https://bugs.gentoo.org/936310 https://github.com/getsolus/packages/commit/50b029f7801c2c73da655e7dd5ae826244f9ae2c https://gitlab.freedesktop.org/mesa/mesa/-/issues/11533#note_2495206 The backport here is from Solus who resolved a conflict for 4.4.x as it doesn't apply cleanly there. From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001 From: David Rosca Date: Wed, 8 May 2024 09:11:11 +0200 Subject: [PATCH 1/2] lavc/vaapi_decode: Make it possible to send multiple slice params buffers Reviewed-by: Neal Gompa Signed-off-by: David Rosca Signed-off-by: Haihao Xiang --- a/libavcodec/vaapi_av1.c +++ b/libavcodec/vaapi_av1.c @@ -419,7 +419,7 @@ static int vaapi_av1_decode_slice(AVCodecContext *avctx, .tg_end = s->tg_end, }; - err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &slice_param, + err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &slice_param, 1, sizeof(VASliceParameterBufferAV1), buffer, size); --- a/libavcodec/vaapi_decode.c +++ b/libavcodec/vaapi_decode.c @@ -59,6 +59,7 @@ int ff_vaapi_decode_make_param_buffer(AVCodecContext *avctx, int ff_vaapi_decode_make_slice_buffer(AVCodecContext *avctx, VAAPIDecodePicture *pic, const void *params_data, + int nb_params, size_t params_size, const void *slice_data, size_t slice_size) @@ -87,7 +88,7 @@ int ff_vaapi_decode_make_slice_buffer(AVCodecContext *avctx, vas = vaCreateBuffer(ctx->hwctx->display, ctx->va_context, VASliceParameterBufferType, - params_size, 1, (void*)params_data, + params_size, nb_params, (void*)params_data, &pic->slice_buffers[index]); if (vas != VA_STATUS_SUCCESS) { av_log(avctx, AV_LOG_ERROR, "Failed to create slice " --- a/libavcodec/vaapi_decode.h +++ b/libavcodec/vaapi_decode.h @@ -86,6 +86,7 @@ int ff_vaapi_decode_make_param_buffer(AVCodecContext *avctx, int ff_vaapi_decode_make_slice_buffer(AVCodecContext *avctx, VAAPIDecodePicture *pic, const void *params_data, + int nb_params, size_t params_size, const void *slice_data, size_t slice_size); --- a/libavcodec/vaapi_h264.c +++ b/libavcodec/vaapi_h264.c @@ -375,7 +375,7 @@ static int vaapi_h264_decode_slice(AVCodecContext *avctx, slice_param.chroma_offset_l1); err = ff_vaapi_decode_make_slice_buffer(avctx, pic, - &slice_param, sizeof(slice_param), + &slice_param, 1, sizeof(slice_param), buffer, size); if (err) { ff_vaapi_decode_cancel(avctx, pic); --- a/libavcodec/vaapi_hevc.c +++ b/libavcodec/vaapi_hevc.c @@ -305,7 +305,7 @@ static int vaapi_hevc_end_frame(AVCodecContext *avctx) if (pic->last_size) { last_slice_param->LongSliceFlags.fields.LastSliceOfPic = 1; ret = ff_vaapi_decode_make_slice_buffer(avctx, &pic->pic, - &pic->last_slice_param, slice_param_size, + &pic->last_slice_param, 1, slice_param_size, pic->last_buffer, pic->last_size); if (ret < 0) goto fail; @@ -410,7 +410,7 @@ static int vaapi_hevc_decode_slice(AVCodecContext *avctx, if (!sh->first_slice_in_pic_flag) { err = ff_vaapi_decode_make_slice_buffer(avctx, &pic->pic, - &pic->last_slice_param, slice_param_size, + &pic->last_slice_param, 1, slice_param_size, pic->last_buffer, pic->last_size); pic->last_buffer = NULL; pic->last_size = 0; --- a/libavcodec/vaapi_mjpeg.c +++ b/libavcodec/vaapi_mjpeg.c @@ -131,7 +131,7 @@ static int vaapi_mjpeg_decode_slice(AVCodecContext *avctx, sp.components[i].ac_table_selector = s->ac_index[i]; } - err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &sp, sizeof(sp), buffer, size); + err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &sp, 1, sizeof(sp), buffer, size); if (err) goto fail; --- a/libavcodec/vaapi_mpeg2.c +++ b/libavcodec/vaapi_mpeg2.c @@ -162,7 +162,7 @@ static int vaapi_mpeg2_decode_slice(AVCodecContext *avctx, const uint8_t *buffer }; err = ff_vaapi_decode_make_slice_buffer(avctx, pic, - &slice_param, sizeof(slice_param), + &slice_param, 1, sizeof(slice_param), buffer, size); if (err < 0) { ff_vaapi_decode_cancel(avctx, pic); --- a/libavcodec/vaapi_mpeg4.c +++ b/libavcodec/vaapi_mpeg4.c @@ -167,7 +167,7 @@ static int vaapi_mpeg4_decode_slice(AVCodecContext *avctx, const uint8_t *buffer }; err = ff_vaapi_decode_make_slice_buffer(avctx, pic, - &slice_param, sizeof(slice_param), + &slice_param, 