Add JPEG decoder for VAAPI JPEG decode acceleration

content/common/gpu/media/vaapi_jpeg_decoder.cc

Index: content/common/gpu/media/vaapi_jpeg_decoder.cc
diff --git a/content/common/gpu/media/vaapi_jpeg_decoder.cc b/content/common/gpu/media/vaapi_jpeg_decoder.cc
new file mode 100644
index 0000000000000000000000000000000000000000..5bef6869161208a81e1c8c52bee3dff92ff14c65
--- /dev/null
+++ b/content/common/gpu/media/vaapi_jpeg_decoder.cc
@@ -0,0 +1,243 @@
+// Copyright 2014 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "content/common/gpu/media/vaapi_jpeg_decoder.h"
+
+#include <algorithm>

[Owen Lin, 2015/01/13 06:17:58]

not used?

[kcwu, 2015/01/16 15:12:54]

Done.

+#include <limits>

[Owen Lin, 2015/01/13 06:17:58]

not used?

[kcwu, 2015/01/16 15:12:54]

Done.

+
+namespace media {
+
+// K.3.3.1 "Specification of typical tables for DC difference coding"
+static JpegHuffmanTable default_dc_table[kJpegMaxHuffmanTableNum_baseline] = {
+    // luminance DC coefficients
+    {
+     true,
+     {0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0},
+     {0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0a, 0x0b},
+    },
+    // chrominance DC coefficients
+    {
+     true,
+     {0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0},
+     {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 0xa, 0xb},
+    },
+};
+
+// K.3.3.2 "Specification of typical tables for AC coefficient coding"
+static JpegHuffmanTable default_ac_table[kJpegMaxHuffmanTableNum_baseline] = {
+    // luminance AC coefficients
+    {
+     true,
+     {0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d},
+     {0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
+      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
+      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
+      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
+      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
+      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
+      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
+      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
+      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
+      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
+      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa},
+    },
+    // chrominance AC coefficients
+    {
+     true,
+     {0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77},
+     {0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
+      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
+      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
+      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
+      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
+      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
+      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
+      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
+      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
+      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
+      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
+      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa},
+    },
+};
+
+}  // namespace media
+
+namespace content {
+
+VaapiJpegDecoder::VaapiJpegDecoder(
+    VaapiWrapper* vaapi_wrapper,

[wuchengli, 2015/01/12 09:08:51]

VaapiWrapper creation should be done in this class (like VaapiVDA) so the
profile won't mismatch. For example, what if the caller passes a vaapi_wrapper
that uses H264 profile?

[kcwu, 2015/01/16 15:12:54]

This class is modeled after VaapiH264Decoder which create VaapiWrapper outside.

+    const ReportErrorToUmaCB& report_error_to_uma_cb)
+    : vaapi_wrapper_(vaapi_wrapper),
+      report_error_to_uma_cb_(report_error_to_uma_cb) {
+}
+
+VaapiJpegDecoder::~VaapiJpegDecoder() {
+}
+
+VaapiJpegDecoder::DecodeStatus VaapiJpegDecoder::Decode(
+    const uint8* buffer,
+    size_t buffer_size,
+    gfx::Size* output_size,
+    VASurfaceID* output_surface_id) {
+  media::JpegParseResult parse_result;
+  if (!media::JpegParser::Parse(buffer, buffer_size, &parse_result))

[wuchengli, 2015/01/12 09:08:52]

add a DLOG(ERROR) here.

[kcwu, 2015/01/16 15:12:54]

Done.

+    return PARSE_ERROR;
+
+  // Create surfaces

[wuchengli, 2015/01/12 09:08:51]

s/surfaces/a surface/. This only creates one surface.

[kcwu, 2015/01/16 15:12:54]

Done.

