state->ref_ystride[_pli];
+  dst=_state->ref_frame_data[OC_FRAME_SELF]+frag_buf_off;
+  if(refi==OC_FRAME_SELF)oc_frag_recon_intra_mmx(dst,ystride,_dct_coeffs+64);
+  else{
+    const unsigned char *ref;
+    int                  mvoffsets[2];
+    ref=_state->ref_frame_data[refi]+frag_buf_off;
+    if(oc_state_get_mv_offsets(_state,mvoffsets,_pli,
+     _state->frag_mvs[_fragi])>1){
+      oc_frag_recon_inter2_mmx(dst,ref+mvoffsets[0],ref+mvoffsets[1],ystride,
+       _dct_coeffs+64);
+    }
+    else oc_frag_recon_inter_mmx(dst,ref+mvoffsets[0],ystride,_dct_coeffs+64);
+  }
+}
+
+/*We copy these entire function to inline the actual MMX routines so that we
+   use only a single indirect call.*/
+
+void oc_loop_filter_init_mmx(signed char _bv[256],int _flimit){
+  memset(_bv,~(_flimit<<1),8);
+}
+
+/*Apply the loop filter to a given set of fragment rows in the given plane.
+  The filter may be run on the bottom edge, affecting pixels in the next row of
+   fragments, so this row also needs to be available.
+  _bv:        The bounding values array.
+  _refi:      The index of the frame buffer to filter.
+  _pli:       The color plane to filter.
+  _fragy0:    The Y coordinate of the first fragment row to filter.
+  _fragy_end: The Y coordinate of the fragment row to stop filtering at.*/
+void oc_state_loop_filter_frag_rows_mmx(const oc_theora_state *_state,
+ signed char _bv[256],int _refi,int _pli,int _fragy0,int _fragy_end){
+  const oc_fragment_plane *fplane;
+  const oc_fragment       *frags;
+  const ptrdiff_t         *frag_buf_offs;
+  unsigned char           *ref_frame_data;
+  ptrdiff_t                fragi_top;
+  ptrdiff_t                fragi_bot;
+  ptrdiff_t                fragi0;
+  ptrdiff_t                fragi0_end;
+  int                      ystride;
+  int                      nhfrags;
+  fplane=_state->fplanes+_pli;
+  nhfrags=fplane->nhfrags;
+  fragi_top=fplane->froffset;
+  fragi_bot=fragi_top+fplane->nfrags;
+  fragi0=fragi_top+_fragy0*(ptrdiff_t)nhfrags;
+  fragi0_end=fragi_top+_fragy_end*(ptrdiff_t)nhfrags;
+  ystride=_state->ref_ystride[_pli];
+  frags=_state->frags;
+  frag_buf_offs=_state->frag_buf_offs;
+  ref_frame_data=_state->ref_frame_data[_refi];
+  /*The following loops are constructed somewhat non-intuitively on purpose.
+    The main idea is: if a block boundary has at least one coded fragment on
+     it, the filter is applied to it.
+    However, the order that the filters are applied in matters, and VP3 chose
+     the somewhat strange ordering used below.*/
+  while(fragi0<fragi0_end){
+    ptrdiff_t fragi;
+    ptrdiff_t fragi_end;
+    fragi=fragi0;
+    fragi_end=fragi+nhfrags;
+    while(fragi<fragi_end){
+      if(frags[fragi].coded){
+        unsigned char *ref;
+        ref=ref_frame_data+frag_buf_offs[fragi];
+#define PIX eax
+#define YSTRIDE3 edi
+#define YSTRIDE ecx
+#define LL edx
+#define D esi
+#define D_WORD si
+        if(fragi>fragi0)OC_LOOP_FILTER_H_MMX(ref,ystride,_bv);
+        if(fragi0>fragi_top)OC_LOOP_FILTER_V_MMX(ref,ystride,_bv);
+        if(fragi+1<fragi_end&&!frags[fragi+1].coded){
+          OC_LOOP_FILTER_H_MMX(ref+8,ystride,_bv);
+        }
+        if(fragi+nhfrags<fragi_bot&&!frags[fragi+nhfrags].coded){
+          OC_LOOP_FILTER_V_MMX(ref+(ystride<<3),ystride,_bv);
+        }
+#undef PIX
+#undef YSTRIDE3
+#undef YSTRIDE
+#undef LL
+#undef D
+#undef D_WORD
+      }
+      fragi++;
+    }
+    fragi0+=nhfrags;
+  }
+}
+
+#endif

lib/theoratest/lib/theora/x86_vc/x86cpu.c

@@ -0,0 +1,192 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+ CPU capability detection for x86 processors.
+  Originally written by Rudolf Marek.
+
+ function:
+  last mod: $Id$
+
+ ********************************************************************/
+
+#include "x86cpu.h"
+
+#if !defined(OC_X86_ASM)
+ogg_uint32_t oc_cpu_flags_get(void){
+  return 0;
+}
+#else
+/*Why does MSVC need this complicated rigamarole?
+  At this point I honestly do not care.*/
+
+/*Visual C cpuid helper function.
+  For VS2005 we could as well use the _cpuid builtin, but that wouldn't work
+   for VS2003 users, so we do it in inline assembler.*/
+static void oc_cpuid_helper(ogg_uint32_t _cpu_info[4],ogg_uint32_t _op){
+  _asm{
+    mov eax,[_op]
+    mov esi,_cpu_info
+    cpuid
+    mov [esi+0],eax
+    mov [esi+4],ebx
+    mov [esi+8],ecx
+    mov [esi+12],edx
+  }
+}
+
+#  define cpuid(_op,_eax,_ebx,_ecx,_edx) \
+  do{ \
+    ogg_uint32_t cpu_info[4]; \
+    oc_cpuid_helper(cpu_info,_op); \
+    (_eax)=cpu_info[0]; \
+    (_ebx)=cpu_info[1]; \
+    (_ecx)=cpu_info[2]; \
+    (_edx)=cpu_info[3]; \
+  }while(0)
+
+static void oc_detect_cpuid_helper(ogg_uint32_t *_eax,ogg_uint32_t *_ebx){
+  _asm{
+    pushfd
+    pushfd
+    pop eax
+    mov ebx,eax
+    xor eax,200000h
+    push eax
+    popfd
+    pushfd
+    pop eax
+    popfd
+    mov ecx,_eax
+    mov [ecx],eax
+    mov ecx,_ebx
+    mov [ecx],ebx
+  }
+}
+
+static ogg_uint32_t oc_parse_intel_flags(ogg_uint32_t _edx,ogg_uint32_t _ecx){
+  ogg_uint32_t flags;
+  /*If there isn't even MMX, give up.*/
+  if(!(_edx&0x00800000))return 0;
+  flags=OC_CPU_X86_MMX;
+  if(_edx&0x02000000)flags|=OC_CPU_X86_MMXEXT|OC_CPU_X86_SSE;
+  if(_edx&0x04000000)flags|=OC_CPU_X86_SSE2;
+  if(_ecx&0x00000001)flags|=OC_CPU_X86_PNI;
+  if(_ecx&0x00000100)flags|=OC_CPU_X86_SSSE3;
+  if(_ecx&0x00080000)flags|=OC_CPU_X86_SSE4_1;
+  if(_ecx&0x00100000)flags|=OC_CPU_X86_SSE4_2;
+  return flags;
+}
+
+static ogg_uint32_t oc_parse_amd_flags(ogg_uint32_t _edx,ogg_uint32_t _ecx){
+  ogg_uint32_t flags;
+  /*If there isn't even MMX, give up.*/
+  if(!(_edx&0x00800000))return 0;
+  flags=OC_CPU_X86_MMX;
+  if(_edx&0x00400000)flags|=OC_CPU_X86_MMXEXT;
+  if(_edx&0x80000000)flags|=OC_CPU_X86_3DNOW;
+  if(_edx&0x40000000)flags|=OC_CPU_X86_3DNOWEXT;
+  if(_ecx&0x00000040)flags|=OC_CPU_X86_SSE4A;
+  if(_ecx&0x00000800)flags|=OC_CPU_X86_SSE5;
+  return flags;
+}
+
+ogg_uint32_t oc_cpu_flags_get(void){
+  ogg_uint32_t flags;
+  ogg_uint32_t eax;
+  ogg_uint32_t ebx;
+  ogg_uint32_t ecx;
+  ogg_uint32_t edx;
+# if !defined(__amd64__)&&!defined(__x86_64__)
+  /*Not all x86-32 chips support cpuid, so we have to check.*/
+  oc_detect_cpuid_helper(&eax,&ebx);
+  /*No cpuid.*/
+  if(eax==ebx)return 0;
+# endif
+  cpuid(0,eax,ebx,ecx,edx);
+  /*         l e t n          I e n i          u n e G*/
+  if(ecx==0x6C65746E&&edx==0x49656E69&&ebx==0x756E6547||
+   /*      6 8 x M          T e n i          u n e G*/
+   ecx==0x3638784D&&edx==0x54656E69&&ebx==0x756E6547){
+    int family;
+    int model;
+    /*Intel, Transmeta (tested with Crusoe TM5800):*/
+    cpuid(1,eax,ebx,ecx,edx);
+    flags=oc_parse_intel_flags(edx,ecx);
+    family=(eax>>8)&0xF;
+    model=(eax>>4)&0xF;
+    /*The SSE unit on the Pentium M and Core Duo is much slower than the MMX
+       unit, so don't use it.*/
+    if(family==6&&(model==9||model==13||model==14)){
+      flags&=~(OC_CPU_X86_SSE2|OC_CPU_X86_PNI);
+    }
+  }
+  /*              D M A c          i t n e          h t u A*/
+  else if(ecx==0x444D4163&&edx==0x69746E65&&ebx==0x68747541||
+   /*      C S N            y b   e          d o e G*/
+   ecx==0x43534e20&&edx==0x79622065&&ebx==0x646f6547){
+    /*AMD, Geode:*/
+    cpuid(0x80000000,eax,ebx,ecx,edx);
+    if(eax<0x80000001)flags=0;
+    else{
+      cpuid(0x80000001,eax,ebx,ecx,edx);
+      flags=oc_parse_amd_flags(edx,ecx);
+    }
+    /*Also check for SSE.*/
+    cpuid(1,eax,ebx,ecx,edx);
+    flags|=oc_parse_intel_flags(edx,ecx);
+  }
+  /*Technically some VIA chips can be configured in the BIOS to return any
+     string here the user wants.
+    There is a special detection method that can be used to identify such
+     processors, but in my opinion, if the user really wants to change it, they
+     deserve what they get.*/
+  /*              s l u a          H r u a          t n e C*/
+  else if(ecx==0x736C7561&&edx==0x48727561&&ebx==0x746E6543){
+    /*VIA:*/
+    /*I only have documentation for the C7 (Esther) and Isaiah (forthcoming)
+       chips (thanks to the engineers from Centaur Technology who provided it).
+      These chips support Intel-like cpuid info.
+      The C3-2 (Nehemiah) cores appear to, as well.*/
+    cpuid(1,eax,ebx,ecx,edx);
+    flags=oc_parse_intel_flags(edx,ecx);
+    if(eax>=0x80000001){
+      /*The (non-Nehemiah) C3 processors support AMD-like cpuid info.
+        We need to check this even if the Intel test succeeds to pick up 3DNow!
+         support on these processors.
+        Unlike actual AMD processors, we cannot _rely_ on this info, since
+         some cores (e.g., the 693 stepping of the Nehemiah) claim to support
+         this function, yet return edx=0, despite the Intel test indicating
+         MMX support.
+        Therefore the features detected here are strictly added to those
+         detected by the Intel test.*/
+      /*TODO: How about earlier chips?*/
+      cpuid(0x80000001,eax,ebx,ecx,edx);
+      /*Note: As of the C7, this function returns Intel-style extended feature
+         flags, not AMD-style.
+        Currently, this only defines bits 11, 20, and 29 (0x20100800), which
+         do not conflict with any of the AMD flags we inspect.
+        For the remaining bits, Intel tells us, "Do not count on their value",
+         but VIA assures us that they will all be zero (at least on the C7 and
+         Isaiah chips).
+        In the (unlikely) event a future processor uses bits 18, 19, 30, or 31
+         (0xC0C00000) for something else, we will have to add code to detect
+         the model to decide when it is appropriate to inspect them.*/
+      flags|=oc_parse_amd_flags(edx,ecx);
+    }
+  }
+  else{
+    /*Implement me.*/
+    flags=0;
+  }
+  return flags;
+}
+#endif