1, sizeof(slice_param), buffer, size); if (err < 0) { ff_vaapi_decode_cancel(avctx, pic); --- a/libavcodec/vaapi_vc1.c +++ b/libavcodec/vaapi_vc1.c @@ -490,7 +490,7 @@ static int vaapi_vc1_decode_slice(AVCodecContext *avctx, const uint8_t *buffer, }; err = ff_vaapi_decode_make_slice_buffer(avctx, pic, - &slice_param, sizeof(slice_param), + &slice_param, 1, sizeof(slice_param), buffer, size); if (err < 0) { ff_vaapi_decode_cancel(avctx, pic); --- a/libavcodec/vaapi_vp8.c +++ b/libavcodec/vaapi_vp8.c @@ -209,7 +209,7 @@ static int vaapi_vp8_decode_slice(AVCodecContext *avctx, for (i = 0; i < 8; i++) sp.partition_size[i+1] = s->coeff_partition_size[i]; - err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &sp, sizeof(sp), data, data_size); + err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &sp, 1, sizeof(sp), data, data_size); if (err) goto fail; --- a/libavcodec/vaapi_vp9.c +++ b/libavcodec/vaapi_vp9.c @@ -158,7 +158,7 @@ static int vaapi_vp9_decode_slice(AVCodecContext *avctx, } err = ff_vaapi_decode_make_slice_buffer(avctx, pic, - &slice_param, sizeof(slice_param), + &slice_param, 1, sizeof(slice_param), buffer, size); if (err) { ff_vaapi_decode_cancel(avctx, pic); From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001 From: David Rosca Date: Wed, 8 May 2024 09:11:13 +0200 Subject: [PATCH 2/2] lavc/vaapi_av1: Avoid sending the same slice buffer multiple times When there are multiple tiles in one slice buffer, use multiple slice params to avoid sending the same slice buffer multiple times and thus increasing the bitstream size the driver will need to upload to hw. Reviewed-by: Neal Gompa Signed-off-by: David Rosca Signed-off-by: Haihao Xiang --- a/libavcodec/vaapi_av1.c +++ b/libavcodec/vaapi_av1.c @@ -19,6 +19,8 @@ */ #include "libavutil/pixdesc.h" +#include "libavutil/frame.h" +#include "libavutil/mem.h" #include "hwconfig.h" #include "vaapi_decode.h" #include "internal.h" @@ -41,6 +43,9 @@ typedef struct VAAPIAV1DecContext { */ VAAPIAV1FrameRef ref_tab[AV1_NUM_REF_FRAMES]; ThreadFrame tmp_frame; + + int nb_slice_params; + VASliceParameterBufferAV1 *slice_params; } VAAPIAV1DecContext; static VASurfaceID vaapi_av1_surface_id(AV1Frame *vf) @@ -107,6 +112,8 @@ static int vaapi_av1_decode_uninit(AVCodecContext *avctx) av_frame_free(&ctx->ref_tab[i].frame.f); } + av_freep(&ctx->slice_params); + return ff_vaapi_decode_uninit(avctx); } @@ -403,13 +410,24 @@ static int vaapi_av1_decode_slice(AVCodecContext *avctx, { const AV1DecContext *s = avctx->priv_data; VAAPIDecodePicture *pic = s->cur_frame.hwaccel_picture_private; - VASliceParameterBufferAV1 slice_param; - int err = 0; + VAAPIAV1DecContext *ctx = avctx->internal->hwaccel_priv_data; + int err, nb_params; + + nb_params = s->tg_end - s->tg_start + 1; + if (ctx->nb_slice_params < nb_params) { + ctx->slice_params = av_realloc_array(ctx->slice_params, + nb_params, + sizeof(*ctx->slice_params)); + if (!ctx->slice_params) { + ctx->nb_slice_params = 0; + err = AVERROR(ENOMEM); + goto fail; + } + ctx->nb_slice_params = nb_params; + } for (int i = s->tg_start; i <= s->tg_end; i++) { - memset(&slice_param, 0, sizeof(VASliceParameterBufferAV1)); - - slice_param = (VASliceParameterBufferAV1) { + ctx->slice_params[i - s->tg_start] = (VASliceParameterBufferAV1) { .slice_data_size = s->tile_group_info[i].tile_size, .slice_data_offset = s->tile_group_info[i].tile_offset, .slice_data_flag = VA_SLICE_DATA_FLAG_ALL, @@ -418,18 +436,20 @@ static int vaapi_av1_decode_slice(AVCodecContext *avctx, .tg_start = s->tg_start, .tg_end = s->tg_end, }; - - err = ff_vaapi_decode_make_slice_buffer(avctx, pic, &slice_param, 1, - sizeof(VASliceParameterBufferAV1), - buffer, - size); - if (err) { - ff_vaapi_decode_cancel(avctx, pic); - return err; - } } + err = ff_vaapi_decode_make_slice_buffer(avctx, pic, ctx->slice_params, nb_params, + sizeof(VASliceParameterBufferAV1), + buffer, + size); + if (err) + goto fail; + return 0; + +fail: + ff_vaapi_decode_cancel(avctx, pic); + return err; } const AVHWAccel ff_av1_vaapi_hwaccel = {