static int hwframe_pool_prealloc(AVBufferRef *ref)
{
AVHWFramesContext *ctx = (AVHWFramesContext*)ref->data;
AVFrame **frames;
int i, ret = 0;
frames = av_mallocz_array(ctx->initial_pool_size, sizeof(*frames));
if (!frames)
return AVERROR(ENOMEM);
for (i = 0; i < ctx->initial_pool_size; i++) {
frames[i] = av_frame_alloc();
if (!frames[i])
goto fail;
ret = av_hwframe_get_buffer(ref, frames[i], 0);
if (ret < 0)
goto fail;
}
fail:
for (i = 0; i < ctx->initial_pool_size; i++)
av_frame_free(&frames[i]);
av_freep(&frames);
return ret;
}
static int dxva2_get_buffer(AVCodecContext *s, AVFrame *frame, int flags)
{
InputStream *ist = s->opaque;
DXVA2Context *ctx = ist->hwaccel_ctx;
return av_hwframe_get_buffer(ctx->hw_frames_ctx, frame, 0);
}
AVFrame *ff_default_get_video_buffer(AVFilterLink *link, int w, int h)
{
AVFrame *frame = NULL;
int pool_width = 0;
int pool_height = 0;
int pool_align = 0;
enum AVPixelFormat pool_format = AV_PIX_FMT_NONE;
if (link->hw_frames_ctx &&
((AVHWFramesContext*)link->hw_frames_ctx->data)->format == link->format) {
int ret;
AVFrame *frame = av_frame_alloc();
if (!frame)
return NULL;
ret = av_hwframe_get_buffer(link->hw_frames_ctx, frame, 0);
if (ret < 0)
av_frame_free(&frame);
return frame;
}
if (!link->frame_pool) {
link->frame_pool = ff_frame_pool_video_init(av_buffer_allocz, w, h,
link->format, BUFFER_ALIGN);
if (!link->frame_pool)
return NULL;
} else {
if (ff_frame_pool_get_video_config(link->frame_pool,
&pool_width, &pool_height,
&pool_format, &pool_align) < 0) {
return NULL;
}
if (pool_width != w || pool_height != h ||
pool_format != link->format || pool_align != BUFFER_ALIGN) {
ff_frame_pool_uninit((FFFramePool **)&link->frame_pool);
link->frame_pool = ff_frame_pool_video_init(av_buffer_allocz, w, h,
link->format, BUFFER_ALIGN);
if (!link->frame_pool)
return NULL;
}
}
frame = ff_frame_pool_get(link->frame_pool);
if (!frame)
return NULL;
frame->sample_aspect_ratio = link->sample_aspect_ratio;
return frame;
}
int av_hwframe_get_buffer(AVBufferRef *hwframe_ref, AVFrame *frame, int flags)
{
AVHWFramesContext *ctx = (AVHWFramesContext*)hwframe_ref->data;
int ret;
if (ctx->internal->source_frames) {
// This is a derived frame context, so we allocate in the source
// and map the frame immediately.
AVFrame *src_frame;
src_frame = av_frame_alloc();
if (!src_frame)
return AVERROR(ENOMEM);
ret = av_hwframe_get_buffer(ctx->internal->source_frames,
src_frame, 0);
if (ret < 0)
return ret;
ret = av_hwframe_map(frame, src_frame, 0);
if (ret) {
av_log(ctx, AV_LOG_ERROR, "Failed to map frame into derived "
"frame context: %d.\n", ret);
av_frame_free(&src_frame);
return ret;
}
// Free the source frame immediately - the mapped frame still
// contains a reference to it.
av_frame_free(&src_frame);
return 0;
}
if (!ctx->internal->hw_type->frames_get_buffer)
return AVERROR(ENOSYS);
if (!ctx->pool)
return AVERROR(EINVAL);
frame->hw_frames_ctx = av_buffer_ref(hwframe_ref);
if (!frame->hw_frames_ctx)
return AVERROR(ENOMEM);
ret = ctx->internal->hw_type->frames_get_buffer(ctx, frame);
if (ret < 0) {
av_buffer_unref(&frame->hw_frames_ctx);
return ret;
}
return 0;
}
static av_cold int init_stage(CUDAScaleContext *s, AVBufferRef *device_ctx)
{
AVBufferRef *out_ref = NULL;
AVHWFramesContext *out_ctx;
int in_sw, in_sh, out_sw, out_sh;
int ret, i;
av_pix_fmt_get_chroma_sub_sample(s->in_fmt, &in_sw, &in_sh);
av_pix_fmt_get_chroma_sub_sample(s->out_fmt, &out_sw, &out_sh);
if (!s->planes_out[0].width) {
s->planes_out[0].width = s->planes_in[0].width;
s->planes_out[0].height = s->planes_in[0].height;
}
for (i = 1; i < FF_ARRAY_ELEMS(s->planes_in); i++) {
s->planes_in[i].width = s->planes_in[0].width >> in_sw;
s->planes_in[i].height = s->planes_in[0].height >> in_sh;
s->planes_out[i].width = s->planes_out[0].width >> out_sw;
s->planes_out[i].height = s->planes_out[0].height >> out_sh;
}
out_ref = av_hwframe_ctx_alloc(device_ctx);
if (!out_ref)
return AVERROR(ENOMEM);
out_ctx = (AVHWFramesContext*)out_ref->data;
out_ctx->format = AV_PIX_FMT_CUDA;
out_ctx->sw_format = s->out_fmt;
out_ctx->width = FFALIGN(s->planes_out[0].width, 32);
out_ctx->height = FFALIGN(s->planes_out[0].height, 32);
ret = av_hwframe_ctx_init(out_ref);
if (ret < 0)
goto fail;
av_frame_unref(s->frame);
ret = av_hwframe_get_buffer(out_ref, s->frame, 0);
if (ret < 0)
goto fail;
s->frame->width = s->planes_out[0].width;
s->frame->height = s->planes_out[0].height;
av_buffer_unref(&s->frames_ctx);
s->frames_ctx = out_ref;
return 0;
fail:
av_buffer_unref(&out_ref);
return ret;
}
/* get the output surface */
static QSVFrame *query_frame(QSVVPPContext *s, AVFilterLink *outlink)
{
AVFilterContext *ctx = outlink->src;
QSVFrame *out_frame;
int ret;
clear_unused_frames(s->out_frame_list);
out_frame = get_free_frame(&s->out_frame_list);
if (!out_frame)
return NULL;
/* For video memory, get a hw frame;
* For system memory, get a sw frame and map it into a mfx_surface. */
if (!IS_SYSTEM_MEMORY(s->out_mem_mode)) {
out_frame->frame = av_frame_alloc();
if (!out_frame->frame)
return NULL;
ret = av_hwframe_get_buffer(outlink->hw_frames_ctx, out_frame->frame, 0);
if (ret < 0) {
av_log(ctx, AV_LOG_ERROR, "Can't allocate a surface.\n");
return NULL;
}
out_frame->surface = (mfxFrameSurface1 *)out_frame->frame->data[3];
} else {
/* Get a frame with aligned dimensions.
