jellyfin/jellyfin-ffmpeg
1
0
Fork 0
mirror of https://github.com/jellyfin/jellyfin-ffmpeg.git synced 2026-09-02 21:03:08 +03:00
Code Issues Packages Projects Releases Wiki Activity

Update patches for 8.1.1

Browse Source 
Signed-off-by: nyanmisaka <nst799610810@gmail.com>
This commit is contained in:
nyanmisaka
2026-05-28 10:53:47 +08:00
parent 959574d802
commit 6f10ceb986
107 changed files with 6058 additions and 6563 deletions
Download Patch File Download Diff File Expand all files Collapse all files
debian/patches/0002-update-cuda-func-header.patch
Vendored
+3 -2
View File
@@ -165,7 +165,7 @@ Index: FFmpeg/compat/cuda/cuda_runtime.h
{
float4 ret;
asm("tex.2d.v4.f32.f32 {%0, %1, %2, %3}, [%4, {%5, %6}];" :
@@ -156,37 +147,42 @@ inline __device__ float4 tex2D<float4>(c
@@ -156,38 +147,43 @@ inline __device__ float4 tex2D<float4>(c
}
template<>
@@ -228,4 +228,5 @@ Index: FFmpeg/compat/cuda/cuda_runtime.h
+static __inline__ __device__ float __log10f(float a) { return __nvvm_lg2_approx_f(a) * 0.30103f; }
+static __inline__ __device__ float __powf(float a, float b) { return __nvvm_ex2_approx_f(__nvvm_lg2_approx_f(a) * b); }
#endif /* COMPAT_CUDA_CUDA_RUNTIME_H */
// Misc helper functions
extern "C" __device__ int printf(const char*, ...);
debian/patches/0003-add-enhanced-cuda-pixfmt-converter-impl.patch
Vendored
+120 -112
View File
@@ -327,7 +327,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.c
#include "filters.h"
#include "scale_eval.h"
#include "video.h"
@@ -108,6 +110,9 @@ typedef struct CUDAScaleContext {
@@ -121,6 +123,9 @@ typedef struct CUDAScaleContext {
int interp_as_integer;
float param;
@@ -337,7 +337,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.c
} CUDAScaleContext;
static av_cold int cudascale_init(AVFilterContext *ctx)
@@ -129,13 +134,23 @@ static av_cold void cudascale_uninit(AVF
@@ -142,13 +147,23 @@ static av_cold void cudascale_uninit(AVF
{
CUDAScaleContext *s = ctx->priv;
@@ -364,7 +364,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.c
CHECK_CU(cu->cuCtxPopCurrent(&dummy));
}
@@ -275,6 +290,69 @@ static av_cold int init_processing_chain
@@ -294,6 +309,67 @@ static av_cold int init_processing_chain
return 0;
}
@@ -387,30 +387,28 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.c
+ .srcPitch = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]),
+ .dstPitch = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]),
+ .WidthInBytes = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]),
+ .Height = ff_fruit_dither_size,
+ .Height = ff_fruit_dither_size
+ };
+
+#ifndef CU_TRSF_NORMALIZED_COORDINATES
+ #define CU_TRSF_NORMALIZED_COORDINATES 2
+#endif
+ CUDA_TEXTURE_DESC tex_desc = {
+ .addressMode = { CU_TR_ADDRESS_MODE_WRAP,
+ CU_TR_ADDRESS_MODE_WRAP },
+ .filterMode = CU_TR_FILTER_MODE_POINT,
+ .flags = CU_TRSF_NORMALIZED_COORDINATES,
+ .filterMode = CU_TR_FILTER_MODE_POINT,
+ .flags = 2 /* CU_TRSF_NORMALIZED_COORDINATES */
+ };
+
+ CUDA_RESOURCE_DESC res_desc = {
+ .resType = CU_RESOURCE_TYPE_PITCH2D,
+ .res.pitch2D.format = CU_AD_FORMAT_UNSIGNED_INT16,
+ .res.pitch2D.numChannels = 1,
+ .res.pitch2D.width = ff_fruit_dither_size,
+ .res.pitch2D.height = ff_fruit_dither_size,
+ .resType = CU_RESOURCE_TYPE_PITCH2D,
+ .res.pitch2D.format = CU_AD_FORMAT_UNSIGNED_INT16,
+ .res.pitch2D.numChannels = 1,
+ .res.pitch2D.width = ff_fruit_dither_size,
+ .res.pitch2D.height = ff_fruit_dither_size,
+ .res.pitch2D.pitchInBytes = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]),
+ .res.pitch2D.devPtr = 0,
+ .res.pitch2D.devPtr = 0
+ };
+
+ av_assert0(sizeof(ff_fruit_dither_matrix) == sizeof(ff_fruit_dither_matrix[0]) * ff_fruit_dither_size * ff_fruit_dither_size);
+ av_assert0(sizeof(ff_fruit_dither_matrix) ==
+ sizeof(ff_fruit_dither_matrix[0]) * ff_fruit_dither_size * ff_fruit_dither_size);
+
+ if ((ret = CHECK_CU(cu->cuCtxPushCurrent(cuda_ctx))) < 0)
+ return ret;
@@ -434,7 +432,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.c
static av_cold int cudascale_load_functions(AVFilterContext *ctx)
{
CUDAScaleContext *s = ctx->priv;
@@ -383,6 +461,11 @@ static av_cold int cudascale_config_prop
@@ -409,6 +485,11 @@ static av_cold int cudascale_config_prop
s->hwctx = device_hwctx;
s->cu_stream = s->hwctx->stream;
@@ -443,27 +441,22 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.c
+ goto fail;
+ }
+
if (inlink->sample_aspect_ratio.num) {
outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h*inlink->w,
outlink->w*inlink->h},
@@ -418,11 +501,15 @@ static int call_resize_kernel(AVFilterCo
(CUdeviceptr)out_frame->data[2], (CUdeviceptr)out_frame->data[3]
if (s->reset_sar)
outlink->sample_aspect_ratio = (AVRational){1, 1};
else if (inlink->sample_aspect_ratio.num) {
@@ -457,7 +538,10 @@ static int call_resize_kernel(AVFilterCo
.src_width = src_width,
.src_height = src_height,
.param = s->param,
- .mpeg_range = mpeg_range
+ .mpeg_range = mpeg_range,
+ .dither_tex = s->dither_tex,
+ .dither_size = (float)ff_fruit_dither_size,
+ .dither_quantization = (float)((1 << s->out_desc->comp[0].depth) - 1)
};
+ float dither_size = (float)ff_fruit_dither_size;
+ float dither_quantization = (float)((1 << s->out_desc->comp[0].depth) - 1);
+
void *args_uchar[] = {
&src_tex[0], &src_tex[1], &src_tex[2], &src_tex[3],
&dst_devptr[0], &dst_devptr[1], &dst_devptr[2], &dst_devptr[3],
&dst_width, &dst_height, &dst_pitch,
- &src_width, &src_height, &s->param
+ &src_width, &src_height, &s->param,
+ &s->dither_tex, &dither_size, &dither_quantization
};
return CHECK_CU(cu->cuLaunchKernel(func,
@@ -446,6 +533,7 @@ static int scalecuda_resize(AVFilterCont
void *args[] = { &params };
@@ -487,6 +571,7 @@ static int scalecuda_resize(AVFilterCont
for (i = 0; i < s->in_planes; i++) {
CUDA_TEXTURE_DESC tex_desc = {
@@ -475,7 +468,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
===================================================================
--- FFmpeg.orig/libavfilter/vf_scale_cuda.cu
+++ FFmpeg/libavfilter/vf_scale_cuda.cu
@@ -29,6 +29,19 @@ using subsample_function_t = T (*)(cudaT
@@ -30,6 +30,19 @@ using subsample_function_t = T (*)(cudaT
int src_width, int src_height,
int bit_depth, float param);
@@ -495,54 +488,42 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
// --- CONVERSION LOGIC ---
static const ushort mask_10bit = 0xFFC0;
@@ -64,7 +77,9 @@ static inline __device__ ushort conv_16t
subsample_function_t<in_T_uv> subsample_func_uv> \
@@ -90,7 +103,9 @@ static inline __device__ ushort conv_16t
__device__ static inline void N(cudaTextureObject_t src_tex[4], T *dst[4], int xo, int yo, \
int dst_width, int dst_height, int dst_pitch, \
- int src_width, int src_height, float param)
+ int src_width, int src_height, float param, \
int src_left, int src_top, int src_width, int src_height, \
- float param, int mpeg_range)
+ float param, int mpeg_range, \
+ cudaTextureObject_t dither_tex, \
+ float dither_size, float dither_quantization)
#define SUB_F(m, plane) \
subsample_func_##m(src_tex[plane], xo, yo, \
@@ -477,7 +492,10 @@ struct Convert_p010le_yuv420p
@@ -248,7 +263,10 @@ struct Convert_planar10_planar8
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_10to8(SUB_F(y, 0));
- DEFAULT_DST(0) = conv_10to8pl(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_10bit);
+ DEFAULT_DST(0) = conv_10to8(res);
+ dither_size, dither_quantization, (float)0x3FF);
+ DEFAULT_DST(0) = conv_10to8pl(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -498,7 +516,10 @@ struct Convert_p010le_nv12
@@ -308,7 +326,10 @@ struct Convert_planar10_semiplanar8
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_10to8(SUB_F(y, 0));
- DEFAULT_DST(0) = conv_10to8pl(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_10bit);
+ DEFAULT_DST(0) = conv_10to8(res);
+ dither_size, dither_quantization, (float)0x3FF);
+ DEFAULT_DST(0) = conv_10to8pl(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -521,7 +542,10 @@ struct Convert_p010le_yuv444p
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_10to8(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_10bit);
+ DEFAULT_DST(0) = conv_10to8(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -607,7 +631,10 @@ struct Convert_p016le_yuv420p
@@ -376,7 +397,10 @@ struct Convert_planar16_planar8
DEF_F(Convert, out_T)
{
@@ -554,7 +535,19 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
}
DEF_F(Convert_uv, out_T_uv)
@@ -628,7 +655,10 @@ struct Convert_p016le_nv12
@@ -396,7 +420,10 @@ struct Convert_planar16_planar10
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_16to10pl(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_16bit);
+ DEFAULT_DST(0) = conv_16to10pl(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -436,7 +463,10 @@ struct Convert_planar16_semiplanar8
DEF_F(Convert, out_T)
{
@@ -566,19 +559,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
}
DEF_F(Convert_uv, out_T_uv)
@@ -651,7 +681,10 @@ struct Convert_p016le_yuv444p
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_16to8(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_16bit);
+ DEFAULT_DST(0) = conv_16to8(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -672,7 +705,10 @@ struct Convert_p016le_p010le
@@ -458,7 +488,10 @@ struct Convert_planar16_semiplanar10
DEF_F(Convert, out_T)
{
@@ -590,7 +571,31 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
}
DEF_F(Convert_uv, out_T_uv)
@@ -737,7 +773,10 @@ struct Convert_yuv444p16le_yuv420p
@@ -634,7 +667,10 @@ struct Convert_semiplanar10_planar8
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_10to8(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_10bit);
+ DEFAULT_DST(0) = conv_10to8(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -697,7 +733,10 @@ struct Convert_semiplanar10_semiplanar8
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_10to8(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_10bit);
+ DEFAULT_DST(0) = conv_10to8(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -765,7 +804,10 @@ struct Convert_semiplanar16_planar8
DEF_F(Convert, out_T)
{
@@ -602,7 +607,19 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
}
DEF_F(Convert_uv, out_T_uv)
@@ -757,7 +796,10 @@ struct Convert_yuv444p16le_nv12
@@ -786,7 +828,10 @@ struct Convert_semiplanar16_planar10
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_16to10pl(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_16bit);
+ DEFAULT_DST(0) = conv_16to10pl(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -828,7 +873,10 @@ struct Convert_semiplanar16_semiplanar8
DEF_F(Convert, out_T)
{
@@ -614,19 +631,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
}
DEF_F(Convert_uv, out_T_uv)
@@ -779,7 +821,10 @@ struct Convert_yuv444p16le_yuv444p
DEF_F(Convert, out_T)
{
- DEFAULT_DST(0) = conv_16to8(SUB_F(y, 0));
+ in_T res = SUB_F(y, 0);
+ res = (in_T)get_dithered_y((float)res, read_dither(dither_tex, dither_size, xo, yo),
+ dither_size, dither_quantization, (float)mask_16bit);
+ DEFAULT_DST(0) = conv_16to8(res);
}
DEF_F(Convert_uv, out_T_uv)
@@ -799,7 +844,10 @@ struct Convert_yuv444p16le_p010le
@@ -851,7 +899,10 @@ struct Convert_semiplanar16_semiplanar10
DEF_F(Convert, out_T)
{
@@ -638,7 +643,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
}
DEF_F(Convert_uv, out_T_uv)
@@ -1015,6 +1063,9 @@ typedef float4 (*coeffs_function_t)(floa
@@ -1045,6 +1096,9 @@ typedef float4 (*coeffs_function_t)(floa
__device__ static inline float4 lanczos_coeffs(float x, float param)
{
@@ -648,7 +653,7 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
const float pi = 3.141592654f;
float4 res = make_float4(
@@ -1129,12 +1180,15 @@ __device__ static inline T Subsample_Bic
@@ -1162,12 +1216,15 @@ __device__ static inline T Subsample_Bic
#define PIX(x, y) tex2D<floatT>(tex, (x), (y))
return from_floatN<T, floatT>(
@@ -670,25 +675,28 @@ Index: FFmpeg/libavfilter/vf_scale_cuda.cu
);
#undef PIX
@@ -1147,7 +1201,9 @@ __device__ static inline T Subsample_Bic
cudaTextureObject_t src_tex_2, cudaTextureObject_t src_tex_3, \
T *dst_0, T *dst_1, T *dst_2, T *dst_3, \
int dst_width, int dst_height, int dst_pitch, \
- int src_width, int src_height, float param
+ int src_width, int src_height, float param, \
+ cudaTextureObject_t dither_tex, \
+ float dither_size, float dither_quantization
#define SUBSAMPLE(Convert, T) \
cudaTextureObject_t src_tex[4] = \
@@ -1159,7 +1215,9 @@ __device__ static inline T Subsample_Bic
Convert( \
src_tex, dst, xo, yo, \
dst_width, dst_height, dst_pitch, \
- src_width, src_height, param);
+ src_width, src_height, param, \
+ dither_tex, \
+ dither_size, dither_quantization);
@@ -1194,7 +1251,8 @@ __device__ static inline T Subsample_Bic
params.dst_width, params.dst_height, params.dst_pitch, \
params.src_left, params.src_top, \
params.src_width, params.src_height, \
- params.param, params.mpeg_range);
+ params.param, params.mpeg_range, \
+ params.dither_tex, params.dither_size, params.dither_quantization);
extern "C" {
Index: FFmpeg/libavfilter/vf_scale_cuda.h
===================================================================
--- FFmpeg.orig/libavfilter/vf_scale_cuda.h
+++ FFmpeg/libavfilter/vf_scale_cuda.h
@@ -45,6 +45,10 @@ typedef struct {
int src_height;
float param;
int mpeg_range;
+
+ CUtexObject dither_tex;
+ float dither_size;
+ float dither_quantization;
} CUDAScaleKernelParams;
#endif
debian/patches/0004-add-cuda-tonemap-impl.patch
Vendored
+160 -151
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3316,6 +3316,8 @@ scale_cuda_filter_deps="ffnvcodec"
@@ -3522,6 +3522,8 @@ scale_cuda_filter_deps="ffnvcodec"
scale_cuda_filter_deps_any="cuda_nvcc cuda_llvm"
thumbnail_cuda_filter_deps="ffnvcodec"
thumbnail_cuda_filter_deps_any="cuda_nvcc cuda_llvm"
@@ -11,9 +11,9 @@ Index: FFmpeg/configure
transpose_npp_filter_deps="ffnvcodec libnpp"
overlay_cuda_filter_deps="ffnvcodec"
overlay_cuda_filter_deps_any="cuda_nvcc cuda_llvm"
@@ -4707,6 +4709,15 @@ if enabled cuda_nvcc; then
@@ -4987,6 +4989,15 @@ if enabled cuda_nvcc; then
if $nvcc $nvccflags_default 2>&1 | grep -qi unsupported; then
nvccflags_default="-gencode arch=compute_60,code=sm_60 -O2"
nvccflags_default="-gencode arch=compute_75,code=sm_75 -O2"
fi
+else
+ clang_version=$($nvcc -dumpversion 2>/dev/null || echo 0)
@@ -27,7 +27,7 @@ Index: FFmpeg/configure
fi
set_default arch cc cxx doxygen pkg_config ranlib strip sysinclude \
@@ -6774,7 +6785,7 @@ fi
@@ -7117,7 +7128,7 @@ fi
if enabled cuda_nvcc; then
nvccflags="$nvccflags -ptx"
else
@@ -40,9 +40,9 @@ Index: FFmpeg/ffbuild/common.mak
===================================================================
--- FFmpeg.orig/ffbuild/common.mak
+++ FFmpeg/ffbuild/common.mak
@@ -44,6 +44,7 @@ ASFLAGS := $(CPPFLAGS) $(ASFLAGS)
@@ -53,6 +53,7 @@ ASFLAGS := $(CPPFLAGS) $(ASFLAGS)
# end up in CXXFLAGS.
$(call PREPEND,CXXFLAGS, CPPFLAGS CFLAGS)
$(call PREPEND,CXXFLAGS, CPPFLAGS)
X86ASMFLAGS += $(IFLAGS:%=%/) -I$(<D)/ -Pconfig.asm
+NVCCFLAGS += $(IFLAGS)
@@ -52,7 +52,7 @@ Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -524,6 +524,8 @@ OBJS-$(CONFIG_TMEDIAN_FILTER)
@@ -541,6 +541,8 @@ OBJS-$(CONFIG_TMEDIAN_FILTER)
OBJS-$(CONFIG_TMIDEQUALIZER_FILTER) += vf_tmidequalizer.o
OBJS-$(CONFIG_TMIX_FILTER) += vf_mix.o framesync.o
OBJS-$(CONFIG_TONEMAP_FILTER) += vf_tonemap.o
@@ -65,14 +65,14 @@ Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -494,6 +494,7 @@ extern const AVFilter ff_vf_tmedian;
extern const AVFilter ff_vf_tmidequalizer;
extern const AVFilter ff_vf_tmix;
extern const AVFilter ff_vf_tonemap;
+extern const AVFilter ff_vf_tonemap_cuda;
extern const AVFilter ff_vf_tonemap_opencl;
extern const AVFilter ff_vf_tonemap_vaapi;
extern const AVFilter ff_vf_tpad;
@@ -509,6 +509,7 @@ extern const FFFilter ff_vf_tmedian;
extern const FFFilter ff_vf_tmidequalizer;
extern const FFFilter ff_vf_tmix;
extern const FFFilter ff_vf_tonemap;
+extern const FFFilter ff_vf_tonemap_cuda;
extern const FFFilter ff_vf_tonemap_opencl;
extern const FFFilter ff_vf_tonemap_vaapi;
extern const FFFilter ff_vf_tpad;
Index: FFmpeg/libavfilter/colorspace.c
===================================================================
--- FFmpeg.orig/libavfilter/colorspace.c
@@ -128,7 +128,7 @@ Index: FFmpeg/libavfilter/colorspace.c
+ return peak;
+}
+
+void ff_map_dovi_metadata(struct DoviMetadata *out, const AVDOVIMetadata *data)
+void ff_map_dovi_metadata(struct FFDOVIMetadataRemap *out, const AVDOVIMetadata *data)
+{
+ int c, i, j, k;
+ const AVDOVIRpuDataHeader *header;
@@ -152,7 +152,7 @@ Index: FFmpeg/libavfilter/colorspace.c
+ }
+ for (c = 0; c < 3; c++) {
+ const AVDOVIReshapingCurve *csrc = &mapping->curves[c];
+ struct ReshapeData *cdst = &out->comp[c];
+ struct FFDOVIReshapeData *cdst = &out->comp[c];
+ cdst->num_pivots = csrc->num_pivots;
+ for (i = 0; i < csrc->num_pivots; i++) {
+ const float scale = 1.0f / ((1 << header->bl_bit_depth) - 1);
@@ -183,7 +183,7 @@ Index: FFmpeg/libavfilter/colorspace.c
+}
+
+// linearizer for PQ/ST2084
+float eotf_st2084_common(float x)
+float ff_eotf_st2084_common(float x)
+{
+ float xpow = powf(FFMAX(x, 0.0f), 1.0f / ST2084_M2);
+ float num = FFMAX(xpow - ST2084_C1, 0.0f);
@@ -192,13 +192,13 @@ Index: FFmpeg/libavfilter/colorspace.c
+ return x;
+}
+
+float eotf_st2084(float x, float ref_white)
+float ff_eotf_st2084(float x, float ref_white)
+{
+ return eotf_st2084_common(x) * ST2084_MAX_LUMINANCE / ref_white;
+ return ff_eotf_st2084_common(x) * ST2084_MAX_LUMINANCE / ref_white;
+}
+
+// delinearizer for PQ/ST2084
+float inverse_eotf_st2084_common(float x)
+float ff_inverse_eotf_st2084_common(float x)
+{
+ float xpow = powf(FFMAX(x, 0.0f), ST2084_M1);
+#if 0
@@ -214,28 +214,28 @@ Index: FFmpeg/libavfilter/colorspace.c
+#endif
+}
+
+float inverse_eotf_st2084(float x, float ref_white)
+float ff_inverse_eotf_st2084(float x, float ref_white)
+{
+ x *= ref_white / ST2084_MAX_LUMINANCE;
+ return inverse_eotf_st2084_common(x);
+ return ff_inverse_eotf_st2084_common(x);
+}
+
+float ootf_1_2(float x) {
+float ff_ootf_1_2(float x) {
+ return x > 0.0f ? powf(x, 1.2f) : x;
+}
+
+float inverse_ootf_1_2(float x) {
+float ff_inverse_ootf_1_2(float x) {
+ return x > 0.0f ? powf(x, 1.0f / 1.2f) : x;
+}
+
+float oetf_arib_b67(float x) {
+float ff_oetf_arib_b67(float x) {
+ x = FFMAX(x, 0.0f);
+ return x <= (1.0f / 12.0f)
+ ? sqrtf(3.0f * x)
+ : (ARIB_B67_A * logf(12.0f * x - ARIB_B67_B) + ARIB_B67_C);
+}
+
+float inverse_oetf_arib_b67(float x) {
+float ff_inverse_oetf_arib_b67(float x) {
+ x = FFMAX(x, 0.0f);
+ return x <= 0.5f
+ ? (x * x) * (1.0f / 3.0f)
@@ -243,17 +243,17 @@ Index: FFmpeg/libavfilter/colorspace.c
+}
+
+// linearizer for HLG/ARIB-B67
+float eotf_arib_b67(float x) {
+ return ootf_1_2(inverse_oetf_arib_b67(x)) * (12.0f / REFERENCE_WHITE_HLG);
+float ff_eotf_arib_b67(float x) {
+ return ff_ootf_1_2(ff_inverse_oetf_arib_b67(x)) * (12.0f / REFERENCE_WHITE_HLG);
+}
+
+// delinearizer for HLG/ARIB-B67
+float inverse_eotf_arib_b67(float x) {
+ return oetf_arib_b67(inverse_ootf_1_2(x / (12.0f / REFERENCE_WHITE_HLG)));
+float ff_inverse_eotf_arib_b67(float x) {
+ return ff_oetf_arib_b67(ff_inverse_ootf_1_2(x / (12.0f / REFERENCE_WHITE_HLG)));
+}
+
+// delinearizer for BT709, BT2020-10
+float inverse_eotf_bt1886(float x) {
+float ff_inverse_eotf_bt1886(float x) {
+ return x > 0.0f ? powf(x, 1.0f / 2.4f) : 0.0f;
+}
Index: FFmpeg/libavfilter/colorspace.h
@@ -305,13 +305,13 @@ Index: FFmpeg/libavfilter/colorspace.h
+#define QUANTIZATION_OFFSET(n) ((double)(1 << (16 - (n))) / ((1 << 16) - 1))
+
+// Parsed metadata from the Dolby Vision RPU
+struct DoviMetadata {
+struct FFDOVIMetadataRemap {
+ float nonlinear_offset[3]; // input offset ("ycc_to_rgb_offset")
+ double nonlinear[3][3]; // before PQ, also called "ycc_to_rgb"
+ double linear[3][3]; // after PQ, also called "rgb_to_lms"
+
+ // Reshape data, grouped by component
+ struct ReshapeData {
+ struct FFDOVIReshapeData {
+ uint8_t num_pivots;
+ float pivots[9]; // normalized to [0.0, 1.0] based on BL bit depth
+ uint8_t method[8]; // 0 = polynomial, 1 = MMR
@@ -333,19 +333,19 @@ Index: FFmpeg/libavfilter/colorspace.h
void ff_update_hdr_metadata(AVFrame *in, double peak);
+double ff_determine_dovi_signal_peak(const AVDOVIMetadata *data, int l0_only);
+void ff_map_dovi_metadata(struct DoviMetadata *out, const AVDOVIMetadata *data);
+void ff_map_dovi_metadata(struct FFDOVIMetadataRemap *out, const AVDOVIMetadata *data);
+
+float eotf_st2084_common(float x);
+float eotf_st2084(float x, float ref_white);
+float inverse_eotf_st2084_common(float x);
+float inverse_eotf_st2084(float x, float ref_white);
+float ootf_1_2(float x);
+float inverse_ootf_1_2(float x);
+float oetf_arib_b67(float x);
+float inverse_oetf_arib_b67(float x);
+float eotf_arib_b67(float x);
+float inverse_eotf_arib_b67(float x);
+float inverse_eotf_bt1886(float x);
+float ff_eotf_st2084_common(float x);
+float ff_eotf_st2084(float x, float ref_white);
+float ff_inverse_eotf_st2084_common(float x);
+float ff_inverse_eotf_st2084(float x, float ref_white);
+float ff_ootf_1_2(float x);
+float ff_inverse_ootf_1_2(float x);
+float ff_oetf_arib_b67(float x);
+float ff_inverse_oetf_arib_b67(float x);
+float ff_eotf_arib_b67(float x);
+float ff_inverse_eotf_arib_b67(float x);
+float ff_inverse_eotf_bt1886(float x);
+
#endif
Index: FFmpeg/libavfilter/cuda/colorspace_common.h
@@ -500,12 +500,12 @@ Index: FFmpeg/libavfilter/cuda/colorspace_common.h
+
+// linearizer for HLG/ARIB-B67
+static __inline__ __device__ float eotf_arib_b67(float x) {
+ return ootf_1_2(inverse_oetf_arib_b67(x)) * (12.0f / REFERENCE_WHITE_ALT);
+ return ootf_1_2(inverse_oetf_arib_b67(x)) * (12.0f / REFERENCE_WHITE_HLG);
+}
+
+// delinearizer for HLG/ARIB-B67
+static __inline__ __device__ float inverse_eotf_arib_b67(float x) {
+ return oetf_arib_b67(inverse_ootf_1_2(x / (12.0f / REFERENCE_WHITE_ALT)));
+ return oetf_arib_b67(inverse_ootf_1_2(x / (12.0f / REFERENCE_WHITE_HLG)));
+}
+
+// delinearizer for BT709, BT2020-10
@@ -698,7 +698,7 @@ Index: FFmpeg/libavfilter/cuda/host_util.c
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/cuda/host_util.c
@@ -0,0 +1,77 @@
@@ -0,0 +1,74 @@
+/*
+ * This file is part of FFmpeg.