+  std::vector<VASurfaceID> va_surfaces;
+  gfx::Size size(parse_result.visible_width, parse_result.visible_height);
+  if (!vaapi_wrapper_->CreateSurfaces(size, 1, &va_surfaces))

[wuchengli, 2015/01/12 13:48:04]

We should reuse the surface to optimize for MJPEG. Each Jpeg has the same size.
Have an Initialize do CreateSurface and move vaapi_wrapper_->DestroySurfaces to
destructor?

[Owen Lin, 2015/01/13 06:17:58]

I believe the difficulty is that we need to parse the buffer to know the picture
size.
So, maybe it would be nice if we parse the the buffer outside the class.

class VaapiJpegDecoder {
  VaapiJpegDecoder(const gfx::Size& size);
  DecodeStatus Decode(const media::JpegParseResult parse_result, VASurfaceID
output);
  VASurfaceID CreateSurface(gfx::Size size);
  DestroySurface(VASurfaceID id);
}

[kcwu, 2015/01/16 15:12:54]

Done.
I moved the responsibility of surface creation to the client.

+    return VAAPI_ERROR;
+
+  // Set picture parameter

[wuchengli, 2015/01/12 09:08:52]

Add periods for all comments. s/parameter/parameters/

[kcwu, 2015/01/16 15:12:54]

Done.

+  VAPictureParameterBufferJPEGBaseline pic_param;
+  memset(&pic_param, 0, sizeof(pic_param));
+  pic_param.picture_width = parse_result.visible_width;
+  pic_param.picture_height = parse_result.visible_height;
+  pic_param.num_components = parse_result.num_components;
+
+  for (int i = 0; i < pic_param.num_components; i++) {
+    pic_param.components[i].component_id = parse_result.components[i].id;
+    pic_param.components[i].h_sampling_factor =
+        parse_result.components[i].horizontal_sampling_factor;
+    pic_param.components[i].v_sampling_factor =
+        parse_result.components[i].vertical_sampling_factor;
+    pic_param.components[i].quantiser_table_selector =
+        parse_result.components[i].quantization_table_selector;
+  }
+
+  if (!vaapi_wrapper_->SubmitBuffer(VAPictureParameterBufferType,
+                                    sizeof(pic_param), &pic_param))
+    return VAAPI_ERROR;
+
+  // Set quantization table

[wuchengli, 2015/01/12 09:08:52]

Break this function into smaller functions. This function is big.

[kcwu, 2015/01/16 15:12:54]

Done.