* Libmfx need system memory being 128x64 aligned */
out_frame->frame = ff_get_video_buffer(outlink,
FFALIGN(outlink->w, 128),
FFALIGN(outlink->h, 64));
if (!out_frame->frame)
return NULL;
out_frame->frame->width = outlink->w;
out_frame->frame->height = outlink->h;
ret = map_frame_to_surface(out_frame->frame,
&out_frame->surface_internal);
if (ret < 0)
return NULL;
out_frame->surface = &out_frame->surface_internal;
}
out_frame->surface->Info = s->vpp_param.vpp.Out;
return out_frame;
}
int CDecD3D11::get_d3d11_buffer(struct AVCodecContext *c, AVFrame *frame, int flags)
{
CDecD3D11 *pDec = (CDecD3D11 *)c->opaque;
HRESULT hr = S_OK;
if (frame->format != AV_PIX_FMT_D3D11) {
DbgLog((LOG_ERROR, 10, L"D3D11 buffer request, but not D3D11 pixfmt"));
pDec->m_bFailHWDecode = TRUE;
return -1;
}
hr = pDec->ReInitD3D11Decoder(c);
if (FAILED(hr)) {
pDec->m_bFailHWDecode = TRUE;
return -1;
}
if (pDec->m_bReadBackFallback && pDec->m_pFramesCtx)
{
int ret = av_hwframe_get_buffer(pDec->m_pFramesCtx, frame, 0);
frame->width = c->coded_width;
frame->height = c->coded_height;
return ret;
}
else if (pDec->m_bReadBackFallback == false && pDec->m_pAllocator)
{
IMediaSample *pSample = nullptr;
hr = pDec->m_pAllocator->GetBuffer(&pSample, nullptr, nullptr, 0);
if (SUCCEEDED(hr))
{
CD3D11MediaSample *pD3D11Sample = dynamic_cast<CD3D11MediaSample *>(pSample);
// fill the frame from the sample, including a reference to the sample
pD3D11Sample->GetAVFrameBuffer(frame);
frame->width = c->coded_width;
frame->height = c->coded_height;
// the frame holds the sample now, can release the direct interface
pD3D11Sample->Release();
return 0;
}
}
return -1;
}
int avcodec_default_get_buffer2(AVCodecContext *avctx, AVFrame *frame, int flags)
{
int ret;
if (avctx->hw_frames_ctx) {
ret = av_hwframe_get_buffer(avctx->hw_frames_ctx, frame, 0);
frame->width = avctx->coded_width;
frame->height = avctx->coded_height;
return ret;
}
if ((ret = update_frame_pool(avctx, frame)) < 0)
return ret;
switch (avctx->codec_type) {
case AVMEDIA_TYPE_VIDEO:
return video_get_buffer(avctx, frame);
case AVMEDIA_TYPE_AUDIO:
return audio_get_buffer(avctx, frame);
default:
return -1;
}
}
static int cudascale_scale(AVFilterContext *ctx, AVFrame *out, AVFrame *in)
{
CUDAScaleContext *s = ctx->priv;
AVFrame *src = in;
int ret;
ret = scalecuda_resize(ctx, s->frame, src);
if (ret < 0)
return ret;
src = s->frame;
ret = av_hwframe_get_buffer(src->hw_frames_ctx, s->tmp_frame, 0);
if (ret < 0)
return ret;
av_frame_move_ref(out, s->frame);
av_frame_move_ref(s->frame, s->tmp_frame);
ret = av_frame_copy_props(out, in);
if (ret < 0)
return ret;
return 0;
}
static int vdpau_get_buffer(AVCodecContext *s, AVFrame *frame, int flags)
{
VDPAUContext *ctx = s->opaque;
return av_hwframe_get_buffer(ctx->hw_frames_ctx, frame, 0);
}
static int cuvid_output_frame(AVCodecContext *avctx, AVFrame *frame)
{
CuvidContext *ctx = avctx->priv_data;
AVHWDeviceContext *device_ctx = (AVHWDeviceContext*)ctx->hwdevice->data;
AVCUDADeviceContext *device_hwctx = device_ctx->hwctx;
CUcontext dummy, cuda_ctx = device_hwctx->cuda_ctx;
CUdeviceptr mapped_frame = 0;
int ret = 0, eret = 0;
av_log(avctx, AV_LOG_TRACE, "cuvid_output_frame\n");
if (ctx->decoder_flushing) {
ret = cuvid_decode_packet(avctx, NULL);
if (ret < 0 && ret != AVERROR_EOF)
return ret;
}
ret = CHECK_CU(cuCtxPushCurrent(cuda_ctx));
if (ret < 0)
return ret;
if (av_fifo_size(ctx->frame_queue)) {
CuvidParsedFrame parsed_frame;
CUVIDPROCPARAMS params;
unsigned int pitch = 0;
int offset = 0;
int i;
av_fifo_generic_read(ctx->frame_queue, &parsed_frame, sizeof(CuvidParsedFrame), NULL);