+ *
@@ -725,7 +725,7 @@ Index: FFmpeg/libavfilter/cuda/host_util.c
+#define CHECK_CU(x) FF_CUDA_CHECK_DL(ctx, cu, x)
+#define DEPTH_BYTES(depth) (((depth) + 7) / 8)
+
+int ff_make_cuda_frame(AVFilterContext *ctx, CudaFunctions *cu, int make_cuTex,
+int ff_make_cuda_frame(AVFilterContext *ctx, CudaFunctions *cu, int make_cu_tex,
+ FFCUDAFrame *dst, const AVFrame *src, const AVPixFmtDescriptor *src_desc)
+{
+ int i, ret = 0;
@@ -738,26 +738,23 @@ Index: FFmpeg/libavfilter/cuda/host_util.c
+ dst->tex[i] = 0;
+ }
+
+ for (i = 0; make_cuTex && (i < dst->planes); i++) {
+#ifndef CU_TRSF_NORMALIZED_COORDINATES
+ #define CU_TRSF_NORMALIZED_COORDINATES 2
+#endif
+ for (i = 0; make_cu_tex && (i < dst->planes); i++) {
+ CUDA_TEXTURE_DESC tex_desc = {
+ .addressMode = { CU_TR_ADDRESS_MODE_CLAMP, CU_TR_ADDRESS_MODE_CLAMP },
+ .filterMode = i == 0 ? CU_TR_FILTER_MODE_POINT : CU_TR_FILTER_MODE_LINEAR,
+ .flags = i == 0 ? 0 : CU_TRSF_NORMALIZED_COORDINATES,
+ .filterMode = i ? CU_TR_FILTER_MODE_LINEAR : CU_TR_FILTER_MODE_POINT,
+ .flags = i ? 2 /* CU_TRSF_NORMALIZED_COORDINATES */ : 0
+ };
+
+ CUDA_RESOURCE_DESC res_desc = {
+ .resType = CU_RESOURCE_TYPE_PITCH2D,
+ .res.pitch2D.format = DEPTH_BYTES(src_desc->comp[i].depth) == 1 ?
+ CU_AD_FORMAT_UNSIGNED_INT8 :
+ CU_AD_FORMAT_UNSIGNED_INT16,
+ .res.pitch2D.numChannels = i == 0 ? 1 : (dst->planes == 2 ? 2 : 1),
+ .res.pitch2D.width = i == 0 ? src->width : AV_CEIL_RSHIFT(src->width, src_desc->log2_chroma_w),
+ .res.pitch2D.height = i == 0 ? src->height : AV_CEIL_RSHIFT(src->height, src_desc->log2_chroma_h),
+ .resType = CU_RESOURCE_TYPE_PITCH2D,
+ .res.pitch2D.format = DEPTH_BYTES(src_desc->comp[i].depth) == 1 ?
+ CU_AD_FORMAT_UNSIGNED_INT8 :
+ CU_AD_FORMAT_UNSIGNED_INT16,
+ .res.pitch2D.numChannels = i ? (dst->planes == 2 ? 2 : 1) : 1,
+ .res.pitch2D.width = AV_CEIL_RSHIFT(src->width, i ? src_desc->log2_chroma_w : 0),
+ .res.pitch2D.height = AV_CEIL_RSHIFT(src->height, i ? src_desc->log2_chroma_h : 0),
+ .res.pitch2D.pitchInBytes = src->linesize[i],
+ .res.pitch2D.devPtr = (CUdeviceptr)src->data[i],
+ .res.pitch2D.devPtr = (CUdeviceptr)src->data[i]
+ };
+
+ if ((ret = CHECK_CU(cu->cuTexObjectCreate(&dst->tex[i], &res_desc, &tex_desc, NULL))) < 0)
@@ -807,7 +804,7 @@ Index: FFmpeg/libavfilter/cuda/host_util.h
+#include "libavfilter/avfilter.h"
+#include "shared.h"
+
+int ff_make_cuda_frame(AVFilterContext *ctx, CudaFunctions *cu, int make_cuTex,
+int ff_make_cuda_frame(AVFilterContext *ctx, CudaFunctions *cu, int make_cu_tex,
+ FFCUDAFrame *dst, const AVFrame *src, const AVPixFmtDescriptor *src_desc);
+
+#endif /* AVFILTER_CUDA_HOST_UTIL_H */
@@ -815,29 +812,29 @@ Index: FFmpeg/libavfilter/cuda/load_helper.c
===================================================================
--- FFmpeg.orig/libavfilter/cuda/load_helper.c
+++ FFmpeg/libavfilter/cuda/load_helper.c
@@ -32,11 +32,9 @@
@@ -31,11 +31,9 @@
#define CHECK_CU(x) FF_CUDA_CHECK_DL(avctx, cu, x)
-int ff_cuda_load_module(void *avctx, AVCUDADeviceContext *hwctx, CUmodule *cu_module,
- const unsigned char *data, const unsigned int length)
+static int decompress_cuda_ptx(void *avctx, uint8_t **data_out, uint64_t *length_out,
+static int decompress_cuda_ptx(void *avctx, uint8_t **data_out, size_t *length_out,
+ const unsigned char *data, const unsigned int length)
{
- CudaFunctions *cu = hwctx->internal->cuda_dl;
-
#if CONFIG_PTX_COMPRESSION
z_stream stream = { 0 };
uint8_t *buf, *tmp;
@@ -88,10 +86,51 @@ int ff_cuda_load_module(void *avctx, AVC
#if CONFIG_SHADER_COMPRESSION
uint8_t *out;
size_t out_len;
@@ -44,10 +42,51 @@ int ff_cuda_load_module(void *avctx, AVC
if (ret < 0)
return ret;
inflateEnd(&stream);
- ret = CHECK_CU(cu->cuModuleLoadData(cu_module, buf));
- av_free(buf);
- ret = CHECK_CU(cu->cuModuleLoadData(cu_module, out));
- av_free(out);
- return ret;
+ *data_out = buf;
+ *length_out = stream.total_out;
+ *data_out = out;
+ *length_out = out_len;
#else
- return CHECK_CU(cu->cuModuleLoadData(cu_module, data));
+ *data_out = NULL;
@@ -851,7 +848,7 @@ Index: FFmpeg/libavfilter/cuda/load_helper.c
+{
+ CudaFunctions *cu = hwctx->internal->cuda_dl;
+ uint8_t *data_out = NULL;
+ uint64_t length_out = 0;
+ size_t length_out = 0;
+ int ret;
+
+ if ((ret = decompress_cuda_ptx(avctx, &data_out, &length_out, data, length)) < 0)
@@ -870,7 +867,7 @@ Index: FFmpeg/libavfilter/cuda/load_helper.c
+{
+ CudaFunctions *cu = hwctx->internal->cuda_dl;
+ uint8_t *data_out = NULL;
+ uint64_t length_out = 0;
+ size_t length_out = 0;
+ int ret;
+
+ if ((ret = decompress_cuda_ptx(avctx, &data_out, &length_out, data, length)) < 0)
@@ -1173,7 +1170,7 @@ Index: FFmpeg/libavfilter/cuda/tonemap.cu
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/cuda/tonemap.cu
@@ -0,0 +1,661 @@
@@ -0,0 +1,663 @@
+/*
+ * This file is part of FFmpeg.
+ *
@@ -1200,6 +1197,7 @@ Index: FFmpeg/libavfilter/cuda/tonemap.cu
+extern __constant__ const enum TonemapAlgorithm tonemap_func;
+extern __constant__ const float tone_param;
+extern __constant__ const float desat_param;
+extern __constant__ const int lut_size;
+extern __constant__ const int enable_dither;
+extern __constant__ const float dither_size;
+extern __constant__ const float dither_quantization;
@@ -1210,11 +1208,10 @@ Index: FFmpeg/libavfilter/cuda/tonemap.cu
+#define dot3(a, b) ((a).z * (b).z + ((a).y * (b).y + (a).x * (b).x))
+#define dot4(a, b) ((a).w * (b).w + ((a).z * (b).z + ((a).y * (b).y + (a).x * (b).x)))
+
+#define LUT_SIZE 65
+
+template <typename T, typename S>
+static __inline__ __device__
+float3 clamp3(const float3 a, const float min_val, const float max_val) {
+ float3 result;
+T clamp3(const T a, const S min_val, const S max_val) {
+ T result;
+ result.x = clamp(a.x, min_val, max_val);
+ result.y = clamp(a.y, min_val, max_val);
+ result.z = clamp(a.z, min_val, max_val);
@@ -1573,56 +1570,58 @@ Index: FFmpeg/libavfilter/cuda/tonemap.cu
+ float4 lut_val;
+ color = clamp3(color, 0.0f, 1.0f);
+
+ // Scale the color to the LUT grid.
+ float3 pos = color * (float)(LUT_SIZE - 1);
+ // Scale the color to the LUT grid
+ float3 pos = color * (float)(lut_size - 1);
+
+ // Get the integer base indices in the LUT.
+ int3 base = make_int3((int)floorf(pos.x), (int)floorf(pos.y), (int)floorf(pos.z));
+ // Compute the fractional part within the cell.
+ // Get the integer base indices in the LUT
+ int3 base = clamp3(make_int3((int)floorf(pos.x),
+ (int)floorf(pos.y),
+ (int)floorf(pos.z)), 0, lut_size - 2);
+
+ // Compute the fractional part within the cell
+ float3 f = pos - make_float3((float)base.x, (float)base.y, (float)base.z);
+
+ // Compute the base linear index.
+ unsigned base_idx = base.x + base.y * LUT_SIZE + base.z * LUT_SIZE * LUT_SIZE;
+ // Sort the fraction offsets, so that we always have f_max>=f_mid>=f_min
+ float f_max = max(f.x, max(f.y, f.z));
+ float f_min = min(f.x, min(f.y, f.z));
+ float f_mid = f.x + f.y + f.z - f_max - f_min;
+
+ // Sort the fraction offsets, so that we always have a>=b>=c
+ float a = max(f.x, max(f.y, f.z));
+ float c = min(f.x, min(f.y, f.z));
+ float b = f.x + f.y + f.z - a - c;
+
+#define LUT_IDX_MAX (LUT_SIZE * LUT_SIZE * LUT_SIZE - 1)
+ // Compute the base linear index
+ unsigned base_idx = base.x + base.y * lut_size + base.z * lut_size * lut_size;
+ unsigned last_idx = base_idx + 1 + lut_size + lut_size * lut_size;
+
+ // The initial and the last corner values of current cube will always be fetched
+ lut_val = lut[min(base_idx, (unsigned)LUT_IDX_MAX)];
+ lut_val = lut[base_idx];
+ float3 c000 = make_float3(lut_val.x, lut_val.y, lut_val.z);
+ lut_val = lut[min(base_idx + 1 + LUT_SIZE + LUT_SIZE * LUT_SIZE, (unsigned)LUT_IDX_MAX)];
+ lut_val = lut[last_idx];
+ float3 c111 = make_float3(lut_val.x, lut_val.y, lut_val.z);
+
+ // Select the index for vertices of the tetrahedron.
+ // Select the index for vertices of the tetrahedron
+ unsigned idx100 = base_idx + 1;
+ unsigned idx010 = base_idx + LUT_SIZE;
+ unsigned idx110 = base_idx + 1 + LUT_SIZE;
+ unsigned idx001 = base_idx + LUT_SIZE * LUT_SIZE;
+ unsigned idx101 = base_idx + 1 + LUT_SIZE * LUT_SIZE;
+ unsigned idx011 = base_idx + LUT_SIZE + LUT_SIZE * LUT_SIZE;
+ unsigned idx010 = base_idx + lut_size;
+ unsigned idx110 = base_idx + 1 + lut_size;
+ unsigned idx001 = base_idx + lut_size * lut_size;
+ unsigned idx101 = base_idx + 1 + lut_size * lut_size;
+ unsigned idx011 = base_idx + lut_size + lut_size * lut_size;
+
+ // Although we have a and c as max and min value, we cannot use them in the
+ // following selection as float equality comparison is not accurate on GPU.
+ unsigned idx0 = select(select(idx001, idx010, (f.y >= f.z && f.y >= f.x)), idx100, (f.x >= f.y && f.x >= f.z));
+ unsigned idx1 = select(select(idx110, idx101, (f.y <= f.z && f.y <= f.x)), idx011, (f.x <= f.y && f.x <= f.z));
+ // following selection as float equality comparison is not accurate on GPU
+ unsigned y_max = f.y >= f.z && f.y >= f.x;
+ unsigned x_max = f.x >= f.y && f.x >= f.z;
+ unsigned idx0 = select(select(idx001, idx010, y_max), idx100, x_max);
+ unsigned y_min = f.y <= f.z && f.y <= f.x;
+ unsigned x_min = f.x <= f.y && f.x <= f.z;
+ unsigned idx1 = select(select(idx110, idx101, y_min), idx011, x_min);
+
+ // Fetch LUT value with determined tetrahedron
+ lut_val = lut[min(idx0, (unsigned)LUT_IDX_MAX)];
+ lut_val = lut[idx0];
+ float3 c0 = make_float3(lut_val.x, lut_val.y, lut_val.z);
+ lut_val = lut[min(idx1, (unsigned)LUT_IDX_MAX)];
+ lut_val = lut[idx1];
+ float3 c1 = make_float3(lut_val.x, lut_val.y, lut_val.z);
+
+ float3 ca = c0 - c000;
+ float3 cb = c1 - c0;
+ float3 cc = c111 - c1;
+
+ float3 result = c000 + a * ca + b * cb + c * cc;
+
+ return clamp3(result, 0.0f, 1.0f);
+ return clamp3(c000 + f_max * (c0 - c000)
+ + f_mid * (c1 - c0)
+ + f_min * (c111 - c1), 0.0f, 1.0f);
+}
+
+extern "C" {
@@ -1775,16 +1774,16 @@ Index: FFmpeg/libavfilter/cuda/tonemap.cu
+ int skip_tonemap,
+ int dovi_reshape)
+{
+ const int total_entries = LUT_SIZE * LUT_SIZE * LUT_SIZE;
+ const int total_entries = lut_size * lut_size * lut_size;
+ int idx = blockIdx.x * blockDim.x + threadIdx.x;
+ if (idx >= total_entries) return;
+ int z = idx / (LUT_SIZE * LUT_SIZE);
+ int rem = idx - (z * LUT_SIZE * LUT_SIZE);
+ int y = rem / LUT_SIZE;
+ int x = rem % LUT_SIZE;
+ float fx = (float)x / (LUT_SIZE - 1);
+ float fy = (float)y / (LUT_SIZE - 1);
+ float fz = (float)z / (LUT_SIZE - 1);
+ int z = idx / (lut_size * lut_size);
+ int rem = idx - (z * lut_size * lut_size);
+ int y = rem / lut_size;
+ int x = rem % lut_size;
+ float fx = (float)x / (lut_size - 1);
+ float fy = (float)y / (lut_size - 1);
+ float fz = (float)z / (lut_size - 1);
+ float3 c = make_float3(fx, fy, fz);
+ if (tonemap_mode == TONEMAP_MODE_ITP) {
+ c = dovi_reshape
@@ -1994,7 +1993,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_tonemap_cuda.c
@@ -0,0 +1,1271 @@
@@ -0,0 +1,1281 @@
+/*
+ * This file is part of FFmpeg.
