Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input during the encoding process
> -Original Message- > From: libav-devel [mailto:libav-devel-boun...@libav.org] On Behalf Of Luca > Barbato > Sent: Monday, April 16, 2018 5:33 AM > To: libav-devel@libav.org > Subject: Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D > frames used as input during the encoding process > > On 13/04/2018 00:41, Alexander Kravchenko wrote: > > This fixes frame corruption issue when decoder started reusing frames > > while they are still in use of encoding process > > > > Issue with frame corruption was reproduced using: > > avconv.exe -y -hwaccel d3d11va -hwaccel_output_format d3d11 -i > > input.h264 -an -c:v h264_amf output.mkv > > > > it is recommended to use -extra_hw_frames 16 option in case if hw > > frames number in pool is not enough > > > > Seems good enough, I do not have the hardware at hand to see if would be > possible to make it simpler than that. > > I'll merge in 8 hours or such. > > lu > Hi Luca, A soft reminder about merging the patch :) Thanks Alexander ___ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel
Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input during the encoding process
On 13/04/2018 00:41, Alexander Kravchenko wrote: > This fixes frame corruption issue when decoder started reusing frames while > they are still in use of encoding process > > Issue with frame corruption was reproduced using: > avconv.exe -y -hwaccel d3d11va -hwaccel_output_format d3d11 -i input.h264 -an > -c:v h264_amf output.mkv > > it is recommended to use -extra_hw_frames 16 option in case if hw frames > number in pool is not enough > Seems good enough, I do not have the hardware at hand to see if would be possible to make it simpler than that. I'll merge in 8 hours or such. lu ___ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel
Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input during the encoding process
Hi Guys, Did you have a chance to review the patch? Is it ok? Thanks, Alexander > -Original Message- > From: Alexander Kravchenko [mailto:akravchenko...@gmail.com] > Sent: Thursday, April 12, 2018 6:42 PM > To: libav-devel@libav.org > Cc: Alexander Kravchenko> Subject: [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input > during the encoding process > > This fixes frame corruption issue when decoder started reusing frames while > they are still in use of encoding process > > Issue with frame corruption was reproduced using: > avconv.exe -y -hwaccel d3d11va -hwaccel_output_format d3d11 -i input.h264 -an > -c:v h264_amf output.mkv > > it is recommended to use -extra_hw_frames 16 option in case if hw frames > number in pool is not enough > > > --- > libavcodec/amfenc.c | 95 > - > libavcodec/amfenc.h | 3 ++ > 2 files changed, 97 insertions(+), 1 deletion(-) > > diff --git a/libavcodec/amfenc.c b/libavcodec/amfenc.c index > 74b020b4d..9a60050bc 100644 > --- a/libavcodec/amfenc.c > +++ b/libavcodec/amfenc.c > @@ -162,6 +162,9 @@ static int amf_init_context(AVCodecContext *avctx) > AmfContext *ctx = avctx->priv_data; > AMF_RESULT res = AMF_OK; > > +ctx->hwsurfaces_in_queue = 0; > +ctx->hwsurfaces_in_queue_max = 16; > + > // configure AMF logger > // the return of these functions indicates old state and do not affect > behaviour > ctx->trace->pVtbl->EnableWriter(ctx->trace, > AMF_TRACE_WRITER_DEBUG_OUTPUT, ctx->log_to_dbg != 0 ); @@ -192,6 +195,8 > @@ static int amf_init_context(AVCodecContext *avctx) > if (!ctx->hw_frames_ctx) { > return AVERROR(ENOMEM); > } > +if (device_ctx->initial_pool_size > 0) > +ctx->hwsurfaces_in_queue_max = > + device_ctx->initial_pool_size - 1; > } else { > if(res == AMF_NOT_SUPPORTED) > av_log(avctx, AV_LOG_INFO, "avctx->hw_frames_ctx > has D3D11 device which doesn't have D3D11VA interface, > switching to default\n"); @@ -447,6 +452,75 @@ int > ff_amf_encode_init(AVCodecContext *avctx) > return ret; > } > > +static AMF_RESULT amf_set_property_buffer(AMFSurface *object, const > +wchar_t *name, AMFBuffer *val) { > +AMF_RESULT res; > +AMFVariantStruct var; > +res = AMFVariantInit(); > +if (res == AMF_OK) { > +AMFGuid guid_AMFInterface = IID_AMFInterface(); > +AMFInterface *amf_interface; > +res = val->pVtbl->QueryInterface(val, _AMFInterface, > +(void**)_interface); > + > +if (res == AMF_OK) { > +res = AMFVariantAssignInterface(, amf_interface); > +amf_interface->pVtbl->Release(amf_interface); > +} > +if (res == AMF_OK) { > +res = object->pVtbl->SetProperty(object, name, var); > +} > +AMFVariantClear(); > +} > +return res; > +} > + > +static AMF_RESULT amf_get_property_buffer(AMFData *object, const > +wchar_t *name, AMFBuffer **val) { > +AMF_RESULT res; > +AMFVariantStruct var; > +res = AMFVariantInit(); > +if (res == AMF_OK) { > +res = object->pVtbl->GetProperty(object, name, ); > +if (res == AMF_OK) { > +if (var.type == AMF_VARIANT_INTERFACE) { > +AMFGuid guid_AMFBuffer = IID_AMFBuffer(); > +AMFInterface *amf_interface = AMFVariantInterface(); > +res = amf_interface->pVtbl->QueryInterface(amf_interface, > _AMFBuffer, (void**)val); > +} else { > +res = AMF_INVALID_DATA_TYPE; > +} > +} > +AMFVariantClear(); > +} > +return res; > +} > + > +static AMFBuffer *amf_create_buffer_with_frame_ref(const AVFrame > +*frame, AMFContext *context) { > +AVFrame *frame_ref; > +AMFBuffer *frame_ref_storage_buffer = NULL; > +AMF_RESULT res; > + > +res = context->pVtbl->AllocBuffer(context, AMF_MEMORY_HOST, > sizeof(frame_ref), _ref_storage_buffer); > +if (res == AMF_OK) { > +frame_ref = av_frame_clone(frame); > +if (frame_ref) { > + > memcpy(frame_ref_storage_buffer->pVtbl->GetNative(frame_ref_storage_buffer), > _ref, sizeof(frame_ref)); > +} else { > + > frame_ref_storage_buffer->pVtbl->Release(frame_ref_storage_buffer); > +frame_ref_storage_buffer = NULL; > +} > +} > +return frame_ref_storage_buffer; > +} > + > +static void amf_release_buffer_with_frame_ref(AMFBuffer > +*frame_ref_storage_buffer) { > +AVFrame *av_frame_ref; > +memcpy(_frame_ref, > frame_ref_storage_buffer->pVtbl->GetNative(frame_ref_storage_buffer), > sizeof(av_frame_ref)); > +av_frame_free(_frame_ref); > +
Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input during the encoding process
Hi guys, thanks for your advice. I have sent one more time using git send-email чт, 12 апр. 2018 г. в 17:56, Diego Biurrun: > On Wed, Apr 11, 2018 at 07:31:19PM +0300, Alexander Kravchenko wrote: > > > > > > Your mailer mangled the patch, could you please resend it attached? > > > > It would be nice if you recommend some good simple mailer or to do > > I use git-send-email, works fine here. > > Diego > ___ > libav-devel mailing list > libav-devel@libav.org > https://lists.libav.org/mailman/listinfo/libav-devel ___ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel
Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input during the encoding process
On Wed, Apr 11, 2018 at 07:31:19PM +0300, Alexander Kravchenko wrote: > > > > Your mailer mangled the patch, could you please resend it attached? > > It would be nice if you recommend some good simple mailer or to do I use git-send-email, works fine here. Diego ___ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel
Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input during the encoding process
On 12/04/2018 01:31, Alexander Kravchenko wrote: >> >> Your mailer mangled the patch, could you please resend it attached? >> >> Diego > > Hi Diego, > Sending plain text mail format (probably I switched from Rich text to plain > after paste patch test) > Hopefully it is ok. > Last time I tried to send attached mail in outlook. It was rejected as binary. > It would be nice if you recommend some good simple mailer or to do > http://trojita.flaska.net/download.html with git-format-patch might work fine. Otherwise you could configure git-send-email to use directly gmail as the end of the man page suggests but it can be relatively annoying IMHO. lu ___ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel
Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input during the encoding process