memset(¶ms, 0, sizeof(params));
params.progressive_frame = parsed_frame.dispinfo.progressive_frame;
params.second_field = parsed_frame.second_field;
params.top_field_first = parsed_frame.dispinfo.top_field_first;
ret = CHECK_CU(cuvidMapVideoFrame(ctx->cudecoder, parsed_frame.dispinfo.picture_index, &mapped_frame, &pitch, ¶ms));
if (ret < 0)
goto error;
if (avctx->pix_fmt == AV_PIX_FMT_CUDA) {
ret = av_hwframe_get_buffer(ctx->hwframe, frame, 0);
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "av_hwframe_get_buffer failed\n");
goto error;
}
ret = ff_decode_frame_props(avctx, frame);
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "ff_decode_frame_props failed\n");
goto error;
}
for (i = 0; i < 2; i++) {
CUDA_MEMCPY2D cpy = {
.srcMemoryType = CU_MEMORYTYPE_DEVICE,
.dstMemoryType = CU_MEMORYTYPE_DEVICE,
.srcDevice = mapped_frame,
.dstDevice = (CUdeviceptr)frame->data[i],
.srcPitch = pitch,
.dstPitch = frame->linesize[i],
.srcY = offset,
.WidthInBytes = FFMIN(pitch, frame->linesize[i]),
.Height = avctx->height >> (i ? 1 : 0),
};
ret = CHECK_CU(cuMemcpy2D(&cpy));
if (ret < 0)
goto error;
offset += avctx->coded_height;
}
} else if (avctx->pix_fmt == AV_PIX_FMT_NV12) {
static int vaapi_encode_issue(AVCodecContext *avctx,
VAAPIEncodePicture *pic)
{
VAAPIEncodeContext *ctx = avctx->priv_data;
VAAPIEncodeSlice *slice;
VAStatus vas;
int err, i;
char data[MAX_PARAM_BUFFER_SIZE];
size_t bit_len;
av_log(avctx, AV_LOG_DEBUG, "Issuing encode for pic %"PRId64"/%"PRId64" "
"as type %s.\n", pic->display_order, pic->encode_order,
picture_type_name[pic->type]);
if (pic->nb_refs == 0) {
av_log(avctx, AV_LOG_DEBUG, "No reference pictures.\n");
} else {
av_log(avctx, AV_LOG_DEBUG, "Refers to:");
for (i = 0; i < pic->nb_refs; i++) {
av_log(avctx, AV_LOG_DEBUG, " %"PRId64"/%"PRId64,
pic->refs[i]->display_order, pic->refs[i]->encode_order);
}
av_log(avctx, AV_LOG_DEBUG, ".\n");
}
av_assert0(pic->input_available && !pic->encode_issued);
for (i = 0; i < pic->nb_refs; i++) {
av_assert0(pic->refs[i]);
// If we are serialised then the references must have already
// completed. If not, they must have been issued but need not
// have completed yet.
if (ctx->issue_mode == ISSUE_MODE_SERIALISE_EVERYTHING)
av_assert0(pic->refs[i]->encode_complete);
else
av_assert0(pic->refs[i]->encode_issued);
}
av_log(avctx, AV_LOG_DEBUG, "Input surface is %#x.\n", pic->input_surface);
pic->recon_image = av_frame_alloc();
if (!pic->recon_image) {
err = AVERROR(ENOMEM);
goto fail;
}
err = av_hwframe_get_buffer(ctx->recon_frames_ref, pic->recon_image, 0);
if (err < 0) {
err = AVERROR(ENOMEM);
goto fail;
}
pic->recon_surface = (VASurfaceID)(uintptr_t)pic->recon_image->data[3];
av_log(avctx, AV_LOG_DEBUG, "Recon surface is %#x.\n", pic->recon_surface);
vas = vaCreateBuffer(ctx->hwctx->display, ctx->va_context,
VAEncCodedBufferType,
MAX_OUTPUT_BUFFER_SIZE, 1, 0,
&pic->output_buffer);
if (vas != VA_STATUS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "Failed to create bitstream "
"output buffer: %d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(ENOMEM);
goto fail;
}
av_log(avctx, AV_LOG_DEBUG, "Output buffer is %#x.\n",
pic->output_buffer);
if (ctx->codec->picture_params_size > 0) {
pic->codec_picture_params = av_malloc(ctx->codec->picture_params_size);
if (!pic->codec_picture_params)
goto fail;
memcpy(pic->codec_picture_params, ctx->codec_picture_params,
ctx->codec->picture_params_size);
} else {
av_assert0(!ctx->codec_picture_params);
}
pic->nb_param_buffers = 0;
if (pic->encode_order == 0) {
// Global parameter buffers are set on the first picture only.
for (i = 0; i < ctx->nb_global_params; i++) {
err = vaapi_encode_make_param_buffer(avctx, pic,
VAEncMiscParameterBufferType,
(char*)ctx->global_params[i],
ctx->global_params_size[i]);
if (err < 0)
goto fail;
}
}
if (pic->type == PICTURE_TYPE_IDR && ctx->codec->init_sequence_params) {
err = vaapi_encode_make_param_buffer(avctx, pic,
VAEncSequenceParameterBufferType,
ctx->codec_sequence_params,
ctx->codec->sequence_params_size);
if (err < 0)
goto fail;
}
if (ctx->codec->init_picture_params) {
err = ctx->codec->init_picture_params(avctx, pic);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to initialise picture "
"parameters: %d.\n", err);
goto fail;
}
err = vaapi_encode_make_param_buffer(avctx, pic,
VAEncPictureParameterBufferType,
pic->codec_picture_params,
ctx->codec->picture_params_size);
if (err < 0)
goto fail;
}
if (pic->type == PICTURE_TYPE_IDR) {
if (ctx->codec->write_sequence_header) {
bit_len = 8 * sizeof(data);
err = ctx->codec->write_sequence_header(avctx, data, &bit_len);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to write per-sequence "
"header: %d.\n", err);
goto fail;
}
err = vaapi_encode_make_packed_header(avctx, pic,
ctx->codec->sequence_header_type,
data, bit_len);
if (err < 0)
goto fail;
}
}
if (ctx->codec->write_picture_header) {
bit_len = 8 * sizeof(data);
err = ctx->codec->write_picture_header(avctx, pic, data, &bit_len);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to write per-picture "
"header: %d.\n", err);
goto fail;
}
err = vaapi_encode_make_packed_header(avctx, pic,
ctx->codec->picture_header_type,
data, bit_len);
if (err < 0)
goto fail;
}
if (ctx->codec->write_extra_buffer) {
for (i = 0;; i++) {
size_t len = sizeof(data);
int type;
err = ctx->codec->write_extra_buffer(avctx, pic, i, &type,
data, &len);
if (err == AVERROR_EOF)
break;
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to write extra "
"buffer %d: %d.\n", i, err);
goto fail;
}
err = vaapi_encode_make_param_buffer(avctx, pic, type,
data, len);
if (err < 0)
goto fail;
}
}
av_assert0(pic->nb_slices <= MAX_PICTURE_SLICES);
for (i = 0; i < pic->nb_slices; i++) {
slice = av_mallocz(sizeof(*slice));
if (!slice) {
err = AVERROR(ENOMEM);
goto fail;
}
pic->slices[i] = slice;
if (ctx->codec->slice_params_size > 0) {
slice->codec_slice_params = av_mallocz(ctx->codec->slice_params_size);
if (!slice->codec_slice_params) {
err = AVERROR(ENOMEM);
goto fail;
}
}
if (ctx->codec->init_slice_params) {
err = ctx->codec->init_slice_params(avctx, pic, slice);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to initalise slice "
"parameters: %d.\n", err);
goto fail;
}
}
if (ctx->codec->write_slice_header) {
bit_len = 8 * sizeof(data);
err = ctx->codec->write_slice_header(avctx, pic, slice,
data, &bit_len);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to write per-slice "
"header: %d.\n", err);
goto fail;
}
err = vaapi_encode_make_packed_header(avctx, pic,
ctx->codec->slice_header_type,
data, bit_len);
if (err < 0)
goto fail;
}
if (ctx->codec->init_slice_params) {
err = vaapi_encode_make_param_buffer(avctx, pic,
VAEncSliceParameterBufferType,
slice->codec_slice_params,
ctx->codec->slice_params_size);
if (err < 0)
goto fail;
}
}
vas = vaBeginPicture(ctx->hwctx->display, ctx->va_context,
pic->input_surface);
if (vas != VA_STATUS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "Failed to begin picture encode issue: "
"%d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail_with_picture;
}
vas = vaRenderPicture(ctx->hwctx->display, ctx->va_context,
pic->param_buffers, pic->nb_param_buffers);
if (vas != VA_STATUS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "Failed to upload encode parameters: "
"%d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail_with_picture;
}
vas = vaEndPicture(ctx->hwctx->display, ctx->va_context);
if (vas != VA_STATUS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "Failed to end picture encode issue: "
"%d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail_at_end;
}
pic->encode_issued = 1;
if (ctx->issue_mode == ISSUE_MODE_SERIALISE_EVERYTHING)
return vaapi_encode_wait(avctx, pic);
else
return 0;
fail_with_picture:
vaEndPicture(ctx->hwctx->display, ctx->va_context);
fail:
for(i = 0; i < pic->nb_param_buffers; i++)
vaDestroyBuffer(ctx->hwctx->display, pic->param_buffers[i]);
fail_at_end:
av_freep(&pic->codec_picture_params);
av_frame_free(&pic->recon_image);
return err;
}
static int cuvid_output_frame(AVCodecContext *avctx, AVFrame *frame)
{
CuvidContext *ctx = avctx->priv_data;
AVHWDeviceContext *device_ctx = (AVHWDeviceContext*)ctx->hwdevice->data;
AVCUDADeviceContext *device_hwctx = device_ctx->hwctx;
CUcontext dummy, cuda_ctx = device_hwctx->cuda_ctx;
CUdeviceptr mapped_frame = 0;
int ret = 0, eret = 0;
av_log(avctx, AV_LOG_TRACE, "cuvid_output_frame\n");
if (ctx->decoder_flushing) {
ret = cuvid_decode_packet(avctx, NULL);
if (ret < 0 && ret != AVERROR_EOF)
return ret;
}
if (!cuvid_is_buffer_full(avctx)) {
AVPacket pkt = {0};
ret = ff_decode_get_packet(avctx, &pkt);
if (ret < 0 && ret != AVERROR_EOF)
return ret;
ret = cuvid_decode_packet(avctx, &pkt);
av_packet_unref(&pkt);
// cuvid_is_buffer_full() should avoid this.
if (ret == AVERROR(EAGAIN))
ret = AVERROR_EXTERNAL;
if (ret < 0 && ret != AVERROR_EOF)
return ret;
}
ret = CHECK_CU(ctx->cudl->cuCtxPushCurrent(cuda_ctx));
if (ret < 0)
return ret;
if (av_fifo_size(ctx->frame_queue)) {
const AVPixFmtDescriptor *pixdesc;
CuvidParsedFrame parsed_frame;
CUVIDPROCPARAMS params;
unsigned int pitch = 0;
int offset = 0;
int i;
av_fifo_generic_read(ctx->frame_queue, &parsed_frame, sizeof(CuvidParsedFrame), NULL);
memset(¶ms, 0, sizeof(params));
params.progressive_frame = parsed_frame.dispinfo.progressive_frame;
params.second_field = parsed_frame.second_field;
params.top_field_first = parsed_frame.dispinfo.top_field_first;
ret = CHECK_CU(ctx->cvdl->cuvidMapVideoFrame(ctx->cudecoder, parsed_frame.dispinfo.picture_index, &mapped_frame, &pitch, ¶ms));
if (ret < 0)
goto error;
if (avctx->pix_fmt == AV_PIX_FMT_CUDA) {
ret = av_hwframe_get_buffer(ctx->hwframe, frame, 0);
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "av_hwframe_get_buffer failed\n");
goto error;
}
ret = ff_decode_frame_props(avctx, frame);
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "ff_decode_frame_props failed\n");
goto error;
}
pixdesc = av_pix_fmt_desc_get(avctx->sw_pix_fmt);
for (i = 0; i < pixdesc->nb_components; i++) {
int height = avctx->height >> (i ? pixdesc->log2_chroma_h : 0);
CUDA_MEMCPY2D cpy = {
.srcMemoryType = CU_MEMORYTYPE_DEVICE,
.dstMemoryType = CU_MEMORYTYPE_DEVICE,
.srcDevice = mapped_frame,
.dstDevice = (CUdeviceptr)frame->data[i],
.srcPitch = pitch,
.dstPitch = frame->linesize[i],
.srcY = offset,
.WidthInBytes = FFMIN(pitch, frame->linesize[i]),
.Height = height,
};
ret = CHECK_CU(ctx->cudl->cuMemcpy2DAsync(&cpy, device_hwctx->stream));
if (ret < 0)
goto error;
offset += height;
}
} else if (avctx->pix_fmt == AV_PIX_FMT_NV12 ||
avctx->pix_fmt == AV_PIX_FMT_P010 ||
avctx->pix_fmt == AV_PIX_FMT_P016 ||
avctx->pix_fmt == AV_PIX_FMT_YUV444P ||
avctx->pix_fmt == AV_PIX_FMT_YUV444P16) {
unsigned int offset = 0;
AVFrame *tmp_frame = av_frame_alloc();
if (!tmp_frame) {
av_log(avctx, AV_LOG_ERROR, "av_frame_alloc failed\n");
ret = AVERROR(ENOMEM);
goto error;
}
pixdesc = av_pix_fmt_desc_get(avctx->sw_pix_fmt);
tmp_frame->format = AV_PIX_FMT_CUDA;
tmp_frame->hw_frames_ctx = av_buffer_ref(ctx->hwframe);
tmp_frame->width = avctx->width;
tmp_frame->height = avctx->height;
/*
* Note that the following logic would not work for three plane
* YUV420 because the pitch value is different for the chroma
* planes.
*/
for (i = 0; i < pixdesc->nb_components; i++) {
tmp_frame->data[i] = (uint8_t*)mapped_frame + offset;
tmp_frame->linesize[i] = pitch;
offset += pitch * (avctx->height >> (i ? pixdesc->log2_chroma_h : 0));
}
ret = ff_get_buffer(avctx, frame, 0);
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "ff_get_buffer failed\n");
av_frame_free(&tmp_frame);
goto error;
}
ret = av_hwframe_transfer_data(frame, tmp_frame, 0);
if (ret) {
av_log(avctx, AV_LOG_ERROR, "av_hwframe_transfer_data failed\n");
av_frame_free(&tmp_frame);
goto error;
}
av_frame_free(&tmp_frame);
} else {
static int qsv_get_buffer(AVCodecContext *s, AVFrame *frame, int flags)
{
InputStream *ist = s->opaque;
return av_hwframe_get_buffer(ist->hw_frames_ctx, frame, 0);
}
static int cuvid_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, AVPacket *avpkt)
{
CuvidContext *ctx = avctx->priv_data;
AVHWDeviceContext *device_ctx = (AVHWDeviceContext*)ctx->hwdevice->data;
AVCUDADeviceContext *device_hwctx = device_ctx->hwctx;
CUcontext dummy, cuda_ctx = device_hwctx->cuda_ctx;
AVFrame *frame = data;
CUVIDSOURCEDATAPACKET cupkt;
AVPacket filter_packet = { 0 };
AVPacket filtered_packet = { 0 };
CUdeviceptr mapped_frame = 0;
int ret = 0, eret = 0;
if (ctx->bsf && avpkt->size) {
if ((ret = av_packet_ref(&filter_packet, avpkt)) < 0) {
av_log(avctx, AV_LOG_ERROR, "av_packet_ref failed\n");
return ret;
}
if ((ret = av_bsf_send_packet(ctx->bsf, &filter_packet)) < 0) {
av_log(avctx, AV_LOG_ERROR, "av_bsf_send_packet failed\n");
av_packet_unref(&filter_packet);
return ret;
}
if ((ret = av_bsf_receive_packet(ctx->bsf, &filtered_packet)) < 0) {
av_log(avctx, AV_LOG_ERROR, "av_bsf_receive_packet failed\n");
return ret;
}
avpkt = &filtered_packet;
}
ret = CHECK_CU(cuCtxPushCurrent(cuda_ctx));
if (ret < 0) {
av_packet_unref(&filtered_packet);
return ret;
}
memset(&cupkt, 0, sizeof(cupkt));
if (avpkt->size) {
cupkt.payload_size = avpkt->size;
cupkt.payload = avpkt->data;
if (avpkt->pts != AV_NOPTS_VALUE) {
cupkt.flags = CUVID_PKT_TIMESTAMP;
if (avctx->pkt_timebase.num && avctx->pkt_timebase.den)
cupkt.timestamp = av_rescale_q(avpkt->pts, avctx->pkt_timebase, (AVRational){1, 10000000});
else
cupkt.timestamp = avpkt->pts;
}
} else {
cupkt.flags = CUVID_PKT_ENDOFSTREAM;
}
ret = CHECK_CU(cuvidParseVideoData(ctx->cuparser, &cupkt));
av_packet_unref(&filtered_packet);
if (ret < 0) {
goto error;
}
// cuvidParseVideoData doesn't return an error just because stuff failed...