+ *
@@ -2095,7 +2094,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+ CUfunction cu_func_dovi;
+ CUfunction cu_func_dovi_pq;
+
+#define LUT_SIZE (65 * 65 * 65)
+#define LUT_SIZE 65
+ CUdeviceptr lut_buffer;
+
+ CUdeviceptr dither_buffer;
@@ -2110,11 +2109,13 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+#define coeffs_sz coeffs_cnt*sizeof(float)
+#define mmr_sz mmr_cnt*sizeof(float)
+ CUdeviceptr dovi_buffer;
+ struct DoviMetadata *dovi;
+ struct FFDOVIMetadataRemap *dovi;
+ float *dovi_pbuf;
+
+ enum TonemapAlgorithm tonemap;
+ enum TonemapMode tonemap_mode;
+ /* enum TonemapAlgorithm */
+ int tonemap;
+ /* enum TonemapMode */
+ int tonemap_mode;
+ int apply_dovi;
+ int tradeoff;
+ int init_with_dovi;
@@ -2257,30 +2258,28 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+ .srcPitch = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]),
+ .dstPitch = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]),
+ .WidthInBytes = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]),
+ .Height = ff_fruit_dither_size,
+ .Height = ff_fruit_dither_size
+ };
+
+#ifndef CU_TRSF_NORMALIZED_COORDINATES
+ #define CU_TRSF_NORMALIZED_COORDINATES 2
+#endif
+ CUDA_TEXTURE_DESC tex_desc = {
+ .addressMode = { CU_TR_ADDRESS_MODE_WRAP,
+ CU_TR_ADDRESS_MODE_WRAP },
+ .filterMode = CU_TR_FILTER_MODE_POINT,
+ .flags = CU_TRSF_NORMALIZED_COORDINATES,
+ .filterMode = CU_TR_FILTER_MODE_POINT,
+ .flags = 2 /* CU_TRSF_NORMALIZED_COORDINATES */
+ };
+
+ CUDA_RESOURCE_DESC res_desc = {
+ .resType = CU_RESOURCE_TYPE_PITCH2D,
+ .res.pitch2D.format = CU_AD_FORMAT_UNSIGNED_INT16,
+ .res.pitch2D.numChannels = 1,
+ .res.pitch2D.width = ff_fruit_dither_size,
+ .res.pitch2D.height = ff_fruit_dither_size,
+ .resType = CU_RESOURCE_TYPE_PITCH2D,
+ .res.pitch2D.format = CU_AD_FORMAT_UNSIGNED_INT16,
+ .res.pitch2D.numChannels = 1,
+ .res.pitch2D.width = ff_fruit_dither_size,
+ .res.pitch2D.height = ff_fruit_dither_size,
+ .res.pitch2D.pitchInBytes = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]),
+ .res.pitch2D.devPtr = 0,
+ .res.pitch2D.devPtr = 0
+ };
+
+ av_assert0(sizeof(ff_fruit_dither_matrix) == sizeof(ff_fruit_dither_matrix[0]) * ff_fruit_dither_size * ff_fruit_dither_size);
+ av_assert0(sizeof(ff_fruit_dither_matrix) ==
+ sizeof(ff_fruit_dither_matrix[0]) * ff_fruit_dither_size * ff_fruit_dither_size);
+
+ if ((ret = CHECK_CU(cu->cuCtxPushCurrent(cuda_ctx))) < 0)
+ return ret;
@@ -2444,7 +2443,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+ for (c = 0; c < 3; c++) {
+ int has_poly = 0, has_mmr = 0, mmr_single = 1;
+ int mmr_idx = 0, min_order = 3, max_order = 1;
+ const struct ReshapeData *comp = &s->dovi->comp[c];
+ const struct FFDOVIReshapeData *comp = &s->dovi->comp[c];
+ if (!comp->num_pivots)
+ continue;
+ av_assert0(comp->num_pivots >= 2 && comp->num_pivots <= 9);
@@ -2714,6 +2713,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+ CONSTANT(".u32 chroma_loc_src = %i", (int)s->in_chroma_loc);
+ CONSTANT(".u32 chroma_loc_dst = %i", (int)s->out_chroma_loc);
+ CONSTANT(".u32 tonemap_func = %i", (int)s->tonemap);
+ CONSTANT(".u32 lut_size = %i", (int)(LUT_SIZE));
+ CONSTANT(".u32 enable_dither = %i", (int)(s->in_desc->comp[0].depth > s->out_desc->comp[0].depth));
+ CONSTANT(".f32 dither_size = %.1f", (float)ff_fruit_dither_size);
+ CONSTANT(".f32 dither_quantization = %.1f", (float)((1 << s->out_desc->comp[0].depth) - 1));
@@ -2783,6 +2783,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+ if (ret < 0)
+ goto fail;
+
+ s->cu_func_build_lut = NULL;
+ s->cu_func_tm = NULL;
+ s->cu_func_dovi = NULL;
+ s->cu_func_dovi_pq = NULL;
@@ -2813,7 +2814,8 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+
+ if (s->tradeoff == 1) {
+ const size_t lut_size = LUT_SIZE;
+ const size_t lut_buffer_size = lut_size * sizeof(float) * 4;
+ const size_t lut_size_3d = lut_size * lut_size * lut_size;
+ const size_t lut_buffer_size = lut_size_3d * sizeof(float) * 4;
+ float peak = (float)s->src_peak;
+ float dst_peak = (float)s->dst_peak;
+ int skip_tonemap = s->out_trc == AVCOL_TRC_SMPTE2084;
@@ -2840,7 +2842,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+ void *args[] = { &s->lut_buffer, &peak, &dst_peak, &s->tonemap_mode, &skip_tonemap, &dovi_reshape };
+
+ ret = CHECK_CU(cu->cuLaunchKernel(s->cu_func_build_lut,
+ lut_size, 1, 1,
+ lut_size_3d, 1, 1,
+ 1, 1, 1, 0, s->hwctx->stream, args, NULL));
+ if (ret < 0)
+ goto fail2;
@@ -2917,6 +2919,13 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+
+ outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
+
+ if (s->trc != AVCOL_TRC_SMPTE2084) {
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
+
+ return 0;
+}
+
@@ -3049,7 +3058,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+ const AVDOVIRpuDataHeader *rpu = av_dovi_get_header(metadata);
+ // only map dovi rpus that don't require an EL
+ if (rpu->disable_residual_flag) {
+ struct DoviMetadata *dovi = av_malloc(sizeof(*dovi));
+ struct FFDOVIMetadataRemap *dovi = av_malloc(sizeof(*dovi));
+ s->dovi = dovi;
+ if (!s->dovi)
+ goto fail;
@@ -3248,16 +3257,16 @@ Index: FFmpeg/libavfilter/vf_tonemap_cuda.c
+ },
+};
+
+const AVFilter ff_vf_tonemap_cuda = {
+ .name = "tonemap_cuda",
+ .description = NULL_IF_CONFIG_SMALL("GPU accelerated HDR to SDR tonemapping"),
+const FFFilter ff_vf_tonemap_cuda = {
+ .p.name = "tonemap_cuda",
+ .p.description = NULL_IF_CONFIG_SMALL("GPU accelerated HDR to SDR tonemapping"),
+
+ .preinit = preinit,
+ .init = init,
+ .uninit = uninit,
+
+ .priv_size = sizeof(TonemapCUDAContext),
+ .priv_class = &tonemap_cuda_class,
+ .p.priv_class = &tonemap_cuda_class,
+
+ FILTER_INPUTS(tonemap_cuda_inputs),
+ FILTER_OUTPUTS(tonemap_cuda_outputs),
debian/patches/0005-add-amf-fixes-from-upstream-and-custom-tunings.patch
Vendored
+397 -580
View File
File diff suppressed because it is too large Load Diff
debian/patches/0006-add-opencl-scaler-and-pixfmt-converter-impl.patch
Vendored
+486 -702
View File
File diff suppressed because it is too large Load Diff
debian/patches/0007-add-bt2390-eetf-and-code-refactor-to-opencl-tonemap.patch
Vendored
+328 -175
View File
@@ -501,7 +501,7 @@ Index: FFmpeg/libavfilter/opencl/tonemap.cl
===================================================================
--- FFmpeg.orig/libavfilter/opencl/tonemap.cl
+++ FFmpeg/libavfilter/opencl/tonemap.cl
@@ -16,54 +16,89 @@
@@ -16,54 +16,88 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
@@ -528,7 +528,6 @@ Index: FFmpeg/libavfilter/opencl/tonemap.cl
+#endif
+
+#define FLOAT_EPS 1e-6f
+#define LUT_SIZE 65
+
extern float3 lrgb2yuv(float3);
extern float lrgb2y(float3);
@@ -612,7 +611,7 @@ Index: FFmpeg/libavfilter/opencl/tonemap.cl
float j = tone_param;
float a, b;
@@ -71,202 +106,1016 @@ float mobius(float s, float peak) {
@@ -71,202 +105,1040 @@ float mobius(float s, float peak) {
return s;
a = -j * j * (peak - 1.0f) / (j * j - 2.0f * j + peak);
@@ -1565,59 +1564,83 @@ Index: FFmpeg/libavfilter/opencl/tonemap.cl
+#endif
+}
+
+float3 apply_lut3d(__global const float3 *restrict lut, float3 color)
+#undef lut3d_read_t
+#ifdef LUT_PERF_IMAGE3D
+ #define lut3d_read_t __read_only image3d_t
+#else
+ #define lut3d_read_t __global const float4 *restrict
+#endif
+
+float3 apply_lut3d(lut3d_read_t lut, float3 color)
+{
+ color = clamp(color, 0.0f, 1.0f);
+
+ // Scale the color to the LUT grid.
+ // Scale the color to the LUT grid
+ float3 pos = color * (float)(LUT_SIZE - 1);
+
+ // Get the integer base indices in the LUT.
+ int3 base = convert_int3(floor(pos));
+ // Compute the fractional part within the cell.
+ // Get the integer base indices in the LUT
+ int3 base = clamp(convert_int3(floor(pos)), 0, LUT_SIZE - 2);
+
+ // Compute the fractional part within the cell
+ float3 f = pos - convert_float3(base);
+
+ // Compute the base linear index.
+ uint base_idx = base.x + base.y * LUT_SIZE + base.z * LUT_SIZE * LUT_SIZE;
+
+ // Sort the fraction offsets, so that we always have a>=b>=c
+ float a = fmax(f.x, fmax(f.y, f.z));
+ float c = fmin(f.x, fmin(f.y, f.z));
+ float b = f.x + f.y + f.z - a - c;
+
+#define LUT_IDX_MAX (LUT_SIZE * LUT_SIZE * LUT_SIZE - 1)
+ // Sort the fraction offsets, so that we always have f_max>=f_mid>=f_min
+ float f_max = fmax(f.x, fmax(f.y, f.z));
+ float f_min = fmin(f.x, fmin(f.y, f.z));
+ float f_mid = f.x + f.y + f.z - f_max - f_min;
+
+ // The initial and the last corner values of current cube will always be fetched
+ float3 c000 = lut[min(base_idx, (uint)LUT_IDX_MAX)];
+ float3 c111 = lut[min(base_idx + 1 + LUT_SIZE + LUT_SIZE * LUT_SIZE, (uint)LUT_IDX_MAX)];
+#ifdef LUT_PERF_IMAGE3D
+ float3 c000 = read_imagef(lut, n_sampler, (int4)(base + 0, 0)).xyz;
+ float3 c111 = read_imagef(lut, n_sampler, (int4)(base + 1, 0)).xyz;
+#else
+ // Compute the base linear index
+ uint base_idx = base.x + base.y * LUT_SIZE + base.z * LUT_SIZE * LUT_SIZE;
+ uint last_idx = base_idx + 1 + LUT_SIZE + LUT_SIZE * LUT_SIZE;
+ float3 c000 = lut[base_idx].xyz;
+ float3 c111 = lut[last_idx].xyz;
+#endif
+
+ // Select the index for vertices of the tetrahedron.
+ // Select the index for vertices of the tetrahedron:
+ // Although we have f_max and f_min values, we cannot use them in the
+ // following selection as float equality comparison is not accurate on GPU
+#ifdef LUT_PERF_IMAGE3D
+ int3 y_max = (int3)(-(f.y >= f.z && f.y >= f.x));
+ int3 x_max = (int3)(-(f.x >= f.y && f.x >= f.z));
+ int3 d0 = select(select((int3)(0, 0, 1), (int3)(0, 1, 0), y_max), (int3)(1, 0, 0), x_max);
+ int3 y_min = (int3)(-(f.y <= f.z && f.y <= f.x));
+ int3 x_min = (int3)(-(f.x <= f.y && f.x <= f.z));
+ int3 d1 = select(select((int3)(1, 1, 0), (int3)(1, 0, 1), y_min), (int3)(0, 1, 1), x_min);
+#else
+ uint idx100 = base_idx + 1;
+ uint idx010 = base_idx + LUT_SIZE;
+ uint idx110 = base_idx + 1 + LUT_SIZE;
+ uint idx001 = base_idx + LUT_SIZE * LUT_SIZE;
+ uint idx101 = base_idx + 1 + LUT_SIZE * LUT_SIZE;
+ uint idx011 = base_idx + LUT_SIZE + LUT_SIZE * LUT_SIZE;
+
+ // Although we have a and c as max and min value, we cannot use them in the
+ // following selection as float equality comparison is not accurate on GPU.
+ uint idx0 = select(select(idx001, idx010, (f.y >= f.z && f.y >= f.x)), idx100, (f.x >= f.y && f.x >= f.z));
+ uint idx1 = select(select(idx110, idx101, (f.y <= f.z && f.y <= f.x)), idx011, (f.x <= f.y && f.x <= f.z));
+ uint y_max = f.y >= f.z && f.y >= f.x;
+ uint x_max = f.x >= f.y && f.x >= f.z;
+ uint idx0 = select(select(idx001, idx010, y_max), idx100, x_max);
+ uint y_min = f.y <= f.z && f.y <= f.x;
+ uint x_min = f.x <= f.y && f.x <= f.z;
+ uint idx1 = select(select(idx110, idx101, y_min), idx011, x_min);
+#endif
+
+ // Fetch LUT value with determined tetrahedron
+ float3 c0 = lut[min(idx0, (uint)LUT_IDX_MAX)];
+ float3 c1 = lut[min(idx1, (uint)LUT_IDX_MAX)];
+#ifdef LUT_PERF_IMAGE3D
+ float3 c0 = read_imagef(lut, n_sampler, (int4)(base + d0, 0)).xyz;
+ float3 c1 = read_imagef(lut, n_sampler, (int4)(base + d1, 0)).xyz;
+#else
+ float3 c0 = lut[idx0].xyz;
+ float3 c1 = lut[idx1].xyz;
+#endif
+
+ float3 ca = c0 - c000;
+ float3 cb = c1 - c0;
+ float3 cc = c111 - c1;
+
+ float3 result = c000 + a * ca + b * cb + c * cc;
+
+ return clamp(result, 0.0f, 1.0f);
+ return clamp(c000 + f_max * (c0 - c000)
+ + f_mid * (c1 - c0)
+ + f_min * (c111 - c1), 0.0f, 1.0f);
+}
+
+__kernel void tonemap_lut(__global const float3 *restrict lut,
+__kernel void tonemap_lut( lut3d_read_t lut,
+ __write_only image2d_t dst1,
+ __read_only image2d_t src1,
+ __write_only image2d_t dst2,
@@ -1753,7 +1776,7 @@ Index: FFmpeg/libavfilter/opencl/tonemap.cl
+#endif
+}
+
+__kernel void build_lut(__global float3 *lut, float peak)
+__kernel void build_lut(__global float4 *lut, float peak)
+{
+ const int total_entries = LUT_SIZE * LUT_SIZE * LUT_SIZE;
+ int idx = get_global_id(0);
@@ -1790,7 +1813,7 @@ Index: FFmpeg/libavfilter/opencl/tonemap.cl
+ c = clamp(c, 0.0f, 1.0f);
+#endif
+ c = lrgb2yuv(c);
+ lut[idx] = clamp(c, 0.0f, 1.0f);
+ lut[idx] = clamp((float4)(c, 0.0f), 0.0f, 1.0f);
}
Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
===================================================================
@@ -1802,13 +1825,14 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
@@ -15,27 +15,49 @@
@@ -15,27 +15,50 @@
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
+
#include <float.h>
+#define CL_USE_DEPRECATED_OPENCL_1_2_APIS
+#ifdef __APPLE__
+#include <OpenCL/cl_ext.h>
+#else
@@ -1858,7 +1882,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
enum TonemapAlgorithm {
TONEMAP_NONE,
@@ -45,7 +67,17 @@ enum TonemapAlgorithm {
@@ -45,7 +68,17 @@ enum TonemapAlgorithm {
TONEMAP_REINHARD,
TONEMAP_HABLE,
TONEMAP_MOBIUS,
@@ -1877,7 +1901,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
};
typedef struct TonemapOpenCLContext {
@@ -56,23 +88,48 @@ typedef struct TonemapOpenCLContext {
@@ -56,24 +89,52 @@ typedef struct TonemapOpenCLContext {
enum AVColorPrimaries primaries, primaries_in, primaries_out;
enum AVColorRange range, range_in, range_out;
enum AVChromaLocation chroma_loc;
@@ -1893,14 +1917,16 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+#define pivots_sz pivots_cnt*sizeof(cl_float)
+#define coeffs_sz coeffs_cnt*sizeof(cl_float)
+#define mmr_sz mmr_cnt*sizeof(cl_float)
+ struct DoviMetadata *dovi;
+ struct FFDOVIMetadataRemap *dovi;
+ cl_mem dovi_buf;
+ unsigned dovi_use_fp16;
+ unsigned is_pure_dovi;
+#define LUT_SIZE (65 * 65 * 65)
enum TonemapAlgorithm tonemap;
+ enum TonemapMode tonemap_mode;
+#define LUT_SIZE 65
/* enum TonemapAlgorithm */
int tonemap;
+ /* enum TonemapMode */
+ int tonemap_mode;
enum AVPixelFormat format;
+ int apply_dovi;
double peak;
@@ -1912,25 +1938,27 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
- double target_peak;
double scene_threshold;
+ int tradeoff;
+ int use_image3d;
int initialised;
+ int init_with_dovi;
cl_kernel kernel;
+ cl_kernel lut_generation_kernel;
+ cl_mem dither_image;
+ cl_mem lut_buffer;
+ cl_mem lut_image;
cl_command_queue command_queue;
- cl_mem util_mem;
} TonemapOpenCLContext;
static const char *const linearize_funcs[AVCOL_TRC_NB] = {
static const char *const linearize_funcs[] = {
- [AVCOL_TRC_SMPTE2084] = "eotf_st2084",
- [AVCOL_TRC_ARIB_STD_B67] = "inverse_oetf_hlg",
+ [AVCOL_TRC_SMPTE2084] = "eotf_st2084",
+ [AVCOL_TRC_ARIB_STD_B67] = "eotf_arib_b67",
};
static const char *const delinearize_funcs[AVCOL_TRC_NB] = {
@@ -80,7 +137,7 @@ static const char *const delinearize_fun
static const char *const delinearize_funcs[] = {
@@ -81,7 +142,7 @@ static const char *const delinearize_fun
[AVCOL_TRC_BT2020_10] = "inverse_eotf_bt1886",
};
@@ -1939,7 +1967,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
[TONEMAP_NONE] = "direct",
[TONEMAP_LINEAR] = "linear",
[TONEMAP_GAMMA] = "gamma",
@@ -88,6 +145,14 @@ static const char *const tonemap_func[TO
@@ -89,6 +150,14 @@ static const char *const tonemap_func[TO
[TONEMAP_REINHARD] = "reinhard",
[TONEMAP_HABLE] = "hable",
[TONEMAP_MOBIUS] = "mobius",
@@ -1954,7 +1982,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
};
static int get_rgb2rgb_matrix(enum AVColorPrimaries in, enum AVColorPrimaries out,
@@ -108,90 +173,453 @@ static int get_rgb2rgb_matrix(enum AVCol
@@ -109,90 +178,510 @@ static int get_rgb2rgb_matrix(enum AVCol
return 0;
}
@@ -1995,7 +2023,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ cl_float mmr_packed_dataf[8*6][4] = {0};
+ cl_half coeffs_datah[8][4] = {0};
+ cl_half mmr_packed_datah[8*6][4] = {0};
+ int c, i, j, k, err;
+ int c, i, j, k, err av_unused;
+ cl_int cle;
+
+ pbuf = clEnqueueMapBuffer(ctx->command_queue, ctx->dovi_buf,
@@ -2008,7 +2036,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ for (c = 0; c < 3; c++) {
+ int has_poly = 0, has_mmr = 0, mmr_single = 1;
+ int mmr_idx = 0, min_order = 3, max_order = 1;
+ const struct ReshapeData *comp = &ctx->dovi->comp[c];
+ const struct FFDOVIReshapeData *comp = &ctx->dovi->comp[c];
+ if (!comp->num_pivots)
+ continue;
+ av_assert0(comp->num_pivots >= 2 && comp->num_pivots <= 9);
@@ -2150,24 +2178,20 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
TonemapOpenCLContext *ctx = avctx->priv;
+ AVBPrint header;
+ const char *opencl_sources[OPENCL_SOURCE_NB];
+ size_t m_origin[3] = {0};
+ size_t m_region[3] = {ff_fruit_dither_size, ff_fruit_dither_size, 1};
+ size_t m_row_pitch = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]);
int rgb2rgb_passthrough = 1;
double rgb2rgb[3][3], rgb2yuv[3][3], yuv2rgb[3][3];
const AVLumaCoefficients *luma_src, *luma_dst;
+ cl_event event = NULL;
+ cl_bool device_is_integrated;
+ cl_uint max_compute_units, device_vendor_id;
+ cl_event event_in = NULL, event_out = NULL;
+ cl_mem_flags dovi_buf_flags = CL_MEM_ALLOC_HOST_PTR | CL_MEM_HOST_WRITE_ONLY | CL_MEM_READ_ONLY;
+ cl_uint device_vendor_id;
cl_int cle;
- int err;
- AVBPrint header;
- const char *opencl_sources[OPENCL_SOURCE_NB];
+ cl_mem_flags dovi_buf_flags = CL_MEM_ALLOC_HOST_PTR | CL_MEM_HOST_WRITE_ONLY | CL_MEM_READ_ONLY;
+ char *device_vendor = NULL;
+ char *device_name = NULL;
+ char *device_exts = NULL;
+ int is_device_qualcomm = 0;
+ int is_qcom_proprietary = 0;
+ int i, j, err;
- av_bprint_init(&header, 1024, AV_BPRINT_SIZE_AUTOMATIC);
@@ -2214,54 +2238,120 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
ctx->target_peak = 1.0f;
- av_log(ctx, AV_LOG_DEBUG, "tone mapping transfer from %s to %s\n",
+
+ cle = clGetDeviceInfo(ctx->ocf.hwctx->device_id, CL_DEVICE_VENDOR_ID,
+ sizeof(cl_uint), &device_vendor_id,
+ NULL);
+ cle = clGetDeviceInfo(ctx->ocf.hwctx->device_id,
+ CL_DEVICE_VENDOR_ID,
+ sizeof(cl_uint), &device_vendor_id, NULL);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to check OpenCL "
+ "device vendor id %d.\n", cle);
+
+ if (ctx->tradeoff == -1) {
+ ctx->tradeoff = 1;
+ cle = clGetDeviceInfo(ctx->ocf.hwctx->device_id, CL_DEVICE_HOST_UNIFIED_MEMORY,
+ sizeof(cl_bool), &device_is_integrated,
+ NULL);
+ device_exts = check_opencl_device_str(ctx->ocf.hwctx->device_id,
+ CL_DEVICE_EXTENSIONS);
+
+ ctx->use_image3d = 0;
+ if (ctx->tradeoff) {
+ cl_bool device_is_uma = 0;
+ int is_intel = 0, is_arm = 0, is_qcom = 0;
+ int is_tradeoff_auto = ctx->tradeoff == -1;
+
+ cle = clGetDeviceInfo(ctx->ocf.hwctx->device_id,
+ CL_DEVICE_HOST_UNIFIED_MEMORY,
+ sizeof(cl_bool), &device_is_uma, NULL);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to check if OpenCL "
+ "device is integrated %d.\n", cle);
+ cle = clGetDeviceInfo(ctx->ocf.hwctx->device_id, CL_DEVICE_MAX_COMPUTE_UNITS,
+ sizeof(cl_uint), &max_compute_units,
+ NULL);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to check OpenCL "
+ "device max compute units %d.\n", cle);
+ if (device_vendor_id == 0x8086 && device_is_integrated == CL_TRUE) {
+ "device is UMA %d.\n", cle);
+
+ device_vendor = check_opencl_device_str(ctx->ocf.hwctx->device_id,
+ CL_DEVICE_VENDOR);
+ device_name = check_opencl_device_str(ctx->ocf.hwctx->device_id,
+ CL_DEVICE_NAME);
+
+ is_intel = device_vendor_id == 0x8086;
+ is_arm = device_vendor_id == 0x13b5 ||
+ (device_vendor && strstr(device_vendor, "ARM")) ||
+ (device_name && strstr(device_name, "Mali"));
+ is_qcom = device_vendor_id == 0x5143 ||
+ device_vendor_id == MKTAG('Q', 'C', 'O', 'M');
+
+ ctx->tradeoff = 1;
+ if (is_intel && device_is_uma && is_tradeoff_auto) {
+ // Use tradeoff on low perf Intel iGPUs
+ cl_uint max_compute_units = 0;
+
+ cle = clGetDeviceInfo(ctx->ocf.hwctx->device_id,
+ CL_DEVICE_MAX_COMPUTE_UNITS,
+ sizeof(cl_uint), &max_compute_units, NULL);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to check OpenCL "
+ "device max compute units %d.\n", cle);
+
+ if (max_compute_units >= 40)
+ ctx->tradeoff = 0;
+ if (device_name = check_opencl_device_str(ctx->ocf.hwctx->device_id, CL_DEVICE_NAME)) {
+ else if (device_name) {
+ const char *excluded_devices[5] = { "Arc", "Iris", "Xe", "770", "750" };
+ for (i = 0; i < FF_ARRAY_ELEMS(excluded_devices); i++) {
+ if (strstr(device_name, excluded_devices[i])) {
+ ctx->tradeoff = 0;
+ break;
+ ctx->tradeoff = 0; break;
+ }
+ }
+ av_freep(&device_name);
+ }
+ } else if (device_vendor_id == 0x5143 || device_vendor_id == MKTAG('Q', 'C', 'O', 'M')) {
+ } else if (is_arm && device_is_uma) {
+ // Use tradeoff and check image3d_t support for lut on ARM Mali Valhall+
+ cl_uint nb_formats = 0;
+ cl_image_format *formats = NULL;
+
+ if (!(device_exts && strstr(device_exts, "cl_arm_job_slot_selection"))) {
+ cle = clGetSupportedImageFormats(ctx->ocf.hwctx->context,
+ CL_MEM_READ_ONLY,
+ CL_MEM_OBJECT_IMAGE3D,
+ 0, NULL, &nb_formats);
+ if (cle == CL_SUCCESS && nb_formats > 0) {
+ formats = av_malloc_array(nb_formats, sizeof(*formats));
+ if (!formats) {
+ err = AVERROR(ENOMEM);
+ goto fail;
+ }
+ cle = clGetSupportedImageFormats(ctx->ocf.hwctx->context,
+ CL_MEM_READ_ONLY,
+ CL_MEM_OBJECT_IMAGE3D,
+ nb_formats, formats, NULL);
+ for (i = 0; cle == CL_SUCCESS && i < nb_formats; i++) {
+ if (formats[i].image_channel_order == CL_RGBA &&
+ formats[i].image_channel_data_type == CL_FLOAT) {
+ ctx->use_image3d = 1; break;
+ }
+ }
+ }
+ av_freep(&formats);
+ }
+ if (ctx->use_image3d) {
+ size_t value = 0;
+ cl_device_info params[] = {
+ CL_DEVICE_IMAGE3D_MAX_WIDTH,
+ CL_DEVICE_IMAGE3D_MAX_HEIGHT,
+ CL_DEVICE_IMAGE3D_MAX_DEPTH
+ };
+
+ for (i = 0; i < FF_ARRAY_ELEMS(params); i++) {
+ cle = clGetDeviceInfo(ctx->ocf.hwctx->device_id, params[i],
+ sizeof(value), &value, NULL);
+ if (cle != CL_SUCCESS || value < LUT_SIZE) {
+ ctx->use_image3d = 0; break;
+ }
+ }
+ }
+ if (!ctx->use_image3d)
+ av_log(avctx, AV_LOG_DEBUG,
+ "Disabled image3d for lut due to lack of support.\n");
+ } else if (is_qcom) {
