Add VDA supported profile to GPUInfo.

content/common/gpu/media/gpu_video_decode_accelerator.cc

Index: content/common/gpu/media/gpu_video_decode_accelerator.cc
diff --git a/content/common/gpu/media/gpu_video_decode_accelerator.cc b/content/common/gpu/media/gpu_video_decode_accelerator.cc
index 436428acd4076cee6a11cfa776bb687f4e970263..29e30ef6af764e641c7eb48081ac1abc61ae4593 100644
--- a/content/common/gpu/media/gpu_video_decode_accelerator.cc
+++ b/content/common/gpu/media/gpu_video_decode_accelerator.cc
@@ -15,6 +15,7 @@
 
 #include "content/common/gpu/gpu_channel.h"
 #include "content/common/gpu/gpu_messages.h"
+#include "content/common/gpu/media/gpu_video_accelerator_util.h"
 #include "content/public/common/content_switches.h"
 #include "gpu/command_buffer/common/command_buffer.h"
 #include "ipc/ipc_message_macros.h"
@@ -249,7 +250,8 @@ void GpuVideoDecodeAccelerator::Initialize(
 #endif
 
   // Array of Create..VDA() function pointers, maybe applicable to the current
-  // platform. This list is ordered by priority of use.
+  // platform. This list is ordered by priority of use and it should be the
+  // same as the order of querying supported profiles of VDAs.

[Pawel Osciak, 2015/03/26 08:36:39]

Why do we have to query in the same order?

Wouldn't it be enough to have a set of supported profiles queried in any order,
and keep Create()ing until one matches? We know one will match eventually, and
we keep the preferred order by just doing Create() in preferred order?

[henryhsu_tw, 2015/03/26 09:38:34]

For example, V4L2 supports H264 with 1080p and vaapi supports H264 with 4k.
If GetSupportedProfiles adds vaapi before V4L2, the profile of v4l2 will be
removed.
Then client wants to decode a H264 video with 4k. According to Create order, VDA
will choose V4L2VDA and initialize successfully. When decoding,
gpu_video_decoder will return error and change to software decoder. Therefore,
we keep the same order to ensure the returned profile can use corresponding
hardware decoder.

We can add resolution as parameter in Initialize function in the future.

   const GpuVideoDecodeAccelerator::CreateVDAFp create_vda_fps[] = {
       &GpuVideoDecodeAccelerator::CreateDXVAVDA,
       &GpuVideoDecodeAccelerator::CreateV4L2VDA,
@@ -383,6 +385,35 @@ GpuVideoDecodeAccelerator::CreateAndroidVDA() {
   return decoder.Pass();
 }
 
+// static
+std::vector<gpu::VideoDecodeAcceleratorSupportedProfile>
+GpuVideoDecodeAccelerator::GetSupportedProfiles() {
+  std::vector<media::VideoDecodeAccelerator::SupportedProfile> profiles;
+  // Query supported profiles for each VDA. The order of querying VDAs should
+  // be the same as the order of initializing VDAs. Then the returned profile
+  // can be initialized by corresponding VDA successfully.
+#if defined(OS_WIN)
+  profiles = DXVAVideoDecodeAccelerator::GetSupportedProfiles();
+#elif defined(OS_CHROMEOS)
+#if defined(USE_V4L2_CODEC)
+  std::vector<media::VideoDecodeAccelerator::SupportedProfile> vda_profiles;
+  vda_profiles = V4L2VideoDecodeAccelerator::GetSupportedProfiles();
+  GpuVideoAcceleratorUtil::InsertUniqueDecodeProfiles(vda_profiles, &profiles);
+  vda_profiles = V4L2SliceVideoDecodeAccelerator::GetSupportedProfiles();
+  GpuVideoAcceleratorUtil::InsertUniqueDecodeProfiles(vda_profiles, &profiles);
+#endif
+#if defined(ARCH_CPU_X86_FAMILY)
+  vda_profiles = VaapiVideoDecodeAccelerator::GetSupportedProfiles();
+  GpuVideoAcceleratorUtil::InsertUniqueDecodeProfiles(vda_profiles, &profiles);
+#endif
+#elif defined(OS_MACOSX)
+  profiles = VTVideoDecodeAccelerator::GetSupportedProfiles();
+#elif defined(OS_ANDROID)
+  profiles = AndroidVideoDecodeAccelerator::GetSupportedProfiles();
+#endif
+  return GpuVideoAcceleratorUtil::ConvertMediaToGpuDecodeProfiles(profiles);
+}
+
 // Runs on IO thread if video_decode_accelerator_->CanDecodeOnIOThread() is
 // true, otherwise on the main thread.
 void GpuVideoDecodeAccelerator::OnDecode(