if (ctx->internal_error) {
av_log(avctx, AV_LOG_ERROR, "cuvid decode callback error\n");
ret = ctx->internal_error;
goto error;
}
if (av_fifo_size(ctx->frame_queue)) {
CUVIDPARSERDISPINFO dispinfo;
CUVIDPROCPARAMS params;
unsigned int pitch = 0;
int offset = 0;
int i;
av_fifo_generic_read(ctx->frame_queue, &dispinfo, sizeof(CUVIDPARSERDISPINFO), NULL);
memset(¶ms, 0, sizeof(params));
params.progressive_frame = dispinfo.progressive_frame;
params.second_field = 0;
params.top_field_first = dispinfo.top_field_first;
ret = CHECK_CU(cuvidMapVideoFrame(ctx->cudecoder, dispinfo.picture_index, &mapped_frame, &pitch, ¶ms));
if (ret < 0)
goto error;
if (avctx->pix_fmt == AV_PIX_FMT_CUDA) {
ret = av_hwframe_get_buffer(ctx->hwframe, frame, 0);
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "av_hwframe_get_buffer failed\n");
goto error;
}
ret = ff_decode_frame_props(avctx, frame);
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "ff_decode_frame_props failed\n");
goto error;
}
for (i = 0; i < 2; i++) {
CUDA_MEMCPY2D cpy = {
.srcMemoryType = CU_MEMORYTYPE_DEVICE,
.dstMemoryType = CU_MEMORYTYPE_DEVICE,
.srcDevice = mapped_frame,
.dstDevice = (CUdeviceptr)frame->data[i],
.srcPitch = pitch,
.dstPitch = frame->linesize[i],
.srcY = offset,
.WidthInBytes = FFMIN(pitch, frame->linesize[i]),
.Height = avctx->coded_height >> (i ? 1 : 0),
};
ret = CHECK_CU(cuMemcpy2D(&cpy));
if (ret < 0)
goto error;
offset += avctx->coded_height;
}
} else if (avctx->pix_fmt == AV_PIX_FMT_NV12) {
static int scale_vaapi_filter_frame(AVFilterLink *inlink, AVFrame *input_frame)
{
AVFilterContext *avctx = inlink->dst;
AVFilterLink *outlink = avctx->outputs[0];
ScaleVAAPIContext *ctx = avctx->priv;
AVFrame *output_frame = NULL;
VASurfaceID input_surface, output_surface;
VAProcPipelineParameterBuffer params;
VABufferID params_id;
VARectangle input_region;
VAStatus vas;
int err;
av_log(ctx, AV_LOG_DEBUG, "Filter input: %s, %ux%u (%"PRId64").\n",
av_get_pix_fmt_name(input_frame->format),
input_frame->width, input_frame->height, input_frame->pts);
if (ctx->va_context == VA_INVALID_ID)
return AVERROR(EINVAL);
input_surface = (VASurfaceID)(uintptr_t)input_frame->data[3];
av_log(ctx, AV_LOG_DEBUG, "Using surface %#x for scale input.\n",
input_surface);
output_frame = av_frame_alloc();
if (!output_frame) {
av_log(ctx, AV_LOG_ERROR, "Failed to allocate output frame.");
err = AVERROR(ENOMEM);
goto fail;
}
err = av_hwframe_get_buffer(ctx->output_frames_ref, output_frame, 0);
if (err < 0) {
av_log(ctx, AV_LOG_ERROR, "Failed to get surface for "
"output: %d\n.", err);
}
output_surface = (VASurfaceID)(uintptr_t)output_frame->data[3];
av_log(ctx, AV_LOG_DEBUG, "Using surface %#x for scale output.\n",
output_surface);
memset(¶ms, 0, sizeof(params));
// If there were top/left cropping, it could be taken into
// account here.
input_region = (VARectangle) {
.x = 0,
.y = 0,
.width = input_frame->width,
.height = input_frame->height,
};
params.surface = input_surface;
params.surface_region = &input_region;
params.surface_color_standard =
vaapi_proc_colour_standard(input_frame->colorspace);
params.output_region = 0;
params.output_background_color = 0xff000000;
params.output_color_standard = params.surface_color_standard;
params.pipeline_flags = 0;
params.filter_flags = VA_FILTER_SCALING_HQ;
vas = vaBeginPicture(ctx->hwctx->display,
ctx->va_context, output_surface);
if (vas != VA_STATUS_SUCCESS) {
av_log(ctx, AV_LOG_ERROR, "Failed to attach new picture: "
"%d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail;
}
vas = vaCreateBuffer(ctx->hwctx->display, ctx->va_context,
VAProcPipelineParameterBufferType,
sizeof(params), 1, ¶ms, ¶ms_id);
if (vas != VA_STATUS_SUCCESS) {
av_log(ctx, AV_LOG_ERROR, "Failed to create parameter buffer: "
"%d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail_after_begin;
}
av_log(ctx, AV_LOG_DEBUG, "Pipeline parameter buffer is %#x.\n",
params_id);
vas = vaRenderPicture(ctx->hwctx->display, ctx->va_context,
¶ms_id, 1);
if (vas != VA_STATUS_SUCCESS) {
av_log(ctx, AV_LOG_ERROR, "Failed to render parameter buffer: "
"%d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail_after_begin;
}
vas = vaEndPicture(ctx->hwctx->display, ctx->va_context);
if (vas != VA_STATUS_SUCCESS) {
av_log(ctx, AV_LOG_ERROR, "Failed to start picture processing: "
"%d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail_after_render;
}
if (ctx->hwctx->driver_quirks &
AV_VAAPI_DRIVER_QUIRK_RENDER_PARAM_BUFFERS) {
vas = vaDestroyBuffer(ctx->hwctx->display, params_id);
if (vas != VA_STATUS_SUCCESS) {
av_log(ctx, AV_LOG_ERROR, "Failed to free parameter buffer: "
"%d (%s).\n", vas, vaErrorStr(vas));
// And ignore.