lib/theoratest/lib/theora/x86_vc/x86cpu.h

@@ -0,0 +1,36 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+ function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#if !defined(_x86_vc_x86cpu_H)
+# define _x86_vc_x86cpu_H (1)
+#include "../internal.h"
+
+#define OC_CPU_X86_MMX      (1<<0)
+#define OC_CPU_X86_3DNOW    (1<<1)
+#define OC_CPU_X86_3DNOWEXT (1<<2)
+#define OC_CPU_X86_MMXEXT   (1<<3)
+#define OC_CPU_X86_SSE      (1<<4)
+#define OC_CPU_X86_SSE2     (1<<5)
+#define OC_CPU_X86_PNI      (1<<6)
+#define OC_CPU_X86_SSSE3    (1<<7)
+#define OC_CPU_X86_SSE4_1   (1<<8)
+#define OC_CPU_X86_SSE4_2   (1<<9)
+#define OC_CPU_X86_SSE4A    (1<<10)
+#define OC_CPU_X86_SSE5     (1<<11)
+
+ogg_uint32_t oc_cpu_flags_get(void);
+
+#endif

lib/theoratest/lib/theora/x86_vc/x86int.h

@@ -0,0 +1,49 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#if !defined(_x86_vc_x86int_H)
+# define _x86_vc_x86int_H (1)
+# include "../internal.h"
+# if defined(OC_X86_ASM)
+#  define oc_state_accel_init oc_state_accel_init_x86
+#  define OC_STATE_USE_VTABLE (1)
+# endif
+# include "../state.h"
+# include "x86cpu.h"
+
+void oc_state_accel_init_x86(oc_theora_state *_state);
+
+void oc_frag_copy_mmx(unsigned char *_dst,
+ const unsigned char *_src,int _ystride);
+void oc_frag_copy_list_mmx(unsigned char *_dst_frame,
+ const unsigned char *_src_frame,int _ystride,
+ const ptrdiff_t *_fragis,ptrdiff_t _nfragis,const ptrdiff_t *_frag_buf_offs);
+void oc_frag_recon_intra_mmx(unsigned char *_dst,int _ystride,
+ const ogg_int16_t *_residue);
+void oc_frag_recon_inter_mmx(unsigned char *_dst,
+ const unsigned char *_src,int _ystride,const ogg_int16_t *_residue);
+void oc_frag_recon_inter2_mmx(unsigned char *_dst,const unsigned char *_src1,
+ const unsigned char *_src2,int _ystride,const ogg_int16_t *_residue);
+void oc_idct8x8_mmx(ogg_int16_t _y[64],ogg_int16_t _x[64],int _last_zzi);
+void oc_state_frag_recon_mmx(const oc_theora_state *_state,ptrdiff_t _fragi,
+ int _pli,ogg_int16_t _dct_coeffs[128],int _last_zzi,ogg_uint16_t _dc_quant);
+void oc_loop_filter_init_mmx(signed char _bv[256],int _flimit);
+void oc_state_loop_filter_frag_rows_mmx(const oc_theora_state *_state,
+ signed char _bv[256],int _refi,int _pli,int _fragy0,int _fragy_end);
+void oc_restore_fpu_mmx(void);
+
+#endif

lib/theoratest/lib/theora/x86_vc/x86state.c

@@ -0,0 +1,61 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2009                *
+ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
+ *                                                                  *
+ ********************************************************************
+
+  function:
+    last mod: $Id$
+
+ ********************************************************************/
+
+#include "x86int.h"
+
+#if defined(OC_X86_ASM)
+
+/*This table has been modified from OC_FZIG_ZAG by baking a 4x4 transpose into
+   each quadrant of the destination.*/
+static const unsigned char OC_FZIG_ZAG_MMX[128]={
+   0, 8, 1, 2, 9,16,24,17,
+  10, 3,32,11,18,25, 4,12,
+   5,26,19,40,33,34,41,48,
+  27, 6,13,20,28,21,14, 7,
+  56,49,42,35,43,50,57,36,
+  15,22,29,30,23,44,37,58,
+  51,59,38,45,52,31,60,53,
+  46,39,47,54,61,62,55,63,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+  64,64,64,64,64,64,64,64,
+};
+
+void oc_state_accel_init_x86(oc_theora_state *_state){
+  _state->cpu_flags=oc_cpu_flags_get();
+  if(_state->cpu_flags&OC_CPU_X86_MMX){
+    _state->opt_vtable.frag_copy=oc_frag_copy_mmx;
+    _state->opt_vtable.frag_copy_list=oc_frag_copy_list_mmx;
+    _state->opt_vtable.frag_recon_intra=oc_frag_recon_intra_mmx;
+    _state->opt_vtable.frag_recon_inter=oc_frag_recon_inter_mmx;
+    _state->opt_vtable.frag_recon_inter2=oc_frag_recon_inter2_mmx;
+    _state->opt_vtable.idct8x8=oc_idct8x8_mmx;
+    _state->opt_vtable.state_frag_recon=oc_state_frag_recon_mmx;
+    _state->opt_vtable.loop_filter_init=oc_loop_filter_init_mmx;
+    _state->opt_vtable.state_loop_filter_frag_rows=
+     oc_state_loop_filter_frag_rows_mmx;
+    _state->opt_vtable.restore_fpu=oc_restore_fpu_mmx;
+    _state->opt_data.dct_fzig_zag=OC_FZIG_ZAG_MMX;
+  }
+  else oc_state_accel_init_c(_state);
+}
+#endif

lib/theoratest/lib/vorbis/analysis.c

@@ -0,0 +1,119 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2007             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: single-block PCM analysis mode dispatch
+
+ ********************************************************************/
+
+#include <stdio.h>
+#include <string.h>
+#include <math.h>
+#include <ogg/ogg.h>
+#include "vorbis/codec.h"
+#include "codec_internal.h"
+#include "registry.h"
+#include "scales.h"
+#include "os.h"
+#include "misc.h"
+
+/* decides between modes, dispatches to the appropriate mapping. */
+int vorbis_analysis(vorbis_block *vb, ogg_packet *op){
+  int ret,i;
+  vorbis_block_internal *vbi=vb->internal;
+
+  vb->glue_bits=0;
+  vb->time_bits=0;
+  vb->floor_bits=0;
+  vb->res_bits=0;
+
+  /* first things first.  Make sure encode is ready */
+  for(i=0;i<PACKETBLOBS;i++)
+    oggpack_reset(vbi->packetblob[i]);
+
+  /* we only have one mapping type (0), and we let the mapping code
+     itself figure out what soft mode to use.  This allows easier
+     bitrate management */
+
+  if((ret=_mapping_P[0]->forward(vb)))
+    return(ret);
+
+  if(op){
+    if(vorbis_bitrate_managed(vb))
+      /* The app is using a bitmanaged mode... but not using the
+         bitrate management interface. */
+      return(OV_EINVAL);
+
+    op->packet=oggpack_get_buffer(&vb->opb);
+    op->bytes=oggpack_bytes(&vb->opb);
+    op->b_o_s=0;
+    op->e_o_s=vb->eofflag;
+    op->granulepos=vb->granulepos;
+    op->packetno=vb->sequence; /* for sake of completeness */
+  }
+  return(0);
+}
+
+#ifdef ANALYSIS
+int analysis_noisy=1;
+
+/* there was no great place to put this.... */
+void _analysis_output_always(char *base,int i,float *v,int n,int bark,int dB,ogg_int64_t off){
+  int j;
+  FILE *of;
+  char buffer[80];
+
+  sprintf(buffer,"%s_%d.m",base,i);
+  of=fopen(buffer,"w");
+
+  if(!of)perror("failed to open data dump file");
+
+  for(j=0;j<n;j++){
+    if(bark){
+      float b=toBARK((4000.f*j/n)+.25);
+      fprintf(of,"%f ",b);
+    }else
+      if(off!=0)
+        fprintf(of,"%f ",(double)(j+off)/8000.);
+      else
+        fprintf(of,"%f ",(double)j);
+
+    if(dB){
+      float val;
+      if(v[j]==0.)
+        val=-140.;
+      else
+        val=todB(v+j);
+      fprintf(of,"%f\n",val);
+    }else{
+      fprintf(of,"%f\n",v[j]);
+    }
+  }
+  fclose(of);
+}
+
+void _analysis_output(char *base,int i,float *v,int n,int bark,int dB,
+                      ogg_int64_t off){
+  if(analysis_noisy)_analysis_output_always(base,i,v,n,bark,dB,off);
+}
+
+#endif
+
+
+
+
+
+
+
+
+
+
+

lib/theoratest/lib/vorbis/backends.h

@@ -0,0 +1,143 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: libvorbis backend and mapping structures; needed for
+           static mode headers
+
+ ********************************************************************/
+
+/* this is exposed up here because we need it for static modes.
+   Lookups for each backend aren't exposed because there's no reason
+   to do so */
+
+#ifndef _vorbis_backend_h_
+#define _vorbis_backend_h_
+
+#include "codec_internal.h"
+
+/* this would all be simpler/shorter with templates, but.... */
+/* Floor backend generic *****************************************/
+typedef struct{
+  void                   (*pack)  (vorbis_info_floor *,oggpack_buffer *);
+  vorbis_info_floor     *(*unpack)(vorbis_info *,oggpack_buffer *);
+  vorbis_look_floor     *(*look)  (vorbis_dsp_state *,vorbis_info_floor *);
+  void (*free_info) (vorbis_info_floor *);
+  void (*free_look) (vorbis_look_floor *);
+  void *(*inverse1)  (struct vorbis_block *,vorbis_look_floor *);
+  int   (*inverse2)  (struct vorbis_block *,vorbis_look_floor *,
+                     void *buffer,float *);
+} vorbis_func_floor;
+
+typedef struct{
+  int   order;
+  long  rate;
+  long  barkmap;
+
+  int   ampbits;
+  int   ampdB;
+
+  int   numbooks; /* <= 16 */
+  int   books[16];
+
+  float lessthan;     /* encode-only config setting hacks for libvorbis */
+  float greaterthan;  /* encode-only config setting hacks for libvorbis */
+
+} vorbis_info_floor0;
+
+
+#define VIF_POSIT 63
+#define VIF_CLASS 16
+#define VIF_PARTS 31
+typedef struct{
+  int   partitions;                /* 0 to 31 */
+  int   partitionclass[VIF_PARTS]; /* 0 to 15 */
+
+  int   class_dim[VIF_CLASS];        /* 1 to 8 */
+  int   class_subs[VIF_CLASS];       /* 0,1,2,3 (bits: 1<<n poss) */
+  int   class_book[VIF_CLASS];       /* subs ^ dim entries */
+  int   class_subbook[VIF_CLASS][8]; /* [VIF_CLASS][subs] */
+
+
+  int   mult;                      /* 1 2 3 or 4 */
+  int   postlist[VIF_POSIT+2];    /* first two implicit */
+
+
+  /* encode side analysis parameters */
+  float maxover;
+  float maxunder;
+  float maxerr;
+
+  float twofitweight;
+  float twofitatten;
+
+  int   n;
+
+} vorbis_info_floor1;
+
+/* Residue backend generic *****************************************/
+typedef struct{
+  void                 (*pack)  (vorbis_info_residue *,oggpack_buffer *);
+  vorbis_info_residue *(*unpack)(vorbis_info *,oggpack_buffer *);
+  vorbis_look_residue *(*look)  (vorbis_dsp_state *,
+                                 vorbis_info_residue *);
+  void (*free_info)    (vorbis_info_residue *);
+  void (*free_look)    (vorbis_look_residue *);
+  long **(*class)      (struct vorbis_block *,vorbis_look_residue *,
+                        int **,int *,int);
+  int  (*forward)      (oggpack_buffer *,struct vorbis_block *,
+                        vorbis_look_residue *,
+                        int **,int *,int,long **,int);
+  int  (*inverse)      (struct vorbis_block *,vorbis_look_residue *,
+                        float **,int *,int);
+} vorbis_func_residue;
+
+typedef struct vorbis_info_residue0{
+/* block-partitioned VQ coded straight residue */
+  long  begin;
+  long  end;
+
+  /* first stage (lossless partitioning) */
+  int    grouping;         /* group n vectors per partition */
+  int    partitions;       /* possible codebooks for a partition */
+  int    partvals;         /* partitions ^ groupbook dim */
+  int    groupbook;        /* huffbook for partitioning */
+  int    secondstages[64]; /* expanded out to pointers in lookup */
+  int    booklist[512];    /* list of second stage books */
+
+  const int classmetric1[64];
+  const int classmetric2[64];
+} vorbis_info_residue0;
+
+/* Mapping backend generic *****************************************/
+typedef struct{
+  void                 (*pack)  (vorbis_info *,vorbis_info_mapping *,
+                                 oggpack_buffer *);
+  vorbis_info_mapping *(*unpack)(vorbis_info *,oggpack_buffer *);
+  void (*free_info)    (vorbis_info_mapping *);
+  int  (*forward)      (struct vorbis_block *vb);
+  int  (*inverse)      (struct vorbis_block *vb,vorbis_info_mapping *);
+} vorbis_func_mapping;
+
+typedef struct vorbis_info_mapping0{
+  int   submaps;  /* <= 16 */
+  int   chmuxlist[256];   /* up to 256 channels in a Vorbis stream */
+
+  int   floorsubmap[16];   /* [mux] submap to floors */
+  int   residuesubmap[16]; /* [mux] submap to residue */
+
+  int   coupling_steps;
+  int   coupling_mag[256];
+  int   coupling_ang[256];
+
+} vorbis_info_mapping0;
+
+#endif