+ // Always use tradeoff on Qualcomm due to inconsistent performance
+ ctx->tradeoff = 1;
+ } else if (device_is_integrated == CL_TRUE) {
+ device_vendor = check_opencl_device_str(ctx->ocf.hwctx->device_id, CL_DEVICE_VENDOR);
+ device_name = check_opencl_device_str(ctx->ocf.hwctx->device_id, CL_DEVICE_NAME);
+ if (!strstr(device_vendor, "ARM") &&
+ !strstr(device_name, "Mali"))
+ ctx->tradeoff = 0;
+ av_freep(&device_vendor);
+ av_freep(&device_name);
+ } else {
+ } else if (is_tradeoff_auto) {
+ ctx->tradeoff = 0;
+ }
+
+ if (!ctx->tradeoff)
+ av_log(avctx, AV_LOG_DEBUG, "Disabled tradeoffs on high performance device.\n");
+ if (is_tradeoff_auto && !ctx->tradeoff)
+ av_log(avctx, AV_LOG_DEBUG,
+ "Disabled tradeoffs on high performance device.\n");
+
+ av_freep(&device_vendor);
+ av_freep(&device_name);
+ }
+
+ // for low perf device, only do reshaping for pure dovi
@@ -2270,17 +2360,19 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ ctx->apply_dovi = 0;
+ }
+
+ device_exts = check_opencl_device_str(ctx->ocf.hwctx->device_id, CL_DEVICE_EXTENSIONS);
+
+ // use FP16 for dovi reshaping only when tradeoff is enabled and it's supported
+ ctx->dovi_use_fp16 = 0;
+ if (ctx->tradeoff && ctx->dovi) {
+ if (device_exts && strstr(device_exts, "cl_khr_fp16")) {
+ ctx->dovi_use_fp16 = 1;
+ av_log(avctx, AV_LOG_DEBUG, "FP16 is enabled for DOVI reshaping.\n");
+ }
+ if (ctx->tradeoff && ctx->dovi && device_exts && strstr(device_exts, "cl_khr_fp16")) {
+ ctx->dovi_use_fp16 = 1;
+ av_log(avctx, AV_LOG_DEBUG, "FP16 is enabled for DOVI reshaping.\n");
+ }
+
+ // zero-copy buffer requires this extension on Intel dGPUs
+ if (device_vendor_id == 0x8086 && device_exts && strstr(device_exts, "cl_intel_mem_force_host_memory"))
+ dovi_buf_flags |= (1 << 20); /* CL_MEM_FORCE_HOST_MEMORY_INTEL */
+
+ av_freep(&device_exts);
+
+ if (device_vendor_id == 0x5143) {
+ // Qualcomm has two device IDs: 0x5143 and 0x4d4f4351 ('Q' | 'C' << 8 | 'O' << 16 | 'M' << 24)
+ // The former is reported by Qualcomm's official OpenCL driver
@@ -2288,21 +2380,10 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ // The former has better performance if the kernel is written in a way its compiler handles properly
+ // The latter one has more predictable performance and compiler behaves a bit more like other GPU
+ // Only use the workaround on Qualcomm native OpenCL driver
+ is_qcom_proprietary = 1;
+ av_log(avctx, AV_LOG_DEBUG, "Qualcomm driver in use, vendor specific workarounds applied.\n");
+ is_device_qualcomm = 1;
+ }
+
+#ifndef CL_MEM_FORCE_HOST_MEMORY_INTEL
+ #define CL_MEM_FORCE_HOST_MEMORY_INTEL (1 << 20)
+#endif
+ // zero-copy buffer requires this extension on Intel dGPUs
+ if (device_vendor_id == 0x8086) {
+ if (device_exts && strstr(device_exts, "cl_intel_mem_force_host_memory"))
+ dovi_buf_flags |= CL_MEM_FORCE_HOST_MEMORY_INTEL;
+ }
+
+ av_freep(&device_exts);
+
+ av_log(ctx, AV_LOG_DEBUG, "Tonemapping transfer from %s to %s\n",
av_color_transfer_name(ctx->trc_in),
av_color_transfer_name(ctx->trc_out));
@@ -2319,7 +2400,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ av_log(ctx, AV_LOG_DEBUG, "Mapping range from %s to %s\n",
av_color_range_name(ctx->range_in),
av_color_range_name(ctx->range_out));
- // checking valid value just because of limited implementaion
- // checking valid value just because of limited implementation
- // please remove when more functionalities are implemented
+
av_assert0(ctx->trc_out == AVCOL_TRC_BT709 ||
@@ -2353,9 +2434,13 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+
+ av_bprint_init(&header, 2048, AV_BPRINT_SIZE_UNLIMITED);
+
+ if (is_device_qualcomm)
+ if (is_qcom_proprietary)
+ av_bprintf(&header, "#define IS_QCOM_GPU\n");
+
+ av_bprintf(&header, "#define LUT_SIZE %d\n", LUT_SIZE);
+ if (ctx->use_image3d)
+ av_bprintf(&header, "#define LUT_PERF_IMAGE3D\n");
+
+ av_bprintf(&header, "__constant float tone_param = %.4ff;\n",
+ ctx->final_param);
+ av_bprintf(&header, "__constant float desat_param = %.4ff;\n",
@@ -2437,7 +2522,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
av_bprintf(&header, "#define chroma_loc %d\n", (int)ctx->chroma_loc);
if (rgb2rgb_passthrough)
@@ -199,19 +627,41 @@ static int tonemap_opencl_init(AVFilterC
@@ -200,19 +689,44 @@ static int tonemap_opencl_init(AVFilterC
else
ff_opencl_print_const_matrix_3x3(&header, "rgb2rgb", rgb2rgb);
@@ -2451,6 +2536,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
- ctx->colorspace_in, av_color_space_name(ctx->colorspace_in));
- goto fail;
+ if (ctx->dovi) {
+ const size_t buf_sz = 3*(params_sz+pivots_sz+coeffs_sz+mmr_sz) >> !!ctx->dovi_use_fp16;
+ double ycc2rgb_offset[3] = {0};
+ double lms2rgb[3][3];
+ av_bprintf(&header, "#define DOVI_RESHAPE\n");
@@ -2465,6 +2551,8 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ ff_matrix_mul_3x3(lms2rgb, dovi_lms2rgb_matrix, ctx->dovi->linear);
+ ff_opencl_print_const_matrix_3x3(&header, "rgb_matrix", ctx->dovi->nonlinear); //ycc2rgb
+ ff_opencl_print_const_matrix_3x3(&header, "lms2rgb_matrix", lms2rgb); //lms2rgb
+
+ CL_CREATE_BUFFER_FLAGS(ctx, dovi_buf, dovi_buf_flags, buf_sz, NULL);
+ } else {
+ luma_src = av_csp_luma_coeffs_from_avcsp(ctx->colorspace_in);
+ if (!luma_src) {
@@ -2486,7 +2574,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
ctx->colorspace_out, av_color_space_name(ctx->colorspace_out));
goto fail;
}
@@ -219,24 +669,13 @@ static int tonemap_opencl_init(AVFilterC
@@ -220,24 +734,13 @@ static int tonemap_opencl_init(AVFilterC
ff_fill_rgb2yuv_table(luma_dst, rgb2yuv);
ff_opencl_print_const_matrix_3x3(&header, "yuv_matrix", rgb2yuv);
@@ -2516,13 +2604,17 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
av_log(avctx, AV_LOG_DEBUG, "Generated OpenCL header:\n%s\n", header.str);
opencl_sources[0] = header.str;
@@ -254,46 +693,231 @@ static int tonemap_opencl_init(AVFilterC
@@ -255,50 +758,291 @@ static int tonemap_opencl_init(AVFilterC
CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to create OpenCL "
"command queue %d.\n", cle);
- ctx->kernel = clCreateKernel(ctx->ocf.program, "tonemap", &cle);
- CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to create kernel %d.\n", cle);
+ if (ctx->in_desc->comp[0].depth > ctx->out_desc->comp[0].depth) {
+ const size_t m_origin[3] = { 0 };
+ const size_t m_region[3] = { ff_fruit_dither_size, ff_fruit_dither_size, 1 };
+ const size_t m_row_pitch = ff_fruit_dither_size * sizeof(ff_fruit_dither_matrix[0]);
+
+ cl_image_format image_format = {
+ .image_channel_data_type = CL_UNORM_INT16,
+ .image_channel_order = CL_R,
@@ -2534,7 +2626,8 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ .image_row_pitch = 0,
+ };
+
+ av_assert0(sizeof(ff_fruit_dither_matrix) == sizeof(ff_fruit_dither_matrix[0]) * ff_fruit_dither_size * ff_fruit_dither_size);
+ av_assert0(sizeof(ff_fruit_dither_matrix) ==
+ sizeof(ff_fruit_dither_matrix[0]) * ff_fruit_dither_size * ff_fruit_dither_size);
+
+ ctx->dither_image = clCreateImage(ctx->ocf.hwctx->context, CL_MEM_READ_ONLY,
+ &image_format, &image_desc, NULL, &cle);
@@ -2550,53 +2643,86 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ CL_FALSE, m_origin, m_region,
+ m_row_pitch, 0,
+ ff_fruit_dither_matrix,
+ 0, NULL, &event);
+ 0, NULL, &event_out);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to enqueue write of dither matrix image: %d.\n", cle);
+
+ cle = clWaitForEvents(1, &event);
+ cle = clWaitForEvents(1, &event_out);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to wait for event completion: %d.\n", cle);
+ if (event) {
+ clReleaseEvent(event);
+ event = NULL;
+ if (event_out) {
+ clReleaseEvent(event_out);
+ event_out = NULL;
+ }
+ }
+
+ if (ctx->tradeoff) {
+ const size_t lut_size = LUT_SIZE;
+ const size_t lut_buffer_size = lut_size * sizeof(cl_float3);
+ float peak = (float)ctx->src_peak;
+
+ ctx->lut_generation_kernel = clCreateKernel(ctx->ocf.program, "build_lut", &cle);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to create kernel %d.\n", cle);
+
+ CL_CREATE_BUFFER_FLAGS(ctx, lut_buffer, CL_MEM_HOST_NO_ACCESS | CL_MEM_READ_WRITE, lut_buffer_size, NULL);
+ CL_SET_KERNEL_ARG(ctx->lut_generation_kernel, 0, cl_mem, &ctx->lut_buffer);
+ CL_SET_KERNEL_ARG(ctx->lut_generation_kernel, 1, cl_float, &peak);
+
+ cle = clEnqueueNDRangeKernel(ctx->command_queue, ctx->lut_generation_kernel, 1, NULL,
+ &lut_size, NULL,
+ 0, NULL, NULL);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to enqueue build_lut kernel: %d.\n", cle);
+
+ cle = clFinish(ctx->command_queue);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to finish command queue: %d.\n", cle);
+
+ ctx->kernel = clCreateKernel(ctx->ocf.program, "tonemap_lut", &cle);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to create kernel %d.\n", cle);
+ } else {
+ ctx->kernel = clCreateKernel(ctx->ocf.program, "tonemap", &cle);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to create kernel %d.\n", cle);
+ }
- ctx->util_mem =
- clCreateBuffer(ctx->ocf.hwctx->context, 0,
- (2 * DETECTION_FRAMES + 7) * sizeof(unsigned),
- NULL, &cle);
- CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to create util buffer: %d.\n", cle);
+ if (ctx->dovi) {
+ const size_t buf_sz = 3*(params_sz+pivots_sz+coeffs_sz+mmr_sz) >> !!ctx->dovi_use_fp16;
+ CL_CREATE_BUFFER_FLAGS(ctx, dovi_buf, dovi_buf_flags, buf_sz, NULL);
+ if (ctx->tradeoff) {
+ const size_t lut_size = LUT_SIZE;
+ const size_t lut_size_3d = lut_size * lut_size * lut_size;
+ const size_t lut_buffer_size = lut_size_3d * sizeof(cl_float4);
+ const float peak = (float)ctx->src_peak;
+ const size_t m_origin[3] = { 0 };
+ const size_t m_region[3] = { lut_size, lut_size, lut_size };
+ cl_mem_flags mem_flags = CL_MEM_HOST_NO_ACCESS;
+
+ cl_image_format image_format = {
+ .image_channel_order = CL_RGBA,
+ .image_channel_data_type = CL_FLOAT,
+ };
+ cl_image_desc image_desc = {
+ .image_type = CL_MEM_OBJECT_IMAGE3D,
+ .image_width = lut_size,
+ .image_height = lut_size,
+ .image_depth = lut_size,
+ .image_row_pitch = 0,
+ .image_slice_pitch = 0,
+ };
+
+ ctx->lut_generation_kernel = clCreateKernel(ctx->ocf.program, "build_lut", &cle);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to create kernel %d.\n", cle);
+
+ CL_CREATE_BUFFER_FLAGS(ctx, lut_buffer, CL_MEM_READ_WRITE | mem_flags, lut_buffer_size, NULL);
+
+ CL_SET_KERNEL_ARG(ctx->lut_generation_kernel, 0, cl_mem, &ctx->lut_buffer);
+ CL_SET_KERNEL_ARG(ctx->lut_generation_kernel, 1, cl_float, &peak);
+ cle = clEnqueueNDRangeKernel(ctx->command_queue, ctx->lut_generation_kernel, 1, NULL,
+ &lut_size_3d, NULL,
+ 0, NULL, ctx->use_image3d ? &event_in : &event_out);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to enqueue build_lut kernel: %d.\n", cle);
+
+ if (ctx->use_image3d) {
+ ctx->lut_image = clCreateImage(ctx->ocf.hwctx->context, CL_MEM_READ_ONLY | mem_flags,
+ &image_format, &image_desc, NULL, &err);
+ if (!ctx->lut_image) {
+ av_log(avctx, AV_LOG_ERROR, "Failed to create image for "
+ "lut image: %d.\n", cle);
+ err = AVERROR(EIO);
+ goto fail;
+ }
+ cle = clEnqueueCopyBufferToImage(ctx->command_queue,
+ ctx->lut_buffer, ctx->lut_image,
+ 0, m_origin, m_region,
+ 1, &event_in, &event_out);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to enqueue copy of lut buffer to image: %d.\n", cle);
+ }
+
+ cle = clWaitForEvents(1, &event_out);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to wait for event completion: %d.\n", cle);
+ if (event_in) {
+ clReleaseEvent(event_in);
+ event_in = NULL;
+ }
+ if (event_out) {
+ clReleaseEvent(event_out);
+ event_out = NULL;
+ }
+ }
+
+ ctx->kernel = clCreateKernel(ctx->ocf.program, ctx->tradeoff ? "tonemap_lut" : "tonemap", &cle);
+ CL_FAIL_ON_ERROR(AVERROR(EIO), "Failed to create kernel %d.\n", cle);
ctx->initialised = 1;
return 0;
@@ -2610,8 +2736,10 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ av_freep(&device_vendor);
+ av_freep(&device_name);
+ av_freep(&device_exts);
+ if (event)
+ clReleaseEvent(event);
+ if (event_in)
+ clReleaseEvent(event_in);
+ if (event_out)
+ clReleaseEvent(event_out);
if (ctx->kernel)
clReleaseKernel(ctx->kernel);
+ if (ctx->lut_generation_kernel)
@@ -2620,6 +2748,8 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ clReleaseMemObject(ctx->dither_image);
+ if (ctx->lut_buffer)
+ clReleaseMemObject(ctx->lut_buffer);
+ if (ctx->lut_image)
+ clReleaseMemObject(ctx->lut_image);
+ if (ctx->dovi_buf)
+ clReleaseMemObject(ctx->dovi_buf);
+ if (ctx->command_queue)
@@ -2686,6 +2816,13 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ "lut buffer: %d.\n", cle);
+ }
+
+ if (ctx->lut_image) {
+ cle = clReleaseMemObject(ctx->lut_image);
+ if (cle != CL_SUCCESS)
+ av_log(avctx, AV_LOG_ERROR, "Failed to release "
+ "lut image: %d.\n", cle);
+ }
+
+ if (ctx->command_queue) {
+ cle = clReleaseCommandQueue(ctx->command_queue);
+ if (cle != CL_SUCCESS)
@@ -2751,13 +2888,14 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ av_log(ctx, AV_LOG_ERROR, "Unsupported output format: %s\n",
+ av_get_pix_fmt_name(out_format));
+ return AVERROR(ENOSYS);
}
+ }
+ if (in_desc->comp[0].depth != 10 && in_desc->comp[0].depth != 16) {
+ av_log(ctx, AV_LOG_ERROR, "Unsupported input format depth: %d\n",
+ in_desc->comp[0].depth);
+ return AVERROR(ENOSYS);
+ }
+
}
- s->ocf.output_format = s->format == AV_PIX_FMT_NONE ? AV_PIX_FMT_NV12 : s->format;
+ ctx->in_fmt = in_format;
+ ctx->out_fmt = out_format;
+ ctx->in_desc = in_desc;
@@ -2765,12 +2903,22 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ ctx->in_planes = av_pix_fmt_count_planes(in_format);
+ ctx->out_planes = av_pix_fmt_count_planes(out_format);
+ ctx->ocf.output_format = out_format;
- s->ocf.output_format = s->format == AV_PIX_FMT_NONE ? AV_PIX_FMT_NV12 : s->format;
+
ret = ff_opencl_filter_config_output(outlink);
if (ret < 0)
return ret;
@@ -308,13 +932,49 @@ static int launch_kernel(AVFilterContext
+ if (ctx->trc != AVCOL_TRC_SMPTE2084) {
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
+
return 0;
}
@@ -309,13 +1053,50 @@ static int launch_kernel(AVFilterContext
size_t global_work[2];
size_t local_work[2];
cl_int cle;
@@ -2785,6 +2933,11 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ err = AVERROR(EIO);
+ goto fail;
+ }
+
+ if (ctx->in_planes > 2 && !input->data[2]) {
+ err = AVERROR(EIO);
+ goto fail;
+ }
- CL_SET_KERNEL_ARG(kernel, 0, cl_mem, &output->data[0]);
- CL_SET_KERNEL_ARG(kernel, 1, cl_mem, &input->data[0]);
@@ -2792,14 +2945,10 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
- CL_SET_KERNEL_ARG(kernel, 3, cl_mem, &input->data[1]);
- CL_SET_KERNEL_ARG(kernel, 4, cl_mem, &ctx->util_mem);
- CL_SET_KERNEL_ARG(kernel, 5, cl_float, &peak);
+ if (ctx->in_planes > 2 && !input->data[2]) {
+ err = AVERROR(EIO);
+ goto fail;
+ }
+
+ idx_arg = 0;
+ if (ctx->tradeoff) {
+ CL_SET_KERNEL_ARG(kernel, idx_arg++, cl_mem, &ctx->lut_buffer);
+ CL_SET_KERNEL_ARG(kernel, idx_arg++, cl_mem,
+ ctx->use_image3d ? &ctx->lut_image : &ctx->lut_buffer);
+ }
+ CL_SET_KERNEL_ARG(kernel, idx_arg++, cl_mem, &output->data[0]);
+ CL_SET_KERNEL_ARG(kernel, idx_arg++, cl_mem, &input->data[0]);
@@ -2826,7 +2975,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
local_work[0] = 16;
local_work[1] = 16;
@@ -338,12 +998,10 @@ static int tonemap_opencl_filter_frame(A
@@ -339,12 +1120,10 @@ static int tonemap_opencl_filter_frame(A
AVFilterContext *avctx = inlink->dst;
AVFilterLink *outlink = avctx->outputs[0];
TonemapOpenCLContext *ctx = avctx->priv;
@@ -2840,7 +2989,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
av_log(ctx, AV_LOG_DEBUG, "Filter input: %s, %ux%u (%"PRId64").\n",
av_get_pix_fmt_name(input->format),
@@ -351,7 +1009,6 @@ static int tonemap_opencl_filter_frame(A
@@ -352,7 +1131,6 @@ static int tonemap_opencl_filter_frame(A
if (!input->hw_frames_ctx)
return AVERROR(EINVAL);
@@ -2848,7 +2997,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
output = ff_get_video_buffer(outlink, outlink->w, outlink->h);
if (!output) {
@@ -363,17 +1020,65 @@ static int tonemap_opencl_filter_frame(A
@@ -364,17 +1142,65 @@ static int tonemap_opencl_filter_frame(A
if (err < 0)
goto fail;
@@ -2890,7 +3039,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
+ const AVDOVIRpuDataHeader *rpu = av_dovi_get_header(metadata);
+ // only map dovi rpus that don't require an EL
+ if (rpu->disable_residual_flag) {
+ struct DoviMetadata *dovi = av_malloc(sizeof(*dovi));
+ struct FFDOVIMetadataRemap *dovi = av_malloc(sizeof(*dovi));
+ ctx->dovi = dovi;
+ if (!ctx->dovi)
+ goto fail;
@@ -2920,7 +3069,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
ctx->trc_in = input->color_trc;
ctx->trc_out = output->color_trc;
@@ -385,72 +1090,50 @@ static int tonemap_opencl_filter_frame(A
@@ -386,72 +1212,50 @@ static int tonemap_opencl_filter_frame(A
ctx->range_out = output->color_range;
ctx->chroma_loc = output->chroma_location;
@@ -3016,7 +3165,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
av_frame_free(&input);
av_frame_free(&output);
return err;
@@ -458,62 +1141,101 @@ fail:
@@ -459,62 +1263,101 @@ fail:
static av_cold void tonemap_opencl_uninit(AVFilterContext *avctx)
{
@@ -3165,10 +3314,14 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
{ NULL }
};
@@ -541,11 +1263,12 @@ const AVFilter ff_vf_tonemap_opencl = {
.description = NULL_IF_CONFIG_SMALL("Perform HDR to SDR conversion with tonemapping."),
@@ -540,13 +1383,14 @@ static const AVFilterPad tonemap_opencl_
const FFFilter ff_vf_tonemap_opencl = {
.p.name = "tonemap_opencl",
.p.description = NULL_IF_CONFIG_SMALL("Perform HDR to SDR conversion with tonemapping."),
- .p.priv_class = &tonemap_opencl_class,
- .p.flags = AVFILTER_FLAG_HWDEVICE,
.priv_size = sizeof(TonemapOpenCLContext),
.priv_class = &tonemap_opencl_class,
+ .p.priv_class = &tonemap_opencl_class,
+ .preinit = &tonemap_opencl_preinit,
.init = &ff_opencl_filter_init,
.uninit = &tonemap_opencl_uninit,
@@ -3177,5 +3330,5 @@ Index: FFmpeg/libavfilter/vf_tonemap_opencl.c
- FILTER_SINGLE_PIXFMT(AV_PIX_FMT_OPENCL),
+ FILTER_QUERY_FUNC(tonemap_opencl_query_formats),
.flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
.flags = AVFILTER_FLAG_HWDEVICE,
+ .p.flags = AVFILTER_FLAG_HWDEVICE,
};
debian/patches/0008-add-pgs-support-to-opencl-overlay.patch
Vendored
+20 -3
View File
@@ -321,7 +321,8 @@ Index: FFmpeg/libavfilter/vf_overlay_opencl.c
- av_log(avctx, AV_LOG_DEBUG, "Using kernel %s.\n", kernel);
+ av_bprint_init(&header, 2048, AV_BPRINT_SIZE_UNLIMITED);
+
- err = ff_opencl_filter_load_program(avctx, &source, 1);
+ if (ctx->alpha && ctx->alpha_format == 1)
+ av_bprintf(&header, "#define NEED_UNPREMUL\n");
+
@@ -329,8 +330,7 @@ Index: FFmpeg/libavfilter/vf_overlay_opencl.c
+ opencl_sources[0] = header.str;
+ opencl_sources[1] = ff_source_overlay_cl;
+ err = ff_opencl_filter_load_program(avctx, opencl_sources, OPENCL_SOURCE_NB);
- err = ff_opencl_filter_load_program(avctx, &source, 1);
+
+ av_bprint_finalize(&header, NULL);
if (err < 0)
goto fail;
@@ -685,3 +685,20 @@ Index: FFmpeg/libavfilter/vf_overlay_opencl.c
{ NULL },
};
@@ -313,9 +514,8 @@ static const AVFilterPad overlay_opencl_
const FFFilter ff_vf_overlay_opencl = {
.p.name = "overlay_opencl",
.p.description = NULL_IF_CONFIG_SMALL("Overlay one video on top of another"),
- .p.priv_class = &overlay_opencl_class,
- .p.flags = AVFILTER_FLAG_HWDEVICE,
.priv_size = sizeof(OverlayOpenCLContext),
+ .p.priv_class = &overlay_opencl_class,
.init = &overlay_opencl_init,
.uninit = &overlay_opencl_uninit,
.activate = &overlay_opencl_activate,
@@ -323,4 +523,5 @@ const FFFilter ff_vf_overlay_opencl = {
FILTER_OUTPUTS(overlay_opencl_outputs),
FILTER_SINGLE_PIXFMT(AV_PIX_FMT_OPENCL),
.flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+ .p.flags = AVFILTER_FLAG_HWDEVICE,
};
debian/patches/0010-add-a-hack-for-opencl-reverse-mapping.patch
Vendored
+42 -10
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/libavfilter/avfilter.h
===================================================================
--- FFmpeg.orig/libavfilter/avfilter.h
+++ FFmpeg/libavfilter/avfilter.h
@@ -579,6 +579,8 @@ struct AVFilterLink {
@@ -417,6 +417,8 @@ struct AVFilterLink {
enum AVColorSpace colorspace; ///< agreed upon YUV color space
enum AVColorRange color_range; ///< agreed upon YUV color range
@@ -15,7 +15,7 @@ Index: FFmpeg/libavfilter/opencl.c
===================================================================
--- FFmpeg.orig/libavfilter/opencl.c
+++ FFmpeg/libavfilter/opencl.c
@@ -77,6 +77,9 @@ int ff_opencl_filter_config_input(AVFilt
@@ -78,6 +78,9 @@ int ff_opencl_filter_config_input(AVFilt
if (!ctx->output_height)
ctx->output_height = inlink->h;
@@ -25,7 +25,7 @@ Index: FFmpeg/libavfilter/opencl.c
return 0;
}
@@ -126,6 +129,9 @@ int ff_opencl_filter_config_output(AVFil
@@ -127,6 +130,9 @@ int ff_opencl_filter_config_output(AVFil
outlink->w = ctx->output_width;
outlink->h = ctx->output_height;
@@ -35,6 +35,29 @@ Index: FFmpeg/libavfilter/opencl.c
return 0;
fail:
av_buffer_unref(&output_frames_ref);
Index: FFmpeg/libavfilter/vf_crop.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_crop.c
+++ FFmpeg/libavfilter/vf_crop.c
@@ -152,6 +152,8 @@ static int config_input(AVFilterLink *li
if (pix_desc->flags & AV_PIX_FMT_FLAG_HWACCEL) {
s->hsub = 1;
s->vsub = 1;
+ if (ctx->nb_outputs > 0)
+ ctx->outputs[0]->fixed_pool_size = link->fixed_pool_size;
} else {
s->hsub = pix_desc->log2_chroma_w;
s->vsub = pix_desc->log2_chroma_h;
@@ -238,6 +240,9 @@ static int config_output(AVFilterLink *l
if (desc->flags & AV_PIX_FMT_FLAG_HWACCEL) {
// Hardware frames adjust the cropping regions rather than
// changing the frame size.
