Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit dee634ee authored by Vignesh Venkatasubramanian's avatar Vignesh Venkatasubramanian Committed by Android (Google) Code Review
Browse files

Merge "codec2: Implement gav1 decoder component"

parents 827b9d69 87f6ede9
Loading
Loading
Loading
Loading
+5 −0
Original line number Diff line number Diff line
@@ -6,4 +6,9 @@ cc_library_shared {
    ],

    srcs: ["C2SoftGav1Dec.cpp"],
    static_libs: ["libgav1"],

    include_dirs: [
        "external/libgav1/libgav1/",
    ],
}
+487 −11
Original line number Diff line number Diff line
@@ -256,20 +256,496 @@ C2SoftGav1Dec::C2SoftGav1Dec(const char *name, c2_node_id_t id,
                             const std::shared_ptr<IntfImpl> &intfImpl)
    : SimpleC2Component(
          std::make_shared<SimpleInterface<IntfImpl>>(name, id, intfImpl)),
      mIntf(intfImpl) {}

c2_status_t C2SoftGav1Dec::onInit() { return C2_OK; }
c2_status_t C2SoftGav1Dec::onStop() { return C2_OK; }
void C2SoftGav1Dec::onReset() {}
void C2SoftGav1Dec::onRelease(){};
c2_status_t C2SoftGav1Dec::onFlush_sm() { return C2_OK; }
void C2SoftGav1Dec::process(const std::unique_ptr<C2Work> & /*work*/,
                            const std::shared_ptr<C2BlockPool> & /*pool*/) {}
c2_status_t C2SoftGav1Dec::drain(
    uint32_t /*drainMode*/, const std::shared_ptr<C2BlockPool> & /*pool*/) {
      mIntf(intfImpl),
      mCodecCtx(nullptr) {
  gettimeofday(&mTimeStart, nullptr);
  gettimeofday(&mTimeEnd, nullptr);
}

C2SoftGav1Dec::~C2SoftGav1Dec() { onRelease(); }

c2_status_t C2SoftGav1Dec::onInit() {
  return initDecoder() ? C2_OK : C2_CORRUPTED;
}

c2_status_t C2SoftGav1Dec::onStop() {
  mSignalledError = false;
  mSignalledOutputEos = false;
  return C2_OK;
}

void C2SoftGav1Dec::onReset() {
  (void)onStop();
  c2_status_t err = onFlush_sm();
  if (err != C2_OK) {
    ALOGW("Failed to flush the av1 decoder. Trying to hard reset.");
    destroyDecoder();
    if (!initDecoder()) {
      ALOGE("Hard reset failed.");
    }
  }
}

void C2SoftGav1Dec::onRelease() { destroyDecoder(); }

c2_status_t C2SoftGav1Dec::onFlush_sm() {
  Libgav1StatusCode status =
      mCodecCtx->EnqueueFrame(/*data=*/nullptr, /*size=*/0,
                              /*user_private_data=*/0);
  if (status != kLibgav1StatusOk) {
    ALOGE("Failed to flush av1 decoder. status: %d.", status);
    return C2_CORRUPTED;
  }

  // Dequeue frame (if any) that was enqueued previously.
  const libgav1::DecoderBuffer *buffer;
  status = mCodecCtx->DequeueFrame(&buffer);
  if (status != kLibgav1StatusOk) {
    ALOGE("Failed to dequeue frame after flushing the av1 decoder. status: %d",
          status);
    return C2_CORRUPTED;
  }

  mSignalledError = false;
  mSignalledOutputEos = false;

  return C2_OK;
}

static int GetCPUCoreCount() {
  int cpuCoreCount = 1;
#if defined(_SC_NPROCESSORS_ONLN)
  cpuCoreCount = sysconf(_SC_NPROCESSORS_ONLN);
#else
  // _SC_NPROC_ONLN must be defined...
  cpuCoreCount = sysconf(_SC_NPROC_ONLN);
#endif
  CHECK(cpuCoreCount >= 1);
  ALOGV("Number of CPU cores: %d", cpuCoreCount);
  return cpuCoreCount;
}

bool C2SoftGav1Dec::initDecoder() {
  mSignalledError = false;
  mSignalledOutputEos = false;
  mCodecCtx.reset(new libgav1::Decoder());

  if (mCodecCtx == nullptr) {
    ALOGE("mCodecCtx is null");
    return false;
  }

  libgav1::DecoderSettings settings = {};
  settings.threads = GetCPUCoreCount();

