STDMETHODIMP CDecD3D11::AllocateFramesContext(int width, int height, AVPixelFormat format, int nSurfaces, AVBufferRef **ppFramesCtx)
{
ASSERT(m_pAVCtx);
ASSERT(m_pDevCtx);
ASSERT(ppFramesCtx);
// unref any old buffer
av_buffer_unref(ppFramesCtx);
SafeRelease(&m_pD3D11StagingTexture);
// allocate a new frames context for the device context
*ppFramesCtx = av_hwframe_ctx_alloc(m_pDevCtx);
if (*ppFramesCtx == nullptr)
return E_OUTOFMEMORY;
AVHWFramesContext *pFrames = (AVHWFramesContext *)(*ppFramesCtx)->data;
pFrames->format = AV_PIX_FMT_D3D11;
pFrames->sw_format = (format == AV_PIX_FMT_YUV420P10) ? AV_PIX_FMT_P010 : AV_PIX_FMT_NV12;
pFrames->width = width;
pFrames->height = height;
pFrames->initial_pool_size = nSurfaces;
AVD3D11VAFramesContext *pFramesHWContext = (AVD3D11VAFramesContext *)pFrames->hwctx;
pFramesHWContext->BindFlags |= D3D11_BIND_DECODER | D3D11_BIND_SHADER_RESOURCE;
pFramesHWContext->MiscFlags |= D3D11_RESOURCE_MISC_SHARED;
int ret = av_hwframe_ctx_init(*ppFramesCtx);
if (ret < 0)
{
av_buffer_unref(ppFramesCtx);
return E_FAIL;
}
return S_OK;
}
static int cudaupload_config_output(AVFilterLink *outlink)
{
AVFilterContext *ctx = outlink->src;
AVFilterLink *inlink = ctx->inputs[0];
CudaUploadContext *s = ctx->priv;
AVHWFramesContext *hwframe_ctx;
int ret;
av_buffer_unref(&s->hwframe);
s->hwframe = av_hwframe_ctx_alloc(s->hwdevice);
if (!s->hwframe)
return AVERROR(ENOMEM);
hwframe_ctx = (AVHWFramesContext*)s->hwframe->data;
hwframe_ctx->format = AV_PIX_FMT_CUDA;
hwframe_ctx->sw_format = inlink->format;
hwframe_ctx->width = inlink->w;
hwframe_ctx->height = inlink->h;
ret = av_hwframe_ctx_init(s->hwframe);
if (ret < 0)
return ret;
outlink->hw_frames_ctx = av_buffer_ref(s->hwframe);
if (!outlink->hw_frames_ctx)
return AVERROR(ENOMEM);
return 0;
}
int ff_decode_get_hw_frames_ctx(AVCodecContext *avctx,
enum AVHWDeviceType dev_type)
{
AVHWDeviceContext *device_ctx;
AVHWFramesContext *frames_ctx;
int ret;
if (!avctx->hwaccel)
return AVERROR(ENOSYS);
if (avctx->hw_frames_ctx)
return 0;
if (!avctx->hw_device_ctx) {
av_log(avctx, AV_LOG_ERROR, "A hardware frames or device context is "
"required for hardware accelerated decoding.\n");
return AVERROR(EINVAL);
}
device_ctx = (AVHWDeviceContext *)avctx->hw_device_ctx->data;
if (device_ctx->type != dev_type) {
av_log(avctx, AV_LOG_ERROR, "Device type %s expected for hardware "
"decoding, but got %s.\n", av_hwdevice_get_type_name(dev_type),
av_hwdevice_get_type_name(device_ctx->type));
return AVERROR(EINVAL);
}
ret = avcodec_get_hw_frames_parameters(avctx,
avctx->hw_device_ctx,
avctx->hwaccel->pix_fmt,
avctx->hw_frames_ctx);
if (ret < 0) {
av_buffer_unref(&avctx->hw_frames_ctx);
return ret;
}
frames_ctx = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
if (frames_ctx->initial_pool_size) {
// We guarantee 4 base work surfaces. The function above guarantees 1
// (the absolute minimum), so add the missing count.
frames_ctx->initial_pool_size += 3;
// Add an additional surface per thread is frame threading is enabled.
if (avctx->active_thread_type & FF_THREAD_FRAME)
frames_ctx->initial_pool_size += avctx->thread_count;
}
ret = av_hwframe_ctx_init(avctx->hw_frames_ctx);
if (ret < 0) {
av_buffer_unref(&avctx->hw_frames_ctx);
return ret;
}
return 0;
}
static int init_out_pool(AVFilterContext *ctx,
int out_width, int out_height)
{
QSVScaleContext *s = ctx->priv;
AVFilterLink *outlink = ctx->outputs[0];
AVHWFramesContext *in_frames_ctx;
AVHWFramesContext *out_frames_ctx;
AVQSVFramesContext *in_frames_hwctx;
AVQSVFramesContext *out_frames_hwctx;
enum AVPixelFormat in_format;
enum AVPixelFormat out_format;
int i, ret;
/* check that we have a hw context */
if (!ctx->inputs[0]->hw_frames_ctx) {
av_log(ctx, AV_LOG_ERROR, "No hw context provided on input\n");
return AVERROR(EINVAL);
}
in_frames_ctx = (AVHWFramesContext*)ctx->inputs[0]->hw_frames_ctx->data;
in_frames_hwctx = in_frames_ctx->hwctx;
in_format = in_frames_ctx->sw_format;
out_format = (s->format == AV_PIX_FMT_NONE) ? in_format : s->format;
outlink->hw_frames_ctx = av_hwframe_ctx_alloc(in_frames_ctx->device_ref);
if (!outlink->hw_frames_ctx)
return AVERROR(ENOMEM);
out_frames_ctx = (AVHWFramesContext*)outlink->hw_frames_ctx->data;
out_frames_hwctx = out_frames_ctx->hwctx;
out_frames_ctx->format = AV_PIX_FMT_QSV;
out_frames_ctx->width = FFALIGN(out_width, 32);
out_frames_ctx->height = FFALIGN(out_height, 32);
out_frames_ctx->sw_format = out_format;
out_frames_ctx->initial_pool_size = 4;
out_frames_hwctx->frame_type = in_frames_hwctx->frame_type;
ret = ff_filter_init_hw_frames(ctx, outlink, 32);
if (ret < 0)
return ret;
ret = av_hwframe_ctx_init(outlink->hw_frames_ctx);
if (ret < 0)
return ret;
for (i = 0; i < out_frames_hwctx->nb_surfaces; i++) {
mfxFrameInfo *info = &out_frames_hwctx->surfaces[i].Info;
info->CropW = out_width;
info->CropH = out_height;
}
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;
}
int ff_opencl_filter_config_output(AVFilterLink *outlink)
{
AVFilterContext *avctx = outlink->src;
OpenCLFilterContext *ctx = avctx->priv;
AVBufferRef *output_frames_ref = NULL;
AVHWFramesContext *output_frames;
int err;
av_buffer_unref(&outlink->hw_frames_ctx);
if (!ctx->device_ref) {
if (!avctx->hw_device_ctx) {
av_log(avctx, AV_LOG_ERROR, "OpenCL filtering requires an "
"OpenCL device.\n");
return AVERROR(EINVAL);
}
err = opencl_filter_set_device(avctx, avctx->hw_device_ctx);
if (err < 0)
return err;
}
output_frames_ref = av_hwframe_ctx_alloc(ctx->device_ref);
if (!output_frames_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
output_frames = (AVHWFramesContext*)output_frames_ref->data;
output_frames->format = AV_PIX_FMT_OPENCL;
output_frames->sw_format = ctx->output_format;
output_frames->width = ctx->output_width;
output_frames->height = ctx->output_height;
err = av_hwframe_ctx_init(output_frames_ref);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to initialise output "
"frames: %d.\n", err);
goto fail;
}
outlink->hw_frames_ctx = output_frames_ref;
outlink->w = ctx->output_width;
outlink->h = ctx->output_height;
return 0;
fail:
av_buffer_unref(&output_frames_ref);
return err;
}
int qsv_init(AVCodecContext *s)
{
InputStream *ist = s->opaque;
AVHWFramesContext *frames_ctx;
AVQSVFramesContext *frames_hwctx;
int ret;
if (!hw_device_ctx) {
ret = qsv_device_init(ist);
if (ret < 0)
return ret;
}
av_buffer_unref(&ist->hw_frames_ctx);
ist->hw_frames_ctx = av_hwframe_ctx_alloc(hw_device_ctx);
if (!ist->hw_frames_ctx)
return AVERROR(ENOMEM);
frames_ctx = (AVHWFramesContext*)ist->hw_frames_ctx->data;
frames_hwctx = frames_ctx->hwctx;
frames_ctx->width = FFALIGN(s->coded_width, 32);
frames_ctx->height = FFALIGN(s->coded_height, 32);
frames_ctx->format = AV_PIX_FMT_QSV;
frames_ctx->sw_format = s->sw_pix_fmt;
frames_ctx->initial_pool_size = 64;
frames_hwctx->frame_type = MFX_MEMTYPE_VIDEO_MEMORY_DECODER_TARGET;
ret = av_hwframe_ctx_init(ist->hw_frames_ctx);
if (ret < 0) {
av_log(NULL, AV_LOG_ERROR, "Error initializing a QSV frame pool\n");
return ret;
}
ist->hwaccel_get_buffer = qsv_get_buffer;
ist->hwaccel_uninit = qsv_uninit;
return 0;
}
av_cold int ff_vaapi_encode_init(AVCodecContext *avctx,
const VAAPIEncodeType *type)
{
VAAPIEncodeContext *ctx = avctx->priv_data;
AVVAAPIFramesContext *recon_hwctx = NULL;
AVVAAPIHWConfig *hwconfig = NULL;
AVHWFramesConstraints *constraints = NULL;
enum AVPixelFormat recon_format;
VAStatus vas;
int err, i;
if (!avctx->hw_frames_ctx) {
av_log(avctx, AV_LOG_ERROR, "A hardware frames reference is "
"required to associate the encoding device.\n");
return AVERROR(EINVAL);
}
ctx->codec = type;
ctx->codec_options = ctx->codec_options_data;
ctx->priv_data = av_mallocz(type->priv_data_size);
if (!ctx->priv_data) {
err = AVERROR(ENOMEM);
goto fail;
}
ctx->input_frames_ref = av_buffer_ref(avctx->hw_frames_ctx);
if (!ctx->input_frames_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
ctx->input_frames = (AVHWFramesContext*)ctx->input_frames_ref->data;
ctx->device_ref = av_buffer_ref(ctx->input_frames->device_ref);
if (!ctx->device_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
ctx->device = (AVHWDeviceContext*)ctx->device_ref->data;
ctx->hwctx = ctx->device->hwctx;
err = ctx->codec->init(avctx);
if (err < 0)
goto fail;
vas = vaCreateConfig(ctx->hwctx->display,
ctx->va_profile, ctx->va_entrypoint,
ctx->config_attributes, ctx->nb_config_attributes,
&ctx->va_config);
if (vas != VA_STATUS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "Failed to create encode pipeline "
"configuration: %d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail;
}
hwconfig = av_hwdevice_hwconfig_alloc(ctx->device_ref);
if (!hwconfig) {
err = AVERROR(ENOMEM);
goto fail;
}
hwconfig->config_id = ctx->va_config;
constraints = av_hwdevice_get_hwframe_constraints(ctx->device_ref,
hwconfig);
if (!constraints) {
err = AVERROR(ENOMEM);
goto fail;
}
// Probably we can use the input surface format as the surface format
// of the reconstructed frames. If not, we just pick the first (only?)