}
}
av_frame_copy_props(output_frame, input_frame);
av_frame_free(&input_frame);
av_log(ctx, AV_LOG_DEBUG, "Filter output: %s, %ux%u (%"PRId64").\n",
av_get_pix_fmt_name(output_frame->format),
output_frame->width, output_frame->height, output_frame->pts);
return ff_filter_frame(outlink, output_frame);
// We want to make sure that if vaBeginPicture has been called, we also
// call vaRenderPicture and vaEndPicture. These calls may well fail or
// do something else nasty, but once we're in this failure case there
// isn't much else we can do.
fail_after_begin:
vaRenderPicture(ctx->hwctx->display, ctx->va_context, ¶ms_id, 1);
fail_after_render:
vaEndPicture(ctx->hwctx->display, ctx->va_context);
fail:
av_frame_free(&input_frame);
av_frame_free(&output_frame);
return err;
}
static av_cold int scale_vaapi_init(AVFilterContext *avctx)
{
ScaleVAAPIContext *ctx = avctx->priv;
ctx->va_config = VA_INVALID_ID;
ctx->va_context = VA_INVALID_ID;
ctx->valid_ids = 1;
if (ctx->output_format_string) {
ctx->output_format = av_get_pix_fmt(ctx->output_format_string);
if (ctx->output_format == AV_PIX_FMT_NONE) {
av_log(ctx, AV_LOG_ERROR, "Invalid output format.\n");
return AVERROR(EINVAL);
}
} else {
// Use the input format once that is configured.
ctx->output_format = AV_PIX_FMT_NONE;
}
return 0;
}
static bool vaapi_encode(void *data, struct encoder_frame *frame,
struct encoder_packet *packet, bool *received_packet)
{
struct vaapi_encoder *enc = data;
AVFrame * hwframe = NULL;
AVPacket av_pkt;
int got_packet;
int ret;
hwframe = av_frame_alloc();
if (!hwframe) {
warn("vaapi_encode: failed to allocate hw frame");
return false;
}
ret = av_hwframe_get_buffer(enc->vaframes_ref, hwframe, 0);
if (ret < 0) {
warn("vaapi_encode: failed to get buffer for hw frame: %s",
av_err2str(ret));
goto fail;
}
copy_data(enc->vframe, frame, enc->height, enc->context->pix_fmt);
enc->vframe->pts = frame->pts;
hwframe->pts = frame->pts;
hwframe->width = enc->vframe->width;
hwframe->height = enc->vframe->height;
ret = av_hwframe_transfer_data(hwframe, enc->vframe, 0);
if (ret < 0) {
warn("vaapi_encode: failed to upload hw frame: %s",
av_err2str(ret));
goto fail;
}
ret = av_frame_copy_props(hwframe, enc->vframe);
if (ret < 0) {
warn("vaapi_encode: failed to copy props to hw frame: %s",
av_err2str(ret));
goto fail;
}
av_init_packet(&av_pkt);
#if LIBAVFORMAT_VERSION_INT >= AV_VERSION_INT(57, 40, 101)
ret = avcodec_send_frame(enc->context, hwframe);
if (ret == 0)
ret = avcodec_receive_packet(enc->context, &av_pkt);
got_packet = (ret == 0);
if (ret == AVERROR_EOF || ret == AVERROR(EAGAIN))
ret = 0;
#else
ret = avcodec_encode_video2(
enc->context, &av_pkt, hwframe, &got_packet);
#endif
if (ret < 0) {
warn("vaapi_encode: Error encoding: %s", av_err2str(ret));
goto fail;
}
if (got_packet && av_pkt.size) {
if (enc->first_packet) {
uint8_t *new_packet;
size_t size;
enc->first_packet = false;
obs_extract_avc_headers(av_pkt.data, av_pkt.size,
&new_packet, &size, &enc->header,
&enc->header_size, &enc->sei,
&enc->sei_size);
da_copy_array(enc->buffer, new_packet, size);
bfree(new_packet);
} else {
da_copy_array(enc->buffer, av_pkt.data, av_pkt.size);
}
packet->pts = av_pkt.pts;
packet->dts = av_pkt.dts;
packet->data = enc->buffer.array;
packet->size = enc->buffer.num;
packet->type = OBS_ENCODER_VIDEO;
packet->keyframe = obs_avc_keyframe(packet->data, packet->size);
*received_packet = true;
} else {
*received_packet = false;
}
av_packet_unref(&av_pkt);
av_frame_free(&hwframe);
return true;
fail:
av_frame_free(&hwframe);
return false;
}
bool ADM_ffVAEncHEVC::preEncode(void)
{
uint32_t nb;
if(source->getNextFrame(&nb,image)==false)
{
ADM_warning("[ffVAEncHEVC] Cannot get next image\n");
return false;
}
swFrame=av_frame_alloc();
if(!swFrame)
{
ADM_error("Could not allocate sw frame\n");
return false;
}
swFrame->width=source->getInfo()->width;
swFrame->height=source->getInfo()->height;
swFrame->format=AV_PIX_FMT_NV12;
int err=av_frame_get_buffer(swFrame, 32);
if(err<0)
{
CLEARTEXT(err)
ADM_warning("get buffer for sw frame failed with error code %d (%s)\n",err,buf);
return false;
}
swFrame->linesize[0] = swFrame->linesize[1] = image->GetPitch(PLANAR_Y);
swFrame->linesize[2] = 0;
swFrame->data[2] = NULL;