lib/theoratest/lib/vorbis/bitrate.c

@@ -0,0 +1,252 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: bitrate tracking and management
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+#include <ogg/ogg.h>
+#include "vorbis/codec.h"
+#include "codec_internal.h"
+#include "os.h"
+#include "misc.h"
+#include "bitrate.h"
+
+/* compute bitrate tracking setup  */
+void vorbis_bitrate_init(vorbis_info *vi,bitrate_manager_state *bm){
+  codec_setup_info *ci=vi->codec_setup;
+  bitrate_manager_info *bi=&ci->bi;
+
+  memset(bm,0,sizeof(*bm));
+
+  if(bi && (bi->reservoir_bits>0)){
+    long ratesamples=vi->rate;
+    int  halfsamples=ci->blocksizes[0]>>1;
+
+    bm->short_per_long=ci->blocksizes[1]/ci->blocksizes[0];
+    bm->managed=1;
+
+    bm->avg_bitsper= rint(1.*bi->avg_rate*halfsamples/ratesamples);
+    bm->min_bitsper= rint(1.*bi->min_rate*halfsamples/ratesamples);
+    bm->max_bitsper= rint(1.*bi->max_rate*halfsamples/ratesamples);
+
+    bm->avgfloat=PACKETBLOBS/2;
+
+    /* not a necessary fix, but one that leads to a more balanced
+       typical initialization */
+    {
+      long desired_fill=bi->reservoir_bits*bi->reservoir_bias;
+      bm->minmax_reservoir=desired_fill;
+      bm->avg_reservoir=desired_fill;
+    }
+
+  }
+}
+
+void vorbis_bitrate_clear(bitrate_manager_state *bm){
+  memset(bm,0,sizeof(*bm));
+  return;
+}
+
+int vorbis_bitrate_managed(vorbis_block *vb){
+  vorbis_dsp_state      *vd=vb->vd;
+  private_state         *b=vd->backend_state;
+  bitrate_manager_state *bm=&b->bms;
+
+  if(bm && bm->managed)return(1);
+  return(0);
+}
+
+/* finish taking in the block we just processed */
+int vorbis_bitrate_addblock(vorbis_block *vb){
+  vorbis_block_internal *vbi=vb->internal;
+  vorbis_dsp_state      *vd=vb->vd;
+  private_state         *b=vd->backend_state;
+  bitrate_manager_state *bm=&b->bms;
+  vorbis_info           *vi=vd->vi;
+  codec_setup_info      *ci=vi->codec_setup;
+  bitrate_manager_info  *bi=&ci->bi;
+
+  int  choice=rint(bm->avgfloat);
+  long this_bits=oggpack_bytes(vbi->packetblob[choice])*8;
+  long min_target_bits=(vb->W?bm->min_bitsper*bm->short_per_long:bm->min_bitsper);
+  long max_target_bits=(vb->W?bm->max_bitsper*bm->short_per_long:bm->max_bitsper);
+  int  samples=ci->blocksizes[vb->W]>>1;
+  long desired_fill=bi->reservoir_bits*bi->reservoir_bias;
+  if(!bm->managed){
+    /* not a bitrate managed stream, but for API simplicity, we'll
+       buffer the packet to keep the code path clean */
+
+    if(bm->vb)return(-1); /* one has been submitted without
+                             being claimed */
+    bm->vb=vb;
+    return(0);
+  }
+
+  bm->vb=vb;
+
+  /* look ahead for avg floater */
+  if(bm->avg_bitsper>0){
+    double slew=0.;
+    long avg_target_bits=(vb->W?bm->avg_bitsper*bm->short_per_long:bm->avg_bitsper);
+    double slewlimit= 15./bi->slew_damp;
+
+    /* choosing a new floater:
+       if we're over target, we slew down
+       if we're under target, we slew up
+
+       choose slew as follows: look through packetblobs of this frame
+       and set slew as the first in the appropriate direction that
+       gives us the slew we want.  This may mean no slew if delta is
+       already favorable.
+
+       Then limit slew to slew max */
+
+    if(bm->avg_reservoir+(this_bits-avg_target_bits)>desired_fill){
+      while(choice>0 && this_bits>avg_target_bits &&
+            bm->avg_reservoir+(this_bits-avg_target_bits)>desired_fill){
+        choice--;
+        this_bits=oggpack_bytes(vbi->packetblob[choice])*8;
+      }
+    }else if(bm->avg_reservoir+(this_bits-avg_target_bits)<desired_fill){
+      while(choice+1<PACKETBLOBS && this_bits<avg_target_bits &&
+            bm->avg_reservoir+(this_bits-avg_target_bits)<desired_fill){
+        choice++;
+        this_bits=oggpack_bytes(vbi->packetblob[choice])*8;
+      }
+    }
+
+    slew=rint(choice-bm->avgfloat)/samples*vi->rate;
+    if(slew<-slewlimit)slew=-slewlimit;
+    if(slew>slewlimit)slew=slewlimit;
+    choice=rint(bm->avgfloat+= slew/vi->rate*samples);
+    this_bits=oggpack_bytes(vbi->packetblob[choice])*8;
+  }
+
+
+
+  /* enforce min(if used) on the current floater (if used) */
+  if(bm->min_bitsper>0){
+    /* do we need to force the bitrate up? */
+    if(this_bits<min_target_bits){
+      while(bm->minmax_reservoir-(min_target_bits-this_bits)<0){
+        choice++;
+        if(choice>=PACKETBLOBS)break;
+        this_bits=oggpack_bytes(vbi->packetblob[choice])*8;
+      }
+    }
+  }
+
+  /* enforce max (if used) on the current floater (if used) */
+  if(bm->max_bitsper>0){
+    /* do we need to force the bitrate down? */
+    if(this_bits>max_target_bits){
+      while(bm->minmax_reservoir+(this_bits-max_target_bits)>bi->reservoir_bits){
+        choice--;
+        if(choice<0)break;
+        this_bits=oggpack_bytes(vbi->packetblob[choice])*8;
+      }
+    }
+  }
+
+  /* Choice of packetblobs now made based on floater, and min/max
+     requirements. Now boundary check extreme choices */
+
+  if(choice<0){
+    /* choosing a smaller packetblob is insufficient to trim bitrate.
+       frame will need to be truncated */
+    long maxsize=(max_target_bits+(bi->reservoir_bits-bm->minmax_reservoir))/8;
+    bm->choice=choice=0;
+
+    if(oggpack_bytes(vbi->packetblob[choice])>maxsize){
+
+      oggpack_writetrunc(vbi->packetblob[choice],maxsize*8);
+      this_bits=oggpack_bytes(vbi->packetblob[choice])*8;
+    }
+  }else{
+    long minsize=(min_target_bits-bm->minmax_reservoir+7)/8;
+    if(choice>=PACKETBLOBS)
+      choice=PACKETBLOBS-1;
+
+    bm->choice=choice;
+
+    /* prop up bitrate according to demand. pad this frame out with zeroes */
+    minsize-=oggpack_bytes(vbi->packetblob[choice]);
+    while(minsize-->0)oggpack_write(vbi->packetblob[choice],0,8);
+    this_bits=oggpack_bytes(vbi->packetblob[choice])*8;
+
+  }
+
+  /* now we have the final packet and the final packet size.  Update statistics */
+  /* min and max reservoir */
+  if(bm->min_bitsper>0 || bm->max_bitsper>0){
+
+    if(max_target_bits>0 && this_bits>max_target_bits){
+      bm->minmax_reservoir+=(this_bits-max_target_bits);
+    }else if(min_target_bits>0 && this_bits<min_target_bits){
+      bm->minmax_reservoir+=(this_bits-min_target_bits);
+    }else{
+      /* inbetween; we want to take reservoir toward but not past desired_fill */
+      if(bm->minmax_reservoir>desired_fill){
+        if(max_target_bits>0){ /* logical bulletproofing against initialization state */
+          bm->minmax_reservoir+=(this_bits-max_target_bits);
+          if(bm->minmax_reservoir<desired_fill)bm->minmax_reservoir=desired_fill;
+        }else{
+          bm->minmax_reservoir=desired_fill;
+        }
+      }else{
+        if(min_target_bits>0){ /* logical bulletproofing against initialization state */
+          bm->minmax_reservoir+=(this_bits-min_target_bits);
+          if(bm->minmax_reservoir>desired_fill)bm->minmax_reservoir=desired_fill;
+        }else{
+          bm->minmax_reservoir=desired_fill;
+        }
+      }
+    }
+  }
+
+  /* avg reservoir */
+  if(bm->avg_bitsper>0){
+    long avg_target_bits=(vb->W?bm->avg_bitsper*bm->short_per_long:bm->avg_bitsper);
+    bm->avg_reservoir+=this_bits-avg_target_bits;
+  }
+
+  return(0);
+}
+
+int vorbis_bitrate_flushpacket(vorbis_dsp_state *vd,ogg_packet *op){
+  private_state         *b=vd->backend_state;
+  bitrate_manager_state *bm=&b->bms;
+  vorbis_block          *vb=bm->vb;
+  int                    choice=PACKETBLOBS/2;
+  if(!vb)return 0;
+
+  if(op){
+    vorbis_block_internal *vbi=vb->internal;
+
+    if(vorbis_bitrate_managed(vb))
+      choice=bm->choice;
+
+    op->packet=oggpack_get_buffer(vbi->packetblob[choice]);
+    op->bytes=oggpack_bytes(vbi->packetblob[choice]);
+    op->b_o_s=0;
+    op->e_o_s=vb->eofflag;
+    op->granulepos=vb->granulepos;
+    op->packetno=vb->sequence; /* for sake of completeness */
+  }
+
+  bm->vb=0;
+  return(1);
+}