// format in the valid list and hope that it all works.
recon_format = AV_PIX_FMT_NONE;
if (constraints->valid_sw_formats) {
for (i = 0; constraints->valid_sw_formats[i] != AV_PIX_FMT_NONE; i++) {
if (ctx->input_frames->sw_format ==
constraints->valid_sw_formats[i]) {
recon_format = ctx->input_frames->sw_format;
break;
}
}
if (recon_format == AV_PIX_FMT_NONE)
recon_format = constraints->valid_sw_formats[i];
} else {
// No idea what to use; copy input format.
recon_format = ctx->input_frames->sw_format;
}
av_log(avctx, AV_LOG_DEBUG, "Using %s as format of "
"reconstructed frames.\n", av_get_pix_fmt_name(recon_format));
if (ctx->aligned_width < constraints->min_width ||
ctx->aligned_height < constraints->min_height ||
ctx->aligned_width > constraints->max_width ||
ctx->aligned_height > constraints->max_height) {
av_log(avctx, AV_LOG_ERROR, "Hardware does not support encoding at "
"size %dx%d (constraints: width %d-%d height %d-%d).\n",
ctx->aligned_width, ctx->aligned_height,
constraints->min_width, constraints->max_width,
constraints->min_height, constraints->max_height);
err = AVERROR(EINVAL);
goto fail;
}
av_freep(&hwconfig);
av_hwframe_constraints_free(&constraints);
ctx->recon_frames_ref = av_hwframe_ctx_alloc(ctx->device_ref);
if (!ctx->recon_frames_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
ctx->recon_frames = (AVHWFramesContext*)ctx->recon_frames_ref->data;
ctx->recon_frames->format = AV_PIX_FMT_VAAPI;
ctx->recon_frames->sw_format = recon_format;
ctx->recon_frames->width = ctx->aligned_width;
ctx->recon_frames->height = ctx->aligned_height;
ctx->recon_frames->initial_pool_size = ctx->nb_recon_frames;
err = av_hwframe_ctx_init(ctx->recon_frames_ref);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to initialise reconstructed "
"frame context: %d.\n", err);
goto fail;
}
recon_hwctx = ctx->recon_frames->hwctx;
vas = vaCreateContext(ctx->hwctx->display, ctx->va_config,
ctx->aligned_width, ctx->aligned_height,
VA_PROGRESSIVE,
recon_hwctx->surface_ids,
recon_hwctx->nb_surfaces,
&ctx->va_context);
if (vas != VA_STATUS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "Failed to create encode pipeline "
"context: %d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail;
}
ctx->input_order = 0;
ctx->output_delay = avctx->max_b_frames;
ctx->decode_delay = 1;
ctx->output_order = - ctx->output_delay - 1;
if (ctx->codec->sequence_params_size > 0) {
ctx->codec_sequence_params =
av_mallocz(ctx->codec->sequence_params_size);
if (!ctx->codec_sequence_params) {
err = AVERROR(ENOMEM);
goto fail;
}
}
if (ctx->codec->picture_params_size > 0) {
ctx->codec_picture_params =
av_mallocz(ctx->codec->picture_params_size);
if (!ctx->codec_picture_params) {
err = AVERROR(ENOMEM);
goto fail;
}
}
if (ctx->codec->init_sequence_params) {
err = ctx->codec->init_sequence_params(avctx);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Codec sequence initialisation "
"failed: %d.\n", err);
goto fail;
}
}
// All I are IDR for now.
ctx->i_per_idr = 0;
ctx->p_per_i = ((avctx->gop_size + avctx->max_b_frames) /
(avctx->max_b_frames + 1));
ctx->b_per_p = avctx->max_b_frames;
// This should be configurable somehow. (Needs testing on a machine
// where it actually overlaps properly, though.)
ctx->issue_mode = ISSUE_MODE_MAXIMISE_THROUGHPUT;
return 0;
fail:
av_freep(&hwconfig);
av_hwframe_constraints_free(&constraints);
ff_vaapi_encode_close(avctx);
return err;
}
int av_hwframe_ctx_create_derived(AVBufferRef **derived_frame_ctx,
enum AVPixelFormat format,
AVBufferRef *derived_device_ctx,
AVBufferRef *source_frame_ctx,
int flags)
{
AVBufferRef *dst_ref = NULL;
AVHWFramesContext *dst = NULL;
AVHWFramesContext *src = (AVHWFramesContext*)source_frame_ctx->data;
int ret;
if (src->internal->source_frames) {
AVHWFramesContext *src_src =
(AVHWFramesContext*)src->internal->source_frames->data;
AVHWDeviceContext *dst_dev =
(AVHWDeviceContext*)derived_device_ctx->data;
if (src_src->device_ctx == dst_dev) {
// This is actually an unmapping, so we just return a
// reference to the source frame context.
*derived_frame_ctx =
av_buffer_ref(src->internal->source_frames);
if (!*derived_frame_ctx) {
ret = AVERROR(ENOMEM);
goto fail;
}
return 0;
}
}
dst_ref = av_hwframe_ctx_alloc(derived_device_ctx);
if (!dst_ref) {
ret = AVERROR(ENOMEM);
goto fail;
}
dst = (AVHWFramesContext*)dst_ref->data;
dst->format = format;
dst->sw_format = src->sw_format;
dst->width = src->width;
dst->height = src->height;
dst->internal->source_frames = av_buffer_ref(source_frame_ctx);
if (!dst->internal->source_frames) {
ret = AVERROR(ENOMEM);
goto fail;
}
ret = av_hwframe_ctx_init(dst_ref);
if (ret)
goto fail;
*derived_frame_ctx = dst_ref;
return 0;
fail:
if (dst)
av_buffer_unref(&dst->internal->source_frames);
av_buffer_unref(&dst_ref);
return ret;
}
static int hwmap_config_output(AVFilterLink *outlink)
{
AVFilterContext *avctx = outlink->src;
HWMapContext *ctx = avctx->priv;
AVFilterLink *inlink = avctx->inputs[0];
AVHWFramesContext *hwfc;
AVBufferRef *device;
const AVPixFmtDescriptor *desc;
int err;
av_log(avctx, AV_LOG_DEBUG, "Configure hwmap %s -> %s.\n",
av_get_pix_fmt_name(inlink->format),
av_get_pix_fmt_name(outlink->format));
av_buffer_unref(&ctx->hwframes_ref);
device = avctx->hw_device_ctx;
if (inlink->hw_frames_ctx) {
hwfc = (AVHWFramesContext*)inlink->hw_frames_ctx->data;
if (ctx->derive_device_type) {
enum AVHWDeviceType type;
type = av_hwdevice_find_type_by_name(ctx->derive_device_type);
if (type == AV_HWDEVICE_TYPE_NONE) {
av_log(avctx, AV_LOG_ERROR, "Invalid device type.\n");
goto fail;
}
err = av_hwdevice_ctx_create_derived(&device, type,
hwfc->device_ref, 0);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to created derived "
"device context: %d.\n", err);
goto fail;
}
}
desc = av_pix_fmt_desc_get(outlink->format);
if (!desc) {
err = AVERROR(EINVAL);
goto fail;
}
if (inlink->format == hwfc->format &&
(desc->flags & AV_PIX_FMT_FLAG_HWACCEL) &&
!ctx->reverse) {
// Map between two hardware formats (including the case of
// undoing an existing mapping).
if (!device) {
av_log(avctx, AV_LOG_ERROR, "A device reference is "
"required to map to a hardware format.\n");
err = AVERROR(EINVAL);
goto fail;
}
err = av_hwframe_ctx_create_derived(&ctx->hwframes_ref,
outlink->format,
device,
inlink->hw_frames_ctx, 0);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to create derived "
"frames context: %d.\n", err);
goto fail;
}
} else if (inlink->format == hwfc->format &&
(desc->flags & AV_PIX_FMT_FLAG_HWACCEL) &&
ctx->reverse) {
// Map between two hardware formats, but do it in reverse.
// Make a new hwframe context for the target type, and then
// overwrite the input hwframe context with a derived context
// mapped from that back to the source type.
AVBufferRef *source;
AVHWFramesContext *frames;
ctx->hwframes_ref = av_hwframe_ctx_alloc(device);
if (!ctx->hwframes_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
frames = (AVHWFramesContext*)ctx->hwframes_ref->data;
frames->format = outlink->format;
frames->sw_format = hwfc->sw_format;
frames->width = hwfc->width;
frames->height = hwfc->height;
frames->initial_pool_size = 64;
err = av_hwframe_ctx_init(ctx->hwframes_ref);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to initialise "
"target frames context: %d.\n", err);
goto fail;
}
err = av_hwframe_ctx_create_derived(&source,
inlink->format,
hwfc->device_ref,
ctx->hwframes_ref,
ctx->mode);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to create "
"derived source frames context: %d.\n", err);
goto fail;
}
// Here is the naughty bit. This overwriting changes what
// ff_get_video_buffer() in the previous filter returns -
// it will now give a frame allocated here mapped back to
// the format it expects. If there were any additional
// constraints on the output frames there then this may
// break nastily.
av_buffer_unref(&inlink->hw_frames_ctx);
inlink->hw_frames_ctx = source;
} else if ((outlink->format == hwfc->format &&
inlink->format == hwfc->sw_format) ||
inlink->format == hwfc->format) {
// Map from a hardware format to a software format, or
// undo an existing such mapping.
ctx->hwframes_ref = av_buffer_ref(inlink->hw_frames_ctx);
if (!ctx->hwframes_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
} else {
// Non-matching formats - not supported.
av_log(avctx, AV_LOG_ERROR, "Unsupported formats for "
"hwmap: from %s (%s) to %s.\n",
av_get_pix_fmt_name(inlink->format),
av_get_pix_fmt_name(hwfc->format),
av_get_pix_fmt_name(outlink->format));
err = AVERROR(EINVAL);
goto fail;
}
} else if (avctx->hw_device_ctx) {
// Map from a software format to a hardware format. This
// creates a new hwframe context like hwupload, but then
// returns frames mapped from that to the previous link in
// order to fill them without an additional copy.
