Support NV12 format in video decoding (#2330)

Summary:
Support NV12 format in Streamer API.

NV12 is a biplanar format with a full sized Y plane followed by a single chroma plane with weaved U and V values.
https://chromium.googlesource.com/libyuv/libyuv/+/HEAD/docs/formats.md#nv12-and-nv21

The original UV plane is smaller than Y plane, so in this implmentation,
UV plane is upsampled to match the size of Y plane.

Pull Request resolved: #2330

Reviewed By: hwangjeff

Differential Revision: D35632351

Pulled By: mthrok

fbshipit-source-id: aab4fbc0ce2bb7a1fb67264c27208b610fb56e27

torchaudio/csrc/ffmpeg/buffer.cpp

@@ -2,6 +2,10 @@
 #include <stdexcept>
 #include <vector>
 
+#ifdef USE_CUDA
+#include <c10/cuda/CUDAStream.h>
+#endif
+
 namespace torchaudio {
 namespace ffmpeg {
 
@@ -213,6 +217,94 @@ torch::Tensor convert_yuv420p(AVFrame* pFrame) {
   return t.permute({0, 3, 1, 2}); // NCHW
 }
 
+torch::Tensor convert_nv12_cpu(AVFrame* pFrame) {
+  int width = pFrame->width;
+  int height = pFrame->height;
+
+  auto options = torch::TensorOptions()
+                     .dtype(torch::kUInt8)
+                     .layout(torch::kStrided)
+                     .device(torch::kCPU);
+
+  torch::Tensor y = torch::empty({1, height, width, 1}, options);
+  {
+    uint8_t* tgt = y.data_ptr<uint8_t>();
+    uint8_t* src = pFrame->data[0];
+    int linesize = pFrame->linesize[0];
+    for (int h = 0; h < height; ++h) {
+      memcpy(tgt, src, width);
+      tgt += width;
+      src += linesize;
+    }
+  }
+  torch::Tensor uv = torch::empty({1, height / 2, width / 2, 2}, options);
+  {
+    uint8_t* tgt = uv.data_ptr<uint8_t>();
+    uint8_t* src = pFrame->data[1];
+    int linesize = pFrame->linesize[1];
+    for (int h = 0; h < height / 2; ++h) {
+      memcpy(tgt, src, width);
+      tgt += width;
+      src += linesize;
+    }
+  }
+  // Upsample width and height
+  uv = uv.repeat_interleave(2, -2).repeat_interleave(2, -3);
+  torch::Tensor t = torch::cat({y, uv}, -1);
+  return t.permute({0, 3, 1, 2}); // NCHW
+}
+
+#ifdef USE_CUDA
+torch::Tensor convert_nv12_cuda(AVFrame* pFrame) {
+  int width = pFrame->width;
+  int height = pFrame->height;
+
+  auto options = torch::TensorOptions()
+                     .dtype(torch::kUInt8)
+                     .layout(torch::kStrided)
+                     .device(torch::kCUDA);
+
+  torch::Tensor y = torch::empty({1, height, width, 1}, options);
+  {
+    uint8_t* tgt = y.data_ptr<uint8_t>();
+    CUdeviceptr src = (CUdeviceptr)pFrame->data[0];
+    int linesize = pFrame->linesize[0];
+    if (cudaSuccess !=
+        cudaMemcpy2D(
+            (void*)tgt,
+            width,
+            (const void*)src,
+            linesize,
+            width,
+            height,
+            cudaMemcpyDeviceToDevice)) {
+      throw std::runtime_error("Failed to copy Y plane to Cuda tensor.");
+    }
+  }
+  torch::Tensor uv = torch::empty({1, height / 2, width / 2, 2}, options);
+  {
+    uint8_t* tgt = uv.data_ptr<uint8_t>();
+    CUdeviceptr src = (CUdeviceptr)pFrame->data[1];
+    int linesize = pFrame->linesize[1];
+    if (cudaSuccess !=
+        cudaMemcpy2D(
+            (void*)tgt,
+            width,
+            (const void*)src,
+            linesize,
+            width,
+            height / 2,
+            cudaMemcpyDeviceToDevice)) {
+      throw std::runtime_error("Failed to copy UV plane to Cuda tensor.");
+    }
+  }
+  // Upsample width and height
+  uv = uv.repeat_interleave(2, -2).repeat_interleave(2, -3);
+  torch::Tensor t = torch::cat({y, uv}, -1);
+  return t.permute({0, 3, 1, 2}); // NCHW
+}
+#endif
+
 torch::Tensor convert_image_tensor(AVFrame* pFrame) {
   // ref:
   // https://ffmpeg.org/doxygen/4.1/filtering__video_8c_source.html#l00179
@@ -240,9 +332,35 @@ torch::Tensor convert_image_tensor(AVFrame* pFrame) {
       break;
     case AV_PIX_FMT_YUV420P:
       return convert_yuv420p(pFrame);
+    case AV_PIX_FMT_NV12:
+      return convert_nv12_cpu(pFrame);
+#ifdef USE_CUDA
+    case AV_PIX_FMT_CUDA: {
+      AVHWFramesContext* hwctx =
+          (AVHWFramesContext*)pFrame->hw_frames_ctx->data;
+      AVPixelFormat sw_format = hwctx->sw_format;
+      // cuvid decoder (nvdec frontend of ffmpeg) only supports the following
+      // output formats
+      // https://github.com/FFmpeg/FFmpeg/blob/072101bd52f7f092ee976f4e6e41c19812ad32fd/libavcodec/cuviddec.c#L1121-L1124
+      switch (sw_format) {
+        case AV_PIX_FMT_NV12:
+          return convert_nv12_cuda(pFrame);
+        case AV_PIX_FMT_P010:
+        case AV_PIX_FMT_P016:
+          throw std::runtime_error(
+              "Unsupported video format found in CUDA HW: " +
+              std::string(av_get_pix_fmt_name(sw_format)));
+        default:
+          throw std::runtime_error(
+              "Unexpected video format found in CUDA HW: " +
+              std::string(av_get_pix_fmt_name(sw_format)));
+      }
+    }
+#endif
     default:
       throw std::runtime_error(
-          "Unexpected format: " + std::string(av_get_pix_fmt_name(format)));
+          "Unexpected video format: " +
+          std::string(av_get_pix_fmt_name(format)));
   }
 
   torch::Tensor t;