lib/theoratest/lib/vorbis/bitrate.h

@@ -0,0 +1,58 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2007             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: bitrate tracking and management
+
+ ********************************************************************/
+
+#ifndef _V_BITRATE_H_
+#define _V_BITRATE_H_
+
+#include "vorbis/codec.h"
+#include "codec_internal.h"
+#include "os.h"
+
+/* encode side bitrate tracking */
+typedef struct bitrate_manager_state {
+  int            managed;
+
+  long           avg_reservoir;
+  long           minmax_reservoir;
+  long           avg_bitsper;
+  long           min_bitsper;
+  long           max_bitsper;
+
+  long           short_per_long;
+  double         avgfloat;
+
+  vorbis_block  *vb;
+  int            choice;
+} bitrate_manager_state;
+
+typedef struct bitrate_manager_info{
+  long           avg_rate;
+  long           min_rate;
+  long           max_rate;
+  long           reservoir_bits;
+  double         reservoir_bias;
+
+  double         slew_damp;
+
+} bitrate_manager_info;
+
+extern void vorbis_bitrate_init(vorbis_info *vi,bitrate_manager_state *bs);
+extern void vorbis_bitrate_clear(bitrate_manager_state *bs);
+extern int vorbis_bitrate_managed(vorbis_block *vb);
+extern int vorbis_bitrate_addblock(vorbis_block *vb);
+extern int vorbis_bitrate_flushpacket(vorbis_dsp_state *vd, ogg_packet *op);
+
+#endif

lib/theoratest/lib/vorbis/codebook.c

@@ -0,0 +1,461 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2015             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: basic codebook pack/unpack/code/decode operations
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+#include <ogg/ogg.h>
+#include "vorbis/codec.h"
+#include "codebook.h"
+#include "scales.h"
+#include "misc.h"
+#include "os.h"
+
+/* packs the given codebook into the bitstream **************************/
+
+int vorbis_staticbook_pack(const static_codebook *c,oggpack_buffer *opb){
+  long i,j;
+  int ordered=0;
+
+  /* first the basic parameters */
+  oggpack_write(opb,0x564342,24);
+  oggpack_write(opb,c->dim,16);
+  oggpack_write(opb,c->entries,24);
+
+  /* pack the codewords.  There are two packings; length ordered and
+     length random.  Decide between the two now. */
+
+  for(i=1;i<c->entries;i++)
+    if(c->lengthlist[i-1]==0 || c->lengthlist[i]<c->lengthlist[i-1])break;
+  if(i==c->entries)ordered=1;
+
+  if(ordered){
+    /* length ordered.  We only need to say how many codewords of
+       each length.  The actual codewords are generated
+       deterministically */
+
+    long count=0;
+    oggpack_write(opb,1,1);  /* ordered */
+    oggpack_write(opb,c->lengthlist[0]-1,5); /* 1 to 32 */
+
+    for(i=1;i<c->entries;i++){
+      char this=c->lengthlist[i];
+      char last=c->lengthlist[i-1];
+      if(this>last){
+        for(j=last;j<this;j++){
+          oggpack_write(opb,i-count,ov_ilog(c->entries-count));
+          count=i;
+        }
+      }
+    }
+    oggpack_write(opb,i-count,ov_ilog(c->entries-count));
+
+  }else{
+    /* length random.  Again, we don't code the codeword itself, just
+       the length.  This time, though, we have to encode each length */
+    oggpack_write(opb,0,1);   /* unordered */
+
+    /* algortihmic mapping has use for 'unused entries', which we tag
+       here.  The algorithmic mapping happens as usual, but the unused
+       entry has no codeword. */
+    for(i=0;i<c->entries;i++)
+      if(c->lengthlist[i]==0)break;
+
+    if(i==c->entries){
+      oggpack_write(opb,0,1); /* no unused entries */
+      for(i=0;i<c->entries;i++)
+        oggpack_write(opb,c->lengthlist[i]-1,5);
+    }else{
+      oggpack_write(opb,1,1); /* we have unused entries; thus we tag */
+      for(i=0;i<c->entries;i++){
+        if(c->lengthlist[i]==0){
+          oggpack_write(opb,0,1);
+        }else{
+          oggpack_write(opb,1,1);
+          oggpack_write(opb,c->lengthlist[i]-1,5);
+        }
+      }
+    }
+  }
+
+  /* is the entry number the desired return value, or do we have a
+     mapping? If we have a mapping, what type? */
+  oggpack_write(opb,c->maptype,4);
+  switch(c->maptype){
+  case 0:
+    /* no mapping */
+    break;
+  case 1:case 2:
+    /* implicitly populated value mapping */
+    /* explicitly populated value mapping */
+
+    if(!c->quantlist){
+      /* no quantlist?  error */
+      return(-1);
+    }
+
+    /* values that define the dequantization */
+    oggpack_write(opb,c->q_min,32);
+    oggpack_write(opb,c->q_delta,32);
+    oggpack_write(opb,c->q_quant-1,4);
+    oggpack_write(opb,c->q_sequencep,1);
+
+    {
+      int quantvals;
+      switch(c->maptype){
+      case 1:
+        /* a single column of (c->entries/c->dim) quantized values for
+           building a full value list algorithmically (square lattice) */
+        quantvals=_book_maptype1_quantvals(c);
+        break;
+      case 2:
+        /* every value (c->entries*c->dim total) specified explicitly */
+        quantvals=c->entries*c->dim;
+        break;
+      default: /* NOT_REACHABLE */
+        quantvals=-1;
+      }
+
+      /* quantized values */
+      for(i=0;i<quantvals;i++)
+        oggpack_write(opb,labs(c->quantlist[i]),c->q_quant);
+
+    }
+    break;
+  default:
+    /* error case; we don't have any other map types now */
+    return(-1);
+  }
+
+  return(0);
+}
+
+/* unpacks a codebook from the packet buffer into the codebook struct,
+   readies the codebook auxiliary structures for decode *************/
+static_codebook *vorbis_staticbook_unpack(oggpack_buffer *opb){
+  long i,j;
+  static_codebook *s=_ogg_calloc(1,sizeof(*s));
+  s->allocedp=1;
+
+  /* make sure alignment is correct */
+  if(oggpack_read(opb,24)!=0x564342)goto _eofout;
+
+  /* first the basic parameters */
+  s->dim=oggpack_read(opb,16);
+  s->entries=oggpack_read(opb,24);
+  if(s->entries==-1)goto _eofout;
+
+  if(ov_ilog(s->dim)+ov_ilog(s->entries)>24)goto _eofout;
+
+  /* codeword ordering.... length ordered or unordered? */
+  switch((int)oggpack_read(opb,1)){
+  case 0:{
+    long unused;
+    /* allocated but unused entries? */
+    unused=oggpack_read(opb,1);
+    if((s->entries*(unused?1:5)+7)>>3>opb->storage-oggpack_bytes(opb))
+      goto _eofout;
+    /* unordered */
+    s->lengthlist=_ogg_malloc(sizeof(*s->lengthlist)*s->entries);
+
+    /* allocated but unused entries? */
+    if(unused){
+      /* yes, unused entries */
+
+      for(i=0;i<s->entries;i++){
+        if(oggpack_read(opb,1)){
+          long num=oggpack_read(opb,5);
+          if(num==-1)goto _eofout;
+          s->lengthlist[i]=num+1;
+        }else
+          s->lengthlist[i]=0;
+      }
+    }else{
+      /* all entries used; no tagging */
+      for(i=0;i<s->entries;i++){
+        long num=oggpack_read(opb,5);
+        if(num==-1)goto _eofout;
+        s->lengthlist[i]=num+1;
+      }
+    }
+
+    break;
+  }
+  case 1:
+    /* ordered */
+    {
+      long length=oggpack_read(opb,5)+1;
+      if(length==0)goto _eofout;
+      s->lengthlist=_ogg_malloc(sizeof(*s->lengthlist)*s->entries);
+
+      for(i=0;i<s->entries;){
+        long num=oggpack_read(opb,ov_ilog(s->entries-i));
+        if(num==-1)goto _eofout;
+        if(length>32 || num>s->entries-i ||
+           (num>0 && (num-1)>>(length-1)>1)){
+          goto _errout;
+        }
+        if(length>32)goto _errout;
+        for(j=0;j<num;j++,i++)
+          s->lengthlist[i]=length;
+        length++;
+      }
+    }
+    break;
+  default:
+    /* EOF */
+    goto _eofout;
+  }
+
+  /* Do we have a mapping to unpack? */
+  switch((s->maptype=oggpack_read(opb,4))){
+  case 0:
+    /* no mapping */
+    break;
+  case 1: case 2:
+    /* implicitly populated value mapping */
+    /* explicitly populated value mapping */
+
+    s->q_min=oggpack_read(opb,32);
+    s->q_delta=oggpack_read(opb,32);
+    s->q_quant=oggpack_read(opb,4)+1;
+    s->q_sequencep=oggpack_read(opb,1);
+    if(s->q_sequencep==-1)goto _eofout;
+
+    {
+      int quantvals=0;
+      switch(s->maptype){
+      case 1:
+        quantvals=(s->dim==0?0:_book_maptype1_quantvals(s));
+        break;
+      case 2:
+        quantvals=s->entries*s->dim;
+        break;
+      }
+
+      /* quantized values */
+      if(((quantvals*s->q_quant+7)>>3)>opb->storage-oggpack_bytes(opb))
+        goto _eofout;
+      s->quantlist=_ogg_malloc(sizeof(*s->quantlist)*quantvals);
+      for(i=0;i<quantvals;i++)
+        s->quantlist[i]=oggpack_read(opb,s->q_quant);
+
+      if(quantvals&&s->quantlist[quantvals-1]==-1)goto _eofout;
+    }
+    break;
+  default:
+    goto _errout;
+  }
+
+  /* all set */
+  return(s);
+
+ _errout:
+ _eofout:
+  vorbis_staticbook_destroy(s);
+  return(NULL);
+}
+
+/* returns the number of bits ************************************************/
+int vorbis_book_encode(codebook *book, int a, oggpack_buffer *b){
+  if(a<0 || a>=book->c->entries)return(0);
+  oggpack_write(b,book->codelist[a],book->c->lengthlist[a]);
+  return(book->c->lengthlist[a]);
+}
+
+/* the 'eliminate the decode tree' optimization actually requires the
+   codewords to be MSb first, not LSb.  This is an annoying inelegancy
+   (and one of the first places where carefully thought out design
+   turned out to be wrong; Vorbis II and future Ogg codecs should go
+   to an MSb bitpacker), but not actually the huge hit it appears to
+   be.  The first-stage decode table catches most words so that
+   bitreverse is not in the main execution path. */
+
+static ogg_uint32_t bitreverse(ogg_uint32_t x){
+  x=    ((x>>16)&0x0000ffff) | ((x<<16)&0xffff0000);
+  x=    ((x>> 8)&0x00ff00ff) | ((x<< 8)&0xff00ff00);
+  x=    ((x>> 4)&0x0f0f0f0f) | ((x<< 4)&0xf0f0f0f0);
+  x=    ((x>> 2)&0x33333333) | ((x<< 2)&0xcccccccc);
+  return((x>> 1)&0x55555555) | ((x<< 1)&0xaaaaaaaa);
+}
+
+STIN long decode_packed_entry_number(codebook *book, oggpack_buffer *b){
+  int  read=book->dec_maxlength;
+  long lo,hi;
+  long lok = oggpack_look(b,book->dec_firsttablen);
+
+  if (lok >= 0) {
+    long entry = book->dec_firsttable[lok];
+    if(entry&0x80000000UL){
+      lo=(entry>>15)&0x7fff;
+      hi=book->used_entries-(entry&0x7fff);
+    }else{
+      oggpack_adv(b, book->dec_codelengths[entry-1]);
+      return(entry-1);
+    }
+  }else{
+    lo=0;
+    hi=book->used_entries;
+  }
+
+  /* Single entry codebooks use a firsttablen of 1 and a
+     dec_maxlength of 1.  If a single-entry codebook gets here (due to
+     failure to read one bit above), the next look attempt will also
+     fail and we'll correctly kick out instead of trying to walk the
+     underformed tree */
+
+  lok = oggpack_look(b, read);
+
+  while(lok<0 && read>1)
+    lok = oggpack_look(b, --read);
+  if(lok<0)return -1;
+
+  /* bisect search for the codeword in the ordered list */
+  {
+    ogg_uint32_t testword=bitreverse((ogg_uint32_t)lok);
+
+    while(hi-lo>1){
+      long p=(hi-lo)>>1;
+      long test=book->codelist[lo+p]>testword;
+      lo+=p&(test-1);
+      hi-=p&(-test);
+      }
+
+    if(book->dec_codelengths[lo]<=read){
+      oggpack_adv(b, book->dec_codelengths[lo]);
+      return(lo);
+    }
+  }
+
+  oggpack_adv(b, read);
+
+  return(-1);
+}
+
+/* Decode side is specced and easier, because we don't need to find
+   matches using different criteria; we simply read and map.  There are
+   two things we need to do 'depending':
+
+   We may need to support interleave.  We don't really, but it's
+   convenient to do it here rather than rebuild the vector later.
+
+   Cascades may be additive or multiplicitive; this is not inherent in
+   the codebook, but set in the code using the codebook.  Like
+   interleaving, it's easiest to do it here.
+   addmul==0 -> declarative (set the value)
+   addmul==1 -> additive
+   addmul==2 -> multiplicitive */
+
+/* returns the [original, not compacted] entry number or -1 on eof *********/
+long vorbis_book_decode(codebook *book, oggpack_buffer *b){
+  if(book->used_entries>0){
+    long packed_entry=decode_packed_entry_number(book,b);
+    if(packed_entry>=0)
+      return(book->dec_index[packed_entry]);
+  }
+
+  /* if there's no dec_index, the codebook unpacking isn't collapsed */
+  return(-1);
+}
+
+/* returns 0 on OK or -1 on eof *************************************/
+/* decode vector / dim granularity gaurding is done in the upper layer */
+long vorbis_book_decodevs_add(codebook *book,float *a,oggpack_buffer *b,int n){
+  if(book->used_entries>0){
+    int step=n/book->dim;
+    long *entry = alloca(sizeof(*entry)*step);
+    float **t = alloca(sizeof(*t)*step);
+    int i,j,o;
+
+    for (i = 0; i < step; i++) {
+      entry[i]=decode_packed_entry_number(book,b);
+      if(entry[i]==-1)return(-1);
+      t[i] = book->valuelist+entry[i]*book->dim;
+    }
+    for(i=0,o=0;i<book->dim;i++,o+=step)
+      for (j=0;o+j<n && j<step;j++)
+        a[o+j]+=t[j][i];
+  }
+  return(0);
+}
+
+/* decode vector / dim granularity gaurding is done in the upper layer */
+long vorbis_book_decodev_add(codebook *book,float *a,oggpack_buffer *b,int n){
+  if(book->used_entries>0){
+    int i,j,entry;
+    float *t;
+
+    for(i=0;i<n;){
+      entry = decode_packed_entry_number(book,b);
+      if(entry==-1)return(-1);
+      t     = book->valuelist+entry*book->dim;
+      for(j=0;i<n && j<book->dim;)
+        a[i++]+=t[j++];
+    }
+  }
+  return(0);
+}
+
+/* unlike the others, we guard against n not being an integer number
+   of <dim> internally rather than in the upper layer (called only by
+   floor0) */
+long vorbis_book_decodev_set(codebook *book,float *a,oggpack_buffer *b,int n){
+  if(book->used_entries>0){
+    int i,j,entry;
+    float *t;
+
+    for(i=0;i<n;){
+      entry = decode_packed_entry_number(book,b);
+      if(entry==-1)return(-1);
+      t     = book->valuelist+entry*book->dim;
+      for (j=0;i<n && j<book->dim;){
+        a[i++]=t[j++];
+      }
+    }
+  }else{
+    int i;
+
+    for(i=0;i<n;){
+      a[i++]=0.f;
+    }
+  }
+  return(0);
+}
+
+long vorbis_book_decodevv_add(codebook *book,float **a,long offset,int ch,
+                              oggpack_buffer *b,int n){
+
+  long i,j,entry;
+  int chptr=0;
+  if(book->used_entries>0){
+    int m=(offset+n)/ch;
+    for(i=offset/ch;i<m;){
+      entry = decode_packed_entry_number(book,b);
+      if(entry==-1)return(-1);
+      {
+        const float *t = book->valuelist+entry*book->dim;
+        for (j=0;i<m && j<book->dim;j++){
+          a[chptr++][i]+=t[j];
+          if(chptr==ch){
+            chptr=0;
+            i++;
+          }
+        }
+      }
+    }
+  }
+  return(0);
+}