if (!device) {
av_log(avctx, AV_LOG_ERROR, "A device reference is "
"required to create new frames with reverse "
"mapping.\n");
err = AVERROR(EINVAL);
goto fail;
}
ctx->reverse = 1;
ctx->hwframes_ref = av_hwframe_ctx_alloc(device);
if (!ctx->hwframes_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
hwfc = (AVHWFramesContext*)ctx->hwframes_ref->data;
hwfc->format = outlink->format;
hwfc->sw_format = inlink->format;
hwfc->width = inlink->w;
hwfc->height = inlink->h;
err = av_hwframe_ctx_init(ctx->hwframes_ref);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to create frame "
"context for reverse mapping: %d.\n", err);
goto fail;
}
} else {
av_log(avctx, AV_LOG_ERROR, "Mapping requires a hardware "
"context (a device, or frames on input).\n");
return AVERROR(EINVAL);
}
outlink->hw_frames_ctx = av_buffer_ref(ctx->hwframes_ref);
if (!outlink->hw_frames_ctx) {
err = AVERROR(ENOMEM);
goto fail;
}
outlink->w = inlink->w;
outlink->h = inlink->h;
return 0;
fail:
av_buffer_unref(&ctx->hwframes_ref);
return err;
}
int qsv_transcode_init(OutputStream *ost)
{
InputStream *ist;
const enum AVPixelFormat *pix_fmt;
int err, i;
AVBufferRef *encode_frames_ref = NULL;
AVHWFramesContext *encode_frames;
AVQSVFramesContext *qsv_frames;
/* check if the encoder supports QSV */
if (!ost->enc->pix_fmts)
return 0;
for (pix_fmt = ost->enc->pix_fmts; *pix_fmt != AV_PIX_FMT_NONE; pix_fmt++)
if (*pix_fmt == AV_PIX_FMT_QSV)
break;
if (*pix_fmt == AV_PIX_FMT_NONE)
return 0;
if (strcmp(ost->avfilter, "null") || ost->source_index < 0)
return 0;
/* check if the decoder supports QSV and the output only goes to this stream */
ist = input_streams[ost->source_index];
if (ist->hwaccel_id != HWACCEL_QSV || !ist->dec || !ist->dec->pix_fmts)
return 0;
for (pix_fmt = ist->dec->pix_fmts; *pix_fmt != AV_PIX_FMT_NONE; pix_fmt++)
if (*pix_fmt == AV_PIX_FMT_QSV)
break;
if (*pix_fmt == AV_PIX_FMT_NONE)
return 0;
for (i = 0; i < nb_output_streams; i++)
if (output_streams[i] != ost &&
output_streams[i]->source_index == ost->source_index)
return 0;
av_log(NULL, AV_LOG_VERBOSE, "Setting up QSV transcoding\n");
if (!hw_device_ctx) {
err = qsv_device_init(ist);
if (err < 0)
goto fail;
}
// This creates a dummy hw_frames_ctx for the encoder to be
// suitably initialised. It only contains one real frame, so
// hopefully doesn't waste too much memory.
encode_frames_ref = av_hwframe_ctx_alloc(hw_device_ctx);
if (!encode_frames_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
encode_frames = (AVHWFramesContext*)encode_frames_ref->data;
qsv_frames = encode_frames->hwctx;
encode_frames->width = FFALIGN(ist->resample_width, 32);
encode_frames->height = FFALIGN(ist->resample_height, 32);
encode_frames->format = AV_PIX_FMT_QSV;
encode_frames->sw_format = AV_PIX_FMT_NV12;
encode_frames->initial_pool_size = 1;
qsv_frames->frame_type = MFX_MEMTYPE_VIDEO_MEMORY_DECODER_TARGET;
err = av_hwframe_ctx_init(encode_frames_ref);
if (err < 0)
goto fail;
ist->dec_ctx->pix_fmt = AV_PIX_FMT_QSV;
ist->resample_pix_fmt = AV_PIX_FMT_QSV;
ost->enc_ctx->pix_fmt = AV_PIX_FMT_QSV;
ost->enc_ctx->hw_frames_ctx = encode_frames_ref;
return 0;
fail:
av_buffer_unref(&encode_frames_ref);
return err;
}
static int scale_vaapi_config_output(AVFilterLink *outlink)
{
AVFilterContext *avctx = outlink->src;
ScaleVAAPIContext *ctx = avctx->priv;
AVVAAPIHWConfig *hwconfig = NULL;
AVHWFramesConstraints *constraints = NULL;
AVVAAPIFramesContext *va_frames;
VAStatus vas;
int err, i;
scale_vaapi_pipeline_uninit(ctx);
ctx->device_ref = av_buffer_ref(ctx->input_frames->device_ref);
ctx->hwctx = ((AVHWDeviceContext*)ctx->device_ref->data)->hwctx;
av_assert0(ctx->va_config == VA_INVALID_ID);
vas = vaCreateConfig(ctx->hwctx->display, VAProfileNone,
VAEntrypointVideoProc, 0, 0, &ctx->va_config);
if (vas != VA_STATUS_SUCCESS) {
av_log(ctx, AV_LOG_ERROR, "Failed to create processing pipeline "
"config: %d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail;
}
hwconfig = av_hwdevice_hwconfig_alloc(ctx->device_ref);
if (!hwconfig) {
err = AVERROR(ENOMEM);
goto fail;
}
hwconfig->config_id = ctx->va_config;
constraints = av_hwdevice_get_hwframe_constraints(ctx->device_ref,
hwconfig);
if (!constraints) {
err = AVERROR(ENOMEM);
goto fail;
}
if (ctx->output_format == AV_PIX_FMT_NONE)
ctx->output_format = ctx->input_frames->sw_format;
if (constraints->valid_sw_formats) {
for (i = 0; constraints->valid_sw_formats[i] != AV_PIX_FMT_NONE; i++) {
if (ctx->output_format == constraints->valid_sw_formats[i])
break;
}
if (constraints->valid_sw_formats[i] == AV_PIX_FMT_NONE) {
av_log(ctx, AV_LOG_ERROR, "Hardware does not support output "
"format %s.\n", av_get_pix_fmt_name(ctx->output_format));
err = AVERROR(EINVAL);
goto fail;
}
}
if (ctx->output_width < constraints->min_width ||
ctx->output_height < constraints->min_height ||
ctx->output_width > constraints->max_width ||
ctx->output_height > constraints->max_height) {
av_log(ctx, AV_LOG_ERROR, "Hardware does not support scaling to "
"size %dx%d (constraints: width %d-%d height %d-%d).\n",
ctx->output_width, ctx->output_height,
constraints->min_width, constraints->max_width,
constraints->min_height, constraints->max_height);
err = AVERROR(EINVAL);
goto fail;
}
ctx->output_frames_ref = av_hwframe_ctx_alloc(ctx->device_ref);
if (!ctx->output_frames_ref) {
av_log(ctx, AV_LOG_ERROR, "Failed to create HW frame context "
"for output.\n");
err = AVERROR(ENOMEM);
goto fail;
}
ctx->output_frames = (AVHWFramesContext*)ctx->output_frames_ref->data;
ctx->output_frames->format = AV_PIX_FMT_VAAPI;
ctx->output_frames->sw_format = ctx->output_format;
ctx->output_frames->width = ctx->output_width;
ctx->output_frames->height = ctx->output_height;
// The number of output frames we need is determined by what follows
// the filter. If it's an encoder with complex frame reference
// structures then this could be very high.
ctx->output_frames->initial_pool_size = 10;
err = av_hwframe_ctx_init(ctx->output_frames_ref);
if (err < 0) {
av_log(ctx, AV_LOG_ERROR, "Failed to initialise VAAPI frame "
"context for output: %d\n", err);
goto fail;
}
va_frames = ctx->output_frames->hwctx;
av_assert0(ctx->va_context == VA_INVALID_ID);
vas = vaCreateContext(ctx->hwctx->display, ctx->va_config,
ctx->output_width, ctx->output_height,
VA_PROGRESSIVE,
va_frames->surface_ids, va_frames->nb_surfaces,
&ctx->va_context);
if (vas != VA_STATUS_SUCCESS) {
av_log(ctx, AV_LOG_ERROR, "Failed to create processing pipeline "
"context: %d (%s).\n", vas, vaErrorStr(vas));
return AVERROR(EIO);
}
outlink->w = ctx->output_width;
outlink->h = ctx->output_height;
outlink->hw_frames_ctx = av_buffer_ref(ctx->output_frames_ref);
if (!outlink->hw_frames_ctx) {
err = AVERROR(ENOMEM);
goto fail;
}
av_freep(&hwconfig);
av_hwframe_constraints_free(&constraints);
return 0;
fail:
av_buffer_unref(&ctx->output_frames_ref);
av_freep(&hwconfig);
av_hwframe_constraints_free(&constraints);
return err;
}
static int CUDAAPI cuvid_handle_video_sequence(void *opaque, CUVIDEOFORMAT* format)
{
AVCodecContext *avctx = opaque;
CuvidContext *ctx = avctx->priv_data;
AVHWFramesContext *hwframe_ctx = (AVHWFramesContext*)ctx->hwframe->data;
CUVIDDECODECREATEINFO cuinfo;
av_log(avctx, AV_LOG_TRACE, "pfnSequenceCallback, progressive_sequence=%d\n", format->progressive_sequence);
ctx->internal_error = 0;
avctx->width = format->display_area.right;
avctx->height = format->display_area.bottom;
ff_set_sar(avctx, av_div_q(
(AVRational){ format->display_aspect_ratio.x, format->display_aspect_ratio.y },
(AVRational){ avctx->width, avctx->height }));
if (!format->progressive_sequence && ctx->deint_mode == cudaVideoDeinterlaceMode_Weave)
avctx->flags |= AV_CODEC_FLAG_INTERLACED_DCT;
else
avctx->flags &= ~AV_CODEC_FLAG_INTERLACED_DCT;
if (format->video_signal_description.video_full_range_flag)
avctx->color_range = AVCOL_RANGE_JPEG;
else
avctx->color_range = AVCOL_RANGE_MPEG;
avctx->color_primaries = format->video_signal_description.color_primaries;
avctx->color_trc = format->video_signal_description.transfer_characteristics;
avctx->colorspace = format->video_signal_description.matrix_coefficients;
if (format->bitrate)
avctx->bit_rate = format->bitrate;
if (format->frame_rate.numerator && format->frame_rate.denominator) {
avctx->framerate.num = format->frame_rate.numerator;