+ AVFilterContext *ctx = link->src;
+ if (ctx->nb_inputs > 0)
+ link->fixed_pool_size = ctx->inputs[0]->fixed_pool_size;
} else {
link->w = s->w;
link->h = s->h;
Index: FFmpeg/libavfilter/vf_hwmap.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_hwmap.c
@@ -50,7 +73,7 @@ Index: FFmpeg/libavfilter/vf_hwmap.c
#include "avfilter.h"
#include "filters.h"
#include "formats.h"
@@ -124,6 +128,12 @@ static int hwmap_config_output(AVFilterL
@@ -126,6 +130,12 @@ static int hwmap_config_output(AVFilterL
goto fail;
}
@@ -63,7 +86,7 @@ Index: FFmpeg/libavfilter/vf_hwmap.c
} else if (inlink->format == hwfc->format &&
(desc->flags & AV_PIX_FMT_FLAG_HWACCEL) &&
ctx->reverse) {
@@ -133,6 +143,9 @@ static int hwmap_config_output(AVFilterL
@@ -135,6 +145,9 @@ static int hwmap_config_output(AVFilterL
// mapped from that back to the source type.
AVBufferRef *source;
AVHWFramesContext *frames;
@@ -73,7 +96,7 @@ Index: FFmpeg/libavfilter/vf_hwmap.c
ctx->hwframes_ref = av_hwframe_ctx_alloc(device);
if (!ctx->hwframes_ref) {
@@ -146,8 +159,19 @@ static int hwmap_config_output(AVFilterL
@@ -148,8 +161,19 @@ static int hwmap_config_output(AVFilterL
frames->width = hwfc->width;
frames->height = hwfc->height;
@@ -110,7 +133,7 @@ Index: FFmpeg/libavfilter/vf_hwupload.c
#include "avfilter.h"
#include "filters.h"
#include "formats.h"
@@ -112,6 +116,9 @@ static int hwupload_config_output(AVFilt
@@ -119,6 +123,9 @@ static int hwupload_config_output(AVFilt
FilterLink *inl = ff_filter_link(inlink);
HWUploadContext *ctx = avctx->priv;
int err;
@@ -120,7 +143,7 @@ Index: FFmpeg/libavfilter/vf_hwupload.c
av_buffer_unref(&ctx->hwframes_ref);
@@ -153,6 +160,11 @@ static int hwupload_config_output(AVFilt
@@ -160,6 +167,11 @@ static int hwupload_config_output(AVFilt
if (avctx->extra_hw_frames >= 0)
ctx->hwframes->initial_pool_size = 2 + avctx->extra_hw_frames;
@@ -136,7 +159,7 @@ Index: FFmpeg/libavutil/hwcontext_d3d11va.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_d3d11va.c
+++ FFmpeg/libavutil/hwcontext_d3d11va.c
@@ -229,7 +229,7 @@ static AVBufferRef *d3d11va_alloc_single
@@ -235,7 +235,7 @@ static AVBufferRef *d3d11va_alloc_single
.ArraySize = 1,
.Usage = D3D11_USAGE_DEFAULT,
.BindFlags = hwctx->BindFlags,
@@ -145,7 +168,7 @@ Index: FFmpeg/libavutil/hwcontext_d3d11va.c
};
hr = ID3D11Device_CreateTexture2D(device_hwctx->device, &texDesc, NULL, &tex);
@@ -293,9 +293,17 @@ static int d3d11va_frames_init(AVHWFrame
@@ -304,9 +304,17 @@ static int d3d11va_frames_init(AVHWFrame
.ArraySize = ctx->initial_pool_size,
.Usage = D3D11_USAGE_DEFAULT,
.BindFlags = hwctx->BindFlags,
@@ -164,3 +187,12 @@ Index: FFmpeg/libavutil/hwcontext_d3d11va.c
if (hwctx->texture) {
D3D11_TEXTURE2D_DESC texDesc2;
ID3D11Texture2D_GetDesc(hwctx->texture, &texDesc2);
@@ -321,7 +329,7 @@ static int d3d11va_frames_init(AVHWFrame
ctx->initial_pool_size = texDesc2.ArraySize;
hwctx->BindFlags = texDesc2.BindFlags;
hwctx->MiscFlags = texDesc2.MiscFlags;
- } else if (texDesc.ArraySize > 0) {
+ } else if (!(texDesc.BindFlags & D3D11_BIND_RENDER_TARGET) && texDesc.ArraySize > 0) {
hr = ID3D11Device_CreateTexture2D(device_hwctx->device, &texDesc, NULL, &hwctx->texture);
if (FAILED(hr)) {
av_log(ctx, AV_LOG_ERROR, "Could not create the texture (%lx)\n", (long)hr);
debian/patches/0011-add-fixes-for-mapping-from-qsv-source-device.patch
Vendored
+14 -14
View File
@@ -2,8 +2,8 @@ Index: FFmpeg/libavutil/hwcontext.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext.c
+++ FFmpeg/libavutil/hwcontext.c
@@ -84,21 +84,6 @@ static const char *const hw_type_names[]
[AV_HWDEVICE_TYPE_VULKAN] = "vulkan",
@@ -92,21 +92,6 @@ static const char *const hw_type_names[]
[AV_HWDEVICE_TYPE_OHCODEC] = "ohcodec",
};
-typedef struct FFHWDeviceContext {
@@ -24,7 +24,7 @@ Index: FFmpeg/libavutil/hwcontext.c
enum AVHWDeviceType av_hwdevice_find_type_by_name(const char *name)
{
int type;
@@ -143,6 +128,7 @@ static void hwdevice_ctx_free(void *opaq
@@ -158,6 +143,7 @@ static void hwdevice_ctx_free(void *opaq
{
FFHWDeviceContext *ctxi = (FFHWDeviceContext*)data;
AVHWDeviceContext *ctx = &ctxi->p;
@@ -32,7 +32,7 @@ Index: FFmpeg/libavutil/hwcontext.c
/* uninit might still want access the hw context and the user
* free() callback might destroy it, so uninit has to be called first */
@@ -153,6 +139,8 @@ static void hwdevice_ctx_free(void *opaq
@@ -168,6 +154,8 @@ static void hwdevice_ctx_free(void *opaq
ctx->free(ctx);
av_buffer_unref(&ctxi->source_device);
@@ -41,7 +41,7 @@ Index: FFmpeg/libavutil/hwcontext.c
av_freep(&ctx->hwctx);
av_freep(&ctx);
@@ -633,6 +621,28 @@ fail:
@@ -648,6 +636,28 @@ fail:
return ret;
}
@@ -70,7 +70,7 @@ Index: FFmpeg/libavutil/hwcontext.c
int av_hwdevice_ctx_create_derived_opts(AVBufferRef **dst_ref_ptr,
enum AVHWDeviceType type,
AVBufferRef *src_ref,
@@ -656,6 +666,16 @@ int av_hwdevice_ctx_create_derived_opts(
@@ -671,6 +681,16 @@ int av_hwdevice_ctx_create_derived_opts(
tmp_ref = tmp_ctx->source_device;
}
@@ -87,7 +87,7 @@ Index: FFmpeg/libavutil/hwcontext.c
dst_ref = av_hwdevice_ctx_alloc(type);
if (!dst_ref) {
ret = AVERROR(ENOMEM);
@@ -676,6 +696,11 @@ int av_hwdevice_ctx_create_derived_opts(
@@ -691,6 +711,11 @@ int av_hwdevice_ctx_create_derived_opts(
ret = AVERROR(ENOMEM);
goto fail;
}
@@ -103,10 +103,10 @@ Index: FFmpeg/libavutil/hwcontext.h
===================================================================
--- FFmpeg.orig/libavutil/hwcontext.h
+++ FFmpeg/libavutil/hwcontext.h
@@ -38,6 +38,7 @@ enum AVHWDeviceType {
AV_HWDEVICE_TYPE_MEDIACODEC,
AV_HWDEVICE_TYPE_VULKAN,
AV_HWDEVICE_TYPE_D3D12VA,
@@ -41,6 +41,7 @@ enum AVHWDeviceType {
AV_HWDEVICE_TYPE_AMF,
/* OpenHarmony Codec device */
AV_HWDEVICE_TYPE_OHCODEC,
+ AV_HWDEVICE_TYPE_NB, ///< number of hw device types, not part of API/ABI.
};
@@ -115,9 +115,9 @@ Index: FFmpeg/libavutil/hwcontext_internal.h
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_internal.h
+++ FFmpeg/libavutil/hwcontext_internal.h
@@ -164,4 +164,25 @@ extern const HWContextType ff_hwcontext_
extern const HWContextType ff_hwcontext_type_mediacodec;
extern const HWContextType ff_hwcontext_type_vulkan;
@@ -166,4 +166,25 @@ extern const HWContextType ff_hwcontext_
extern const HWContextType ff_hwcontext_type_amf;
extern const HWContextType ff_hwcontext_type_oh;
+typedef struct FFHWDeviceContext {
+ /**
debian/patches/0014-add-vaapi-hwupload-filter.patch
Vendored
+27 -25
View File
@@ -2,8 +2,8 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3884,6 +3884,7 @@ gblur_vulkan_filter_deps="vulkan spirv_c
hflip_vulkan_filter_deps="vulkan spirv_compiler"
@@ -4143,6 +4143,7 @@ gblur_vulkan_filter_deps="vulkan spirv_l
hflip_vulkan_filter_deps="vulkan spirv_library"
histeq_filter_deps="gpl"
hqdn3d_filter_deps="gpl"
+hwupload_vaapi_filter_deps="vaapi"
@@ -14,7 +14,7 @@ Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -347,6 +347,7 @@ OBJS-$(CONFIG_HUESATURATION_FILTER)
@@ -355,6 +355,7 @@ OBJS-$(CONFIG_HUESATURATION_FILTER)
OBJS-$(CONFIG_HWDOWNLOAD_FILTER) += vf_hwdownload.o
OBJS-$(CONFIG_HWMAP_FILTER) += vf_hwmap.o
OBJS-$(CONFIG_HWUPLOAD_CUDA_FILTER) += vf_hwupload_cuda.o
@@ -26,19 +26,19 @@ Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -324,6 +324,7 @@ extern const AVFilter ff_vf_hwdownload;
extern const AVFilter ff_vf_hwmap;
extern const AVFilter ff_vf_hwupload;
extern const AVFilter ff_vf_hwupload_cuda;
+extern const AVFilter ff_vf_hwupload_vaapi;
extern const AVFilter ff_vf_hysteresis;
extern const AVFilter ff_vf_iccdetect;
extern const AVFilter ff_vf_iccgen;
@@ -330,6 +330,7 @@ extern const FFFilter ff_vf_hwdownload;
extern const FFFilter ff_vf_hwmap;
extern const FFFilter ff_vf_hwupload;
extern const FFFilter ff_vf_hwupload_cuda;
+extern const FFFilter ff_vf_hwupload_vaapi;
extern const FFFilter ff_vf_hysteresis;
extern const FFFilter ff_vf_iccdetect;
extern const FFFilter ff_vf_iccgen;
Index: FFmpeg/libavfilter/vf_hwupload_vaapi.c
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_hwupload_vaapi.c
@@ -0,0 +1,195 @@
@@ -0,0 +1,197 @@
+/*
+ * This file is part of FFmpeg.
+ *
@@ -89,7 +89,9 @@ Index: FFmpeg/libavfilter/vf_hwupload_vaapi.c
+ av_buffer_unref(&s->hwdevice);
+}
+
+static int vaapiupload_query_formats(AVFilterContext *ctx)
+static int vaapiupload_query_formats(const AVFilterContext *ctx,
+ AVFilterFormatsConfig **cfg_in,
+ AVFilterFormatsConfig **cfg_out)
+{
+ int ret;
+
@@ -105,16 +107,16 @@ Index: FFmpeg/libavfilter/vf_hwupload_vaapi.c
+ static const enum AVPixelFormat output_pix_fmts[] = {
+ AV_PIX_FMT_VAAPI, AV_PIX_FMT_NONE,
+ };
+ AVFilterFormats *in_fmts = ff_make_format_list(input_pix_fmts);
+ AVFilterFormats *in_fmts = ff_make_pixel_format_list(input_pix_fmts);
+ AVFilterFormats *out_fmts;
+
+ ret = ff_formats_ref(in_fmts, &ctx->inputs[0]->outcfg.formats);
+ ret = ff_formats_ref(in_fmts, &cfg_in[0]->formats);
+ if (ret < 0)
+ return ret;
+
+ out_fmts = ff_make_format_list(output_pix_fmts);
+ out_fmts = ff_make_pixel_format_list(output_pix_fmts);
+
+ ret = ff_formats_ref(out_fmts, &ctx->outputs[0]->incfg.formats);
+ ret = ff_formats_ref(out_fmts, &cfg_out[0]->formats);
+ if (ret < 0)
+ return ret;
+
@@ -218,19 +220,19 @@ Index: FFmpeg/libavfilter/vf_hwupload_vaapi.c
+ },
+};
+
+const AVFilter ff_vf_hwupload_vaapi = {
+ .name = "hwupload_vaapi",
+ .description = NULL_IF_CONFIG_SMALL("Upload a system memory frame to a VAAPI device."),
+const FFFilter ff_vf_hwupload_vaapi = {
+ .p.name = "hwupload_vaapi",
+ .p.description = NULL_IF_CONFIG_SMALL("Upload a system memory frame to a VAAPI device."),
+
+ .init = vaapiupload_init,
+ .uninit = vaapiupload_uninit,
+ .init = vaapiupload_init,
+ .uninit = vaapiupload_uninit,
+
+ .priv_size = sizeof(VaapiUploadContext),
+ .priv_class = &vaapiupload_class,
+ .priv_size = sizeof(VaapiUploadContext),
+ .p.priv_class = &vaapiupload_class,
+
+ FILTER_INPUTS(vaapiupload_inputs),
+ FILTER_OUTPUTS(vaapiupload_outputs),
+ FILTER_QUERY_FUNC(vaapiupload_query_formats),
+ FILTER_QUERY_FUNC2(vaapiupload_query_formats),
+
+ .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+};
debian/patches/0015-add-alpha-format-option-to-vaapi-overlay-filter.patch
Vendored
-43
View File
@@ -1,43 +0,0 @@
Index: FFmpeg/libavfilter/vf_overlay_vaapi.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_overlay_vaapi.c
+++ FFmpeg/libavfilter/vf_overlay_vaapi.c
@@ -53,6 +53,7 @@ typedef struct OverlayVAAPIContext {
int ow;
int oh;
float alpha;
+ int alpha_format;
unsigned int blend_flags;
float blend_alpha;
} OverlayVAAPIContext;
@@ -289,6 +290,7 @@ static int overlay_vaapi_config_input_ma
static int overlay_vaapi_config_input_overlay(AVFilterLink *inlink)
{
AVFilterContext *avctx = inlink->dst;
+ AVFilterLink *inlink0 = avctx->inputs[0];
OverlayVAAPIContext *ctx = avctx->priv;
int ret;
@@ -312,7 +314,11 @@ static int overlay_vaapi_config_input_ov
ctx->blend_alpha = ctx->alpha;
}
- if (have_alpha_planar(inlink))
+ // VA_BLEND_PREMULTIPLIED_ALPHA requires both inputs
+ // have alpha planar, otherwise it just fails.