> > Your mailer mangled the patch, could you please resend it attached? > > Diego Hi Diego, Sending plain text mail format (probably I switched from Rich text to plain after paste patch test) Hopefully it is ok. Last time I tried to send attached mail in outlook. It was rejected as binary. It would be nice if you recommend some good simple mailer or to do Thanks, Alexander --- libavcodec/amfenc.c | 95 - libavcodec/amfenc.h | 3 ++ 2 files changed, 97 insertions(+), 1 deletion(-) diff --git a/libavcodec/amfenc.c b/libavcodec/amfenc.c index 74b020b4d..9a60050bc 100644 --- a/libavcodec/amfenc.c +++ b/libavcodec/amfenc.c @@ -162,6 +162,9 @@ static int amf_init_context(AVCodecContext *avctx) AmfContext *ctx = avctx->priv_data; AMF_RESULT res = AMF_OK; +ctx->hwsurfaces_in_queue = 0; +ctx->hwsurfaces_in_queue_max = 16; + // configure AMF logger // the return of these functions indicates old state and do not affect behaviour ctx->trace->pVtbl->EnableWriter(ctx->trace, AMF_TRACE_WRITER_DEBUG_OUTPUT, ctx->log_to_dbg != 0 ); @@ -192,6 +195,8 @@ static int amf_init_context(AVCodecContext *avctx) if (!ctx->hw_frames_ctx) { return AVERROR(ENOMEM); } +if (device_ctx->initial_pool_size > 0) +ctx->hwsurfaces_in_queue_max = device_ctx->initial_pool_size - 1; } else { if(res == AMF_NOT_SUPPORTED) av_log(avctx, AV_LOG_INFO, "avctx->hw_frames_ctx has D3D11 device which doesn't have D3D11VA interface, switching to default\n"); @@ -447,6 +452,75 @@ int ff_amf_encode_init(AVCodecContext *avctx) return ret; } +static AMF_RESULT amf_set_property_buffer(AMFSurface *object, const wchar_t *name, AMFBuffer *val) +{ +AMF_RESULT res; +AMFVariantStruct var; +res = AMFVariantInit(); +if (res == AMF_OK) { +AMFGuid guid_AMFInterface = IID_AMFInterface(); +AMFInterface *amf_interface; +res = val->pVtbl->QueryInterface(val, _AMFInterface, (void**)_interface); + +if (res == AMF_OK) { +res = AMFVariantAssignInterface(, amf_interface); +amf_interface->pVtbl->Release(amf_interface); +} +if (res == AMF_OK) { +res = object->pVtbl->SetProperty(object, name, var); +} +AMFVariantClear(); +} +return res; +} + +static AMF_RESULT amf_get_property_buffer(AMFData *object, const wchar_t *name, AMFBuffer **val) +{ +AMF_RESULT res; +AMFVariantStruct var; +res = AMFVariantInit(); +if (res == AMF_OK) { +res = object->pVtbl->GetProperty(object, name, ); +if (res == AMF_OK) { +if (var.type == AMF_VARIANT_INTERFACE) { +AMFGuid guid_AMFBuffer = IID_AMFBuffer(); +AMFInterface *amf_interface = AMFVariantInterface(); +res = amf_interface->pVtbl->QueryInterface(amf_interface, _AMFBuffer, (void**)val); +} else { +res = AMF_INVALID_DATA_TYPE; +} +} +AMFVariantClear(); +} +return res; +} + +static AMFBuffer *amf_create_buffer_with_frame_ref(const AVFrame *frame, AMFContext *context) +{ +AVFrame *frame_ref; +AMFBuffer *frame_ref_storage_buffer = NULL; +AMF_RESULT res; + +res = context->pVtbl->AllocBuffer(context, AMF_MEMORY_HOST, sizeof(frame_ref), _ref_storage_buffer); +if (res == AMF_OK) { +frame_ref = av_frame_clone(frame); +if (frame_ref) { + memcpy(frame_ref_storage_buffer->pVtbl->GetNative(frame_ref_storage_buffer), _ref, sizeof(frame_ref)); +} else { +frame_ref_storage_buffer->pVtbl->Release(frame_ref_storage_buffer); +frame_ref_storage_buffer = NULL; +} +} +return frame_ref_storage_buffer; +} + +static void amf_release_buffer_with_frame_ref(AMFBuffer *frame_ref_storage_buffer) +{ +AVFrame *av_frame_ref; +memcpy(_frame_ref, frame_ref_storage_buffer->pVtbl->GetNative(frame_ref_storage_buffer), sizeof(av_frame_ref)); +av_frame_free(_frame_ref); +frame_ref_storage_buffer->pVtbl->Release(frame_ref_storage_buffer); +} int ff_amf_send_frame(AVCodecContext *avctx, const AVFrame *frame) { @@ -488,6 +562,8 @@ int ff_amf_send_frame(AVCodecContext *avctx, const AVFrame *frame) (ctx->hw_device_ctx && ((AVHWFramesContext*)frame->hw_frames_ctx->data)->device_ctx == (AVHWDeviceContext*)ctx->hw_device_ctx->data) )) { +AMFBuffer *frame_ref_storage_buffer; + #if CONFIG_D3D11VA static const GUID AMFTextureArrayIndexGUID = { 0x28115527, 0xe7c3, 0x4b66, { 0x99, 0xd3, 0x4f, 0x2a, 0xe6, 0xb4, 0x7f, 0xaf } }; ID3D11Texture2D *texture = (ID3D11Texture2D*)frame->data[0]; // actual texture
Re: [libav-devel] [PATCH] lavc/amfenc: Retain a reference to D3D frames used as input during the encoding process
On Wed, Apr 11, 2018 at 06:02:32PM +0300, Alexander Kravchenko wrote: > This fixes frame corruption issue when decoder started reusing frames while > they are still in use of encoding process > > Issue with frame corruption was reproduced using: > avconv.exe -y -hwaccel d3d11va -hwaccel_output_format d3d11 -i input.h264 > -an -c:v h264_amf output.mkv > > it is recommended to use -extra_hw_frames 16 option in case if hw frames > number in pool is not enough > > --- > libavcodec/amfenc.c | 95 > - > libavcodec/amfenc.h | 3 ++ > 2 files changed, 97 insertions(+), 1 deletion(-) Your mailer mangled the patch, could you please resend it attached? Diego ___ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel