gpu: Add async pixel transfer interface, stub and tests.

gpu/command_buffer/service/gles2_cmd_decoder.cc

Index: gpu/command_buffer/service/gles2_cmd_decoder.cc
diff --git a/gpu/command_buffer/service/gles2_cmd_decoder.cc b/gpu/command_buffer/service/gles2_cmd_decoder.cc
index 3fc39c6dda1c29f787f7a8648d00baa58d85d481..d734af536f472fcabeb857c917e574f6f7bdb2fe 100644
--- a/gpu/command_buffer/service/gles2_cmd_decoder.cc
+++ b/gpu/command_buffer/service/gles2_cmd_decoder.cc
@@ -56,6 +56,7 @@
 #include "gpu/command_buffer/service/texture_manager.h"
 #include "gpu/command_buffer/service/vertex_attrib_manager.h"
 #include "gpu/command_buffer/service/vertex_array_manager.h"
+#include "ui/gl/async_pixel_transfer_delegate.h"
 #include "ui/gl/gl_image.h"
 #include "ui/gl/gl_implementation.h"
 #include "ui/gl/gl_surface.h"
@@ -193,6 +194,61 @@ static inline GLenum GetTexInternalFormat(GLenum internal_format) {
   return internal_format;
 }
 
+// TODO(epenner): Could the above function be merged into this and removed?
+static inline GLenum GetTexInternalFormat(GLenum internal_format,
+                                          GLenum format,
+                                          GLenum type) {
+  GLenum gl_internal_format = GetTexInternalFormat(internal_format);
+
+  if (gfx::GetGLImplementation() == gfx::kGLImplementationEGLGLES2)
+    return gl_internal_format;
+
+  if (type == GL_FLOAT) {
+    switch (format) {
+      case GL_RGBA:
+        gl_internal_format = GL_RGBA32F_ARB;
+        break;
+      case GL_RGB:
+        gl_internal_format = GL_RGB32F_ARB;
+        break;
+      case GL_LUMINANCE_ALPHA:
+        gl_internal_format = GL_LUMINANCE_ALPHA32F_ARB;
+        break;
+      case GL_LUMINANCE:
+        gl_internal_format = GL_LUMINANCE32F_ARB;
+        break;
+      case GL_ALPHA:
+        gl_internal_format = GL_ALPHA32F_ARB;
+        break;
+      default:
+        NOTREACHED();
+        break;
+    }
+  } else if (type == GL_HALF_FLOAT_OES) {
+    switch (format) {
+      case GL_RGBA:
+        gl_internal_format = GL_RGBA16F_ARB;
+        break;
+      case GL_RGB:
+        gl_internal_format = GL_RGB16F_ARB;
+        break;
+      case GL_LUMINANCE_ALPHA:
+        gl_internal_format = GL_LUMINANCE_ALPHA16F_ARB;
+        break;
+      case GL_LUMINANCE:
+        gl_internal_format = GL_LUMINANCE16F_ARB;
+        break;
+      case GL_ALPHA:
+        gl_internal_format = GL_ALPHA16F_ARB;
+        break;
+      default:
+        NOTREACHED();
+        break;
+    }
+  }
+  return gl_internal_format;
+}
+
 static void WrappedTexImage2D(
     GLenum target,
     GLint level,
@@ -203,55 +259,9 @@ static void WrappedTexImage2D(
     GLenum format,
     GLenum type,
     const void* pixels) {
-  GLenum gl_internal_format = GetTexInternalFormat(internal_format);
-  if (gfx::GetGLImplementation() != gfx::kGLImplementationEGLGLES2) {
-    if (type == GL_FLOAT) {
-      switch (format) {
-        case GL_RGBA:
-          gl_internal_format = GL_RGBA32F_ARB;
-          break;
-        case GL_RGB:
-          gl_internal_format = GL_RGB32F_ARB;
-          break;
-        case GL_LUMINANCE_ALPHA:
-          gl_internal_format = GL_LUMINANCE_ALPHA32F_ARB;
-          break;
-        case GL_LUMINANCE:
-          gl_internal_format = GL_LUMINANCE32F_ARB;
-          break;
-        case GL_ALPHA:
-          gl_internal_format = GL_ALPHA32F_ARB;
-          break;
-        default:
-          NOTREACHED();
-          break;
-      }
-    } else if (type == GL_HALF_FLOAT_OES) {
-      switch (format) {
-        case GL_RGBA:
-          gl_internal_format = GL_RGBA16F_ARB;
-          break;
-        case GL_RGB:
-          gl_internal_format = GL_RGB16F_ARB;
-          break;
-        case GL_LUMINANCE_ALPHA:
-          gl_internal_format = GL_LUMINANCE_ALPHA16F_ARB;
-          break;
-        case GL_LUMINANCE:
-          gl_internal_format = GL_LUMINANCE16F_ARB;
-          break;
-        case GL_ALPHA:
-          gl_internal_format = GL_ALPHA16F_ARB;
-          break;
-        default:
-          NOTREACHED();
-          break;
-      }
-    }
-  }
   glTexImage2D(
-      target, level, gl_internal_format, width, height, border, format, type,
-      pixels);
+      target, level, GetTexInternalFormat(internal_format, format, type),
+      width, height, border, format, type, pixels);
 }
 
 // Wrapper for glEnable/glDisable that doesn't suck.
@@ -552,6 +562,12 @@ class GLES2DecoderImpl : public GLES2Decoder {
   virtual void SetMsgCallback(const MsgCallback& callback) OVERRIDE;
 
   virtual void SetStreamTextureManager(StreamTextureManager* manager) OVERRIDE;
+
+  virtual gfx::AsyncPixelTransferDelegate*
+      GetAsyncPixelTransferDelegate() OVERRIDE;
+  virtual void SetAsyncPixelTransferDelegate(
+      gfx::AsyncPixelTransferDelegate* delegate) OVERRIDE;
+
   virtual bool GetServiceTextureId(uint32 client_texture_id,
                                    uint32* service_texture_id) OVERRIDE;
 
@@ -774,6 +790,14 @@ class GLES2DecoderImpl : public GLES2Decoder {
       GLenum type,
       const void * data);
 
+  // Extra validation for async tex(Sub)Image2D.
+  bool ValidateAsyncTransfer(
+      const char* function_name,
+      TextureManager::TextureInfo* info,
+      GLenum target,
+      GLint level,
+      const void * data);
+
   // Wrapper for TexImageIOSurface2DCHROMIUM.
   void DoTexImageIOSurface2DCHROMIUM(
       GLenum target,
@@ -1598,6 +1622,7 @@ class GLES2DecoderImpl : public GLES2Decoder {
   MsgCallback msg_callback_;
 
   StreamTextureManager* stream_texture_manager_;
+  scoped_ptr<gfx::AsyncPixelTransferDelegate> async_pixel_transfer_delegate_;
 
   // The format of the back buffer_
   GLenum back_buffer_color_format_;
@@ -2390,6 +2415,10 @@ bool GLES2DecoderImpl::Initialize(
     glPointParameteri(GL_POINT_SPRITE_COORD_ORIGIN, GL_LOWER_LEFT);
   }
 
+  // Create a delegate to perform async pixel transfers.
+  async_pixel_transfer_delegate_ =
+      gfx::AsyncPixelTransferDelegate::Create(context.get());
+
   return true;
 }
 
@@ -2887,6 +2916,16 @@ void GLES2DecoderImpl::SetStreamTextureManager(StreamTextureManager* manager) {
   stream_texture_manager_ = manager;
 }
 
+gfx::AsyncPixelTransferDelegate*
+    GLES2DecoderImpl::GetAsyncPixelTransferDelegate() {
+  return async_pixel_transfer_delegate_.get();
+}
+
+void GLES2DecoderImpl::SetAsyncPixelTransferDelegate(
+    gfx::AsyncPixelTransferDelegate* delegate) {
+  async_pixel_transfer_delegate_ = make_scoped_ptr(delegate);
+}
+
 bool GLES2DecoderImpl::GetServiceTextureId(uint32 client_texture_id,
                                            uint32* service_texture_id) {
   TextureManager::TextureInfo* texture =
@@ -3629,6 +3668,29 @@ void GLES2DecoderImpl::DoBindTexture(GLenum target, GLuint client_id) {
     texture_manager()->SetInfoTarget(info, target);
   }
   glBindTexture(target, info->service_id());
+
+  // Lazily bind async transfers if they have completed.
+  // TODO(epenner): Should this be delayed to draw time?
+  if (info->GetAsyncTransferState()) {
+    gfx::AsyncTexImage2DParams tex_define_params;
+    if (info->GetAsyncTransferState()->BindTransfer(
+        target, &tex_define_params)) {
+      DCHECK_EQ(target, tex_define_params.target);
+      texture_manager()->SetLevelInfo(
+          info,
+          tex_define_params.target,
+          tex_define_params.level,
+          tex_define_params.internal_format,
+          tex_define_params.width,
+          tex_define_params.height,
+          1, // depth
+          tex_define_params.border,
+          tex_define_params.format,
+          tex_define_params.type,
+          true); // cleared
+    }
+  }
+
   TextureUnit& unit = state_.texture_units[state_.active_texture_unit];
   unit.bind_target = target;
   switch (target) {
@@ -9666,13 +9728,39 @@ void GLES2DecoderImpl::DoTraceEndCHROMIUM() {
   gpu_trace_stack_.pop();
 }
 
+bool GLES2DecoderImpl::ValidateAsyncTransfer(
+    const char* function_name,
+    TextureManager::TextureInfo* info,
+    GLenum target,
+    GLint level,
+    const void * data) {
+  // We only support async uploads to 2D textures for now.
+  if (GL_TEXTURE_2D != target) {
+    SetGLErrorInvalidEnum(function_name, target, "target");
+    return false;
+  }
+  // We only support uploads to level zero for now.
+  if (level != 0) {
+    SetGLError(GL_INVALID_VALUE, function_name, "level != 0");
+    return false;
+  }
+  // A transfer buffer must be bound, even for asyncTexImage2D.
+  if (data == NULL) {
+    SetGLError(GL_INVALID_OPERATION, function_name, "buffer == 0");
+    return false;
+  }
+  // We only support one async transfer in progress.
+  if (!info || info->AsyncTransferIsInProgress()) {
+    SetGLError(GL_INVALID_OPERATION,
+        function_name, "transfer already in progress");
+    return false;
+  }
+  return true;
+}
+
 error::Error GLES2DecoderImpl::HandleAsyncTexImage2DCHROMIUM(
     uint32 immediate_data_size, const gles2::AsyncTexImage2DCHROMIUM& c) {
   TRACE_EVENT0("gpu", "GLES2DecoderImpl::HandleAsyncTexImage2DCHROMIUM");
-
-  // TODO: This is a copy of HandleTexImage2D validation. Merge
-  // as much of it as possible.
-  tex_image_2d_failed_ = true;
   GLenum target = static_cast<GLenum>(c.target);
   GLint level = static_cast<GLint>(c.level);
   GLint internal_format = static_cast<GLint>(c.internalformat);
@@ -9684,6 +9772,9 @@ error::Error GLES2DecoderImpl::HandleAsyncTexImage2DCHROMIUM(
   uint32 pixels_shm_id = static_cast<uint32>(c.pixels_shm_id);
   uint32 pixels_shm_offset = static_cast<uint32>(c.pixels_shm_offset);
   uint32 pixels_size;
+
+  // TODO(epenner): Move this and copies of this memory validation
+  // into ValidateTexImage2D step.
   if (!GLES2Util::ComputeImageDataSizes(
       width, height, format, type, state_.unpack_alignment, &pixels_size, NULL,
       NULL)) {
@@ -9698,19 +9789,62 @@ error::Error GLES2DecoderImpl::HandleAsyncTexImage2DCHROMIUM(
     }
   }
 