+ if (have_alpha_planar(inlink) &&
+ have_alpha_planar(inlink0) &&
+ ctx->alpha_format == 1)
ctx->blend_flags |= VA_BLEND_PREMULTIPLIED_ALPHA;
return 0;
@@ -380,6 +386,9 @@ static const AVOption overlay_vaapi_opti
{ "w", "Overlay width", OFFSET(overlay_ow), AV_OPT_TYPE_STRING, { .str="overlay_iw"}, 0, 255, .flags = FLAGS},
{ "h", "Overlay height", OFFSET(overlay_oh), AV_OPT_TYPE_STRING, { .str="overlay_ih*w/overlay_iw"}, 0, 255, .flags = FLAGS},
{ "alpha", "Overlay global alpha", OFFSET(alpha), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 }, 0.0, 1.0, .flags = FLAGS },
+ { "alpha_format", "alpha format", OFFSET(alpha_format), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, FLAGS, .unit = "alpha_format" },
+ { "straight", "The overlay input is unpremultiplied", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, .flags = FLAGS, .unit = "alpha_format" },
+ { "premultiplied", "The overlay input is premultiplied", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, .flags = FLAGS, .unit = "alpha_format" },
{ "eof_action", "Action to take when encountering EOF from secondary input ",
OFFSET(fs.opt_eof_action), AV_OPT_TYPE_INT, { .i64 = EOF_ACTION_REPEAT },
EOF_ACTION_REPEAT, EOF_ACTION_PASS, .flags = FLAGS, .unit = "eof_action" },
debian/patches/0015-check-alpha-mode-for-vaapi-overlay-filter.patch
Vendored
+46
View File
@@ -0,0 +1,46 @@
Index: FFmpeg/libavfilter/vf_overlay_vaapi.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_overlay_vaapi.c
+++ FFmpeg/libavfilter/vf_overlay_vaapi.c
@@ -194,6 +194,14 @@ static int overlay_vaapi_blend(FFFrameSy
if (err < 0)
goto fail;
+ if (input_overlay) {
+ /* call vaapi_vpp_colour_properties() for overlay */
+ err = ff_vaapi_vpp_init_params(avctx, &params[1],
+ input_overlay, output);
+ if (err < 0)
+ goto fail;
+ }
+
err = ff_vaapi_vpp_init_params(avctx, &params[0],
input_main, output);
if (err < 0)
@@ -228,8 +236,6 @@ static int overlay_vaapi_blend(FFFrameSy
"will crop the overlay image according based on the main image.\n");
}
- memcpy(&params[1], &params[0], sizeof(params[0]));
-
blend_state.flags = ctx->blend_flags;
blend_state.global_alpha = ctx->blend_alpha;
params[1].blend_state = &blend_state;
@@ -237,6 +243,7 @@ static int overlay_vaapi_blend(FFFrameSy
params[1].surface = (VASurfaceID)(uintptr_t)input_overlay->data[3];
params[1].surface_region = NULL;
params[1].output_region = &overlay_region;
+ params[1].output_background_color = 0;
}
err = ff_vaapi_vpp_render_pictures(avctx, params, input_overlay ? 2 : 1, output);
@@ -312,7 +319,8 @@ static int overlay_vaapi_config_input_ov
ctx->blend_alpha = ctx->alpha;
}
- if (have_alpha_planar(inlink))
+ if (have_alpha_planar(inlink) &&
+ inlink->alpha_mode == AVALPHA_MODE_PREMULTIPLIED)
ctx->blend_flags |= VA_BLEND_PREMULTIPLIED_ALPHA;
return 0;
debian/patches/0016-add-fixes-for-nvdec-exceed-32-surfaces-error.patch
Vendored
+6 -6
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/libavcodec/nvdec.c
===================================================================
--- FFmpeg.orig/libavcodec/nvdec.c
+++ FFmpeg/libavcodec/nvdec.c
@@ -399,8 +399,8 @@ int ff_nvdec_decode_init(AVCodecContext
@@ -413,8 +413,8 @@ int ff_nvdec_decode_init(AVCodecContext
params.OutputFormat = output_format;
params.CodecType = cuvid_codec_type;
params.ChromaFormat = cuvid_chroma_format;
@@ -13,20 +13,20 @@ Index: FFmpeg/libavcodec/nvdec.c
ret = nvdec_decoder_create(&ctx->decoder, frames_ctx->device_ref, &params, avctx);
if (ret < 0) {
@@ -424,7 +424,7 @@ int ff_nvdec_decode_init(AVCodecContext
@@ -438,7 +438,7 @@ int ff_nvdec_decode_init(AVCodecContext
ret = AVERROR(ENOMEM);
goto fail;
}
- pool->dpb_size = frames_ctx->initial_pool_size;
+ pool->dpb_size = FFMIN(frames_ctx->initial_pool_size, 32);
ctx->decoder_pool = ff_refstruct_pool_alloc_ext(sizeof(unsigned int), 0, pool,
ctx->decoder_pool = av_refstruct_pool_alloc_ext(sizeof(unsigned int), 0, pool,
nvdec_decoder_frame_init,
@@ -529,7 +529,6 @@ static int nvdec_retrieve_data(void *log
@@ -543,7 +543,6 @@ static int nvdec_retrieve_data(void *log
goto copy_fail;
unmap_data->idx = cf->idx;
- unmap_data->idx_ref = ff_refstruct_ref(cf->idx_ref);
unmap_data->decoder = ff_refstruct_ref(cf->decoder);
- unmap_data->idx_ref = av_refstruct_ref(cf->idx_ref);
unmap_data->decoder = av_refstruct_ref(cf->decoder);
av_pix_fmt_get_chroma_sub_sample(hwctx->sw_format, &shift_h, &shift_v);
debian/patches/0017-backport-fixes-for-qsv-from-upstream.patch
Vendored
+54
View File
@@ -0,0 +1,54 @@
Index: FFmpeg/libavcodec/qsvenc.c
===================================================================
--- FFmpeg.orig/libavcodec/qsvenc.c
+++ FFmpeg/libavcodec/qsvenc.c
@@ -1840,6 +1840,9 @@ int ff_qsv_enc_init(AVCodecContext *avct
return ret;
}
+ // Update AVCodecContext with actual encoding parameters
+ avctx->has_b_frames = q->param.mfx.GopRefDist > 1 ? q->param.mfx.GopRefDist - 1 : 0;
+
q->avctx = avctx;
return 0;
@@ -2497,7 +2500,8 @@ static int encode_frame(AVCodecContext *
pkt.bs->Data = pkt.pkt.data;
pkt.bs->MaxLength = pkt.pkt.size;
- if (avctx->codec_id == AV_CODEC_ID_H264) {
+ if (avctx->codec_id == AV_CODEC_ID_H264 ||
+ avctx->codec_id == AV_CODEC_ID_HEVC) {
enc_info = av_mallocz(sizeof(*enc_info));
if (!enc_info)
goto nomem;
@@ -2563,7 +2567,8 @@ free:
av_freep(&pkt.sync);
av_packet_unref(&pkt.pkt);
av_freep(&pkt.bs);
- if (avctx->codec_id == AV_CODEC_ID_H264) {
+ if (avctx->codec_id == AV_CODEC_ID_H264 ||
+ avctx->codec_id == AV_CODEC_ID_HEVC) {
av_freep(&enc_info);
av_freep(&enc_buf);
}
@@ -2684,7 +2689,8 @@ int ff_qsv_encode(AVCodecContext *avctx,
return AVERROR_INVALIDDATA;
}
- if (avctx->codec_id == AV_CODEC_ID_H264) {
+ if (avctx->codec_id == AV_CODEC_ID_H264 ||
+ avctx->codec_id == AV_CODEC_ID_HEVC) {
enc_buf = qpkt.bs->ExtParam;
enc_info = (mfxExtAVCEncodedFrameInfo *)(*enc_buf);
ff_encode_add_stats_side_data(&qpkt.pkt,
@@ -2728,7 +2734,8 @@ int ff_qsv_enc_close(AVCodecContext *avc
if (q->async_fifo) {
QSVPacket pkt;
while (av_fifo_read(q->async_fifo, &pkt, 1) >= 0) {
- if (avctx->codec_id == AV_CODEC_ID_H264) {
+ if (avctx->codec_id == AV_CODEC_ID_H264 ||
+ avctx->codec_id == AV_CODEC_ID_HEVC) {
mfxExtBuffer **enc_buf = pkt.bs->ExtParam;
mfxExtAVCEncodedFrameInfo *enc_info = (mfxExtAVCEncodedFrameInfo *)(*enc_buf);
av_freep(&enc_info);
debian/patches/0017-backport-fixes-for-vaapi-from-upstream.patch
Vendored
-144
View File
@@ -1,144 +0,0 @@
Index: FFmpeg/libavutil/hwcontext_vaapi.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_vaapi.c
+++ FFmpeg/libavutil/hwcontext_vaapi.c
@@ -1813,6 +1813,7 @@ static int vaapi_device_create(AVHWDevic
"Failed to get DRM device info for device %d.\n", n);
close(priv->drm_fd);
priv->drm_fd = -1;
+ drmFreeVersion(info);
continue;
}
@@ -1824,6 +1825,7 @@ static int vaapi_device_create(AVHWDevic
drmFreeDevice(&device);
close(priv->drm_fd);
priv->drm_fd = -1;
+ drmFreeVersion(info);
continue;
}
av_log(ctx, AV_LOG_VERBOSE, "Trying to use "
@@ -1831,6 +1833,7 @@ static int vaapi_device_create(AVHWDevic
"with matching vendor id (%s).\n",
n, vendor_id->value);
drmFreeDevice(&device);
+ drmFreeVersion(info);
break;
}
drmFreeVersion(info);
Index: FFmpeg/libavcodec/vaapi_encode.c
===================================================================
--- FFmpeg.orig/libavcodec/vaapi_encode.c
+++ FFmpeg/libavcodec/vaapi_encode.c
@@ -1133,6 +1133,68 @@ fail:
return err;
}
+static av_cold int vaapi_encode_surface_alignment(av_unused AVCodecContext *avctx)
+{
+#if VA_CHECK_VERSION(1, 21, 0)
+ VAAPIEncodeContext *ctx = avctx->priv_data;
+ VASurfaceAttrib *attr_list = NULL;
+ unsigned int attr_count = 0;
+ VAConfigID va_config;
+ VAStatus vas;
+ int err = 0;
+
+ vas = vaCreateConfig(ctx->hwctx->display,
+ ctx->va_profile, ctx->va_entrypoint,
+ NULL, 0, &va_config);
+ if (vas != VA_STATUS_SUCCESS) {
+ av_log(avctx, AV_LOG_ERROR, "Failed to create temp encode pipeline "
+ "configuration: %d (%s).\n", vas, vaErrorStr(vas));
+ return AVERROR(EIO);
+ }
+
+ vas = vaQuerySurfaceAttributes(ctx->hwctx->display, va_config,
+ 0, &attr_count);
+ if (vas != VA_STATUS_SUCCESS) {
+ av_log(avctx, AV_LOG_ERROR, "Failed to query surface attributes: "
+ "%d (%s).\n", vas, vaErrorStr(vas));
+ err = AVERROR_EXTERNAL;
+ goto fail;
+ }
+
+ attr_list = av_malloc(attr_count * sizeof(*attr_list));
+ if (!attr_list) {
+ err = AVERROR(ENOMEM);
+ goto fail;
+ }
+
+ vas = vaQuerySurfaceAttributes(ctx->hwctx->display, va_config,
+ attr_list, &attr_count);
+ if (vas != VA_STATUS_SUCCESS) {
+ av_log(avctx, AV_LOG_ERROR, "Failed to query surface attributes: "
+ "%d (%s).\n", vas, vaErrorStr(vas));
+ err = AVERROR_EXTERNAL;
+ goto fail;
+ }
+
+ for (unsigned int i = 0; i < attr_count; i++) {
+ if (attr_list[i].type == VASurfaceAttribAlignmentSize) {
+ ctx->surface_alignment_width =
+ 1 << (attr_list[i].value.value.i & 0xf);
+ ctx->surface_alignment_height =
+ 1 << ((attr_list[i].value.value.i & 0xf0) >> 4);
+ break;
+ }
+ }
+
+fail:
+ av_freep(&attr_list);
+ vaDestroyConfig(ctx->hwctx->display, va_config);
+ return err;
+#else
+ return 0;
+#endif
+}
+
static const VAAPIEncodeRCMode vaapi_encode_rc_modes[] = {
// Bitrate Quality
// | Maxrate | HRD/VBV
@@ -2111,6 +2173,10 @@ av_cold int ff_vaapi_encode_init(AVCodec
if (err < 0)
goto fail;
+ err = vaapi_encode_surface_alignment(avctx);
+ if (err < 0)
+ goto fail;
+
if (ctx->codec->get_encoder_caps) {
err = ctx->codec->get_encoder_caps(avctx);
if (err < 0)
Index: FFmpeg/libavcodec/vaapi_encode.h
===================================================================
--- FFmpeg.orig/libavcodec/vaapi_encode.h
+++ FFmpeg/libavcodec/vaapi_encode.h
@@ -260,6 +260,10 @@ typedef struct VAAPIEncodeContext {
* This is a RefStruct reference.
*/
VABufferID *coded_buffer_ref;
+
+ // Surface alignment required by driver.
+ int surface_alignment_width;
+ int surface_alignment_height;
} VAAPIEncodeContext;
typedef struct VAAPIEncodeType {
Index: FFmpeg/libavcodec/vaapi_encode_h265.c
===================================================================
--- FFmpeg.orig/libavcodec/vaapi_encode_h265.c
+++ FFmpeg/libavcodec/vaapi_encode_h265.c
@@ -951,8 +951,10 @@ static av_cold int vaapi_encode_h265_get
"min CB size %dx%d.\n", priv->ctu_size, priv->ctu_size,
priv->min_cb_size, priv->min_cb_size);
- base_ctx->surface_width = FFALIGN(avctx->width, priv->min_cb_size);
- base_ctx->surface_height = FFALIGN(avctx->height, priv->min_cb_size);
+ base_ctx->surface_width = FFALIGN(avctx->width,
+ FFMAX(priv->min_cb_size, priv->common.surface_alignment_width));
+ base_ctx->surface_height = FFALIGN(avctx->height,
+ FFMAX(priv->min_cb_size, priv->common.surface_alignment_height));
base_ctx->slice_block_width = base_ctx->slice_block_height = priv->ctu_size;
debian/patches/0019-backport-fixes-for-qsv-from-cartwheel.patch → debian/patches/0018-backport-fixes-for-qsv-from-cartwheel.patch
Vendored
+2 -2
View File
@@ -30,7 +30,7 @@ Index: FFmpeg/libavcodec/qsvdec.c
#include "qsv.h"
+#include "h264_sei.h"
#include "qsv_internal.h"
#include "refstruct.h"
#include "libavutil/refstruct.h"
@@ -111,8 +114,13 @@ typedef struct QSVContext {
@@ -207,7 +207,7 @@ Index: FFmpeg/libavcodec/qsvdec.c
if ((avctx->export_side_data & AV_CODEC_EXPORT_DATA_FILM_GRAIN) &&
@@ -986,6 +1139,8 @@ static void qsv_decode_close_qsvcontext(
av_buffer_unref(&q->frames_ctx.hw_frames_ctx);
ff_refstruct_unref(&q->frames_ctx.mids);
av_refstruct_unref(&q->frames_ctx.mids);
av_buffer_pool_uninit(&q->pool);
+
+ av_freep(&q->payload.Data);
debian/patches/0018-backport-fixes-for-qsv-from-upstream.patch
Vendored
-81
View File
@@ -1,81 +0,0 @@
Index: FFmpeg/libavcodec/qsvdec.c
===================================================================
--- FFmpeg.orig/libavcodec/qsvdec.c
+++ FFmpeg/libavcodec/qsvdec.c
@@ -202,6 +202,7 @@ static int qsv_init_session(AVCodecConte
MFXClose(q->internal_qs.session);
q->internal_qs.session = NULL;
}
+ ff_refstruct_unref(&q->frames_ctx.mids);
av_buffer_unref(&q->frames_ctx.hw_frames_ctx);
q->frames_ctx.hw_frames_ctx = av_buffer_ref(hw_frames_ref);
Index: FFmpeg/libavcodec/qsvenc.c
===================================================================
--- FFmpeg.orig/libavcodec/qsvenc.c
+++ FFmpeg/libavcodec/qsvenc.c
@@ -1842,6 +1842,9 @@ int ff_qsv_enc_init(AVCodecContext *avct
return ret;
}
+ // Update AVCodecContext with actual encoding parameters
+ avctx->has_b_frames = q->param.mfx.GopRefDist > 1 ? q->param.mfx.GopRefDist - 1 : 0;
+
q->avctx = avctx;
return 0;
Index: FFmpeg/libavfilter/qsvvpp.c
===================================================================
--- FFmpeg.orig/libavfilter/qsvvpp.c
+++ FFmpeg/libavfilter/qsvvpp.c
@@ -471,7 +471,8 @@ static QSVFrame *submit_frame(QSVVPPCont
}
/* get the output surface */
-static QSVFrame *query_frame(QSVVPPContext *s, AVFilterLink *outlink, const AVFrame *in)
+static QSVFrame *query_frame(QSVVPPContext *s, AVFilterLink *outlink, const AVFrame *in,
+ const AVFrame *propref)
{
FilterLink *l = ff_filter_link(outlink);
AVFilterContext *ctx = outlink->src;
@@ -513,6 +514,15 @@ static QSVFrame *query_frame(QSVVPPConte
return NULL;
}
+ if (propref) {
+ ret = av_frame_copy_props(out_frame->frame, propref);
+ if (ret < 0) {
+ av_frame_free(&out_frame->frame);
+ av_log(ctx, AV_LOG_ERROR, "Failed to copy metadata fields from src to dst.\n");
+ return NULL;
+ }
+ }
+
if (l->frame_rate.num && l->frame_rate.den)
out_frame->frame->duration = av_rescale_q(1, av_inv_q(l->frame_rate), outlink->time_base);
else
@@ -985,7 +995,7 @@ int ff_qsvvpp_filter_frame(QSVVPPContext
}
do {
- out_frame = query_frame(s, outlink, in_frame->frame);
+ out_frame = query_frame(s, outlink, in_frame->frame, propref);
if (!out_frame) {
av_log(ctx, AV_LOG_ERROR, "Failed to query an output frame.\n");
return AVERROR(ENOMEM);
@@ -1009,15 +1019,6 @@ int ff_qsvvpp_filter_frame(QSVVPPContext
break;
}
- if (propref) {
- ret1 = av_frame_copy_props(out_frame->frame, propref);
- if (ret1 < 0) {
- av_frame_free(&out_frame->frame);
- av_log(ctx, AV_LOG_ERROR, "Failed to copy metadata fields from src to dst.\n");
- return ret1;
- }
- }
-
out_frame->frame->pts = av_rescale_q(out_frame->surface.Data.TimeStamp,
default_tb, outlink->time_base);
debian/patches/0020-add-fixes-for-qsv-vpp-filters.patch → debian/patches/0019-add-fixes-for-qsv-vpp-filters.patch
Vendored
+21 -7
View File
@@ -33,7 +33,7 @@ Index: FFmpeg/libavfilter/qsvvpp.c
else if (qsv_frame->frame->repeat_pict == 2)
qsv_frame->surface.Info.PicStruct |= MFX_PICSTRUCT_FRAME_DOUBLING;
else if (qsv_frame->frame->repeat_pict == 4)
@@ -913,8 +918,13 @@ static int qsvvpp_init_vpp_session(AVFil
@@ -908,8 +913,13 @@ static int qsvvpp_init_vpp_session(AVFil
/* Query VPP params again, including params for frame */
ret = MFXVideoVPP_Query(s->session, &s->vpp_param, &s->vpp_param);
@@ -150,7 +150,7 @@ Index: FFmpeg/libavfilter/vf_overlay_qsv.c
vpp->qsv_param.out_sw_format = hw_frame0->sw_format;
}
@@ -374,6 +371,7 @@ static int overlay_qsv_query_formats(AVF
@@ -376,6 +373,7 @@ static int overlay_qsv_query_formats(con
static const enum AVPixelFormat main_in_fmts[] = {
AV_PIX_FMT_YUV420P,
AV_PIX_FMT_NV12,
@@ -158,7 +158,7 @@ Index: FFmpeg/libavfilter/vf_overlay_qsv.c
AV_PIX_FMT_YUYV422,
AV_PIX_FMT_RGB32,
AV_PIX_FMT_QSV,
@@ -381,6 +379,7 @@ static int overlay_qsv_query_formats(AVF
@@ -383,6 +381,7 @@ static int overlay_qsv_query_formats(con
};
static const enum AVPixelFormat out_pix_fmts[] = {
AV_PIX_FMT_NV12,
@@ -269,7 +269,21 @@ Index: FFmpeg/libavfilter/vf_vpp_qsv.c
INIT_MFX_EXTBUF(scale_conf, MFX_EXTBUFF_VPP_SCALING);
SET_MFX_PARAM_FIELD(scale_conf, ScalingMode, mode);
@@ -884,19 +924,13 @@ static const AVOption vpp_options[] = {
@@ -706,6 +746,13 @@ static int config_output(AVFilterLink *o
#undef INIT_MFX_EXTBUF
#undef SET_MFX_PARAM_FIELD
+ if (vpp->tonemap) {
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
+
if (vpp->use_frc || vpp->use_crop || vpp->deinterlace || vpp->denoise ||
vpp->detail || vpp->procamp || vpp->rotate || vpp->hflip ||
inlink->w != outlink->w || inlink->h != outlink->h || in_format != vpp->out_format ||
@@ -884,19 +931,13 @@ static const AVOption vpp_options[] = {
{ "height", "Output video height(0=input video height, -1=keep input video aspect)", OFFSET(oh), AV_OPT_TYPE_STRING, { .str="w*ch/cw" }, 0, 255, .flags = FLAGS },
{ "format", "Output pixel format", OFFSET(output_format_str), AV_OPT_TYPE_STRING, { .str = "same" }, .flags = FLAGS },
{ "async_depth", "Internal parallelization depth, the higher the value the higher the latency.", OFFSET(qsv.async_depth), AV_OPT_TYPE_INT, { .i64 = 4 }, 0, INT_MAX, .flags = FLAGS },
@@ -290,7 +304,7 @@ Index: FFmpeg/libavfilter/vf_vpp_qsv.c
{ "rate", "Generate output at frame rate or field rate, available only for deinterlace mode",
OFFSET(field_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, FLAGS, .unit = "rate" },
@@ -927,8 +961,9 @@ static const AVOption vpp_options[] = {
@@ -927,8 +968,9 @@ static const AVOption vpp_options[] = {
{ "out_color_transfer", "Output color transfer characteristics",
OFFSET(color_transfer_str), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS },
@@ -301,7 +315,7 @@ Index: FFmpeg/libavfilter/vf_vpp_qsv.c
{ NULL }
};
@@ -982,19 +1017,14 @@ static const AVOption qsvscale_options[]
@@ -984,19 +1026,14 @@ static const AVOption qsvscale_options[]
{ "h", "Output video height(0=input video height, -1=keep input video aspect)", OFFSET(oh), AV_OPT_TYPE_STRING, { .str = "ih" }, .flags = FLAGS },
{ "format", "Output pixel format", OFFSET(output_format_str), AV_OPT_TYPE_STRING, { .str = "same" }, .flags = FLAGS },
@@ -323,7 +337,7 @@ Index: FFmpeg/libavfilter/vf_vpp_qsv.c
{ NULL },
};
@@ -1019,6 +1049,7 @@ static const AVOption qsvdeint_options[]
@@ -1021,6 +1058,7 @@ static const AVOption qsvdeint_options[]
{ "bob", "bob algorithm", 0, AV_OPT_TYPE_CONST, {.i64 = MFX_DEINTERLACING_BOB}, MFX_DEINTERLACING_BOB, MFX_DEINTERLACING_ADVANCED, FLAGS, .unit = "mode"},
{ "advanced", "Motion adaptive algorithm", 0, AV_OPT_TYPE_CONST, {.i64 = MFX_DEINTERLACING_ADVANCED}, MFX_DEINTERLACING_BOB, MFX_DEINTERLACING_ADVANCED, FLAGS, .unit = "mode"},
debian/patches/0021-add-external-frame-support-for-qsv-enc.patch → debian/patches/0020-add-external-frame-support-for-qsv-enc.patch
Vendored
View File
debian/patches/0022-suppress-overlay-warning-logs.patch → debian/patches/0021-suppress-overlay-warning-logs.patch
Vendored
+4 -4
View File
@@ -2,12 +2,12 @@ Index: FFmpeg/libavfilter/buffersrc.c
===================================================================
--- FFmpeg.orig/libavfilter/buffersrc.c
+++ FFmpeg/libavfilter/buffersrc.c
@@ -74,7 +74,7 @@ typedef struct BufferSourceContext {
@@ -76,7 +76,7 @@ typedef struct BufferSourceContext {
c->prev_delta = c->prev_w != width || c->prev_h != height || c->prev_pix_fmt != format ||\
c->prev_color_space != csp || c->prev_color_range != range;\
c->prev_color_space != csp || c->prev_color_range != range || c->prev_alpha_mode != alpha;\
if (c->link_delta) {\
- int loglevel = c->prev_delta ? AV_LOG_WARNING : AV_LOG_DEBUG;\