image->convertToNV12(swFrame->data[0],swFrame->data[1],swFrame->linesize[0],swFrame->linesize[1]);
if(hwFrame)
{
av_frame_free(&hwFrame);
hwFrame=NULL;
}
hwFrame=av_frame_alloc();
if(!hwFrame)
{
ADM_error("Could not allocate hw frame\n");
return false;
}
hwFrame->width=source->getInfo()->width;
hwFrame->height=source->getInfo()->height;
hwFrame->format=AV_PIX_FMT_VAAPI;
err=av_hwframe_get_buffer(_context->hw_frames_ctx,hwFrame,0);
if(err<0)
{
CLEARTEXT(err)
ADM_warning("get buffer for hw frame failed with error code %d (%s)\n",err,buf);
return false;
}
err=av_hwframe_transfer_data(hwFrame, swFrame, 0);
if(err<0)
{
CLEARTEXT(err)
ADM_warning("data transfer to the hw frame failed with error code %d (%s)\n",err,buf);
return false;
}
uint64_t p=image->Pts;
queueOfDts.push_back(p);
aprintf("Incoming frame PTS=%" PRIu64", delay=%" PRIu64"\n",p,getEncoderDelay());
p+=getEncoderDelay();
hwFrame->pts=timingToLav(p);
if(!hwFrame->pts)
hwFrame->pts=AV_NOPTS_VALUE;
ADM_timeMapping map; // Store real PTS <->lav value mapping
map.realTS=p;
map.internalTS=hwFrame->pts;
mapper.push_back(map);
av_frame_free(&swFrame);
swFrame=NULL;
return true;
}
bool VideoEncoderFFmpeg::encode(const VideoFrame &frame)
{
DPTR_D(VideoEncoderFFmpeg);
QScopedPointer<AVFrame, ScopedAVFrameDeleter> f;
// hwupload
AVPixelFormat pixfmt = AVPixelFormat(frame.pixelFormatFFmpeg());
if (frame.isValid()) {
f.reset(av_frame_alloc());
f->format = pixfmt;
f->width = frame.width();
f->height = frame.height();
// f->quality = d.avctx->global_quality;
switch (timestampMode()) {
case TimestampCopy:
f->pts = int64_t(frame.timestamp()*frameRate()); // TODO: check monotically increase and fix if not. or another mode?
break;
case TimestampMonotonic:
f->pts = d.nb_encoded+1;
break;
default:
break;
}
// pts is set in muxer
const int nb_planes = frame.planeCount();
for (int i = 0; i < nb_planes; ++i) {
f->linesize[i] = frame.bytesPerLine(i);
f->data[i] = (uint8_t*)frame.constBits(i);
}
if (d.avctx->width <= 0) {
d.avctx->width = frame.width();
}
if (d.avctx->height <= 0) {
d.avctx->height = frame.width();
}
#ifdef HAVE_AVHWCTX
if (d.avctx->hw_frames_ctx) {
// TODO: try to map to SourceSurface
// checl valid sw_formats
if (!d.hwframes_ref) {
qWarning("no hw frame context for uploading");
return false;
}
if (pixfmt != d.hwframes->sw_format) {
// reinit or got an unsupported format. assume parameters will not change, so it's the 1st init
// check constraints
bool init_frames_ctx = d.hwframes->sw_format == AVPixelFormat(-1);
if (d.sw_fmts.contains(pixfmt)) { // format changed
init_frames_ctx = true;
} else { // convert to supported sw format
pixfmt = d.sw_fmts[0];
f->format = pixfmt;
VideoFrame converted = frame.to(VideoFormat::pixelFormatFromFFmpeg(pixfmt));
for (int i = 0; i < converted.planeCount(); ++i) {
f->linesize[i] = converted.bytesPerLine(i);
f->data[i] = (uint8_t*)frame.constBits(i);
}
}
if (init_frames_ctx) {
d.hwframes->sw_format = pixfmt;
d.hwframes->width = frame.width();
d.hwframes->height = frame.height();
AV_ENSURE(av_hwframe_ctx_init(d.hwframes_ref), false);
}
}
// upload
QScopedPointer<AVFrame, ScopedAVFrameDeleter> hwf( av_frame_alloc());
AV_ENSURE(av_hwframe_get_buffer(d.hwframes_ref, hwf.data(), 0), false);
//hwf->format = d.hwframes->format; // not necessary
//hwf->width = f->width;
//hwf->height = f->height;
AV_ENSURE(av_hwframe_transfer_data(hwf.data(), f.data(), 0), false);
AV_ENSURE(av_frame_copy_props(hwf.data(), f.data()), false);
av_frame_unref(f.data());
av_frame_move_ref(f.data(), hwf.data());
}
#endif //HAVE_AVHWCTX
}
AVPacket pkt;
av_init_packet(&pkt);
pkt.data = (uint8_t*)d.buffer.constData();
pkt.size = d.buffer.size();
int got_packet = 0;
int ret = avcodec_encode_video2(d.avctx, &pkt, f.data(), &got_packet);
if (ret < 0) {
qWarning("error avcodec_encode_video2: %s" ,av_err2str(ret));
return false; //false
}
d.nb_encoded++;
if (!got_packet) {
qWarning("no packet got");
d.packet = Packet();
// invalid frame means eof
return frame.isValid();
}
// qDebug("enc avpkt.pts: %lld, dts: %lld.", pkt.pts, pkt.dts);
d.packet = Packet::fromAVPacket(&pkt, av_q2d(d.avctx->time_base));
// qDebug("enc packet.pts: %.3f, dts: %.3f.", d.packet.pts, d.packet.dts);
return true;
}