+  VAIQMatrixBufferJPEGBaseline iq_matrix;
+  memset(&iq_matrix, 0, sizeof(iq_matrix));
+  static_assert(media::kJpegMaxQuantizationTableNum ==
+                    arraysize(iq_matrix.load_quantiser_table),
+                "max number of quantization table mismatched");
+  for (size_t i = 0; i < media::kJpegMaxQuantizationTableNum; i++) {
+    if (!parse_result.q_table[i].valid)
+      continue;
+    iq_matrix.load_quantiser_table[i] = 1;
+    for (int j = 0; j < 64; j++)
+      iq_matrix.quantiser_table[i][j] = parse_result.q_table[i].value[j];
+  }
+
+  if (!vaapi_wrapper_->SubmitBuffer(VAIQMatrixBufferType, sizeof(iq_matrix),
+                                    &iq_matrix))
+    return VAAPI_ERROR;
+
+  // Use default huffman tables if not specified in header.
+  const media::JpegHuffmanTable* dc_table = media::default_dc_table;
+  const media::JpegHuffmanTable* ac_table = media::default_ac_table;
+  for (size_t i = 0; i < media::kJpegMaxHuffmanTableNum_baseline; i++) {
+    if (parse_result.dc_table[i].valid || parse_result.ac_table[i].valid) {
+      dc_table = parse_result.dc_table;
+      ac_table = parse_result.ac_table;
+      break;
+    }
+  }
+
+  // Set huffman table
+  VAHuffmanTableBufferJPEGBaseline huffman_table;
+  memset(&huffman_table, 0, sizeof(huffman_table));
+  static_assert(media::kJpegMaxHuffmanTableNum_baseline ==
+                    arraysize(huffman_table.load_huffman_table),
+                "max number of huffman table mismatched");
+  static_assert(sizeof(huffman_table.huffman_table[0].num_dc_codes) ==
+                    sizeof(dc_table[0].code_length),
+                "size of huffman table code length mismatch");
+  static_assert(sizeof(huffman_table.huffman_table[0].dc_values[0]) ==
+                    sizeof(dc_table[0].code_value[0]),
+                "size of huffman table code value mismatch");
+  for (size_t i = 0; i < media::kJpegMaxHuffmanTableNum_baseline; i++) {
+    if (!dc_table[i].valid || !ac_table[i].valid)
+      continue;
+    huffman_table.load_huffman_table[i] = 1;
+
+    memcpy(huffman_table.huffman_table[i].num_dc_codes, dc_table[i].code_length,
+           sizeof(huffman_table.huffman_table[i].num_dc_codes));
+    memcpy(huffman_table.huffman_table[i].dc_values, dc_table[i].code_value,
+           sizeof(huffman_table.huffman_table[i].dc_values));
+    memcpy(huffman_table.huffman_table[i].num_ac_codes, ac_table[i].code_length,
+           sizeof(huffman_table.huffman_table[i].num_ac_codes));
+    memcpy(huffman_table.huffman_table[i].ac_values, ac_table[i].code_value,
+           sizeof(huffman_table.huffman_table[i].ac_values));
+  }
+
+  if (!vaapi_wrapper_->SubmitBuffer(VAHuffmanTableBufferType,
+                                    sizeof(huffman_table), &huffman_table))
+    return VAAPI_ERROR;
+
+  // Set slice parameter
+  VASliceParameterBufferJPEGBaseline slice_param;
+  memset(&slice_param, 0, sizeof(slice_param));
+  int max_h_factor = parse_result.components[0].horizontal_sampling_factor;
+  int max_v_factor = parse_result.components[0].vertical_sampling_factor;
+  slice_param.slice_data_size = parse_result.scan.data_size;
+  slice_param.slice_data_offset = 0;
+  slice_param.slice_data_flag = VA_SLICE_DATA_FLAG_ALL;
+  slice_param.slice_horizontal_position = 0;
+  slice_param.slice_vertical_position = 0;
+  slice_param.num_components = parse_result.scan.num_components;
+  for (int i = 0; i < slice_param.num_components; i++) {
+    slice_param.components[i].component_selector =
+        parse_result.scan.components[i].component_selector;
+    slice_param.components[i].dc_table_selector =
+        parse_result.scan.components[i].dc_selector;
+    slice_param.components[i].ac_table_selector =
+        parse_result.scan.components[i].ac_selector;
+  }
+  slice_param.restart_interval = parse_result.restart_interval;
+  int mcu_cols =
+      (parse_result.visible_width + max_h_factor * 8 - 1) / (max_h_factor * 8);
+  int mcu_rows =
+      (parse_result.visible_height + max_v_factor * 8 - 1) / (max_v_factor * 8);
+  slice_param.num_mcus = mcu_rows * mcu_cols;
+
+  if (!vaapi_wrapper_->SubmitBuffer(VASliceParameterBufferType,
+                                    sizeof(slice_param), &slice_param))
+    return VAAPI_ERROR;
+
+  // Set scan data
+  if (!vaapi_wrapper_->SubmitBuffer(VASliceDataBufferType,
+                                    parse_result.scan.data_size,
+                                    const_cast<char*>(parse_result.scan.data)))
+    return VAAPI_ERROR;
+
+  if (!vaapi_wrapper_->ExecuteAndDestroyPendingBuffers(va_surfaces[0]))
+    return VAAPI_ERROR;
+
+  *output_size = size;
+  *output_surface_id = va_surfaces[0];
+
+  return SUCCESS;
+}
+
+void VaapiJpegDecoder::DecodeDone() {
+  vaapi_wrapper_->DestroySurfaces();
+}
+
+}  // namespace content