lib/theoratest/lib/vorbis/codebook.h

@@ -0,0 +1,117 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2015             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: basic shared codebook operations
+
+ ********************************************************************/
+
+#ifndef _V_CODEBOOK_H_
+#define _V_CODEBOOK_H_
+
+#include <ogg/ogg.h>
+
+/* This structure encapsulates huffman and VQ style encoding books; it
+   doesn't do anything specific to either.
+
+   valuelist/quantlist are nonNULL (and q_* significant) only if
+   there's entry->value mapping to be done.
+
+   If encode-side mapping must be done (and thus the entry needs to be
+   hunted), the auxiliary encode pointer will point to a decision
+   tree.  This is true of both VQ and huffman, but is mostly useful
+   with VQ.
+
+*/
+
+typedef struct static_codebook{
+  long   dim;           /* codebook dimensions (elements per vector) */
+  long   entries;       /* codebook entries */
+  char  *lengthlist;    /* codeword lengths in bits */
+
+  /* mapping ***************************************************************/
+  int    maptype;       /* 0=none
+                           1=implicitly populated values from map column
+                           2=listed arbitrary values */
+
+  /* The below does a linear, single monotonic sequence mapping. */
+  long     q_min;       /* packed 32 bit float; quant value 0 maps to minval */
+  long     q_delta;     /* packed 32 bit float; val 1 - val 0 == delta */
+  int      q_quant;     /* bits: 0 < quant <= 16 */
+  int      q_sequencep; /* bitflag */
+
+  long     *quantlist;  /* map == 1: (int)(entries^(1/dim)) element column map
+                           map == 2: list of dim*entries quantized entry vals
+                        */
+  int allocedp;
+} static_codebook;
+
+typedef struct codebook{
+  long dim;           /* codebook dimensions (elements per vector) */
+  long entries;       /* codebook entries */
+  long used_entries;  /* populated codebook entries */
+  const static_codebook *c;
+
+  /* for encode, the below are entry-ordered, fully populated */
+  /* for decode, the below are ordered by bitreversed codeword and only
+     used entries are populated */
+  float        *valuelist;  /* list of dim*entries actual entry values */
+  ogg_uint32_t *codelist;   /* list of bitstream codewords for each entry */
+
+  int          *dec_index;  /* only used if sparseness collapsed */
+  char         *dec_codelengths;
+  ogg_uint32_t *dec_firsttable;
+  int           dec_firsttablen;
+  int           dec_maxlength;
+
+  /* The current encoder uses only centered, integer-only lattice books. */
+  int           quantvals;
+  int           minval;
+  int           delta;
+} codebook;
+
+extern void vorbis_staticbook_destroy(static_codebook *b);
+extern int vorbis_book_init_encode(codebook *dest,const static_codebook *source);
+extern int vorbis_book_init_decode(codebook *dest,const static_codebook *source);
+extern void vorbis_book_clear(codebook *b);
+
+extern float *_book_unquantize(const static_codebook *b,int n,int *map);
+extern float *_book_logdist(const static_codebook *b,float *vals);
+extern float _float32_unpack(long val);
+extern long   _float32_pack(float val);
+extern int  _best(codebook *book, float *a, int step);
+extern long _book_maptype1_quantvals(const static_codebook *b);
+
+extern int vorbis_book_besterror(codebook *book,float *a,int step,int addmul);
+extern long vorbis_book_codeword(codebook *book,int entry);
+extern long vorbis_book_codelen(codebook *book,int entry);
+
+
+
+extern int vorbis_staticbook_pack(const static_codebook *c,oggpack_buffer *b);
+extern static_codebook *vorbis_staticbook_unpack(oggpack_buffer *b);
+
+extern int vorbis_book_encode(codebook *book, int a, oggpack_buffer *b);
+
+extern long vorbis_book_decode(codebook *book, oggpack_buffer *b);
+extern long vorbis_book_decodevs_add(codebook *book, float *a,
+                                     oggpack_buffer *b,int n);
+extern long vorbis_book_decodev_set(codebook *book, float *a,
+                                    oggpack_buffer *b,int n);
+extern long vorbis_book_decodev_add(codebook *book, float *a,
+                                    oggpack_buffer *b,int n);
+extern long vorbis_book_decodevv_add(codebook *book, float **a,
+                                     long off,int ch,
+                                    oggpack_buffer *b,int n);
+
+
+
+#endif