avctx->framerate.den = format->frame_rate.denominator;
}
if (ctx->cudecoder
&& avctx->coded_width == format->coded_width
&& avctx->coded_height == format->coded_height
&& ctx->chroma_format == format->chroma_format
&& ctx->codec_type == format->codec)
return 1;
if (ctx->cudecoder) {
av_log(avctx, AV_LOG_TRACE, "Re-initializing decoder\n");
ctx->internal_error = CHECK_CU(cuvidDestroyDecoder(ctx->cudecoder));
if (ctx->internal_error < 0)
return 0;
ctx->cudecoder = NULL;
}
if (hwframe_ctx->pool && (
hwframe_ctx->width < avctx->width ||
hwframe_ctx->height < avctx->height ||
hwframe_ctx->format != AV_PIX_FMT_CUDA ||
hwframe_ctx->sw_format != AV_PIX_FMT_NV12)) {
av_log(avctx, AV_LOG_ERROR, "AVHWFramesContext is already initialized with incompatible parameters\n");
ctx->internal_error = AVERROR(EINVAL);
return 0;
}
if (format->chroma_format != cudaVideoChromaFormat_420) {
av_log(avctx, AV_LOG_ERROR, "Chroma formats other than 420 are not supported\n");
ctx->internal_error = AVERROR(EINVAL);
return 0;
}
avctx->coded_width = format->coded_width;
avctx->coded_height = format->coded_height;
ctx->chroma_format = format->chroma_format;
memset(&cuinfo, 0, sizeof(cuinfo));
cuinfo.CodecType = ctx->codec_type = format->codec;
cuinfo.ChromaFormat = format->chroma_format;
cuinfo.OutputFormat = cudaVideoSurfaceFormat_NV12;
cuinfo.ulWidth = avctx->coded_width;
cuinfo.ulHeight = avctx->coded_height;
cuinfo.ulTargetWidth = cuinfo.ulWidth;
cuinfo.ulTargetHeight = cuinfo.ulHeight;
cuinfo.target_rect.left = 0;
cuinfo.target_rect.top = 0;
cuinfo.target_rect.right = cuinfo.ulWidth;
cuinfo.target_rect.bottom = cuinfo.ulHeight;
cuinfo.ulNumDecodeSurfaces = MAX_FRAME_COUNT;
cuinfo.ulNumOutputSurfaces = 1;
cuinfo.ulCreationFlags = cudaVideoCreate_PreferCUVID;
cuinfo.bitDepthMinus8 = format->bit_depth_luma_minus8;
if (format->progressive_sequence) {
ctx->deint_mode = cuinfo.DeinterlaceMode = cudaVideoDeinterlaceMode_Weave;
} else {
cuinfo.DeinterlaceMode = ctx->deint_mode;
}
if (ctx->deint_mode != cudaVideoDeinterlaceMode_Weave)
avctx->framerate = av_mul_q(avctx->framerate, (AVRational){2, 1});
ctx->internal_error = CHECK_CU(cuvidCreateDecoder(&ctx->cudecoder, &cuinfo));
if (ctx->internal_error < 0)
return 0;
if (!hwframe_ctx->pool) {
hwframe_ctx->format = AV_PIX_FMT_CUDA;
hwframe_ctx->sw_format = AV_PIX_FMT_NV12;
hwframe_ctx->width = avctx->width;
hwframe_ctx->height = avctx->height;
if ((ctx->internal_error = av_hwframe_ctx_init(ctx->hwframe)) < 0) {
av_log(avctx, AV_LOG_ERROR, "av_hwframe_ctx_init failed\n");
return 0;
}
}
return 1;
}
static int hwmap_config_output(AVFilterLink *outlink)
{
AVFilterContext *avctx = outlink->src;
HWMapContext *ctx = avctx->priv;
AVFilterLink *inlink = avctx->inputs[0];
AVHWFramesContext *hwfc;
const AVPixFmtDescriptor *desc;
int err;
av_log(avctx, AV_LOG_DEBUG, "Configure hwmap %s -> %s.\n",
av_get_pix_fmt_name(inlink->format),
av_get_pix_fmt_name(outlink->format));
if (inlink->hw_frames_ctx) {
hwfc = (AVHWFramesContext*)inlink->hw_frames_ctx->data;
desc = av_pix_fmt_desc_get(outlink->format);
if (!desc)
return AVERROR(EINVAL);
if (inlink->format == hwfc->format &&
(desc->flags & AV_PIX_FMT_FLAG_HWACCEL)) {
// Map between two hardware formats (including the case of
// undoing an existing mapping).
ctx->hwdevice_ref = av_buffer_ref(avctx->hw_device_ctx);
if (!ctx->hwdevice_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
err = av_hwframe_ctx_create_derived(&ctx->hwframes_ref,
outlink->format,
ctx->hwdevice_ref,
inlink->hw_frames_ctx, 0);
if (err < 0)
goto fail;
} else if ((outlink->format == hwfc->format &&
inlink->format == hwfc->sw_format) ||
inlink->format == hwfc->format) {
// Map from a hardware format to a software format, or
// undo an existing such mapping.
ctx->hwdevice_ref = NULL;
ctx->hwframes_ref = av_buffer_ref(inlink->hw_frames_ctx);
if (!ctx->hwframes_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
} else {
// Non-matching formats - not supported.
av_log(avctx, AV_LOG_ERROR, "Unsupported formats for "
"hwmap: from %s (%s) to %s.\n",
av_get_pix_fmt_name(inlink->format),
av_get_pix_fmt_name(hwfc->format),
av_get_pix_fmt_name(outlink->format));
err = AVERROR(EINVAL);
goto fail;
}
} else if (avctx->hw_device_ctx) {
// Map from a software format to a hardware format. This
// creates a new hwframe context like hwupload, but then
// returns frames mapped from that to the previous link in
// order to fill them without an additional copy.
ctx->map_backwards = 1;
ctx->hwdevice_ref = av_buffer_ref(avctx->hw_device_ctx);
if (!ctx->hwdevice_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
ctx->hwframes_ref = av_hwframe_ctx_alloc(ctx->hwdevice_ref);
if (!ctx->hwframes_ref) {
err = AVERROR(ENOMEM);
goto fail;
}
hwfc = (AVHWFramesContext*)ctx->hwframes_ref->data;
hwfc->format = outlink->format;
hwfc->sw_format = inlink->format;
hwfc->width = inlink->w;
hwfc->height = inlink->h;
err = av_hwframe_ctx_init(ctx->hwframes_ref);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to create frame "
"context for backward mapping: %d.\n", err);
goto fail;
}
} else {
av_log(avctx, AV_LOG_ERROR, "Mapping requires a hardware "
"context (a device, or frames on input).\n");
return AVERROR(EINVAL);
}
outlink->hw_frames_ctx = av_buffer_ref(ctx->hwframes_ref);
if (!outlink->hw_frames_ctx) {
err = AVERROR(ENOMEM);
goto fail;
}
outlink->w = inlink->w;
outlink->h = inlink->h;
return 0;
fail:
av_buffer_unref(&ctx->hwframes_ref);
av_buffer_unref(&ctx->hwdevice_ref);
return err;
}
static int CUDAAPI cuvid_handle_video_sequence(void *opaque, CUVIDEOFORMAT* format)
{
AVCodecContext *avctx = opaque;
CuvidContext *ctx = avctx->priv_data;
AVHWFramesContext *hwframe_ctx = (AVHWFramesContext*)ctx->hwframe->data;
CUVIDDECODECAPS *caps = NULL;
CUVIDDECODECREATEINFO cuinfo;
int surface_fmt;
int old_width = avctx->width;
int old_height = avctx->height;
enum AVPixelFormat pix_fmts[3] = { AV_PIX_FMT_CUDA,
AV_PIX_FMT_NONE, // Will be updated below
AV_PIX_FMT_NONE };
av_log(avctx, AV_LOG_TRACE, "pfnSequenceCallback, progressive_sequence=%d\n", format->progressive_sequence);
memset(&cuinfo, 0, sizeof(cuinfo));
ctx->internal_error = 0;
avctx->coded_width = cuinfo.ulWidth = format->coded_width;
avctx->coded_height = cuinfo.ulHeight = format->coded_height;
// apply cropping
cuinfo.display_area.left = format->display_area.left + ctx->crop.left;
cuinfo.display_area.top = format->display_area.top + ctx->crop.top;
cuinfo.display_area.right = format->display_area.right - ctx->crop.right;
cuinfo.display_area.bottom = format->display_area.bottom - ctx->crop.bottom;
// width and height need to be set before calling ff_get_format
if (ctx->resize_expr) {
avctx->width = ctx->resize.width;
avctx->height = ctx->resize.height;
} else {
avctx->width = cuinfo.display_area.right - cuinfo.display_area.left;
avctx->height = cuinfo.display_area.bottom - cuinfo.display_area.top;
}
// target width/height need to be multiples of two
cuinfo.ulTargetWidth = avctx->width = (avctx->width + 1) & ~1;
cuinfo.ulTargetHeight = avctx->height = (avctx->height + 1) & ~1;
// aspect ratio conversion, 1:1, depends on scaled resolution
cuinfo.target_rect.left = 0;
cuinfo.target_rect.top = 0;
cuinfo.target_rect.right = cuinfo.ulTargetWidth;
cuinfo.target_rect.bottom = cuinfo.ulTargetHeight;
switch (format->bit_depth_luma_minus8) {
case 0: // 8-bit
pix_fmts[1] = AV_PIX_FMT_NV12;
caps = &ctx->caps8;
break;
case 2: // 10-bit
pix_fmts[1] = AV_PIX_FMT_P010;
caps = &ctx->caps10;
break;
case 4: // 12-bit
pix_fmts[1] = AV_PIX_FMT_P016;
caps = &ctx->caps12;
break;
default:
break;
}
if (!caps || !caps->bIsSupported) {
av_log(avctx, AV_LOG_ERROR, "unsupported bit depth: %d\n",
format->bit_depth_luma_minus8 + 8);
ctx->internal_error = AVERROR(EINVAL);
return 0;
}
surface_fmt = ff_get_format(avctx, pix_fmts);
if (surface_fmt < 0) {
av_log(avctx, AV_LOG_ERROR, "ff_get_format failed: %d\n", surface_fmt);
ctx->internal_error = AVERROR(EINVAL);
return 0;
}
av_log(avctx, AV_LOG_VERBOSE, "Formats: Original: %s | HW: %s | SW: %s\n",
av_get_pix_fmt_name(avctx->pix_fmt),
av_get_pix_fmt_name(surface_fmt),
av_get_pix_fmt_name(avctx->sw_pix_fmt));
avctx->pix_fmt = surface_fmt;
// Update our hwframe ctx, as the get_format callback might have refreshed it!