-  // TODO(epenner): Do this via an async task.
-  DoTexImage2D(
-      target, level, internal_format, width, height, border, format, type,
-      pixels, pixels_size);
+  // All the normal glTexSubImage2D validation.
+  if (!ValidateTexImage2D("glAsyncTexImage2D", target, level, internal_format,
+      width, height, border, format, type, pixels, pixels_size)) {
+    return error::kNoError;
+  }
+
+  // Extra async validation.
+  TextureManager::TextureInfo* info = GetTextureInfoForTarget(target);
+  if (!ValidateAsyncTransfer("glAsyncTexImage2D", info, target, level, pixels))
+    return error::kNoError;
+
+  // Don't allow async redefinition of a textures.
+  if (info->IsDefined()) {
+    SetGLError(GL_INVALID_OPERATION, "glAsyncTexImage2D", "already defined");
+    return error::kNoError;
+  }
+
+  if (info->IsAttachedToFramebuffer()) {

[greggman, 2012/12/13 06:29:21]

Just to document our conversation

This code marks says if the texture is used in an fbo then re-validate fbos.

You don't want to ask for fbos to be revalidated until you've actually changed
the texture. Since the texture is changing async this needs to happen later.

The plans is to add some kind of task list. The async code will take a task with
a weakptr to the decoder and to the TextureInfo. Once it's done uploading it
will add the task to the decoder's task list.

In GLES2DecoderImpl::CheckFramebufferValid any tasks on the list will be
executed.

those tasks will

   Call any GL functions needed on this context
   restore any state those functions changed
   call SetTextureInfo
   do the IsAttachedToFramebuffer check here to mark
   the fbo for re-validation

this will also fix the need to bind after async completion.

+    clear_state_dirty_ = true;
+    // TODO(gman): If textures tracked which framebuffers they were attached to
+    // we could just mark those framebuffers as not complete.
+    framebuffer_manager()->IncFramebufferStateChangeCount();
+  }
+
+  // We know the memory/size is safe, so get the real shared memory since
+  // it might need to be duped to prevent use-after-free of the memory.
+  Buffer buffer = GetSharedMemoryBuffer(c.pixels_shm_id);
+  base::SharedMemory* shared_memory = buffer.shared_memory;
+  uint32 shm_size = buffer.size;
+  uint32 shm_data_offset = c.pixels_shm_offset;
+  uint32 shm_data_size = pixels_size;
+
+  // Set up the async state if needed, and make the texture
+  // immutable so the async state stays valid. The level info
+  // is set up lazily when the transfer completes.
+  DCHECK(!info->GetAsyncTransferState());
+  info->SetAsyncTransferState(
+      async_pixel_transfer_delegate_->
+          CreatePixelTransferState(info->service_id()));
+  info->SetImmutable(true);
+
+  // Issue the async call and set up the texture.
+  GLenum gl_internal_format =
+      GetTexInternalFormat(internal_format, format, type);
+  gfx::AsyncTexImage2DParams tex_params = {target, level, gl_internal_format,
+                                           width, height, border, format, type};
+  gfx::AsyncMemoryParams mem_params = {shared_memory, shm_size,
+                                       shm_data_offset, shm_data_size};
+  async_pixel_transfer_delegate_->AsyncTexImage2D(
+      info->GetAsyncTransferState(), tex_params, mem_params);
   return error::kNoError;
 }
 