+ int loglevel = c->prev_delta ? AV_LOG_VERBOSE : AV_LOG_DEBUG;\
av_log(s, loglevel, "Changing video frame properties on the fly is not supported by all filters.\n");\
av_log(s, loglevel, "filter context - w: %d h: %d fmt: %d csp: %s range: %s, incoming frame - w: %d h: %d fmt: %d csp: %s range: %s pts_time: %s\n",\
c->w, c->h, c->pix_fmt, av_color_space_name(c->color_space), av_color_range_name(c->color_range),\
av_log(s, loglevel, "filter context - w: %d h: %d fmt: %d csp: %s range: %s alpha: %s, incoming frame - w: %d h: %d fmt: %d csp: %s range: %s alpha: %s pts_time: %s\n",\
c->w, c->h, c->pix_fmt, av_color_space_name(c->color_space), av_color_range_name(c->color_range), av_alpha_mode_name(c->alpha_mode),\
debian/patches/0024-add-sub2video-option-to-subtitles-filter.patch → debian/patches/0022-add-sub2video-option-to-subtitles-filter.patch
Vendored
+46 -58
View File
@@ -2,16 +2,16 @@ Index: FFmpeg/libavfilter/drawutils.c
===================================================================
--- FFmpeg.orig/libavfilter/drawutils.c
+++ FFmpeg/libavfilter/drawutils.c
@@ -437,7 +437,7 @@ void ff_blend_rectangle(FFDrawContext *d
@@ -461,7 +461,7 @@ void ff_blend_rectangle(FFDrawContext *d
static void blend_pixel16(uint8_t *dst, unsigned src, unsigned alpha,
const uint8_t *mask, int mask_linesize, int l2depth,
- unsigned w, unsigned h, unsigned shift, unsigned xm0)
+ unsigned w, unsigned h, unsigned shift, unsigned xm0, int srca_opaque)
{
unsigned xm, x, y, t = 0;
unsigned t = 0;
unsigned xmshf = 3 - l2depth;
@@ -456,12 +456,13 @@ static void blend_pixel16(uint8_t *dst,
@@ -480,12 +480,13 @@ static void blend_pixel16(uint8_t *dst,
mask += mask_linesize;
}
alpha = (t >> shift) * alpha;
@@ -24,9 +24,9 @@ Index: FFmpeg/libavfilter/drawutils.c
- unsigned w, unsigned h, unsigned shift, unsigned xm0)
+ unsigned w, unsigned h, unsigned shift, unsigned xm0, int srca_opaque)
{
unsigned xm, x, y, t = 0;
unsigned t = 0;
unsigned xmshf = 3 - l2depth;
@@ -479,6 +480,7 @@ static void blend_pixel(uint8_t *dst, un
@@ -503,6 +504,7 @@ static void blend_pixel(uint8_t *dst, un
mask += mask_linesize;
}
alpha = (t >> shift) * alpha;
@@ -34,14 +34,13 @@ Index: FFmpeg/libavfilter/drawutils.c
*dst = ((0x1010101 - alpha) * *dst + alpha * src) >> 24;
}
@@ -486,50 +488,134 @@ static void blend_line_hv16(uint8_t *dst
@@ -510,48 +512,130 @@ static void blend_line_hv16(uint8_t *dst
unsigned src, unsigned alpha,
const uint8_t *mask, int mask_linesize, int l2depth, int w,
unsigned hsub, unsigned vsub,
- int xm, int left, int right, int hband)
+ int xm, int left, int right, int hband, int srca_opaque)
{
int x;
if (left) {
blend_pixel16(dst, src, alpha, mask, mask_linesize, l2depth,
@@ -50,7 +49,7 @@ Index: FFmpeg/libavfilter/drawutils.c
dst += dst_delta;
xm += left;
}
for (x = 0; x < w; x++) {
for (int x = 0; x < w; x++) {
blend_pixel16(dst, src, alpha, mask, mask_linesize, l2depth,
- 1 << hsub, hband, hsub + vsub, xm);
+ 1 << hsub, hband, hsub + vsub, xm, srca_opaque);
@@ -70,7 +69,6 @@ Index: FFmpeg/libavfilter/drawutils.c
- int xm, int left, int right, int hband)
+ int xm, int left, int right, int hband, int srca_opaque)
{
int x;
if (left) {
blend_pixel(dst, src, alpha, mask, mask_linesize, l2depth,
@@ -79,7 +77,7 @@ Index: FFmpeg/libavfilter/drawutils.c
dst += dst_delta;
xm += left;
}
for (x = 0; x < w; x++) {
for (int x = 0; x < w; x++) {
blend_pixel(dst, src, alpha, mask, mask_linesize, l2depth,
- 1 << hsub, hband, hsub + vsub, xm);
+ 1 << hsub, hband, hsub + vsub, xm, srca_opaque);
@@ -99,16 +97,16 @@ Index: FFmpeg/libavfilter/drawutils.c
+ unsigned w, unsigned h, unsigned shift, unsigned xm0,
+ int srca_opaque, int limited)
+{
+ unsigned xm, x, y, t = 0;
+ unsigned t = 0;
+ unsigned xmshf = 3 - l2depth;
+ unsigned xmmod = 7 >> l2depth;
+ unsigned mbits = (1 << (1 << l2depth)) - 1;
+ unsigned mmult = 255 / mbits;
+ const uint8_t offset = limited ? 16 : 0;
+
+ for (y = 0; y < h; y++) {
+ xm = xm0;
+ for (x = 0; x < w; x++) {
+ for (unsigned y = 0; y < h; y++) {
+ unsigned xm = xm0;
+ for (unsigned x = 0; x < w; x++) {
+ t += ((mask[xm >> xmshf] >> ((~xm & xmmod) << l2depth)) & mbits)
+ * mmult;
+ xm++;
@@ -145,8 +143,6 @@ Index: FFmpeg/libavfilter/drawutils.c
+ int xm, int left, int right, int hband,
+ int srca_opaque, int limited)
+{
+ int x;
+
+ if (left) {
+ blend_pixel_unpremul_rgb32(
+ dst0, dst1, dst2, dst3,
@@ -158,7 +154,7 @@ Index: FFmpeg/libavfilter/drawutils.c
+ dst3 += dst_delta;
+ xm += left;
+ }
+ for (x = 0; x < w; x++) {
+ for (int x = 0; x < w; x++) {
+ blend_pixel_unpremul_rgb32(
+ dst0, dst1, dst2, dst3,
+ src, alpha, mask, mask_linesize, l2depth,
@@ -177,7 +173,7 @@ Index: FFmpeg/libavfilter/drawutils.c
}
void ff_blend_mask(FFDrawContext *draw, FFDrawColor *color,
@@ -567,13 +653,79 @@ void ff_blend_mask(FFDrawContext *draw,
@@ -589,13 +673,79 @@ void ff_blend_mask(FFDrawContext *draw,
y_sub = y0;
subsampling_bounds(draw->hsub[plane], &x_sub, &w_sub, &left, &right);
subsampling_bounds(draw->vsub[plane], &y_sub, &h_sub, &top, &bottom);
@@ -195,7 +191,7 @@ Index: FFmpeg/libavfilter/drawutils.c
+ uint8_t *pdst[4];
+ unsigned src[4];
+
+ for (comp = 0; comp < nb_comp; comp++) {
+ for (unsigned comp = 0; comp < nb_comp; comp++) {
+ av_assert0(draw->desc->comp[comp].plane == plane);
+
+ offset[comp] = draw->desc->comp[comp].offset;
@@ -213,11 +209,11 @@ Index: FFmpeg/libavfilter/drawutils.c
+ draw->hsub[plane], draw->vsub[plane],
+ xm0, left, right, top, srca_opaque, limited);
+
+ for (comp = 0; comp < nb_comp; comp++)
+ for (unsigned comp = 0; comp < nb_comp; comp++)
+ pdst[comp] += dst_linesize[plane];
+ m += top * mask_linesize;
+ }
+ for (y = 0; y < h_sub; y++) {
+ for (int y = 0; y < h_sub; y++) {
+ blend_line_hv_unpremul_rgb32(
+ pdst[0], pdst[1], pdst[2], pdst[3], draw->pixelstep[plane],
+ src, alpha,
@@ -225,7 +221,7 @@ Index: FFmpeg/libavfilter/drawutils.c
+ draw->hsub[plane], draw->vsub[plane],
+ xm0, left, right, 1 << draw->vsub[plane], srca_opaque, limited);
+
+ for (comp = 0; comp < nb_comp; comp++)
+ for (unsigned comp = 0; comp < nb_comp; comp++)
+ pdst[comp] += dst_linesize[plane];
+ m += mask_linesize << draw->vsub[plane];
+ }
@@ -240,7 +236,7 @@ Index: FFmpeg/libavfilter/drawutils.c
+ break;
+ }
+
for (comp = 0; comp < nb_comp; comp++) {
for (unsigned comp = 0; comp < nb_comp; comp++) {
const int depth = draw->desc->comp[comp].depth;
const int offset = draw->desc->comp[comp].offset;
const int index = offset / ((depth + 7) / 8);
@@ -257,7 +253,7 @@ Index: FFmpeg/libavfilter/drawutils.c
p = p0 + offset;
m = mask;
if (top) {
@@ -582,13 +734,13 @@ void ff_blend_mask(FFDrawContext *draw,
@@ -604,13 +754,13 @@ void ff_blend_mask(FFDrawContext *draw,
color->comp[plane].u8[index], alpha,
m, mask_linesize, l2depth, w_sub,
draw->hsub[plane], draw->vsub[plane],
@@ -273,7 +269,7 @@ Index: FFmpeg/libavfilter/drawutils.c
}
p += dst_linesize[plane];
m += top * mask_linesize;
@@ -599,7 +751,7 @@ void ff_blend_mask(FFDrawContext *draw,
@@ -621,7 +771,7 @@ void ff_blend_mask(FFDrawContext *draw,
color->comp[plane].u8[index], alpha,
m, mask_linesize, l2depth, w_sub,
draw->hsub[plane], draw->vsub[plane],
@@ -282,7 +278,7 @@ Index: FFmpeg/libavfilter/drawutils.c
p += dst_linesize[plane];
m += mask_linesize << draw->vsub[plane];
}
@@ -609,7 +761,7 @@ void ff_blend_mask(FFDrawContext *draw,
@@ -631,7 +781,7 @@ void ff_blend_mask(FFDrawContext *draw,
color->comp[plane].u16[index], alpha,
m, mask_linesize, l2depth, w_sub,
draw->hsub[plane], draw->vsub[plane],
@@ -291,7 +287,7 @@ Index: FFmpeg/libavfilter/drawutils.c
p += dst_linesize[plane];
m += mask_linesize << draw->vsub[plane];
}
@@ -620,13 +772,13 @@ void ff_blend_mask(FFDrawContext *draw,
@@ -642,13 +792,13 @@ void ff_blend_mask(FFDrawContext *draw,
color->comp[plane].u8[index], alpha,
m, mask_linesize, l2depth, w_sub,
draw->hsub[plane], draw->vsub[plane],
@@ -307,16 +303,11 @@ Index: FFmpeg/libavfilter/drawutils.c
}
}
}
@@ -658,3 +810,4 @@ AVFilterFormats *ff_draw_supported_pixel
return NULL;
return fmts;
}
+
Index: FFmpeg/libavfilter/drawutils.h
===================================================================
--- FFmpeg.orig/libavfilter/drawutils.h
+++ FFmpeg/libavfilter/drawutils.h
@@ -59,7 +59,17 @@ typedef struct FFDrawColor {
@@ -61,7 +61,17 @@ typedef struct FFDrawColor {
/**
* Process alpha pixel component.
*/
@@ -335,11 +326,6 @@ Index: FFmpeg/libavfilter/drawutils.h
/**
* Init a draw context.
@@ -164,3 +174,4 @@ int ff_draw_round_to_sub(FFDrawContext *
AVFilterFormats *ff_draw_supported_pixel_formats(unsigned flags);
#endif /* AVFILTER_DRAWUTILS_H */
+
Index: FFmpeg/libavfilter/vf_subtitles.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_subtitles.c
@@ -376,19 +362,15 @@ Index: FFmpeg/libavfilter/vf_subtitles.c
/* libass supports a log level ranging from 0 to 7 */
static const int ass_libavfilter_log_level_map[] = {
@@ -146,9 +155,25 @@ static int query_formats(AVFilterContext
static int config_input(AVFilterLink *inlink)
@@ -184,12 +193,28 @@ static int config_input(AVFilterLink *in
{
AssContext *ass = inlink->dst->priv;
AVFilterContext *ctx = inlink->dst;
AssContext *ass = ctx->priv;
+ const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+ int draw_flags = ass->alpha ? FF_DRAW_PROCESS_ALPHA : 0;
int ret;
- ff_draw_init2(&ass->draw, inlink->format, inlink->colorspace, inlink->color_range,
- ass->alpha ? FF_DRAW_PROCESS_ALPHA : 0);
+ if (ass->sub2video) {
+ const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+ if (!desc)
+ return AVERROR(ENOSYS);
+
+ if (ass->sub2video && inlink->alpha_mode != AVALPHA_MODE_PREMULTIPLIED) {
+ if ((desc->flags & AV_PIX_FMT_FLAG_ALPHA) &&
+ (desc->flags & (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB)))
+ draw_flags |= FF_DRAW_MASK_SRC_ALPHA_OPAQUE | FF_DRAW_PROCESS_ALPHA;
@@ -400,11 +382,17 @@ Index: FFmpeg/libavfilter/vf_subtitles.c
+ draw_flags |= FF_DRAW_MASK_UNPREMUL_RGB32;
+ }
+
+ ff_draw_init2(&ass->draw, inlink->format, inlink->colorspace, inlink->color_range, draw_flags);
ass_set_frame_size (ass->renderer, inlink->w, inlink->h);
if (ass->original_w && ass->original_h) {
@@ -161,6 +186,8 @@ static int config_input(AVFilterLink *in
ret = ff_draw_init2(&ass->draw, inlink->format,
+ (desc->flags & AV_PIX_FMT_FLAG_RGB) ? inlink->colorspace :
ass_get_color_space(ass->track->YCbCrMatrix, inlink->colorspace),
+ (desc->flags & AV_PIX_FMT_FLAG_RGB) ? inlink->color_range :
ass_get_color_range(ass->track->YCbCrMatrix, inlink->color_range),
- inlink->alpha_mode, ass->alpha ? FF_DRAW_PROCESS_ALPHA : 0);
+ inlink->alpha_mode, draw_flags);
if (ret < 0) {
av_log(ctx, AV_LOG_ERROR, "Failed to initialize FFDrawContext\n");
return ret;
@@ -206,6 +231,8 @@ static int config_input(AVFilterLink *in
if (ass->shaping != -1)
ass_set_shaper(ass->renderer, ass->shaping);
@@ -413,7 +401,7 @@ Index: FFmpeg/libavfilter/vf_subtitles.c
return 0;
}
@@ -189,20 +216,44 @@ static int filter_frame(AVFilterLink *in
@@ -234,20 +261,44 @@ static int filter_frame(AVFilterLink *in
{
AVFilterContext *ctx = inlink->dst;
AVFilterLink *outlink = ctx->outputs[0];
@@ -460,7 +448,7 @@ Index: FFmpeg/libavfilter/vf_subtitles.c
static const AVFilterPad ass_inputs[] = {
{
.name = "default",
@@ -217,10 +268,6 @@ static const AVFilterPad ass_inputs[] =
@@ -262,10 +313,6 @@ static const AVFilterPad ass_inputs[] =
static const AVOption ass_options[] = {
COMMON_OPTIONS
@@ -471,7 +459,7 @@ Index: FFmpeg/libavfilter/vf_subtitles.c
{NULL},
};
@@ -244,6 +291,9 @@ static av_cold int init_ass(AVFilterCont
@@ -289,6 +336,9 @@ static av_cold int init_ass(AVFilterCont
ass->filename);
return AVERROR(EINVAL);
}
@@ -481,7 +469,7 @@ Index: FFmpeg/libavfilter/vf_subtitles.c
return 0;
}
@@ -265,8 +315,8 @@ const AVFilter ff_vf_ass = {
@@ -310,8 +360,8 @@ const FFFilter ff_vf_ass = {
static const AVOption subtitles_options[] = {
COMMON_OPTIONS
{"charenc", "set input character encoding", OFFSET(charenc), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, FLAGS},
@@ -492,7 +480,7 @@ Index: FFmpeg/libavfilter/vf_subtitles.c
{"force_style", "force subtitle style", OFFSET(force_style), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, FLAGS},
#if FF_ASS_FEATURE_WRAP_UNICODE
{"wrap_unicode", "break lines according to the Unicode Line Breaking Algorithm", OFFSET(wrap_unicode), AV_OPT_TYPE_BOOL, { .i64 = -1 }, -1, 1, FLAGS },
@@ -495,6 +545,8 @@ static av_cold int init_subtitles(AVFilt
@@ -540,6 +590,8 @@ static av_cold int init_subtitles(AVFilt
avsubtitle_free(&sub);
}
debian/patches/0025-add-alphasrc-source-video-filter.patch → debian/patches/0023-add-alphasrc-source-video-filter.patch
Vendored
+20 -25
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -593,6 +593,7 @@ OBJS-$(CONFIG_DRAWBOX_VAAPI_FILTER)
@@ -611,6 +611,7 @@ OBJS-$(CONFIG_DRAWBOX_VAAPI_FILTER)
OBJS-$(CONFIG_ALLRGB_FILTER) += vsrc_testsrc.o
OBJS-$(CONFIG_ALLYUV_FILTER) += vsrc_testsrc.o
@@ -14,19 +14,19 @@ Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -556,6 +556,7 @@ extern const AVFilter ff_vf_drawbox_vaap
@@ -571,6 +571,7 @@ extern const FFFilter ff_vf_drawbox_vaap
extern const AVFilter ff_vsrc_allrgb;
extern const AVFilter ff_vsrc_allyuv;
+extern const AVFilter ff_vsrc_alphasrc;
extern const AVFilter ff_vsrc_cellauto;
extern const AVFilter ff_vsrc_color;
extern const AVFilter ff_vsrc_color_vulkan;
extern const FFFilter ff_vsrc_allrgb;
extern const FFFilter ff_vsrc_allyuv;
+extern const FFFilter ff_vsrc_alphasrc;
extern const FFFilter ff_vsrc_amf_capture;
extern const FFFilter ff_vsrc_cellauto;
extern const FFFilter ff_vsrc_color;
Index: FFmpeg/libavfilter/vsrc_alphasrc.c
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vsrc_alphasrc.c
@@ -0,0 +1,165 @@
@@ -0,0 +1,160 @@
+/*
+ * Copyright (c) 2021 NyanMisaka
+ *
@@ -84,17 +84,12 @@ Index: FFmpeg/libavfilter/vsrc_alphasrc.c
+ return 0;
+}
+
+static int alphasrc_query_formats(AVFilterContext *ctx)
+static int alphasrc_query_formats(const AVFilterContext *ctx,
+ AVFilterFormatsConfig **cfg_in,
+ AVFilterFormatsConfig **cfg_out)
+{
+ AVFilterLink *outlink = ctx->outputs[0];
+ AVFilterFormats *formats = ff_formats_pixdesc_filter(AV_PIX_FMT_FLAG_ALPHA, 0);
+ int ret;
+
+ ret = ff_formats_ref(formats, &outlink->incfg.formats);
+ if (ret < 0)
+ return ret;
+
+ return 0;
+ return ff_set_common_formats2(ctx, cfg_in, cfg_out,
+ ff_formats_pixdesc_filter(AV_PIX_FMT_FLAG_ALPHA, 0));
+}
+
+static int alphasrc_config_output(AVFilterLink *outlink)
@@ -181,14 +176,14 @@ Index: FFmpeg/libavfilter/vsrc_alphasrc.c
+ },
+};
+
+const AVFilter ff_vsrc_alphasrc = {
+ .name = "alphasrc",
+ .description = NULL_IF_CONFIG_SMALL("Provide a blank video input with alpha channel."),
+const FFFilter ff_vsrc_alphasrc = {
+ .p.name = "alphasrc",
+ .p.description = NULL_IF_CONFIG_SMALL("Provide a blank video input with alpha channel."),
+ .priv_size = sizeof(AlphaSrc),
+ .priv_class = &alphasrc_class,
+ .p.priv_class = &alphasrc_class,
+ .init = alphasrc_init,
+ .uninit = NULL,
+ .inputs = NULL,
+ .p.inputs = NULL,
+ FILTER_OUTPUTS(alphasrc_outputs),
+ FILTER_QUERY_FUNC(alphasrc_query_formats),
+ FILTER_QUERY_FUNC2(alphasrc_query_formats),
+};
debian/patches/0023-add-fixes-for-hevc-hdr-decoding-in-bsf.patch
Vendored
-41
View File
@@ -1,41 +0,0 @@
Index: FFmpeg/libavcodec/bsf/hevc_mp4toannexb.c
===================================================================
--- FFmpeg.orig/libavcodec/bsf/hevc_mp4toannexb.c
+++ FFmpeg/libavcodec/bsf/hevc_mp4toannexb.c
@@ -127,6 +127,7 @@ static int hevc_mp4toannexb_filter(AVBSF
int got_irap = 0;
int i, ret = 0;
+ int has_sps = 0, has_pps = 0;
ret = ff_bsf_get_packet(ctx, &in);
if (ret < 0)
@@ -158,11 +159,14 @@ static int hevc_mp4toannexb_filter(AVBSF
}
nalu_type = (bytestream2_peek_byte(&gb) >> 1) & 0x3f;
+ has_sps = (has_sps || nalu_type == HEVC_NAL_SPS);
+ has_pps = (has_pps || nalu_type == HEVC_NAL_PPS);
/* prepend extradata to IRAP frames */
is_irap = nalu_type >= HEVC_NAL_BLA_W_LP &&
nalu_type <= HEVC_NAL_RSV_IRAP_VCL23;
- add_extradata = is_irap && !got_irap;
+ /* ignore the extradata if IRAP frame has sps and pps */
+ add_extradata = is_irap && !got_irap && !(has_sps && has_pps);
extra_size = add_extradata * ctx->par_out->extradata_size;
got_irap |= is_irap;
Index: FFmpeg/tests/fate/hevc.mak
===================================================================
--- FFmpeg.orig/tests/fate/hevc.mak
+++ FFmpeg/tests/fate/hevc.mak
@@ -233,7 +233,7 @@ FATE_HEVC-$(call ALLYES, HEVC_DEMUXER MO
fate-hevc-bsf-mp4toannexb: tests/data/hevc-mp4.mov
fate-hevc-bsf-mp4toannexb: CMD = md5 -i $(TARGET_PATH)/tests/data/hevc-mp4.mov -c:v copy -fflags +bitexact -f hevc
fate-hevc-bsf-mp4toannexb: CMP = oneline
-fate-hevc-bsf-mp4toannexb: REF = 73019329ed7f81c24f9af67c34c640c0
+fate-hevc-bsf-mp4toannexb: REF = 7d05a79c7a6665ae22c0043a4d83a811
fate-hevc-skiploopfilter: CMD = framemd5 -skip_loop_filter nokey -i $(TARGET_SAMPLES)/hevc-conformance/SAO_D_Samsung_5.bit -sws_flags bitexact
FATE_HEVC-$(call FRAMEMD5, HEVC, HEVC, HEVC_PARSER) += fate-hevc-skiploopfilter
debian/patches/0026-increase-the-default-async-depth-for-vaapi-encode.patch → debian/patches/0024-increase-the-default-async-depth-for-vaapi-encode.patch
Vendored
View File
debian/patches/0028-add-multiple-values-tags-and-webp-support-for-id3v2.patch → debian/patches/0025-add-multiple-values-tags-and-webp-support-for-id3v2.patch
Vendored
+1 -9
View File
@@ -2,15 +2,7 @@ Index: FFmpeg/libavformat/id3v2.c
===================================================================
--- FFmpeg.orig/libavformat/id3v2.c
+++ FFmpeg/libavformat/id3v2.c
@@ -137,6 +137,7 @@ const CodecMime ff_id3v2_mime_tags[] = {
{ "image/png", AV_CODEC_ID_PNG },
{ "image/tiff", AV_CODEC_ID_TIFF },
{ "image/bmp", AV_CODEC_ID_BMP },
+ { "image/webp", AV_CODEC_ID_WEBP },
{ "JPG", AV_CODEC_ID_MJPEG }, /* ID3v2.2 */
{ "PNG", AV_CODEC_ID_PNG }, /* ID3v2.2 */
{ "", AV_CODEC_ID_NONE },
@@ -326,39 +327,54 @@ static void read_ttag(AVFormatContext *s
@@ -335,39 +335,54 @@ static void read_ttag(AVFormatContext *s
AVDictionary **metadata, const char *key)
{
uint8_t *dst;
debian/patches/0029-remove-fdk-aac-from-nonfree.patch → debian/patches/0026-remove-fdk-aac-from-nonfree.patch
Vendored
+3 -3
View File
@@ -2,15 +2,15 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -1876,7 +1876,6 @@ EXTERNAL_LIBRARY_GPL_LIST="
@@ -2003,7 +2003,6 @@ EXTERNAL_LIBRARY_GPL_LIST="
EXTERNAL_LIBRARY_NONFREE_LIST="
decklink
- libfdk_aac
libtls
libmpeghdec
"
@@ -1916,6 +1915,7 @@ EXTERNAL_LIBRARY_LIST="
@@ -2044,6 +2043,7 @@ EXTERNAL_LIBRARY_LIST="
libcodec2
libdav1d
libdc1394
debian/patches/0027-add-fixes-for-hevc-vaapi-encoding-on-tgl.patch
Vendored
-16
View File
@@ -1,16 +0,0 @@
Index: FFmpeg/libavcodec/hw_base_encode_h265.c
===================================================================
--- FFmpeg.orig/libavcodec/hw_base_encode_h265.c
+++ FFmpeg/libavcodec/hw_base_encode_h265.c
@@ -216,8 +216,9 @@ int ff_hw_base_encode_init_params_h265(F
sps->log2_min_luma_transform_block_size_minus2 = 0;
sps->log2_diff_max_min_luma_transform_block_size = 3;
// Full transform hierarchy allowed (2-5).