if (avctx->hw_frames_ctx) {
av_buffer_unref(&ctx->hwframe);
ctx->hwframe = av_buffer_ref(avctx->hw_frames_ctx);
if (!ctx->hwframe) {
ctx->internal_error = AVERROR(ENOMEM);
return 0;
}
hwframe_ctx = (AVHWFramesContext*)ctx->hwframe->data;
}
ff_set_sar(avctx, av_div_q(
(AVRational){ format->display_aspect_ratio.x, format->display_aspect_ratio.y },
(AVRational){ avctx->width, avctx->height }));
ctx->deint_mode_current = format->progressive_sequence
? cudaVideoDeinterlaceMode_Weave
: ctx->deint_mode;
if (!format->progressive_sequence && ctx->deint_mode_current == cudaVideoDeinterlaceMode_Weave)
avctx->flags |= AV_CODEC_FLAG_INTERLACED_DCT;
else
avctx->flags &= ~AV_CODEC_FLAG_INTERLACED_DCT;
if (format->video_signal_description.video_full_range_flag)
avctx->color_range = AVCOL_RANGE_JPEG;
else
avctx->color_range = AVCOL_RANGE_MPEG;
avctx->color_primaries = format->video_signal_description.color_primaries;
avctx->color_trc = format->video_signal_description.transfer_characteristics;
avctx->colorspace = format->video_signal_description.matrix_coefficients;
if (format->bitrate)
avctx->bit_rate = format->bitrate;
if (format->frame_rate.numerator && format->frame_rate.denominator) {
avctx->framerate.num = format->frame_rate.numerator;
avctx->framerate.den = format->frame_rate.denominator;
}
if (ctx->cudecoder
&& avctx->coded_width == format->coded_width
&& avctx->coded_height == format->coded_height
&& avctx->width == old_width
&& avctx->height == old_height
&& ctx->chroma_format == format->chroma_format
&& ctx->codec_type == format->codec)
return 1;
if (ctx->cudecoder) {
av_log(avctx, AV_LOG_TRACE, "Re-initializing decoder\n");
ctx->internal_error = CHECK_CU(ctx->cvdl->cuvidDestroyDecoder(ctx->cudecoder));
if (ctx->internal_error < 0)
return 0;
ctx->cudecoder = NULL;
}
if (hwframe_ctx->pool && (
hwframe_ctx->width < avctx->width ||
hwframe_ctx->height < avctx->height ||
hwframe_ctx->format != AV_PIX_FMT_CUDA ||
hwframe_ctx->sw_format != avctx->sw_pix_fmt)) {
av_log(avctx, AV_LOG_ERROR, "AVHWFramesContext is already initialized with incompatible parameters\n");
av_log(avctx, AV_LOG_DEBUG, "width: %d <-> %d\n", hwframe_ctx->width, avctx->width);
av_log(avctx, AV_LOG_DEBUG, "height: %d <-> %d\n", hwframe_ctx->height, avctx->height);
av_log(avctx, AV_LOG_DEBUG, "format: %s <-> cuda\n", av_get_pix_fmt_name(hwframe_ctx->format));
av_log(avctx, AV_LOG_DEBUG, "sw_format: %s <-> %s\n",
av_get_pix_fmt_name(hwframe_ctx->sw_format), av_get_pix_fmt_name(avctx->sw_pix_fmt));
ctx->internal_error = AVERROR(EINVAL);
return 0;
}
if (format->chroma_format != cudaVideoChromaFormat_420) {
av_log(avctx, AV_LOG_ERROR, "Chroma formats other than 420 are not supported\n");
ctx->internal_error = AVERROR(EINVAL);
return 0;
}
ctx->chroma_format = format->chroma_format;
cuinfo.CodecType = ctx->codec_type = format->codec;
cuinfo.ChromaFormat = format->chroma_format;
switch (avctx->sw_pix_fmt) {
case AV_PIX_FMT_NV12:
cuinfo.OutputFormat = cudaVideoSurfaceFormat_NV12;
break;
case AV_PIX_FMT_P010:
case AV_PIX_FMT_P016:
cuinfo.OutputFormat = cudaVideoSurfaceFormat_P016;
break;
default:
av_log(avctx, AV_LOG_ERROR, "Output formats other than NV12, P010 or P016 are not supported\n");
ctx->internal_error = AVERROR(EINVAL);
return 0;
}
cuinfo.ulNumDecodeSurfaces = ctx->nb_surfaces;
cuinfo.ulNumOutputSurfaces = 1;
cuinfo.ulCreationFlags = cudaVideoCreate_PreferCUVID;
cuinfo.bitDepthMinus8 = format->bit_depth_luma_minus8;
cuinfo.DeinterlaceMode = ctx->deint_mode_current;
if (ctx->deint_mode_current != cudaVideoDeinterlaceMode_Weave && !ctx->drop_second_field)
avctx->framerate = av_mul_q(avctx->framerate, (AVRational){2, 1});
ctx->internal_error = CHECK_CU(ctx->cvdl->cuvidCreateDecoder(&ctx->cudecoder, &cuinfo));
if (ctx->internal_error < 0)
return 0;
if (!hwframe_ctx->pool) {
hwframe_ctx->format = AV_PIX_FMT_CUDA;
hwframe_ctx->sw_format = avctx->sw_pix_fmt;
hwframe_ctx->width = avctx->width;
hwframe_ctx->height = avctx->height;
if ((ctx->internal_error = av_hwframe_ctx_init(ctx->hwframe)) < 0) {
av_log(avctx, AV_LOG_ERROR, "av_hwframe_ctx_init failed\n");
return 0;
}
}
return 1;
}
static int vdpau_alloc(AVCodecContext *s)
{
int loglevel = AV_LOG_ERROR;
VDPAUContext *ctx;
const char *display, *vendor;
VdpStatus err;
int ret;
VdpDevice device;
VdpGetProcAddress *get_proc_address;
VdpGetInformationString *get_information_string;
VDPAUHWDevicePriv *device_priv = NULL;
AVHWDeviceContext *device_ctx;
AVVDPAUDeviceContext *device_hwctx;
AVHWFramesContext *frames_ctx;
ctx = av_mallocz(sizeof(*ctx));
if (!ctx)
return AVERROR(ENOMEM);
device_priv = av_mallocz(sizeof(*device_priv));
if (!device_priv) {
av_freep(&ctx);
goto fail;
}
device_priv->dpy = XOpenDisplay(":0");
if (!device_priv->dpy) {
av_log(NULL, loglevel, "Cannot open the X11 display %s.\n",
XDisplayName(":0"));
goto fail;
}
display = XDisplayString(device_priv->dpy);
err = vdp_device_create_x11(device_priv->dpy, XDefaultScreen(device_priv->dpy),
&device, &get_proc_address);
if (err != VDP_STATUS_OK) {
av_log(NULL, loglevel, "VDPAU device creation on X11 display %s failed.\n",
display);
goto fail;
}
#define GET_CALLBACK(id, result) \
do { \
void *tmp; \
err = get_proc_address(device, id, &tmp); \
if (err != VDP_STATUS_OK) { \
av_log(NULL, loglevel, "Error getting the " #id " callback.\n"); \
goto fail; \
} \
result = tmp; \
} while (0)
GET_CALLBACK(VDP_FUNC_ID_GET_INFORMATION_STRING, get_information_string);
GET_CALLBACK(VDP_FUNC_ID_DEVICE_DESTROY, device_priv->device_destroy);
device_ref = av_hwdevice_ctx_alloc(AV_HWDEVICE_TYPE_VDPAU);
if (!device_ref)
goto fail;
device_ctx = (AVHWDeviceContext*)device_ref->data;
device_hwctx = device_ctx->hwctx;
device_ctx->user_opaque = device_priv;
device_ctx->free = device_free;
device_hwctx->device = device;
device_hwctx->get_proc_address = get_proc_address;
device_priv = NULL;
ret = av_hwdevice_ctx_init(device_ref);
if (ret < 0)
goto fail;
ctx->hw_frames_ctx = av_hwframe_ctx_alloc(device_ref);
if (!ctx->hw_frames_ctx)
goto fail;
//av_buffer_unref(&device_ref);
frames_ctx = (AVHWFramesContext*)ctx->hw_frames_ctx->data;
frames_ctx->format = AV_PIX_FMT_VDPAU;
frames_ctx->sw_format = s->sw_pix_fmt;
frames_ctx->width = 1920;
frames_ctx->height = 1080;
ret = av_hwframe_ctx_init(ctx->hw_frames_ctx);
if (ret < 0)
goto fail;
if (av_vdpau_bind_context(s, device, get_proc_address, 0))
goto fail;
s->opaque = ctx;
return 0;
fail:
if (device_priv) {
if (device_priv->device_destroy)
device_priv->device_destroy(device);
if (device_priv->dpy)
XCloseDisplay(device_priv->dpy);
}
av_freep(&device_priv);
av_buffer_unref(&device_ref);
vdpau_uninit(s);
return AVERROR(EINVAL);
}
static bool vaapi_init_codec(struct vaapi_encoder *enc, const char *path)
{
int ret;
ret = av_hwdevice_ctx_create(&enc->vadevice_ref, AV_HWDEVICE_TYPE_VAAPI,
path, NULL, 0);
if (ret < 0) {
warn("Failed to create VAAPI device context: %s",
av_err2str(ret));
return false;
}
enc->vaframes_ref = av_hwframe_ctx_alloc(enc->vadevice_ref);
if (!enc->vaframes_ref) {
warn("Failed to alloc HW frames context");
return false;
}
AVHWFramesContext *frames_ctx =
(AVHWFramesContext *)enc->vaframes_ref->data;
frames_ctx->format = AV_PIX_FMT_VAAPI;
frames_ctx->sw_format = AV_PIX_FMT_NV12;
frames_ctx->width = enc->context->width;
frames_ctx->height = enc->context->height;
frames_ctx->initial_pool_size = 20;
ret = av_hwframe_ctx_init(enc->vaframes_ref);
if (ret < 0) {
warn("Failed to init HW frames context: %s", av_err2str(ret));
return false;
}
/* 2. Create software frame and picture */
enc->vframe = av_frame_alloc();
if (!enc->vframe) {
warn("Failed to allocate video frame");
return false;
}
enc->vframe->format = enc->context->pix_fmt;
enc->vframe->width = enc->context->width;
enc->vframe->height = enc->context->height;
enc->vframe->colorspace = enc->context->colorspace;
enc->vframe->color_range = enc->context->color_range;
ret = av_frame_get_buffer(enc->vframe, base_get_alignment());
if (ret < 0) {
warn("Failed to allocate vframe: %s", av_err2str(ret));
return false;
}
/* 3. set up codec */
enc->context->pix_fmt = AV_PIX_FMT_VAAPI;
enc->context->hw_frames_ctx = av_buffer_ref(enc->vaframes_ref);
ret = avcodec_open2(enc->context, enc->vaapi, NULL);
if (ret < 0) {
warn("Failed to open VAAPI codec: %s", av_err2str(ret));
return false;
}
enc->initialized = true;
return true;
}
int ff_vdpau_common_init(AVCodecContext *avctx, VdpDecoderProfile profile,
int level)
{
VDPAUHWContext *hwctx = avctx->hwaccel_context;
VDPAUContext *vdctx = avctx->internal->hwaccel_priv_data;
VdpVideoSurfaceQueryCapabilities *surface_query_caps;
VdpDecoderQueryCapabilities *decoder_query_caps;
VdpDecoderCreate *create;
void *func;
VdpStatus status;
VdpBool supported;
uint32_t max_level, max_mb, max_width, max_height;
VdpChromaType type;
uint32_t width;
uint32_t height;
vdctx->width = UINT32_MAX;
vdctx->height = UINT32_MAX;
if (av_vdpau_get_surface_parameters(avctx, &type, &width, &height))
return AVERROR(ENOSYS);
if (hwctx) {
hwctx->reset = 0;
if (hwctx->context.decoder != VDP_INVALID_HANDLE) {
vdctx->decoder = hwctx->context.decoder;
vdctx->render = hwctx->context.render;
vdctx->device = VDP_INVALID_HANDLE;
return 0; /* Decoder created by user */
}
vdctx->device = hwctx->device;
vdctx->get_proc_address = hwctx->get_proc_address;
if (hwctx->flags & AV_HWACCEL_FLAG_IGNORE_LEVEL)
level = 0;
if (!(hwctx->flags & AV_HWACCEL_FLAG_ALLOW_HIGH_DEPTH) &&
type != VDP_CHROMA_TYPE_420)
return AVERROR(ENOSYS);
} else {
AVHWFramesContext *frames_ctx = NULL;
AVVDPAUDeviceContext *dev_ctx;
// We assume the hw_frames_ctx always survives until ff_vdpau_common_uninit
// is called. This holds true as the user is not allowed to touch
// hw_device_ctx, or hw_frames_ctx after get_format (and ff_get_format
// itself also uninits before unreffing hw_frames_ctx).