lib/theoratest/lib/vorbis/codec.h

@@ -0,0 +1,242 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2001             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+
+ ********************************************************************
+
+ function: libvorbis codec headers
+
+ ********************************************************************/
+
+#ifndef _vorbis_codec_h_
+#define _vorbis_codec_h_
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif /* __cplusplus */
+
+#include <ogg/ogg.h>
+
+typedef struct vorbis_info{
+  int version;
+  int channels;
+  long rate;
+
+  /* The below bitrate declarations are *hints*.
+     Combinations of the three values carry the following implications:
+
+     all three set to the same value:
+       implies a fixed rate bitstream
+     only nominal set:
+       implies a VBR stream that averages the nominal bitrate.  No hard
+       upper/lower limit
+     upper and or lower set:
+       implies a VBR bitstream that obeys the bitrate limits. nominal
+       may also be set to give a nominal rate.
+     none set:
+       the coder does not care to speculate.
+  */
+
+  long bitrate_upper;
+  long bitrate_nominal;
+  long bitrate_lower;
+  long bitrate_window;
+
+  void *codec_setup;
+} vorbis_info;
+
+/* vorbis_dsp_state buffers the current vorbis audio
+   analysis/synthesis state.  The DSP state belongs to a specific
+   logical bitstream ****************************************************/
+typedef struct vorbis_dsp_state{
+  int analysisp;
+  vorbis_info *vi;
+
+  float **pcm;
+  float **pcmret;
+  int      pcm_storage;
+  int      pcm_current;
+  int      pcm_returned;
+
+  int  preextrapolate;
+  int  eofflag;
+
+  long lW;
+  long W;
+  long nW;
+  long centerW;
+
+  ogg_int64_t granulepos;
+  ogg_int64_t sequence;
+
+  ogg_int64_t glue_bits;
+  ogg_int64_t time_bits;
+  ogg_int64_t floor_bits;
+  ogg_int64_t res_bits;
+
+  void       *backend_state;
+} vorbis_dsp_state;
+
+typedef struct vorbis_block{
+  /* necessary stream state for linking to the framing abstraction */
+  float  **pcm;       /* this is a pointer into local storage */
+  oggpack_buffer opb;
+
+  long  lW;
+  long  W;
+  long  nW;
+  int   pcmend;
+  int   mode;
+
+  int         eofflag;
+  ogg_int64_t granulepos;
+  ogg_int64_t sequence;
+  vorbis_dsp_state *vd; /* For read-only access of configuration */
+
+  /* local storage to avoid remallocing; it's up to the mapping to
+     structure it */
+  void               *localstore;
+  long                localtop;
+  long                localalloc;
+  long                totaluse;
+  struct alloc_chain *reap;
+
+  /* bitmetrics for the frame */
+  long glue_bits;
+  long time_bits;
+  long floor_bits;
+  long res_bits;
+
+  void *internal;
+
+} vorbis_block;
+
+/* vorbis_block is a single block of data to be processed as part of
+the analysis/synthesis stream; it belongs to a specific logical
+bitstream, but is independent from other vorbis_blocks belonging to
+that logical bitstream. *************************************************/
+
+struct alloc_chain{
+  void *ptr;
+  struct alloc_chain *next;
+};
+
+/* vorbis_info contains all the setup information specific to the
+   specific compression/decompression mode in progress (eg,
+   psychoacoustic settings, channel setup, options, codebook
+   etc). vorbis_info and substructures are in backends.h.
+*********************************************************************/
+
+/* the comments are not part of vorbis_info so that vorbis_info can be
+   static storage */
+typedef struct vorbis_comment{
+  /* unlimited user comment fields.  libvorbis writes 'libvorbis'
+     whatever vendor is set to in encode */
+  char **user_comments;
+  int   *comment_lengths;
+  int    comments;
+  char  *vendor;
+
+} vorbis_comment;
+
+
+/* libvorbis encodes in two abstraction layers; first we perform DSP
+   and produce a packet (see docs/analysis.txt).  The packet is then
+   coded into a framed OggSquish bitstream by the second layer (see
+   docs/framing.txt).  Decode is the reverse process; we sync/frame
+   the bitstream and extract individual packets, then decode the
+   packet back into PCM audio.
+
+   The extra framing/packetizing is used in streaming formats, such as
+   files.  Over the net (such as with UDP), the framing and
+   packetization aren't necessary as they're provided by the transport
+   and the streaming layer is not used */
+
+/* Vorbis PRIMITIVES: general ***************************************/
+
+extern void     vorbis_info_init(vorbis_info *vi);
+extern void     vorbis_info_clear(vorbis_info *vi);
+extern int      vorbis_info_blocksize(vorbis_info *vi,int zo);
+extern void     vorbis_comment_init(vorbis_comment *vc);
+extern void     vorbis_comment_add(vorbis_comment *vc, const char *comment);
+extern void     vorbis_comment_add_tag(vorbis_comment *vc,
+                                       const char *tag, const char *contents);
+extern char    *vorbis_comment_query(vorbis_comment *vc, const char *tag, int count);
+extern int      vorbis_comment_query_count(vorbis_comment *vc, const char *tag);
+extern void     vorbis_comment_clear(vorbis_comment *vc);
+
+extern int      vorbis_block_init(vorbis_dsp_state *v, vorbis_block *vb);
+extern int      vorbis_block_clear(vorbis_block *vb);
+extern void     vorbis_dsp_clear(vorbis_dsp_state *v);
+extern double   vorbis_granule_time(vorbis_dsp_state *v,
+                                    ogg_int64_t granulepos);
+
+extern const char *vorbis_version_string(void);
+
+/* Vorbis PRIMITIVES: analysis/DSP layer ****************************/
+
+extern int      vorbis_analysis_init(vorbis_dsp_state *v,vorbis_info *vi);
+extern int      vorbis_commentheader_out(vorbis_comment *vc, ogg_packet *op);
+extern int      vorbis_analysis_headerout(vorbis_dsp_state *v,
+                                          vorbis_comment *vc,
+                                          ogg_packet *op,
+                                          ogg_packet *op_comm,
+                                          ogg_packet *op_code);
+extern float  **vorbis_analysis_buffer(vorbis_dsp_state *v,int vals);
+extern int      vorbis_analysis_wrote(vorbis_dsp_state *v,int vals);
+extern int      vorbis_analysis_blockout(vorbis_dsp_state *v,vorbis_block *vb);
+extern int      vorbis_analysis(vorbis_block *vb,ogg_packet *op);
+
+extern int      vorbis_bitrate_addblock(vorbis_block *vb);
+extern int      vorbis_bitrate_flushpacket(vorbis_dsp_state *vd,
+                                           ogg_packet *op);
+
+/* Vorbis PRIMITIVES: synthesis layer *******************************/
+extern int      vorbis_synthesis_idheader(ogg_packet *op);
+extern int      vorbis_synthesis_headerin(vorbis_info *vi,vorbis_comment *vc,
+                                          ogg_packet *op);
+
+extern int      vorbis_synthesis_init(vorbis_dsp_state *v,vorbis_info *vi);
+extern int      vorbis_synthesis_restart(vorbis_dsp_state *v);
+extern int      vorbis_synthesis(vorbis_block *vb,ogg_packet *op);
+extern int      vorbis_synthesis_trackonly(vorbis_block *vb,ogg_packet *op);
+extern int      vorbis_synthesis_blockin(vorbis_dsp_state *v,vorbis_block *vb);
+extern int      vorbis_synthesis_pcmout(vorbis_dsp_state *v,float ***pcm);
+extern int      vorbis_synthesis_lapout(vorbis_dsp_state *v,float ***pcm);
+extern int      vorbis_synthesis_read(vorbis_dsp_state *v,int samples);
+extern long     vorbis_packet_blocksize(vorbis_info *vi,ogg_packet *op);
+
+extern int      vorbis_synthesis_halfrate(vorbis_info *v,int flag);
+extern int      vorbis_synthesis_halfrate_p(vorbis_info *v);
+
+/* Vorbis ERRORS and return codes ***********************************/
+
+#define OV_FALSE      -1
+#define OV_EOF        -2
+#define OV_HOLE       -3
+
+#define OV_EREAD      -128
+#define OV_EFAULT     -129
+#define OV_EIMPL      -130
+#define OV_EINVAL     -131
+#define OV_ENOTVORBIS -132
+#define OV_EBADHEADER -133
+#define OV_EVERSION   -134
+#define OV_ENOTAUDIO  -135
+#define OV_EBADPACKET -136
+#define OV_EBADLINK   -137
+#define OV_ENOSEEK    -138
+
+#ifdef __cplusplus
+}
+#endif /* __cplusplus */
+
+#endif
+

lib/theoratest/lib/vorbis/codec_internal.h

@@ -0,0 +1,166 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: libvorbis codec headers
+
+ ********************************************************************/
+
+#ifndef _V_CODECI_H_
+#define _V_CODECI_H_
+
+#include "envelope.h"
+#include "codebook.h"
+
+#define BLOCKTYPE_IMPULSE    0
+#define BLOCKTYPE_PADDING    1
+#define BLOCKTYPE_TRANSITION 0
+#define BLOCKTYPE_LONG       1
+
+#define PACKETBLOBS 15
+
+typedef struct vorbis_block_internal{
+  float  **pcmdelay;  /* this is a pointer into local storage */
+  float  ampmax;
+  int    blocktype;
+
+  oggpack_buffer *packetblob[PACKETBLOBS]; /* initialized, must be freed;
+                                              blob [PACKETBLOBS/2] points to
+                                              the oggpack_buffer in the
+                                              main vorbis_block */
+} vorbis_block_internal;
+
+typedef void vorbis_look_floor;
+typedef void vorbis_look_residue;
+typedef void vorbis_look_transform;
+
+/* mode ************************************************************/
+typedef struct {
+  int blockflag;
+  int windowtype;
+  int transformtype;
+  int mapping;
+} vorbis_info_mode;
+
+typedef void vorbis_info_floor;
+typedef void vorbis_info_residue;
+typedef void vorbis_info_mapping;
+
+#include "psy.h"
+#include "bitrate.h"
+
+typedef struct private_state {
+  /* local lookup storage */
+  envelope_lookup        *ve; /* envelope lookup */
+  int                     window[2];
+  vorbis_look_transform **transform[2];    /* block, type */
+  drft_lookup             fft_look[2];
+
+  int                     modebits;
+  vorbis_look_floor     **flr;
+  vorbis_look_residue   **residue;
+  vorbis_look_psy        *psy;
+  vorbis_look_psy_global *psy_g_look;
+
+  /* local storage, only used on the encoding side.  This way the
+     application does not need to worry about freeing some packets'
+     memory and not others'; packet storage is always tracked.
+     Cleared next call to a _dsp_ function */
+  unsigned char *header;
+  unsigned char *header1;
+  unsigned char *header2;
+
+  bitrate_manager_state bms;
+
+  ogg_int64_t sample_count;
+} private_state;
+
+/* codec_setup_info contains all the setup information specific to the
+   specific compression/decompression mode in progress (eg,
+   psychoacoustic settings, channel setup, options, codebook
+   etc).
+*********************************************************************/
+
+#include "highlevel.h"
+typedef struct codec_setup_info {
+
+  /* Vorbis supports only short and long blocks, but allows the
+     encoder to choose the sizes */
+
+  long blocksizes[2];
+
+  /* modes are the primary means of supporting on-the-fly different
+     blocksizes, different channel mappings (LR or M/A),
+     different residue backends, etc.  Each mode consists of a
+     blocksize flag and a mapping (along with the mapping setup */
+
+  int        modes;
+  int        maps;
+  int        floors;
+  int        residues;
+  int        books;
+  int        psys;     /* encode only */
+
+  vorbis_info_mode       *mode_param[64];
+  int                     map_type[64];
+  vorbis_info_mapping    *map_param[64];
+  int                     floor_type[64];
+  vorbis_info_floor      *floor_param[64];
+  int                     residue_type[64];
+  vorbis_info_residue    *residue_param[64];
+  static_codebook        *book_param[256];
+  codebook               *fullbooks;
+
+  vorbis_info_psy        *psy_param[4]; /* encode only */
+  vorbis_info_psy_global psy_g_param;
+
+  bitrate_manager_info   bi;
+  highlevel_encode_setup hi; /* used only by vorbisenc.c.  It's a
+                                highly redundant structure, but
+                                improves clarity of program flow. */
+  int         halfrate_flag; /* painless downsample for decode */
+} codec_setup_info;
+
+extern vorbis_look_psy_global *_vp_global_look(vorbis_info *vi);
+extern void _vp_global_free(vorbis_look_psy_global *look);
+
+
+
+typedef struct {
+  int sorted_index[VIF_POSIT+2];
+  int forward_index[VIF_POSIT+2];
+  int reverse_index[VIF_POSIT+2];
+
+  int hineighbor[VIF_POSIT];
+  int loneighbor[VIF_POSIT];
+  int posts;
+
+  int n;
+  int quant_q;
+  vorbis_info_floor1 *vi;
+
+  long phrasebits;
+  long postbits;
+  long frames;
+} vorbis_look_floor1;
+
+
+
+extern int *floor1_fit(vorbis_block *vb,vorbis_look_floor1 *look,
+                          const float *logmdct,   /* in */
+                          const float *logmask);
+extern int *floor1_interpolate_fit(vorbis_block *vb,vorbis_look_floor1 *look,
+                          int *A,int *B,
+                          int del);
+extern int floor1_encode(oggpack_buffer *opb,vorbis_block *vb,
+                  vorbis_look_floor1 *look,
+                  int *post,int *ilogmask);
+#endif