  Libgav1StatusCode status = mCodecCtx->Init(&settings);
  if (status != kLibgav1StatusOk) {
    ALOGE("av1 decoder failed to initialize. status: %d.", status);
    return false;
  }

  return true;
}

void C2SoftGav1Dec::destroyDecoder() { mCodecCtx = nullptr; }

void fillEmptyWork(const std::unique_ptr<C2Work> &work) {
  uint32_t flags = 0;
  if (work->input.flags & C2FrameData::FLAG_END_OF_STREAM) {
    flags |= C2FrameData::FLAG_END_OF_STREAM;
    ALOGV("signalling eos");
  }
  work->worklets.front()->output.flags = (C2FrameData::flags_t)flags;
  work->worklets.front()->output.buffers.clear();
  work->worklets.front()->output.ordinal = work->input.ordinal;
  work->workletsProcessed = 1u;
}

void C2SoftGav1Dec::finishWork(uint64_t index,
                               const std::unique_ptr<C2Work> &work,
                               const std::shared_ptr<C2GraphicBlock> &block) {
  std::shared_ptr<C2Buffer> buffer =
      createGraphicBuffer(block, C2Rect(mWidth, mHeight));
  auto fillWork = [buffer, index](const std::unique_ptr<C2Work> &work) {
    uint32_t flags = 0;
    if ((work->input.flags & C2FrameData::FLAG_END_OF_STREAM) &&
        (c2_cntr64_t(index) == work->input.ordinal.frameIndex)) {
      flags |= C2FrameData::FLAG_END_OF_STREAM;
      ALOGV("signalling eos");
    }
    work->worklets.front()->output.flags = (C2FrameData::flags_t)flags;
    work->worklets.front()->output.buffers.clear();
    work->worklets.front()->output.buffers.push_back(buffer);
    work->worklets.front()->output.ordinal = work->input.ordinal;
    work->workletsProcessed = 1u;
  };
  if (work && c2_cntr64_t(index) == work->input.ordinal.frameIndex) {
    fillWork(work);
  } else {
    finish(index, fillWork);
  }
}

void C2SoftGav1Dec::process(const std::unique_ptr<C2Work> &work,
                            const std::shared_ptr<C2BlockPool> &pool) {
  work->result = C2_OK;
  work->workletsProcessed = 0u;
  work->worklets.front()->output.configUpdate.clear();
  work->worklets.front()->output.flags = work->input.flags;
  if (mSignalledError || mSignalledOutputEos) {
    work->result = C2_BAD_VALUE;
    return;
  }

  size_t inOffset = 0u;
  size_t inSize = 0u;
  C2ReadView rView = mDummyReadView;
  if (!work->input.buffers.empty()) {
    rView = work->input.buffers[0]->data().linearBlocks().front().map().get();
    inSize = rView.capacity();
    if (inSize && rView.error()) {
      ALOGE("read view map failed %d", rView.error());
      work->result = C2_CORRUPTED;
      return;
    }
  }

  bool codecConfig =
      ((work->input.flags & C2FrameData::FLAG_CODEC_CONFIG) != 0);
  bool eos = ((work->input.flags & C2FrameData::FLAG_END_OF_STREAM) != 0);

  ALOGV("in buffer attr. size %zu timestamp %d frameindex %d, flags %x", inSize,
        (int)work->input.ordinal.timestamp.peeku(),
        (int)work->input.ordinal.frameIndex.peeku(), work->input.flags);

  if (codecConfig) {
    fillEmptyWork(work);
    return;
  }

  int64_t frameIndex = work->input.ordinal.frameIndex.peekll();
  if (inSize) {
    uint8_t *bitstream = const_cast<uint8_t *>(rView.data() + inOffset);
    int32_t decodeTime = 0;
    int32_t delay = 0;

    GETTIME(&mTimeStart, nullptr);
    TIME_DIFF(mTimeEnd, mTimeStart, delay);

    const Libgav1StatusCode status =
        mCodecCtx->EnqueueFrame(bitstream, inSize, frameIndex);

    GETTIME(&mTimeEnd, nullptr);
    TIME_DIFF(mTimeStart, mTimeEnd, decodeTime);
    ALOGV("decodeTime=%4d delay=%4d\n", decodeTime, delay);

    if (status != kLibgav1StatusOk) {
      ALOGE("av1 decoder failed to decode frame. status: %d.", status);
      work->result = C2_CORRUPTED;
      work->workletsProcessed = 1u;
      mSignalledError = true;
      return;
    }

  } else {
    const Libgav1StatusCode status =
        mCodecCtx->EnqueueFrame(/*data=*/nullptr, /*size=*/0,
                                /*user_private_data=*/0);
    if (status != kLibgav1StatusOk) {
      ALOGE("Failed to flush av1 decoder. status: %d.", status);
      work->result = C2_CORRUPTED;
      work->workletsProcessed = 1u;
      mSignalledError = true;
      return;
    }
  }

  (void)outputBuffer(pool, work);

  if (eos) {
    drainInternal(DRAIN_COMPONENT_WITH_EOS, pool, work);
    mSignalledOutputEos = true;
  } else if (!inSize) {
    fillEmptyWork(work);
  }
}

static void copyOutputBufferToYV12Frame(uint8_t *dst, const uint8_t *srcY,
                                        const uint8_t *srcU,
                                        const uint8_t *srcV, size_t srcYStride,
                                        size_t srcUStride, size_t srcVStride,
                                        uint32_t width, uint32_t height) {
  const size_t dstYStride = align(width, 16);
  const size_t dstUVStride = align(dstYStride / 2, 16);
  uint8_t *const dstStart = dst;

  for (size_t i = 0; i < height; ++i) {
    memcpy(dst, srcY, width);
    srcY += srcYStride;
    dst += dstYStride;
  }

  dst = dstStart + dstYStride * height;
  for (size_t i = 0; i < height / 2; ++i) {
    memcpy(dst, srcV, width / 2);
    srcV += srcVStride;
    dst += dstUVStride;
  }

  dst = dstStart + (dstYStride * height) + (dstUVStride * height / 2);
  for (size_t i = 0; i < height / 2; ++i) {
    memcpy(dst, srcU, width / 2);
    srcU += srcUStride;
    dst += dstUVStride;
  }
}

static void convertYUV420Planar16ToY410(uint32_t *dst, const uint16_t *srcY,
                                        const uint16_t *srcU,
                                        const uint16_t *srcV, size_t srcYStride,
                                        size_t srcUStride, size_t srcVStride,
                                        size_t dstStride, size_t width,
                                        size_t height) {
  // Converting two lines at a time, slightly faster
  for (size_t y = 0; y < height; y += 2) {
    uint32_t *dstTop = (uint32_t *)dst;
    uint32_t *dstBot = (uint32_t *)(dst + dstStride);
    uint16_t *ySrcTop = (uint16_t *)srcY;
    uint16_t *ySrcBot = (uint16_t *)(srcY + srcYStride);
    uint16_t *uSrc = (uint16_t *)srcU;
    uint16_t *vSrc = (uint16_t *)srcV;

    uint32_t u01, v01, y01, y23, y45, y67, uv0, uv1;
    size_t x = 0;
    for (; x < width - 3; x += 4) {
      u01 = *((uint32_t *)uSrc);
      uSrc += 2;
      v01 = *((uint32_t *)vSrc);
      vSrc += 2;

      y01 = *((uint32_t *)ySrcTop);
      ySrcTop += 2;
      y23 = *((uint32_t *)ySrcTop);
      ySrcTop += 2;
      y45 = *((uint32_t *)ySrcBot);
      ySrcBot += 2;
      y67 = *((uint32_t *)ySrcBot);
      ySrcBot += 2;

      uv0 = (u01 & 0x3FF) | ((v01 & 0x3FF) << 20);
      uv1 = (u01 >> 16) | ((v01 >> 16) << 20);

      *dstTop++ = 3 << 30 | ((y01 & 0x3FF) << 10) | uv0;
      *dstTop++ = 3 << 30 | ((y01 >> 16) << 10) | uv0;
      *dstTop++ = 3 << 30 | ((y23 & 0x3FF) << 10) | uv1;
      *dstTop++ = 3 << 30 | ((y23 >> 16) << 10) | uv1;

      *dstBot++ = 3 << 30 | ((y45 & 0x3FF) << 10) | uv0;
      *dstBot++ = 3 << 30 | ((y45 >> 16) << 10) | uv0;
      *dstBot++ = 3 << 30 | ((y67 & 0x3FF) << 10) | uv1;
      *dstBot++ = 3 << 30 | ((y67 >> 16) << 10) | uv1;
    }