if (avctx->hw_frames_ctx) {
frames_ctx = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
} else if (avctx->hw_device_ctx) {
int ret;
avctx->hw_frames_ctx = av_hwframe_ctx_alloc(avctx->hw_device_ctx);
if (!avctx->hw_frames_ctx)
return AVERROR(ENOMEM);
frames_ctx = (AVHWFramesContext*)avctx->hw_frames_ctx->data;
frames_ctx->format = AV_PIX_FMT_VDPAU;
frames_ctx->sw_format = avctx->sw_pix_fmt;
frames_ctx->width = avctx->coded_width;
frames_ctx->height = avctx->coded_height;
ret = av_hwframe_ctx_init(avctx->hw_frames_ctx);
if (ret < 0) {
av_buffer_unref(&avctx->hw_frames_ctx);
return ret;
}
}
if (!frames_ctx) {
av_log(avctx, AV_LOG_ERROR, "A hardware frames context is "
"required for VDPAU decoding.\n");
return AVERROR(EINVAL);
}
dev_ctx = frames_ctx->device_ctx->hwctx;
vdctx->device = dev_ctx->device;
vdctx->get_proc_address = dev_ctx->get_proc_address;
if (avctx->hwaccel_flags & AV_HWACCEL_FLAG_IGNORE_LEVEL)
level = 0;
}
if (level < 0)
return AVERROR(ENOTSUP);
status = vdctx->get_proc_address(vdctx->device,
VDP_FUNC_ID_VIDEO_SURFACE_QUERY_CAPABILITIES,
&func);
if (status != VDP_STATUS_OK)
return vdpau_error(status);
else
surface_query_caps = func;
status = surface_query_caps(vdctx->device, type, &supported,
&max_width, &max_height);
if (status != VDP_STATUS_OK)
return vdpau_error(status);
if (supported != VDP_TRUE ||
max_width < width || max_height < height)
return AVERROR(ENOTSUP);
status = vdctx->get_proc_address(vdctx->device,
VDP_FUNC_ID_DECODER_QUERY_CAPABILITIES,
&func);
if (status != VDP_STATUS_OK)
return vdpau_error(status);
else
decoder_query_caps = func;
status = decoder_query_caps(vdctx->device, profile, &supported, &max_level,
&max_mb, &max_width, &max_height);
#ifdef VDP_DECODER_PROFILE_H264_CONSTRAINED_BASELINE
if ((status != VDP_STATUS_OK || supported != VDP_TRUE) && profile == VDP_DECODER_PROFILE_H264_CONSTRAINED_BASELINE) {
profile = VDP_DECODER_PROFILE_H264_MAIN;
status = decoder_query_caps(vdctx->device, profile, &supported,
&max_level, &max_mb,
&max_width, &max_height);
}
#endif
if (status != VDP_STATUS_OK)
return vdpau_error(status);
if (supported != VDP_TRUE || max_level < level ||
max_width < width || max_height < height)
return AVERROR(ENOTSUP);
status = vdctx->get_proc_address(vdctx->device, VDP_FUNC_ID_DECODER_CREATE,
&func);
if (status != VDP_STATUS_OK)
return vdpau_error(status);
else
create = func;
status = vdctx->get_proc_address(vdctx->device, VDP_FUNC_ID_DECODER_RENDER,
&func);
if (status != VDP_STATUS_OK)
return vdpau_error(status);
else
vdctx->render = func;
status = create(vdctx->device, profile, width, height, avctx->refs,
&vdctx->decoder);
if (status == VDP_STATUS_OK) {
vdctx->width = avctx->coded_width;
vdctx->height = avctx->coded_height;
}
return vdpau_error(status);
}
static mfxStatus qsv_frame_alloc(mfxHDL pthis, mfxFrameAllocRequest *req,
mfxFrameAllocResponse *resp)
{
QSVFramesContext *ctx = pthis;
int ret;
/* this should only be called from an encoder or decoder and
* only allocates video memory frames */
if (!(req->Type & (MFX_MEMTYPE_VIDEO_MEMORY_DECODER_TARGET |
MFX_MEMTYPE_VIDEO_MEMORY_PROCESSOR_TARGET)) ||
!(req->Type & (MFX_MEMTYPE_FROM_DECODE | MFX_MEMTYPE_FROM_ENCODE)))
return MFX_ERR_UNSUPPORTED;
if (req->Type & MFX_MEMTYPE_EXTERNAL_FRAME) {
/* external frames -- fill from the caller-supplied frames context */
AVHWFramesContext *frames_ctx = (AVHWFramesContext*)ctx->hw_frames_ctx->data;
AVQSVFramesContext *frames_hwctx = frames_ctx->hwctx;
mfxFrameInfo *i = &req->Info;
mfxFrameInfo *i1 = &frames_hwctx->surfaces[0].Info;
if (i->Width > i1->Width || i->Height > i1->Height ||
i->FourCC != i1->FourCC || i->ChromaFormat != i1->ChromaFormat) {
av_log(ctx->logctx, AV_LOG_ERROR, "Mismatching surface properties in an "
"allocation request: %dx%d %d %d vs %dx%d %d %d\n",
i->Width, i->Height, i->FourCC, i->ChromaFormat,
i1->Width, i1->Height, i1->FourCC, i1->ChromaFormat);
return MFX_ERR_UNSUPPORTED;
}
ret = qsv_setup_mids(resp, ctx->hw_frames_ctx, ctx->mids_buf);
if (ret < 0) {
av_log(ctx->logctx, AV_LOG_ERROR,
"Error filling an external frame allocation request\n");
return MFX_ERR_MEMORY_ALLOC;
}
} else if (req->Type & MFX_MEMTYPE_INTERNAL_FRAME) {
/* internal frames -- allocate a new hw frames context */
AVHWFramesContext *ext_frames_ctx = (AVHWFramesContext*)ctx->hw_frames_ctx->data;
mfxFrameInfo *i = &req->Info;
AVBufferRef *frames_ref, *mids_buf;
AVHWFramesContext *frames_ctx;
AVQSVFramesContext *frames_hwctx;
frames_ref = av_hwframe_ctx_alloc(ext_frames_ctx->device_ref);
if (!frames_ref)
return MFX_ERR_MEMORY_ALLOC;
frames_ctx = (AVHWFramesContext*)frames_ref->data;
frames_hwctx = frames_ctx->hwctx;
frames_ctx->format = AV_PIX_FMT_QSV;
frames_ctx->sw_format = qsv_map_fourcc(i->FourCC);
frames_ctx->width = i->Width;
frames_ctx->height = i->Height;
frames_ctx->initial_pool_size = req->NumFrameSuggested;
frames_hwctx->frame_type = req->Type;
ret = av_hwframe_ctx_init(frames_ref);
if (ret < 0) {
av_log(ctx->logctx, AV_LOG_ERROR,
"Error initializing a frames context for an internal frame "
"allocation request\n");
av_buffer_unref(&frames_ref);
return MFX_ERR_MEMORY_ALLOC;
}
mids_buf = qsv_create_mids(frames_ref);
if (!mids_buf) {
av_buffer_unref(&frames_ref);
return MFX_ERR_MEMORY_ALLOC;
}
ret = qsv_setup_mids(resp, frames_ref, mids_buf);
av_buffer_unref(&mids_buf);
av_buffer_unref(&frames_ref);
if (ret < 0) {
av_log(ctx->logctx, AV_LOG_ERROR,
"Error filling an internal frame allocation request\n");
return MFX_ERR_MEMORY_ALLOC;
}
} else {
return MFX_ERR_UNSUPPORTED;
}
return MFX_ERR_NONE;
}
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;
}
static int deint_vaapi_config_output(AVFilterLink *outlink)
{
AVFilterContext *avctx = outlink->src;
AVFilterLink *inlink = avctx->inputs[0];
DeintVAAPIContext *ctx = avctx->priv;
AVVAAPIHWConfig *hwconfig = NULL;
AVHWFramesConstraints *constraints = NULL;
AVVAAPIFramesContext *va_frames;
VAStatus vas;
int err;
deint_vaapi_pipeline_uninit(avctx);
av_assert0(ctx->input_frames);
ctx->device_ref = av_buffer_ref(ctx->input_frames->device_ref);
ctx->hwctx = ((AVHWDeviceContext*)ctx->device_ref->data)->hwctx;
ctx->output_width = ctx->input_frames->width;
ctx->output_height = ctx->input_frames->height;
av_assert0(ctx->va_config == VA_INVALID_ID);
vas = vaCreateConfig(ctx->hwctx->display, VAProfileNone,
VAEntrypointVideoProc, 0, 0, &ctx->va_config);
if (vas != VA_STATUS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "Failed to create processing pipeline "
"config: %d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail;
}
hwconfig = av_hwdevice_hwconfig_alloc(ctx->device_ref);
if (!hwconfig) {
err = AVERROR(ENOMEM);
goto fail;
}
hwconfig->config_id = ctx->va_config;
constraints = av_hwdevice_get_hwframe_constraints(ctx->device_ref,
hwconfig);
if (!constraints) {
err = AVERROR(ENOMEM);
goto fail;
}
if (ctx->output_width < constraints->min_width ||
ctx->output_height < constraints->min_height ||
ctx->output_width > constraints->max_width ||
ctx->output_height > constraints->max_height) {
av_log(avctx, AV_LOG_ERROR, "Hardware does not support "
"deinterlacing to size %dx%d "
"(constraints: width %d-%d height %d-%d).\n",
ctx->output_width, ctx->output_height,
constraints->min_width, constraints->max_width,
constraints->min_height, constraints->max_height);
err = AVERROR(EINVAL);
goto fail;
}
ctx->output_frames_ref = av_hwframe_ctx_alloc(ctx->device_ref);
if (!ctx->output_frames_ref) {
av_log(avctx, AV_LOG_ERROR, "Failed to create HW frame context "
"for output.\n");
err = AVERROR(ENOMEM);
goto fail;
}
ctx->output_frames = (AVHWFramesContext*)ctx->output_frames_ref->data;
ctx->output_frames->format = AV_PIX_FMT_VAAPI;
ctx->output_frames->sw_format = ctx->input_frames->sw_format;
ctx->output_frames->width = ctx->output_width;
ctx->output_frames->height = ctx->output_height;
// The number of output frames we need is determined by what follows
// the filter. If it's an encoder with complex frame reference
// structures then this could be very high.