 error::Error GLES2DecoderImpl::HandleAsyncTexSubImage2DCHROMIUM(
     uint32 immediate_data_size, const gles2::AsyncTexSubImage2DCHROMIUM& c) {
   TRACE_EVENT0("gpu", "GLES2DecoderImpl::HandleAsyncTexSubImage2DCHROMIUM");
-
-  // TODO: This is a copy of HandleTexSubImage2D validation. Merge
-  // as much of it as possible.
   GLenum target = static_cast<GLenum>(c.target);
   GLint level = static_cast<GLint>(c.level);
   GLint xoffset = static_cast<GLint>(c.xoffset);
@@ -9719,6 +9853,9 @@ error::Error GLES2DecoderImpl::HandleAsyncTexSubImage2DCHROMIUM(
   GLsizei height = static_cast<GLsizei>(c.height);
   GLenum format = static_cast<GLenum>(c.format);
   GLenum type = static_cast<GLenum>(c.type);
+
+  // TODO(epenner): Move this and copies of this memory validation
+  // into ValidateTexSubImage2D step.
   uint32 data_size;
   if (!GLES2Util::ComputeImageDataSizes(
       width, height, format, type, state_.unpack_alignment, &data_size,
@@ -9727,13 +9864,58 @@ error::Error GLES2DecoderImpl::HandleAsyncTexSubImage2DCHROMIUM(
   }
   const void* pixels = GetSharedMemoryAs<const void*>(
       c.data_shm_id, c.data_shm_offset, data_size);
-  if (pixels == NULL) {
-    return error::kOutOfBounds;
+
+  // All the normal glTexSubImage2D validation.
+  error::Error error = error::kNoError;
+  if (!ValidateTexSubImage2D(&error, "glAsyncTexSubImage2D", target, level,
+      xoffset, yoffset, width, height, format, type, pixels)) {
+    return error;
   }
 
-  // TODO(epenner): Do this via an async task.
-  return DoTexSubImage2D(
-      target, level, xoffset, yoffset, width, height, format, type, pixels);
+  // Extra async validation.
+  TextureManager::TextureInfo* info = GetTextureInfoForTarget(target);
+  if (!ValidateAsyncTransfer(
+         "glAsyncTexSubImage2D", info, target, level, pixels))
+    return error::kNoError;
+
+  // Guarantee async textures are always 'cleared' as follows:
+  // - AsyncTexImage2D can not redefine an existing texture
+  // - AsyncTexImage2D must initialize the entire image via non-null buffer.
+  // - AsyncTexSubImage2D clears synchronously if not already cleared.
+  // - Textures become immutable after an async call.
+  // This way we know in all cases that an async texture is always clear.
+  if (!info->SafeToRenderFrom()) {
+    if (!texture_manager()->ClearTextureLevel(this, info, target, level)) {
+      SetGLError(
+          GL_OUT_OF_MEMORY, "glAsyncTexSubImage2D", "dimensions too big");
+      return error::kNoError;
+    }
+  }
+
+  // We know the memory/size is safe, so get the real shared memory since
+  // it might need to be duped to prevent use-after-free of the memory.
+  Buffer buffer = GetSharedMemoryBuffer(c.data_shm_id);
+  base::SharedMemory* shared_memory = buffer.shared_memory;
+  uint32 shm_size = buffer.size;
+  uint32 shm_data_offset = c.data_shm_offset;
+  uint32 shm_data_size = data_size;
+
+  if (!info->GetAsyncTransferState()) {
+    // Set up the async state if needed, and make the texture
+    // immutable so the async state stays valid.
+    info->SetAsyncTransferState(
+        async_pixel_transfer_delegate_->
+            CreatePixelTransferState(info->service_id()));
+    info->SetImmutable(true);
+  }
+
+  gfx::AsyncTexSubImage2DParams tex_params = {target, level, xoffset, yoffset,
+                                              width, height, format, type};
+  gfx::AsyncMemoryParams mem_params = {shared_memory, shm_size,
+                                       shm_data_offset, shm_data_size};
+  async_pixel_transfer_delegate_->AsyncTexSubImage2D(
+      info->GetAsyncTransferState(), tex_params, mem_params);
+  return error::kNoError;
 }
 
 // Include the auto-generated part of this file. We split this because it means