    // There should be at most 2 more pixels to process. Note that we don't
    // need to consider odd case as the buffer is always aligned to even.
    if (x < width) {
      u01 = *uSrc;
      v01 = *vSrc;
      y01 = *((uint32_t *)ySrcTop);
      y45 = *((uint32_t *)ySrcBot);
      uv0 = (u01 & 0x3FF) | ((v01 & 0x3FF) << 20);
      *dstTop++ = ((y01 & 0x3FF) << 10) | uv0;
      *dstTop++ = ((y01 >> 16) << 10) | uv0;
      *dstBot++ = ((y45 & 0x3FF) << 10) | uv0;
      *dstBot++ = ((y45 >> 16) << 10) | uv0;
    }

    srcY += srcYStride * 2;
    srcU += srcUStride;
    srcV += srcVStride;
    dst += dstStride * 2;
  }
}

static void convertYUV420Planar16ToYUV420Planar(
    uint8_t *dst, const uint16_t *srcY, const uint16_t *srcU,
    const uint16_t *srcV, size_t srcYStride, size_t srcUStride,
    size_t srcVStride, size_t dstStride, size_t width, size_t height) {
  uint8_t *dstY = (uint8_t *)dst;
  size_t dstYSize = dstStride * height;
  size_t dstUVStride = align(dstStride / 2, 16);
  size_t dstUVSize = dstUVStride * height / 2;
  uint8_t *dstV = dstY + dstYSize;
  uint8_t *dstU = dstV + dstUVSize;

  for (size_t y = 0; y < height; ++y) {
    for (size_t x = 0; x < width; ++x) {
      dstY[x] = (uint8_t)(srcY[x] >> 2);
    }

    srcY += srcYStride;
    dstY += dstStride;
  }

  for (size_t y = 0; y < (height + 1) / 2; ++y) {
    for (size_t x = 0; x < (width + 1) / 2; ++x) {
      dstU[x] = (uint8_t)(srcU[x] >> 2);
      dstV[x] = (uint8_t)(srcV[x] >> 2);
    }

    srcU += srcUStride;
    srcV += srcVStride;
    dstU += dstUVStride;
    dstV += dstUVStride;
  }
}

bool C2SoftGav1Dec::outputBuffer(const std::shared_ptr<C2BlockPool> &pool,
                                 const std::unique_ptr<C2Work> &work) {
  if (!(work && pool)) return false;

  const libgav1::DecoderBuffer *buffer;
  const Libgav1StatusCode status = mCodecCtx->DequeueFrame(&buffer);

  if (status != kLibgav1StatusOk) {
    ALOGE("av1 decoder DequeueFrame failed. status: %d.", status);
    return false;
  }

  // |buffer| can be NULL if status was equal to kLibgav1StatusOk. This is not
  // an error. This could mean one of two things:
  //  - The EnqueueFrame() call was either a flush (called with nullptr).
  //  - The enqueued frame did not have any displayable frames.
  if (!buffer) {
    return false;
  }

  const int width = buffer->displayed_width[0];
  const int height = buffer->displayed_height[0];
  if (width != mWidth || height != mHeight) {
    mWidth = width;
    mHeight = height;

    C2StreamPictureSizeInfo::output size(0u, mWidth, mHeight);
    std::vector<std::unique_ptr<C2SettingResult>> failures;
    c2_status_t err = mIntf->config({&size}, C2_MAY_BLOCK, &failures);
    if (err == C2_OK) {
      work->worklets.front()->output.configUpdate.push_back(
          C2Param::Copy(size));
    } else {
      ALOGE("Config update size failed");
      mSignalledError = true;
      work->result = C2_CORRUPTED;
      work->workletsProcessed = 1u;
      return false;
    }
  }

  // TODO(vigneshv): Add support for monochrome videos since AV1 supports it.
  CHECK(buffer->image_format == libgav1::kImageFormatYuv420);

  std::shared_ptr<C2GraphicBlock> block;
  uint32_t format = HAL_PIXEL_FORMAT_YV12;
  if (buffer->bitdepth == 10) {
    IntfImpl::Lock lock = mIntf->lock();
    std::shared_ptr<C2StreamColorAspectsTuning::output> defaultColorAspects =
        mIntf->getDefaultColorAspects_l();

    if (defaultColorAspects->primaries == C2Color::PRIMARIES_BT2020 &&
        defaultColorAspects->matrix == C2Color::MATRIX_BT2020 &&
        defaultColorAspects->transfer == C2Color::TRANSFER_ST2084) {
      format = HAL_PIXEL_FORMAT_RGBA_1010102;
    }
  }
  C2MemoryUsage usage = {C2MemoryUsage::CPU_READ, C2MemoryUsage::CPU_WRITE};

  c2_status_t err = pool->fetchGraphicBlock(align(mWidth, 16), mHeight, format,
                                            usage, &block);

  if (err != C2_OK) {
    ALOGE("fetchGraphicBlock for Output failed with status %d", err);
    work->result = err;
    return false;
  }