lib/theoratest/lib/vorbis/envelope.c

@@ -0,0 +1,374 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: PCM data envelope analysis
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <math.h>
+#include <ogg/ogg.h>
+#include "vorbis/codec.h"
+#include "codec_internal.h"
+
+#include "os.h"
+#include "scales.h"
+#include "envelope.h"
+#include "mdct.h"
+#include "misc.h"
+
+void _ve_envelope_init(envelope_lookup *e,vorbis_info *vi){
+  codec_setup_info *ci=vi->codec_setup;
+  vorbis_info_psy_global *gi=&ci->psy_g_param;
+  int ch=vi->channels;
+  int i,j;
+  int n=e->winlength=128;
+  e->searchstep=64; /* not random */
+
+  e->minenergy=gi->preecho_minenergy;
+  e->ch=ch;
+  e->storage=128;
+  e->cursor=ci->blocksizes[1]/2;
+  e->mdct_win=_ogg_calloc(n,sizeof(*e->mdct_win));
+  mdct_init(&e->mdct,n);
+
+  for(i=0;i<n;i++){
+    e->mdct_win[i]=sin(i/(n-1.)*M_PI);
+    e->mdct_win[i]*=e->mdct_win[i];
+  }
+
+  /* magic follows */
+  e->band[0].begin=2;  e->band[0].end=4;
+  e->band[1].begin=4;  e->band[1].end=5;
+  e->band[2].begin=6;  e->band[2].end=6;
+  e->band[3].begin=9;  e->band[3].end=8;
+  e->band[4].begin=13;  e->band[4].end=8;
+  e->band[5].begin=17;  e->band[5].end=8;
+  e->band[6].begin=22;  e->band[6].end=8;
+
+  for(j=0;j<VE_BANDS;j++){
+    n=e->band[j].end;
+    e->band[j].window=_ogg_malloc(n*sizeof(*e->band[0].window));
+    for(i=0;i<n;i++){
+      e->band[j].window[i]=sin((i+.5)/n*M_PI);
+      e->band[j].total+=e->band[j].window[i];
+    }
+    e->band[j].total=1./e->band[j].total;
+  }
+
+  e->filter=_ogg_calloc(VE_BANDS*ch,sizeof(*e->filter));
+  e->mark=_ogg_calloc(e->storage,sizeof(*e->mark));
+
+}
+
+void _ve_envelope_clear(envelope_lookup *e){
+  int i;
+  mdct_clear(&e->mdct);
+  for(i=0;i<VE_BANDS;i++)
+    _ogg_free(e->band[i].window);
+  _ogg_free(e->mdct_win);
+  _ogg_free(e->filter);
+  _ogg_free(e->mark);
+  memset(e,0,sizeof(*e));
+}
+
+/* fairly straight threshhold-by-band based until we find something
+   that works better and isn't patented. */
+
+static int _ve_amp(envelope_lookup *ve,
+                   vorbis_info_psy_global *gi,
+                   float *data,
+                   envelope_band *bands,
+                   envelope_filter_state *filters){
+  long n=ve->winlength;
+  int ret=0;
+  long i,j;
+  float decay;
+
+  /* we want to have a 'minimum bar' for energy, else we're just
+     basing blocks on quantization noise that outweighs the signal
+     itself (for low power signals) */
+
+  float minV=ve->minenergy;
+  float *vec=alloca(n*sizeof(*vec));
+
+  /* stretch is used to gradually lengthen the number of windows
+     considered prevoius-to-potential-trigger */
+  int stretch=max(VE_MINSTRETCH,ve->stretch/2);
+  float penalty=gi->stretch_penalty-(ve->stretch/2-VE_MINSTRETCH);
+  if(penalty<0.f)penalty=0.f;
+  if(penalty>gi->stretch_penalty)penalty=gi->stretch_penalty;
+
+  /*_analysis_output_always("lpcm",seq2,data,n,0,0,
+    totalshift+pos*ve->searchstep);*/
+
+ /* window and transform */
+  for(i=0;i<n;i++)
+    vec[i]=data[i]*ve->mdct_win[i];
+  mdct_forward(&ve->mdct,vec,vec);
+
+  /*_analysis_output_always("mdct",seq2,vec,n/2,0,1,0); */
+
+  /* near-DC spreading function; this has nothing to do with
+     psychoacoustics, just sidelobe leakage and window size */
+  {
+    float temp=vec[0]*vec[0]+.7*vec[1]*vec[1]+.2*vec[2]*vec[2];
+    int ptr=filters->nearptr;
+
+    /* the accumulation is regularly refreshed from scratch to avoid
+       floating point creep */
+    if(ptr==0){
+      decay=filters->nearDC_acc=filters->nearDC_partialacc+temp;
+      filters->nearDC_partialacc=temp;
+    }else{
+      decay=filters->nearDC_acc+=temp;
+      filters->nearDC_partialacc+=temp;
+    }
+    filters->nearDC_acc-=filters->nearDC[ptr];
+    filters->nearDC[ptr]=temp;
+
+    decay*=(1./(VE_NEARDC+1));
+    filters->nearptr++;
+    if(filters->nearptr>=VE_NEARDC)filters->nearptr=0;
+    decay=todB(&decay)*.5-15.f;
+  }
+
+  /* perform spreading and limiting, also smooth the spectrum.  yes,
+     the MDCT results in all real coefficients, but it still *behaves*
+     like real/imaginary pairs */
+  for(i=0;i<n/2;i+=2){
+    float val=vec[i]*vec[i]+vec[i+1]*vec[i+1];
+    val=todB(&val)*.5f;
+    if(val<decay)val=decay;
+    if(val<minV)val=minV;
+    vec[i>>1]=val;
+    decay-=8.;
+  }
+
+  /*_analysis_output_always("spread",seq2++,vec,n/4,0,0,0);*/
+
+  /* perform preecho/postecho triggering by band */
+  for(j=0;j<VE_BANDS;j++){
+    float acc=0.;
+    float valmax,valmin;
+
+    /* accumulate amplitude */
+    for(i=0;i<bands[j].end;i++)
+      acc+=vec[i+bands[j].begin]*bands[j].window[i];
+
+    acc*=bands[j].total;
+
+    /* convert amplitude to delta */
+    {
+      int p,this=filters[j].ampptr;
+      float postmax,postmin,premax=-99999.f,premin=99999.f;
+
+      p=this;
+      p--;
+      if(p<0)p+=VE_AMP;
+      postmax=max(acc,filters[j].ampbuf[p]);
+      postmin=min(acc,filters[j].ampbuf[p]);
+
+      for(i=0;i<stretch;i++){
+        p--;
+        if(p<0)p+=VE_AMP;
+        premax=max(premax,filters[j].ampbuf[p]);
+        premin=min(premin,filters[j].ampbuf[p]);
+      }
+
+      valmin=postmin-premin;
+      valmax=postmax-premax;
+
+      /*filters[j].markers[pos]=valmax;*/
+      filters[j].ampbuf[this]=acc;
+      filters[j].ampptr++;
+      if(filters[j].ampptr>=VE_AMP)filters[j].ampptr=0;
+    }
+
+    /* look at min/max, decide trigger */
+    if(valmax>gi->preecho_thresh[j]+penalty){
+      ret|=1;
+      ret|=4;
+    }
+    if(valmin<gi->postecho_thresh[j]-penalty)ret|=2;
+  }
+
+  return(ret);
+}
+
+#if 0
+static int seq=0;
+static ogg_int64_t totalshift=-1024;
+#endif
+
+long _ve_envelope_search(vorbis_dsp_state *v){
+  vorbis_info *vi=v->vi;
+  codec_setup_info *ci=vi->codec_setup;
+  vorbis_info_psy_global *gi=&ci->psy_g_param;
+  envelope_lookup *ve=((private_state *)(v->backend_state))->ve;
+  long i,j;
+
+  int first=ve->current/ve->searchstep;
+  int last=v->pcm_current/ve->searchstep-VE_WIN;
+  if(first<0)first=0;
+
+  /* make sure we have enough storage to match the PCM */
+  if(last+VE_WIN+VE_POST>ve->storage){
+    ve->storage=last+VE_WIN+VE_POST; /* be sure */
+    ve->mark=_ogg_realloc(ve->mark,ve->storage*sizeof(*ve->mark));
+  }
+
+  for(j=first;j<last;j++){
+    int ret=0;
+
+    ve->stretch++;
+    if(ve->stretch>VE_MAXSTRETCH*2)
+      ve->stretch=VE_MAXSTRETCH*2;
+
+    for(i=0;i<ve->ch;i++){
+      float *pcm=v->pcm[i]+ve->searchstep*(j);
+      ret|=_ve_amp(ve,gi,pcm,ve->band,ve->filter+i*VE_BANDS);
+    }
+
+    ve->mark[j+VE_POST]=0;
+    if(ret&1){
+      ve->mark[j]=1;
+      ve->mark[j+1]=1;
+    }
+
+    if(ret&2){
+      ve->mark[j]=1;
+      if(j>0)ve->mark[j-1]=1;
+    }
+
+    if(ret&4)ve->stretch=-1;
+  }
+
+  ve->current=last*ve->searchstep;
+
+  {
+    long centerW=v->centerW;
+    long testW=
+      centerW+
+      ci->blocksizes[v->W]/4+
+      ci->blocksizes[1]/2+
+      ci->blocksizes[0]/4;
+
+    j=ve->cursor;
+
+    while(j<ve->current-(ve->searchstep)){/* account for postecho
+                                             working back one window */
+      if(j>=testW)return(1);
+
+      ve->cursor=j;
+
+      if(ve->mark[j/ve->searchstep]){
+        if(j>centerW){
+
+#if 0
+          if(j>ve->curmark){
+            float *marker=alloca(v->pcm_current*sizeof(*marker));
+            int l,m;
+            memset(marker,0,sizeof(*marker)*v->pcm_current);
+            fprintf(stderr,"mark! seq=%d, cursor:%fs time:%fs\n",
+                    seq,
+                    (totalshift+ve->cursor)/44100.,
+                    (totalshift+j)/44100.);
+            _analysis_output_always("pcmL",seq,v->pcm[0],v->pcm_current,0,0,totalshift);
+            _analysis_output_always("pcmR",seq,v->pcm[1],v->pcm_current,0,0,totalshift);
+
+            _analysis_output_always("markL",seq,v->pcm[0],j,0,0,totalshift);
+            _analysis_output_always("markR",seq,v->pcm[1],j,0,0,totalshift);
+
+            for(m=0;m<VE_BANDS;m++){
+              char buf[80];
+              sprintf(buf,"delL%d",m);
+              for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[m].markers[l]*.1;
+              _analysis_output_always(buf,seq,marker,v->pcm_current,0,0,totalshift);
+            }
+
+            for(m=0;m<VE_BANDS;m++){
+              char buf[80];
+              sprintf(buf,"delR%d",m);
+              for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[m+VE_BANDS].markers[l]*.1;
+              _analysis_output_always(buf,seq,marker,v->pcm_current,0,0,totalshift);
+            }
+
+            for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->mark[l]*.4;
+            _analysis_output_always("mark",seq,marker,v->pcm_current,0,0,totalshift);
+
+
+            seq++;
+
+          }
+#endif
+
+          ve->curmark=j;
+          if(j>=testW)return(1);
+          return(0);
+        }
+      }
+      j+=ve->searchstep;
+    }
+  }
+
+  return(-1);
+}
+
+int _ve_envelope_mark(vorbis_dsp_state *v){
+  envelope_lookup *ve=((private_state *)(v->backend_state))->ve;
+  vorbis_info *vi=v->vi;
+  codec_setup_info *ci=vi->codec_setup;
+  long centerW=v->centerW;
+  long beginW=centerW-ci->blocksizes[v->W]/4;
+  long endW=centerW+ci->blocksizes[v->W]/4;
+  if(v->W){
+    beginW-=ci->blocksizes[v->lW]/4;
+    endW+=ci->blocksizes[v->nW]/4;
+  }else{
+    beginW-=ci->blocksizes[0]/4;
+    endW+=ci->blocksizes[0]/4;
+  }
+
+  if(ve->curmark>=beginW && ve->curmark<endW)return(1);
+  {
+    long first=beginW/ve->searchstep;
+    long last=endW/ve->searchstep;
+    long i;
+    for(i=first;i<last;i++)
+      if(ve->mark[i])return(1);
+  }
+  return(0);
+}
+
+void _ve_envelope_shift(envelope_lookup *e,long shift){
+  int smallsize=e->current/e->searchstep+VE_POST; /* adjust for placing marks
+                                                     ahead of ve->current */
+  int smallshift=shift/e->searchstep;
+
+  memmove(e->mark,e->mark+smallshift,(smallsize-smallshift)*sizeof(*e->mark));
+
+#if 0
+  for(i=0;i<VE_BANDS*e->ch;i++)
+    memmove(e->filter[i].markers,
+            e->filter[i].markers+smallshift,
+            (1024-smallshift)*sizeof(*(*e->filter).markers));
+  totalshift+=shift;
+#endif
+
+  e->current-=shift;
+  if(e->curmark>=0)
+    e->curmark-=shift;
+  e->cursor-=shift;
+}