ctx->output_frames->initial_pool_size = 10;
err = av_hwframe_ctx_init(ctx->output_frames_ref);
if (err < 0) {
av_log(avctx, AV_LOG_ERROR, "Failed to initialise VAAPI frame "
"context for output: %d\n", err);
goto fail;
}
va_frames = ctx->output_frames->hwctx;
av_assert0(ctx->va_context == VA_INVALID_ID);
vas = vaCreateContext(ctx->hwctx->display, ctx->va_config,
ctx->output_width, ctx->output_height, 0,
va_frames->surface_ids, va_frames->nb_surfaces,
&ctx->va_context);
if (vas != VA_STATUS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "Failed to create processing pipeline "
"context: %d (%s).\n", vas, vaErrorStr(vas));
err = AVERROR(EIO);
goto fail;
}
err = deint_vaapi_build_filter_params(avctx);
if (err < 0)
goto fail;
outlink->w = inlink->w;
outlink->h = inlink->h;
outlink->time_base = av_mul_q(inlink->time_base,
(AVRational) { 1, ctx->field_rate });
outlink->frame_rate = av_mul_q(inlink->frame_rate,
(AVRational) { ctx->field_rate, 1 });
outlink->hw_frames_ctx = av_buffer_ref(ctx->output_frames_ref);
if (!outlink->hw_frames_ctx) {
err = AVERROR(ENOMEM);
goto fail;
}
av_freep(&hwconfig);
av_hwframe_constraints_free(&constraints);
return 0;
fail:
av_buffer_unref(&ctx->output_frames_ref);
av_freep(&hwconfig);
av_hwframe_constraints_free(&constraints);
return err;
}
/* create the QSV session */
static int init_vpp_session(AVFilterContext *avctx, QSVVPPContext *s)
{
AVFilterLink *inlink = avctx->inputs[0];
AVFilterLink *outlink = avctx->outputs[0];
AVQSVFramesContext *in_frames_hwctx = NULL;
AVQSVFramesContext *out_frames_hwctx = NULL;
AVBufferRef *device_ref;
AVHWDeviceContext *device_ctx;
AVQSVDeviceContext *device_hwctx;
mfxHDL handle;
mfxHandleType handle_type;
mfxVersion ver;
mfxIMPL impl;
int ret, i;
if (inlink->hw_frames_ctx) {
AVHWFramesContext *frames_ctx = (AVHWFramesContext *)inlink->hw_frames_ctx->data;
device_ref = frames_ctx->device_ref;
in_frames_hwctx = frames_ctx->hwctx;
s->in_mem_mode = in_frames_hwctx->frame_type;
s->surface_ptrs_in = av_mallocz_array(in_frames_hwctx->nb_surfaces,
sizeof(*s->surface_ptrs_in));
if (!s->surface_ptrs_in)
return AVERROR(ENOMEM);
for (i = 0; i < in_frames_hwctx->nb_surfaces; i++)
s->surface_ptrs_in[i] = in_frames_hwctx->surfaces + i;
s->nb_surface_ptrs_in = in_frames_hwctx->nb_surfaces;
} else if (avctx->hw_device_ctx) {
device_ref = avctx->hw_device_ctx;
s->in_mem_mode = MFX_MEMTYPE_SYSTEM_MEMORY;
} else {
av_log(avctx, AV_LOG_ERROR, "No hw context provided.\n");
return AVERROR(EINVAL);
}
device_ctx = (AVHWDeviceContext *)device_ref->data;
device_hwctx = device_ctx->hwctx;
if (outlink->format == AV_PIX_FMT_QSV) {
AVHWFramesContext *out_frames_ctx;
AVBufferRef *out_frames_ref = av_hwframe_ctx_alloc(device_ref);
if (!out_frames_ref)
return AVERROR(ENOMEM);
s->out_mem_mode = IS_OPAQUE_MEMORY(s->in_mem_mode) ?
MFX_MEMTYPE_OPAQUE_FRAME :
MFX_MEMTYPE_VIDEO_MEMORY_DECODER_TARGET;
out_frames_ctx = (AVHWFramesContext *)out_frames_ref->data;
out_frames_hwctx = out_frames_ctx->hwctx;
out_frames_ctx->format = AV_PIX_FMT_QSV;
out_frames_ctx->width = FFALIGN(outlink->w, 32);
out_frames_ctx->height = FFALIGN(outlink->h, 32);
out_frames_ctx->sw_format = s->out_sw_format;
out_frames_ctx->initial_pool_size = 64;
out_frames_hwctx->frame_type = s->out_mem_mode;
ret = av_hwframe_ctx_init(out_frames_ref);
if (ret < 0) {
av_buffer_unref(&out_frames_ref);
av_log(avctx, AV_LOG_ERROR, "Error creating frames_ctx for output pad.\n");
return ret;
}
s->surface_ptrs_out = av_mallocz_array(out_frames_hwctx->nb_surfaces,
sizeof(*s->surface_ptrs_out));
if (!s->surface_ptrs_out) {
av_buffer_unref(&out_frames_ref);
return AVERROR(ENOMEM);
}
for (i = 0; i < out_frames_hwctx->nb_surfaces; i++)
s->surface_ptrs_out[i] = out_frames_hwctx->surfaces + i;
s->nb_surface_ptrs_out = out_frames_hwctx->nb_surfaces;
av_buffer_unref(&outlink->hw_frames_ctx);
outlink->hw_frames_ctx = out_frames_ref;
} else
s->out_mem_mode = MFX_MEMTYPE_SYSTEM_MEMORY;
/* extract the properties of the "master" session given to us */
ret = MFXQueryIMPL(device_hwctx->session, &impl);
if (ret == MFX_ERR_NONE)
ret = MFXQueryVersion(device_hwctx->session, &ver);
if (ret != MFX_ERR_NONE) {
av_log(avctx, AV_LOG_ERROR, "Error querying the session attributes\n");
return AVERROR_UNKNOWN;
}
for (i = 0; i < FF_ARRAY_ELEMS(handle_types); i++) {
ret = MFXVideoCORE_GetHandle(device_hwctx->session, handle_types[i], &handle);
if (ret == MFX_ERR_NONE) {
handle_type = handle_types[i];
break;
}
}
if (ret != MFX_ERR_NONE) {
av_log(avctx, AV_LOG_ERROR, "Error getting the session handle\n");
return AVERROR_UNKNOWN;
}
/* create a "slave" session with those same properties, to be used for vpp */
ret = MFXInit(impl, &ver, &s->session);
if (ret != MFX_ERR_NONE) {
av_log(avctx, AV_LOG_ERROR, "Error initializing a session for scaling\n");
return AVERROR_UNKNOWN;
}
if (handle) {
ret = MFXVideoCORE_SetHandle(s->session, handle_type, handle);
if (ret != MFX_ERR_NONE)
return AVERROR_UNKNOWN;
}
if (QSV_RUNTIME_VERSION_ATLEAST(ver, 1, 25)) {
ret = MFXJoinSession(device_hwctx->session, s->session);
if (ret != MFX_ERR_NONE)
return AVERROR_UNKNOWN;
}
if (IS_OPAQUE_MEMORY(s->in_mem_mode) || IS_OPAQUE_MEMORY(s->out_mem_mode)) {
s->opaque_alloc.In.Surfaces = s->surface_ptrs_in;
s->opaque_alloc.In.NumSurface = s->nb_surface_ptrs_in;
s->opaque_alloc.In.Type = s->in_mem_mode;
s->opaque_alloc.Out.Surfaces = s->surface_ptrs_out;
s->opaque_alloc.Out.NumSurface = s->nb_surface_ptrs_out;
s->opaque_alloc.Out.Type = s->out_mem_mode;
s->opaque_alloc.Header.BufferId = MFX_EXTBUFF_OPAQUE_SURFACE_ALLOCATION;
s->opaque_alloc.Header.BufferSz = sizeof(s->opaque_alloc);
} else if (IS_VIDEO_MEMORY(s->in_mem_mode) || IS_VIDEO_MEMORY(s->out_mem_mode)) {
mfxFrameAllocator frame_allocator = {
.pthis = s,
.Alloc = frame_alloc,
.Lock = frame_lock,
.Unlock = frame_unlock,
.GetHDL = frame_get_hdl,
.Free = frame_free,
};
ret = MFXVideoCORE_SetFrameAllocator(s->session, &frame_allocator);
if (ret != MFX_ERR_NONE)
return AVERROR_UNKNOWN;
}
static int qsv_init_child_ctx(AVHWFramesContext *ctx)
{
AVQSVFramesContext *hwctx = ctx->hwctx;
QSVFramesContext *s = ctx->internal->priv;
QSVDeviceContext *device_priv = ctx->device_ctx->internal->priv;
AVBufferRef *child_device_ref = NULL;
AVBufferRef *child_frames_ref = NULL;
AVHWDeviceContext *child_device_ctx;
AVHWFramesContext *child_frames_ctx;
int i, ret = 0;
if (!device_priv->handle) {
av_log(ctx, AV_LOG_ERROR,
"Cannot create a non-opaque internal surface pool without "
"a hardware handle\n");
return AVERROR(EINVAL);
}
child_device_ref = av_hwdevice_ctx_alloc(device_priv->child_device_type);
if (!child_device_ref)
return AVERROR(ENOMEM);
child_device_ctx = (AVHWDeviceContext*)child_device_ref->data;
#if CONFIG_VAAPI
if (child_device_ctx->type == AV_HWDEVICE_TYPE_VAAPI) {
AVVAAPIDeviceContext *child_device_hwctx = child_device_ctx->hwctx;
child_device_hwctx->display = (VADisplay)device_priv->handle;
}
#endif
#if CONFIG_DXVA2
if (child_device_ctx->type == AV_HWDEVICE_TYPE_DXVA2) {
AVDXVA2DeviceContext *child_device_hwctx = child_device_ctx->hwctx;
child_device_hwctx->devmgr = (IDirect3DDeviceManager9*)device_priv->handle;
}
#endif
ret = av_hwdevice_ctx_init(child_device_ref);
if (ret < 0) {
av_log(ctx, AV_LOG_ERROR, "Error initializing a child device context\n");
goto fail;
}
child_frames_ref = av_hwframe_ctx_alloc(child_device_ref);
if (!child_frames_ref) {
ret = AVERROR(ENOMEM);
goto fail;
}
child_frames_ctx = (AVHWFramesContext*)child_frames_ref->data;
child_frames_ctx->format = device_priv->child_pix_fmt;
child_frames_ctx->sw_format = ctx->sw_format;