  C2GraphicView wView = block->map().get();

  if (wView.error()) {
    ALOGE("graphic view map failed %d", wView.error());
    work->result = C2_CORRUPTED;
    return false;
  }

  ALOGV("provided (%dx%d) required (%dx%d), out frameindex %d", block->width(),
        block->height(), mWidth, mHeight, (int)buffer->user_private_data);

  uint8_t *dst = const_cast<uint8_t *>(wView.data()[C2PlanarLayout::PLANE_Y]);
  size_t srcYStride = buffer->stride[0];
  size_t srcUStride = buffer->stride[1];
  size_t srcVStride = buffer->stride[2];

  if (buffer->bitdepth == 10) {
    const uint16_t *srcY = (const uint16_t *)buffer->plane[0];
    const uint16_t *srcU = (const uint16_t *)buffer->plane[1];
    const uint16_t *srcV = (const uint16_t *)buffer->plane[2];

    if (format == HAL_PIXEL_FORMAT_RGBA_1010102) {
      convertYUV420Planar16ToY410(
          (uint32_t *)dst, srcY, srcU, srcV, srcYStride / 2, srcUStride / 2,
          srcVStride / 2, align(mWidth, 16), mWidth, mHeight);
    } else {
      convertYUV420Planar16ToYUV420Planar(dst, srcY, srcU, srcV, srcYStride / 2,
                                          srcUStride / 2, srcVStride / 2,
                                          align(mWidth, 16), mWidth, mHeight);
    }
  } else {
    const uint8_t *srcY = (const uint8_t *)buffer->plane[0];
    const uint8_t *srcU = (const uint8_t *)buffer->plane[1];
    const uint8_t *srcV = (const uint8_t *)buffer->plane[2];
    copyOutputBufferToYV12Frame(dst, srcY, srcU, srcV, srcYStride, srcUStride,
                                srcVStride, mWidth, mHeight);
  }
  finishWork(buffer->user_private_data, work, std::move(block));
  block = nullptr;
  return true;
}

c2_status_t C2SoftGav1Dec::drainInternal(
    uint32_t drainMode, const std::shared_ptr<C2BlockPool> &pool,
    const std::unique_ptr<C2Work> &work) {
  if (drainMode == NO_DRAIN) {
    ALOGW("drain with NO_DRAIN: no-op");
    return C2_OK;
  }
  if (drainMode == DRAIN_CHAIN) {
    ALOGW("DRAIN_CHAIN not supported");
    return C2_OMITTED;
  }

  Libgav1StatusCode status =
      mCodecCtx->EnqueueFrame(/*data=*/nullptr, /*size=*/0,
                              /*user_private_data=*/0);
  if (status != kLibgav1StatusOk) {
    ALOGE("Failed to flush av1 decoder. status: %d.", status);
    return C2_CORRUPTED;
  }

  while (outputBuffer(pool, work)) {
  }

  if (drainMode == DRAIN_COMPONENT_WITH_EOS && work &&
      work->workletsProcessed == 0u) {
    fillEmptyWork(work);
  }

  return C2_OK;
}

c2_status_t C2SoftGav1Dec::drain(uint32_t drainMode,
                                 const std::shared_ptr<C2BlockPool> &pool) {
  return drainInternal(drainMode, pool, nullptr);
}

class C2SoftGav1Factory : public C2ComponentFactory {
 public:
  C2SoftGav1Factory()
+9 −0
Original line number Diff line number Diff line
@@ -18,6 +18,13 @@
#define ANDROID_C2_SOFT_GAV1_DEC_H_

#include <SimpleC2Component.h>
#include "libgav1/src/decoder.h"
#include "libgav1/src/decoder_settings.h"

#define GETTIME(a, b) gettimeofday(a, b);
#define TIME_DIFF(start, end, diff)     \
    diff = (((end).tv_sec - (start).tv_sec) * 1000000) + \
            ((end).tv_usec - (start).tv_usec);

namespace android {

@@ -26,6 +33,7 @@ struct C2SoftGav1Dec : public SimpleC2Component {

  C2SoftGav1Dec(const char* name, c2_node_id_t id,
                const std::shared_ptr<IntfImpl>& intfImpl);
  ~C2SoftGav1Dec();

  // Begin SimpleC2Component overrides.
  c2_status_t onInit() override;
@@ -41,6 +49,7 @@ struct C2SoftGav1Dec : public SimpleC2Component {

 private:
  std::shared_ptr<IntfImpl> mIntf;
  std::unique_ptr<libgav1::Decoder> mCodecCtx;

  uint32_t mWidth;
  uint32_t mHeight;