lib/theoratest/lib/vorbis/envelope.h

@@ -0,0 +1,79 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: PCM data envelope analysis and manipulation
+
+ ********************************************************************/
+
+#ifndef _V_ENVELOPE_
+#define _V_ENVELOPE_
+
+#include "mdct.h"
+
+#define VE_PRE    16
+#define VE_WIN    4
+#define VE_POST   2
+#define VE_AMP    (VE_PRE+VE_POST-1)
+
+#define VE_BANDS  7
+#define VE_NEARDC 15
+
+#define VE_MINSTRETCH 2   /* a bit less than short block */
+#define VE_MAXSTRETCH 12  /* one-third full block */
+
+typedef struct {
+  float ampbuf[VE_AMP];
+  int   ampptr;
+
+  float nearDC[VE_NEARDC];
+  float nearDC_acc;
+  float nearDC_partialacc;
+  int   nearptr;
+
+} envelope_filter_state;
+
+typedef struct {
+  int begin;
+  int end;
+  float *window;
+  float total;
+} envelope_band;
+
+typedef struct {
+  int ch;
+  int winlength;
+  int searchstep;
+  float minenergy;
+
+  mdct_lookup  mdct;
+  float       *mdct_win;
+
+  envelope_band          band[VE_BANDS];
+  envelope_filter_state *filter;
+  int   stretch;
+
+  int                   *mark;
+
+  long storage;
+  long current;
+  long curmark;
+  long cursor;
+} envelope_lookup;
+
+extern void _ve_envelope_init(envelope_lookup *e,vorbis_info *vi);
+extern void _ve_envelope_clear(envelope_lookup *e);
+extern long _ve_envelope_search(vorbis_dsp_state *v);
+extern void _ve_envelope_shift(envelope_lookup *e,long shift);
+extern int  _ve_envelope_mark(vorbis_dsp_state *v);
+
+
+#endif

lib/theoratest/lib/vorbis/floor0.c

@@ -0,0 +1,223 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2015             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: floor backend 0 implementation
+
+ ********************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+#include <ogg/ogg.h>
+#include "vorbis/codec.h"
+#include "codec_internal.h"
+#include "registry.h"
+#include "lpc.h"
+#include "lsp.h"
+#include "codebook.h"
+#include "scales.h"
+#include "misc.h"
+#include "os.h"
+
+#include "misc.h"
+#include <stdio.h>
+
+typedef struct {
+  int ln;
+  int  m;
+  int **linearmap;
+  int  n[2];
+
+  vorbis_info_floor0 *vi;
+
+  long bits;
+  long frames;
+} vorbis_look_floor0;
+
+
+/***********************************************/
+
+static void floor0_free_info(vorbis_info_floor *i){
+  vorbis_info_floor0 *info=(vorbis_info_floor0 *)i;
+  if(info){
+    memset(info,0,sizeof(*info));
+    _ogg_free(info);
+  }
+}
+
+static void floor0_free_look(vorbis_look_floor *i){
+  vorbis_look_floor0 *look=(vorbis_look_floor0 *)i;
+  if(look){
+
+    if(look->linearmap){
+
+      if(look->linearmap[0])_ogg_free(look->linearmap[0]);
+      if(look->linearmap[1])_ogg_free(look->linearmap[1]);
+
+      _ogg_free(look->linearmap);
+    }
+    memset(look,0,sizeof(*look));
+    _ogg_free(look);
+  }
+}
+
+static vorbis_info_floor *floor0_unpack (vorbis_info *vi,oggpack_buffer *opb){
+  codec_setup_info     *ci=vi->codec_setup;
+  int j;
+
+  vorbis_info_floor0 *info=_ogg_malloc(sizeof(*info));
+  info->order=oggpack_read(opb,8);
+  info->rate=oggpack_read(opb,16);
+  info->barkmap=oggpack_read(opb,16);
+  info->ampbits=oggpack_read(opb,6);
+  info->ampdB=oggpack_read(opb,8);
+  info->numbooks=oggpack_read(opb,4)+1;
+
+  if(info->order<1)goto err_out;
+  if(info->rate<1)goto err_out;
+  if(info->barkmap<1)goto err_out;
+  if(info->numbooks<1)goto err_out;
+
+  for(j=0;j<info->numbooks;j++){
+    info->books[j]=oggpack_read(opb,8);
+    if(info->books[j]<0 || info->books[j]>=ci->books)goto err_out;
+    if(ci->book_param[info->books[j]]->maptype==0)goto err_out;
+    if(ci->book_param[info->books[j]]->dim<1)goto err_out;
+  }
+  return(info);
+
+ err_out:
+  floor0_free_info(info);
+  return(NULL);
+}
+
+/* initialize Bark scale and normalization lookups.  We could do this
+   with static tables, but Vorbis allows a number of possible
+   combinations, so it's best to do it computationally.
+
+   The below is authoritative in terms of defining scale mapping.
+   Note that the scale depends on the sampling rate as well as the
+   linear block and mapping sizes */
+
+static void floor0_map_lazy_init(vorbis_block      *vb,
+                                 vorbis_info_floor *infoX,
+                                 vorbis_look_floor0 *look){
+  if(!look->linearmap[vb->W]){
+    vorbis_dsp_state   *vd=vb->vd;
+    vorbis_info        *vi=vd->vi;
+    codec_setup_info   *ci=vi->codec_setup;
+    vorbis_info_floor0 *info=(vorbis_info_floor0 *)infoX;
+    int W=vb->W;
+    int n=ci->blocksizes[W]/2,j;
+
+    /* we choose a scaling constant so that:
+       floor(bark(rate/2-1)*C)=mapped-1
+     floor(bark(rate/2)*C)=mapped */
+    float scale=look->ln/toBARK(info->rate/2.f);
+
+    /* the mapping from a linear scale to a smaller bark scale is
+       straightforward.  We do *not* make sure that the linear mapping
+       does not skip bark-scale bins; the decoder simply skips them and
+       the encoder may do what it wishes in filling them.  They're
+       necessary in some mapping combinations to keep the scale spacing
+       accurate */
+    look->linearmap[W]=_ogg_malloc((n+1)*sizeof(**look->linearmap));
+    for(j=0;j<n;j++){
+      int val=floor( toBARK((info->rate/2.f)/n*j)
+                     *scale); /* bark numbers represent band edges */
+      if(val>=look->ln)val=look->ln-1; /* guard against the approximation */
+      look->linearmap[W][j]=val;
+    }
+    look->linearmap[W][j]=-1;
+    look->n[W]=n;
+  }
+}
+
+static vorbis_look_floor *floor0_look(vorbis_dsp_state *vd,
+                                      vorbis_info_floor *i){
+  vorbis_info_floor0 *info=(vorbis_info_floor0 *)i;
+  vorbis_look_floor0 *look=_ogg_calloc(1,sizeof(*look));
+
+  (void)vd;
+
+  look->m=info->order;
+  look->ln=info->barkmap;
+  look->vi=info;
+
+  look->linearmap=_ogg_calloc(2,sizeof(*look->linearmap));
+
+  return look;
+}
+
+static void *floor0_inverse1(vorbis_block *vb,vorbis_look_floor *i){
+  vorbis_look_floor0 *look=(vorbis_look_floor0 *)i;
+  vorbis_info_floor0 *info=look->vi;
+  int j,k;
+
+  int ampraw=oggpack_read(&vb->opb,info->ampbits);
+  if(ampraw>0){ /* also handles the -1 out of data case */
+    long maxval=(1<<info->ampbits)-1;
+    float amp=(float)ampraw/maxval*info->ampdB;
+    int booknum=oggpack_read(&vb->opb,ov_ilog(info->numbooks));
+
+    if(booknum!=-1 && booknum<info->numbooks){ /* be paranoid */
+      codec_setup_info  *ci=vb->vd->vi->codec_setup;
+      codebook *b=ci->fullbooks+info->books[booknum];
+      float last=0.f;
+
+      /* the additional b->dim is a guard against any possible stack
+         smash; b->dim is provably more than we can overflow the
+         vector */
+      float *lsp=_vorbis_block_alloc(vb,sizeof(*lsp)*(look->m+b->dim+1));
+
+      if(vorbis_book_decodev_set(b,lsp,&vb->opb,look->m)==-1)goto eop;
+      for(j=0;j<look->m;){
+        for(k=0;j<look->m && k<b->dim;k++,j++)lsp[j]+=last;
+        last=lsp[j-1];
+      }
+
+      lsp[look->m]=amp;
+      return(lsp);
+    }
+  }
+ eop:
+  return(NULL);
+}
+
+static int floor0_inverse2(vorbis_block *vb,vorbis_look_floor *i,
+                           void *memo,float *out){
+  vorbis_look_floor0 *look=(vorbis_look_floor0 *)i;
+  vorbis_info_floor0 *info=look->vi;
+
+  floor0_map_lazy_init(vb,info,look);
+
+  if(memo){
+    float *lsp=(float *)memo;
+    float amp=lsp[look->m];
+
+    /* take the coefficients back to a spectral envelope curve */
+    vorbis_lsp_to_curve(out,
+                        look->linearmap[vb->W],
+                        look->n[vb->W],
+                        look->ln,
+                        lsp,look->m,amp,(float)info->ampdB);
+    return(1);
+  }
+  memset(out,0,sizeof(*out)*look->n[vb->W]);
+  return(0);
+}
+
+/* export hooks */
+const vorbis_func_floor floor0_exportbundle={
+  NULL,&floor0_unpack,&floor0_look,&floor0_free_info,
+  &floor0_free_look,&floor0_inverse1,&floor0_inverse2
+};

lib/theoratest/lib/vorbis/highlevel.h

@@ -0,0 +1,57 @@
+/********************************************************************
+ *                                                                  *
+ * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
+ * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
+ * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
+ * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
+ *                                                                  *
+ * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
+ * by the Xiph.Org Foundation http://www.xiph.org/                  *
+ *                                                                  *
+ ********************************************************************
+
+ function: highlevel encoder setup struct separated out for vorbisenc clarity
+
+ ********************************************************************/
+
+typedef struct highlevel_byblocktype {
+  double tone_mask_setting;
+  double tone_peaklimit_setting;
+  double noise_bias_setting;
+  double noise_compand_setting;
+} highlevel_byblocktype;
+
+typedef struct highlevel_encode_setup {
+  int   set_in_stone;
+  const void *setup;
+  double base_setting;
+
+  double impulse_noisetune;
+
+  /* bitrate management below all settable */
+  float  req;
+  int    managed;
+  long   bitrate_min;
+  long   bitrate_av;
+  double bitrate_av_damp;
+  long   bitrate_max;
+  long   bitrate_reservoir;
+  double bitrate_reservoir_bias;
+
+  int impulse_block_p;
+  int noise_normalize_p;
+  int coupling_p;
+
+  double stereo_point_setting;
+  double lowpass_kHz;
+  int    lowpass_altered;
+
+  double ath_floating_dB;
+  double ath_absolute_dB;
+
+  double amplitude_track_dBpersec;
+  double trigger_setting;
+
+  highlevel_byblocktype block[4]; /* padding, impulse, transition, long */
+
+} highlevel_encode_setup;