child_frames_ctx->initial_pool_size = ctx->initial_pool_size;
child_frames_ctx->width = FFALIGN(ctx->width, 16);
child_frames_ctx->height = FFALIGN(ctx->height, 16);
#if CONFIG_DXVA2
if (child_device_ctx->type == AV_HWDEVICE_TYPE_DXVA2) {
AVDXVA2FramesContext *child_frames_hwctx = child_frames_ctx->hwctx;
if (hwctx->frame_type & MFX_MEMTYPE_VIDEO_MEMORY_PROCESSOR_TARGET)
child_frames_hwctx->surface_type = DXVA2_VideoProcessorRenderTarget;
else
child_frames_hwctx->surface_type = DXVA2_VideoDecoderRenderTarget;
}
#endif
ret = av_hwframe_ctx_init(child_frames_ref);
if (ret < 0) {
av_log(ctx, AV_LOG_ERROR, "Error initializing a child frames context\n");
goto fail;
}
#if CONFIG_VAAPI
if (child_device_ctx->type == AV_HWDEVICE_TYPE_VAAPI) {
AVVAAPIFramesContext *child_frames_hwctx = child_frames_ctx->hwctx;
for (i = 0; i < ctx->initial_pool_size; i++)
s->surfaces_internal[i].Data.MemId = child_frames_hwctx->surface_ids + i;
hwctx->frame_type = MFX_MEMTYPE_VIDEO_MEMORY_DECODER_TARGET;
}
#endif
#if CONFIG_DXVA2
if (child_device_ctx->type == AV_HWDEVICE_TYPE_DXVA2) {
AVDXVA2FramesContext *child_frames_hwctx = child_frames_ctx->hwctx;
for (i = 0; i < ctx->initial_pool_size; i++)
s->surfaces_internal[i].Data.MemId = (mfxMemId)child_frames_hwctx->surfaces[i];
if (child_frames_hwctx->surface_type == DXVA2_VideoProcessorRenderTarget)
hwctx->frame_type = MFX_MEMTYPE_VIDEO_MEMORY_PROCESSOR_TARGET;
else
hwctx->frame_type = MFX_MEMTYPE_VIDEO_MEMORY_DECODER_TARGET;
}
#endif
s->child_frames_ref = child_frames_ref;
child_frames_ref = NULL;
fail:
av_buffer_unref(&child_device_ref);
av_buffer_unref(&child_frames_ref);
return ret;
}
/**
\fn pre-open
*/
bool ADM_ffVAEncHEVC::configureContext(void)
{
ADM_info("Configuring context for VAAPI encoder\n");
ADM_info("Our display: %#x\n",admLibVA::getDisplay());
_context->bit_rate=VaEncHevcSettings.bitrate*1000;
_context->rc_max_rate=VaEncHevcSettings.max_bitrate*1000;
_context->max_b_frames=VaEncHevcSettings.bframes;
_context->pix_fmt =AV_PIX_FMT_VAAPI;
#define CLEARTEXT(x) char buf[AV_ERROR_MAX_STRING_SIZE]={0}; av_make_error_string(buf,AV_ERROR_MAX_STRING_SIZE,x);
hwDeviceCtx = av_hwdevice_ctx_alloc(AV_HWDEVICE_TYPE_VAAPI);
if(!hwDeviceCtx)
{
ADM_error("Cannot allocate hw device context.\n");
return false;
}
AVHWDeviceContext *hwctx = (AVHWDeviceContext *)hwDeviceCtx->data;
AVVAAPIDeviceContext *vactx = (AVVAAPIDeviceContext *)hwctx->hwctx;
vactx->display = admLibVA::getDisplay();
int err = av_hwdevice_ctx_init(hwDeviceCtx);
if(err)
{
CLEARTEXT(err)
ADM_warning("Cannot initialize VAAPI hwdevice (%d, %s)\n",err,buf);
return false;
}
AVBufferRef *hwFramesRef = NULL;
AVHWFramesContext *hwFramesCtx = NULL;
hwFramesRef = av_hwframe_ctx_alloc(hwDeviceCtx);
if(!hwFramesRef)
{
ADM_error("Cannot create VAAPI frame context.\n");
return false;
}
hwFramesCtx=(AVHWFramesContext*)(hwFramesRef->data);
hwFramesCtx->format=AV_PIX_FMT_VAAPI;
hwFramesCtx->sw_format=AV_PIX_FMT_NV12;
hwFramesCtx->width=source->getInfo()->width;
hwFramesCtx->height=source->getInfo()->height;
hwFramesCtx->initial_pool_size=20;
err = av_hwframe_ctx_init(hwFramesRef);
if(err<0)
{
CLEARTEXT(err)
ADM_error("Cannot initialize VAAPI frame context (%d, %s)\n",err,buf);
av_buffer_unref(&hwFramesRef);
return false;
}
_context->hw_frames_ctx = av_buffer_ref(hwFramesRef);
if(!_context->hw_frames_ctx)
{
ADM_error("hw_frames_ctx is NULL!\n");
return false;
}
av_buffer_unref(&hwFramesRef);
return true;
}
static int dxva2_create_decoder(AVCodecContext *s)
{
InputStream *ist = s->opaque;
int loglevel = (ist->hwaccel_id == HWACCEL_AUTO) ? AV_LOG_VERBOSE : AV_LOG_ERROR;
DXVA2Context *ctx = ist->hwaccel_ctx;
struct dxva_context *dxva_ctx = s->hwaccel_context;
GUID *guid_list = NULL;
unsigned guid_count = 0, i, j;
GUID device_guid = GUID_NULL;
const D3DFORMAT surface_format = s->sw_pix_fmt == AV_PIX_FMT_YUV420P10 ?
MKTAG('P', '0', '1', '0') : MKTAG('N', 'V', '1', '2');
D3DFORMAT target_format = 0;
DXVA2_VideoDesc desc = { 0 };
DXVA2_ConfigPictureDecode config;
HRESULT hr;
int surface_alignment, num_surfaces;
int ret;
AVDXVA2FramesContext *frames_hwctx;
AVHWFramesContext *frames_ctx;
hr = IDirectXVideoDecoderService_GetDecoderDeviceGuids(ctx->decoder_service, &guid_count, &guid_list);
if (FAILED(hr)) {
av_log(NULL, loglevel, "Failed to retrieve decoder device GUIDs\n");
goto fail;
}
for (i = 0; dxva2_modes[i].guid; i++) {
D3DFORMAT *target_list = NULL;
unsigned target_count = 0;
const dxva2_mode *mode = &dxva2_modes[i];
if (mode->codec != s->codec_id)
continue;
for (j = 0; j < guid_count; j++) {
if (IsEqualGUID(mode->guid, &guid_list[j]))
break;
}
if (j == guid_count)
continue;
hr = IDirectXVideoDecoderService_GetDecoderRenderTargets(ctx->decoder_service, mode->guid, &target_count, &target_list);
if (FAILED(hr)) {
continue;
}
for (j = 0; j < target_count; j++) {
const D3DFORMAT format = target_list[j];
if (format == surface_format) {
target_format = format;
break;
}
}
CoTaskMemFree(target_list);
if (target_format) {
device_guid = *mode->guid;
break;
}
}
CoTaskMemFree(guid_list);
if (IsEqualGUID(&device_guid, &GUID_NULL)) {
av_log(NULL, loglevel, "No decoder device for codec found\n");
goto fail;
}
desc.SampleWidth = s->coded_width;
desc.SampleHeight = s->coded_height;
desc.Format = target_format;
ret = dxva2_get_decoder_configuration(s, &device_guid, &desc, &config);
if (ret < 0) {
goto fail;
}
/* decoding MPEG-2 requires additional alignment on some Intel GPUs,
but it causes issues for H.264 on certain AMD GPUs..... */
if (s->codec_id == AV_CODEC_ID_MPEG2VIDEO)
surface_alignment = 32;
/* the HEVC DXVA2 spec asks for 128 pixel aligned surfaces to ensure
all coding features have enough room to work with */
else if (s->codec_id == AV_CODEC_ID_HEVC)
surface_alignment = 128;
else
surface_alignment = 16;
/* 4 base work surfaces */
num_surfaces = 4;
/* add surfaces based on number of possible refs */
if (s->codec_id == AV_CODEC_ID_H264 || s->codec_id == AV_CODEC_ID_HEVC)
num_surfaces += 16;
else
num_surfaces += 2;
/* add extra surfaces for frame threading */
if (s->active_thread_type & FF_THREAD_FRAME)
num_surfaces += s->thread_count;
ctx->hw_frames_ctx = av_hwframe_ctx_alloc(ctx->hw_device_ctx);
if (!ctx->hw_frames_ctx)
goto fail;
frames_ctx = (AVHWFramesContext*)ctx->hw_frames_ctx->data;
frames_hwctx = frames_ctx->hwctx;
frames_ctx->format = AV_PIX_FMT_DXVA2_VLD;
frames_ctx->sw_format = s->sw_pix_fmt == AV_PIX_FMT_YUV420P10 ?
AV_PIX_FMT_P010 : AV_PIX_FMT_NV12;
frames_ctx->width = FFALIGN(s->coded_width, surface_alignment);
frames_ctx->height = FFALIGN(s->coded_height, surface_alignment);
frames_ctx->initial_pool_size = num_surfaces;
frames_hwctx->surface_type = DXVA2_VideoDecoderRenderTarget;
ret = av_hwframe_ctx_init(ctx->hw_frames_ctx);
if (ret < 0) {
av_log(NULL, loglevel, "Failed to initialize the HW frames context\n");
goto fail;
}
hr = IDirectXVideoDecoderService_CreateVideoDecoder(ctx->decoder_service, &device_guid,
&desc, &config, frames_hwctx->surfaces,
frames_hwctx->nb_surfaces, &frames_hwctx->decoder_to_release);
if (FAILED(hr)) {
av_log(NULL, loglevel, "Failed to create DXVA2 video decoder\n");
goto fail;
}
ctx->decoder_guid = device_guid;
ctx->decoder_config = config;
dxva_ctx->cfg = &ctx->decoder_config;
dxva_ctx->decoder = frames_hwctx->decoder_to_release;
dxva_ctx->surface = frames_hwctx->surfaces;
dxva_ctx->surface_count = frames_hwctx->nb_surfaces;
if (IsEqualGUID(&ctx->decoder_guid, &DXVADDI_Intel_ModeH264_E))
dxva_ctx->workaround |= FF_DXVA2_WORKAROUND_INTEL_CLEARVIDEO;
return 0;
fail:
av_buffer_unref(&ctx->hw_frames_ctx);
return AVERROR(EINVAL);
}