- sps->max_transform_hierarchy_depth_inter = 3;
- sps->max_transform_hierarchy_depth_intra = 3;
+ // Default to 2 based on Programmer's Reference Manuals of Intel graphics.
+ sps->max_transform_hierarchy_depth_inter = 2;
+ sps->max_transform_hierarchy_depth_intra = 2;
// AMP works.
sps->amp_enabled_flag = 1;
// SAO and temporal MVP do not work.
debian/patches/0030-pass-dovi-sidedata-to-hlsenc-and-mpegtsenc.patch → debian/patches/0027-pass-dovi-sidedata-to-hlsenc-and-mpegtsenc.patch
Vendored
View File
debian/patches/0031-add-pause-support-for-ffmpeg-cli.patch → debian/patches/0028-add-pause-support-for-ffmpeg-cli.patch
Vendored
+9 -9
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/fftools/ffmpeg.c
===================================================================
--- FFmpeg.orig/fftools/ffmpeg.c
+++ FFmpeg/fftools/ffmpeg.c
@@ -95,6 +95,9 @@ typedef struct BenchmarkTimeStamps {
@@ -99,6 +99,9 @@ typedef struct BenchmarkTimeStamps {
static BenchmarkTimeStamps get_benchmark_time_stamps(void);
static int64_t getmaxrss(void);
@@ -12,7 +12,7 @@ Index: FFmpeg/fftools/ffmpeg.c
atomic_uint nb_output_dumped = 0;
@@ -113,6 +116,9 @@ int nb_filtergraphs;
@@ -117,6 +120,9 @@ int nb_filtergraphs;
Decoder **decoders;
int nb_decoders;
@@ -22,7 +22,7 @@ Index: FFmpeg/fftools/ffmpeg.c
#if HAVE_TERMIOS_H
/* init terminal so that we can grab keys */
@@ -785,6 +791,20 @@ static void set_tty_echo(int on)
@@ -818,6 +825,20 @@ static void set_tty_echo(int on)
#endif
}
@@ -43,7 +43,7 @@ Index: FFmpeg/fftools/ffmpeg.c
static int check_keyboard_interaction(int64_t cur_time)
{
int i, key;
@@ -801,6 +821,11 @@ static int check_keyboard_interaction(in
@@ -834,6 +855,11 @@ static int check_keyboard_interaction(in
}
if (key == '+') av_log_set_level(av_log_get_level()+10);
if (key == '-') av_log_set_level(av_log_get_level()-10);
@@ -55,7 +55,7 @@ Index: FFmpeg/fftools/ffmpeg.c
if (key == 'c' || key == 'C'){
char buf[4096], target[64], command[256], arg[256] = {0};
double time;
@@ -840,7 +865,9 @@ static int check_keyboard_interaction(in
@@ -873,7 +899,9 @@ static int check_keyboard_interaction(in
"c Send command to first matching filter supporting it\n"
"C Send/Queue command to all matching filters\n"
"h dump packets/hex press to cycle through the 3 states\n"
@@ -65,7 +65,7 @@ Index: FFmpeg/fftools/ffmpeg.c
"s Show QP histogram\n"
);
}
@@ -870,15 +897,20 @@ static int transcode(Scheduler *sch)
@@ -903,15 +931,20 @@ static int transcode(Scheduler *sch)
timer_start = av_gettime_relative();
while (!sch_wait(sch, stats_period, &transcode_ts)) {
@@ -90,7 +90,7 @@ Index: FFmpeg/fftools/ffmpeg.c
/* dump report by using the output first video and audio streams */
print_report(0, timer_start, cur_time, transcode_ts);
@@ -895,11 +927,17 @@ static int transcode(Scheduler *sch)
@@ -928,11 +961,17 @@ static int transcode(Scheduler *sch)
term_exit();
/* dump report by using the first video and audio streams */
@@ -113,7 +113,7 @@ Index: FFmpeg/fftools/ffmpeg.h
===================================================================
--- FFmpeg.orig/fftools/ffmpeg.h
+++ FFmpeg/fftools/ffmpeg.h
@@ -748,6 +748,9 @@ extern int recast_media;
@@ -802,6 +802,9 @@ extern int recast_media;
extern FILE *vstats_file;
@@ -127,7 +127,7 @@ Index: FFmpeg/fftools/ffmpeg_demux.c
===================================================================
--- FFmpeg.orig/fftools/ffmpeg_demux.c
+++ FFmpeg/fftools/ffmpeg_demux.c
@@ -704,6 +704,11 @@ static int input_thread(void *arg)
@@ -748,6 +748,11 @@ static int input_thread(void *arg)
DemuxStream *ds;
unsigned send_flags = 0;
debian/patches/0032-enable-gcc-vectorization-and-fix-cpuflags.patch → debian/patches/0029-enable-gcc-vectorization-and-fix-cpuflags.patch
Vendored
+12 -12
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -5601,7 +5601,7 @@ elif enabled x86; then
@@ -5914,7 +5914,7 @@ elif enabled x86; then
;;
# everything else should support nopl and conditional mov (cmov)
*)
@@ -11,17 +11,17 @@ Index: FFmpeg/configure
enable i686
enable fast_cmov
;;
@@ -7592,7 +7592,9 @@ if enabled icc; then
disable aligned_stack
fi
@@ -8052,8 +8052,8 @@ if enabled icc; then
elif enabled gcc; then
- check_optflags -fno-tree-vectorize
+ case $gcc_basever in
+ 2|2.*|3.*|4.*|5.*|6.*) check_optflags -fno-tree-vectorize ;;
+ esac
check_cflags -Werror=format-security
check_cflags -Werror=implicit-function-declaration
check_cflags -Werror=missing-prototypes
gcc_version=$($cc -dumpversion)
major_version=${gcc_version%%[!0-9]*}
- if [ $major_version -lt 13 ]; then
- # Disable tree-vectorize for GCC <13 - it has historically been buggy.
+ if [ $major_version -lt 11 ]; then
+ # Disable tree-vectorize for GCC <11 - it has historically been buggy.
check_optflags -fno-tree-vectorize
else
case $arch in
Index: FFmpeg/libavcodec/huffyuvdsp.c
===================================================================
--- FFmpeg.orig/libavcodec/huffyuvdsp.c
@@ -66,7 +66,7 @@ Index: FFmpeg/libavcodec/x86/cabac.h
#else
av_always_inline
#endif
+#ifdef __GNUC__
+#if defined(__GNUC__) && !defined(__clang__)
+__attribute__((optimize("-fno-tree-vectorize")))
+#endif
int get_cabac_inline_x86(CABACContext *c, uint8_t *const state)
debian/patches/0033-tune-dxva-align-for-intel-to-avoid-copy-on-qsv.patch → debian/patches/0030-tune-dxva-align-for-intel-to-avoid-copy-on-qsv.patch
Vendored
+6 -6
View File
@@ -35,7 +35,7 @@ Index: FFmpeg/libavutil/hwcontext_d3d11va.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_d3d11va.c
+++ FFmpeg/libavutil/hwcontext_d3d11va.c
@@ -611,6 +611,35 @@ static int d3d11va_device_find_adapter_b
@@ -622,6 +622,35 @@ static int d3d11va_device_find_adapter_b
return -1;
}
@@ -71,7 +71,7 @@ Index: FFmpeg/libavutil/hwcontext_d3d11va.c
static int d3d11va_device_create(AVHWDeviceContext *ctx, const char *device,
AVDictionary *opts, int flags)
{
@@ -688,6 +717,9 @@ static int d3d11va_device_create(AVHWDev
@@ -699,6 +728,9 @@ static int d3d11va_device_create(AVHWDev
ID3D10Multithread_Release(pMultithread);
}
@@ -85,10 +85,10 @@ Index: FFmpeg/libavutil/hwcontext_d3d11va.h
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_d3d11va.h
+++ FFmpeg/libavutil/hwcontext_d3d11va.h
@@ -94,6 +94,16 @@ typedef struct AVD3D11VADeviceContext {
void (*lock)(void *lock_ctx);
void (*unlock)(void *lock_ctx);
void *lock_ctx;
@@ -108,6 +108,16 @@ typedef struct AVD3D11VADeviceContext {
* It applies globally to all AVD3D11VAFramesContext allocated from this device context.
*/
UINT MiscFlags;
+
+ /**
+ * DXGI adapter description of the device.
debian/patches/0034-add-10bit-support-for-cuda-overlay.patch → debian/patches/0031-add-10bit-support-for-cuda-overlay.patch
Vendored
+11 -11
View File
@@ -11,7 +11,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
AV_PIX_FMT_NONE,
};
@@ -104,12 +106,14 @@ typedef struct OverlayCUDAContext {
@@ -98,12 +100,14 @@ typedef struct OverlayCUDAContext {
enum AVPixelFormat in_format_overlay;
enum AVPixelFormat in_format_main;
@@ -28,7 +28,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
CUstream cu_stream;
FFFrameSync fs;
@@ -117,6 +121,7 @@ typedef struct OverlayCUDAContext {
@@ -111,6 +115,7 @@ typedef struct OverlayCUDAContext {
int eval_mode;
int x_position;
int y_position;
@@ -36,7 +36,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
double var_values[VAR_VARS_NB];
char *x_expr, *y_expr;
@@ -184,7 +189,10 @@ static int set_expr(AVExpr **pexpr, cons
@@ -178,7 +183,10 @@ static int set_expr(AVExpr **pexpr, cons
static int formats_match(const enum AVPixelFormat format_main, const enum AVPixelFormat format_overlay) {
switch(format_main) {
case AV_PIX_FMT_NV12:
@@ -48,7 +48,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
case AV_PIX_FMT_YUV420P:
return format_overlay == AV_PIX_FMT_YUV420P ||
format_overlay == AV_PIX_FMT_YUVA420P;
@@ -199,26 +207,34 @@ static int formats_match(const enum AVPi
@@ -193,26 +201,34 @@ static int formats_match(const enum AVPi
static int overlay_cuda_call_kernel(
OverlayCUDAContext *ctx,
int x_position, int y_position,
@@ -88,7 +88,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
DIV_UP(main_width, BLOCK_X), DIV_UP(main_height, BLOCK_Y), 1,
BLOCK_X, BLOCK_Y, 1,
0, ctx->cu_stream, kernel_args, NULL));
@@ -242,8 +258,6 @@ static int overlay_cuda_blend(FFFrameSyn
@@ -236,8 +252,6 @@ static int overlay_cuda_blend(FFFrameSyn
AVFrame *input_main, *input_overlay;
@@ -97,7 +97,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
// read main and overlay frames from inputs
ret = ff_framesync_dualinput_get(fs, &input_main, &input_overlay);
if (ret < 0)
@@ -300,11 +314,15 @@ FF_ENABLE_DEPRECATION_WARNINGS
@@ -285,11 +299,15 @@ static int overlay_cuda_blend(FFFrameSyn
overlay_cuda_call_kernel(ctx,
ctx->x_position, ctx->y_position,
@@ -116,7 +116,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
// overlay rest planes depending on pixel format
@@ -312,29 +330,46 @@ FF_ENABLE_DEPRECATION_WARNINGS
@@ -297,29 +315,46 @@ static int overlay_cuda_blend(FFFrameSyn
case AV_PIX_FMT_NV12:
overlay_cuda_call_kernel(ctx,
ctx->x_position, ctx->y_position / 2,
@@ -181,7 +181,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
break;
default:
av_log(ctx, AV_LOG_ERROR, "Passed unsupported overlay pixel format\n");
@@ -406,7 +441,7 @@ static av_cold void overlay_cuda_uninit(
@@ -388,7 +423,7 @@ static av_cold void overlay_cuda_uninit(
if (ctx->hwctx && ctx->cu_module) {
CUcontext dummy;
CudaFunctions *cu = ctx->hwctx->internal->cuda_dl;
@@ -190,7 +190,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
CHECK_CU(cu->cuModuleUnload(ctx->cu_module));
CHECK_CU(cu->cuCtxPopCurrent(&dummy));
}
@@ -465,6 +500,8 @@ static int overlay_cuda_config_output(AV
@@ -447,6 +482,8 @@ static int overlay_cuda_config_output(AV
return AVERROR(ENOSYS);
}
@@ -199,7 +199,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
// check overlay input formats
if (!frames_ctx_overlay) {
@@ -518,7 +555,13 @@ static int overlay_cuda_config_output(AV
@@ -500,7 +537,13 @@ static int overlay_cuda_config_output(AV
return err;
}
@@ -214,7 +214,7 @@ Index: FFmpeg/libavfilter/vf_overlay_cuda.c
if (err < 0) {
CHECK_CU(cu->cuCtxPopCurrent(&dummy));
return err;
@@ -543,6 +586,9 @@ static int overlay_cuda_config_output(AV
@@ -525,6 +568,9 @@ static int overlay_cuda_config_output(AV
static const AVOption overlay_cuda_options[] = {
{ "x", "set the x expression of overlay", OFFSET(x_expr), AV_OPT_TYPE_STRING, { .str = "0" }, 0, 0, FLAGS },
{ "y", "set the y expression of overlay", OFFSET(y_expr), AV_OPT_TYPE_STRING, { .str = "0" }, 0, 0, FLAGS },
debian/patches/0035-add-hdr-metadata-for-nvenc-hevc-encoder.patch → debian/patches/0032-add-hdr-metadata-for-nvenc-hevc-encoder.patch
Vendored
+9 -13
View File
@@ -2,27 +2,21 @@ Index: FFmpeg/libavcodec/nvenc.c
===================================================================
--- FFmpeg.orig/libavcodec/nvenc.c
+++ FFmpeg/libavcodec/nvenc.c
@@ -24,6 +24,7 @@
@@ -24,6 +24,9 @@
#include "nvenc.h"
#include "hevc/sei.h"
+#ifndef NVENC_HAVE_HEVC_AND_AV1_MASTERING_METADATA
+#include "put_bits.h"
+#endif
#if CONFIG_AV1_NVENC_ENCODER
#include "av1.h"
#endif
@@ -32,6 +33,7 @@
#include "libavutil/hwcontext.h"
#include "libavutil/cuda_check.h"
#include "libavutil/imgutils.h"
+#include "libavutil/mastering_display_metadata.h"
#include "libavutil/mem.h"
#include "libavutil/pixdesc.h"
#include "libavutil/mathematics.h"
@@ -2639,6 +2641,80 @@ static int prepare_sei_data_array(AVCode
}
@@ -2902,6 +2905,82 @@ static int prepare_sei_data_array(AVCode
}
}
+
+#ifndef NVENC_HAVE_HEVC_AND_AV1_MASTERING_METADATA
+ if (avctx->codec->id == AV_CODEC_ID_HEVC) {
+ AVFrameSideData *sd = av_frame_get_side_data(frame, AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+
@@ -96,6 +90,8 @@ Index: FFmpeg/libavcodec/nvenc.c
+ }
+ }
+ }
+#endif
+
if (!ctx->udu_sei)
return sei_count;
debian/patches/0036-add-d3d11va-dxva2-hevc-422-444-decoding-on-intel.patch → debian/patches/0033-add-d3d11va-dxva2-hevc-422-444-decoding-on-intel.patch
Vendored
+10 -12
View File
@@ -11,7 +11,7 @@ Index: FFmpeg/libavcodec/d3d12va_hevc.c
DXVA_Qmatrix_HEVC qm;
unsigned slice_count;
DXVA_Slice_HEVC_Short slice_short[MAX_SLICES];
@@ -151,12 +151,12 @@ static int d3d12va_hevc_end_frame(AVCode
@@ -154,12 +154,12 @@ static int d3d12va_hevc_end_frame(AVCode
HEVCContext *h = avctx->priv_data;
HEVCDecodePictureContext *ctx_pic = h->cur_frame->hwaccel_picture_private;
@@ -167,7 +167,7 @@ Index: FFmpeg/libavcodec/dxva2.c
}
}
@@ -627,8 +688,7 @@ int ff_dxva2_common_frame_params(AVCodec
@@ -637,8 +698,7 @@ int ff_dxva2_common_frame_params(AVCodec
else
num_surfaces += 2 + 4; /* 4 base work surface in vpp async */
@@ -243,7 +243,7 @@ Index: FFmpeg/libavcodec/dxva2_hevc.c
pp->wFormatAndSequenceInfoFlags = (sps->chroma_format_idc << 0) |
(sps->separate_colour_plane << 2) |
((sps->bit_depth - 8) << 3) |
@@ -413,14 +443,15 @@ static int dxva2_hevc_end_frame(AVCodecC
@@ -414,14 +444,15 @@ static int dxva2_hevc_end_frame(AVCodecC
{
HEVCContext *h = avctx->priv_data;
struct hevc_dxva2_picture_context *ctx_pic = h->cur_frame->hwaccel_picture_private;
@@ -324,22 +324,20 @@ Index: FFmpeg/libavcodec/dxva2_internal.h
#if CONFIG_D3D12VA
unsigned ff_d3d12va_get_surface_index(const AVCodecContext *avctx,
D3D12VADecodeContext *ctx, const AVFrame *frame,
@@ -171,7 +223,9 @@ void ff_dxva2_h264_fill_picture_paramete
@@ -172,7 +224,7 @@ void ff_dxva2_h264_fill_picture_paramete
void ff_dxva2_h264_fill_scaling_lists(const AVCodecContext *avctx, AVDXVAContext *ctx, DXVA_Qmatrix_H264 *qm);
#if CONFIG_HEVC_D3D12VA_HWACCEL || CONFIG_HEVC_D3D11VA_HWACCEL || CONFIG_HEVC_D3D11VA2_HWACCEL || CONFIG_HEVC_DXVA2_HWACCEL
-void ff_dxva2_hevc_fill_picture_parameters(const AVCodecContext *avctx, AVDXVAContext *ctx, DXVA_PicParams_HEVC *pp);
+#if CONFIG_HEVC_D3D12VA_HWACCEL || CONFIG_HEVC_D3D11VA_HWACCEL || CONFIG_HEVC_D3D11VA2_HWACCEL || CONFIG_HEVC_DXVA2_HWACCEL
+void ff_dxva2_hevc_fill_picture_parameters(const AVCodecContext *avctx, AVDXVAContext *ctx, ff_DXVA_PicParams_HEVC_Rext *ppext);
+#endif
void ff_dxva2_hevc_fill_scaling_lists(const AVCodecContext *avctx, AVDXVAContext *ctx, DXVA_Qmatrix_HEVC *qm);
#endif
Index: FFmpeg/libavcodec/hevc/hevcdec.c
===================================================================
--- FFmpeg.orig/libavcodec/hevc/hevcdec.c
+++ FFmpeg/libavcodec/hevc/hevcdec.c
@@ -592,6 +592,13 @@ static enum AVPixelFormat get_format(HEV
@@ -648,6 +648,13 @@ static enum AVPixelFormat get_format(HEV
#endif
break;
case AV_PIX_FMT_YUV444P:
@@ -353,7 +351,7 @@ Index: FFmpeg/libavcodec/hevc/hevcdec.c
#if CONFIG_HEVC_VAAPI_HWACCEL
*fmt++ = AV_PIX_FMT_VAAPI;
#endif
@@ -610,6 +617,13 @@ static enum AVPixelFormat get_format(HEV
@@ -666,6 +673,13 @@ static enum AVPixelFormat get_format(HEV
break;
case AV_PIX_FMT_YUV422P:
case AV_PIX_FMT_YUV422P10LE:
@@ -367,7 +365,7 @@ Index: FFmpeg/libavcodec/hevc/hevcdec.c
#if CONFIG_HEVC_VAAPI_HWACCEL
*fmt++ = AV_PIX_FMT_VAAPI;
#endif
@@ -627,6 +641,13 @@ static enum AVPixelFormat get_format(HEV
@@ -686,6 +700,13 @@ static enum AVPixelFormat get_format(HEV
/* NOTE: fallthrough */
case AV_PIX_FMT_YUV420P12:
case AV_PIX_FMT_YUV444P12:
@@ -381,7 +379,7 @@ Index: FFmpeg/libavcodec/hevc/hevcdec.c
#if CONFIG_HEVC_VAAPI_HWACCEL
*fmt++ = AV_PIX_FMT_VAAPI;
#endif
@@ -641,6 +662,13 @@ static enum AVPixelFormat get_format(HEV
@@ -700,6 +721,13 @@ static enum AVPixelFormat get_format(HEV
#endif
break;
case AV_PIX_FMT_YUV422P12:
debian/patches/0037-default-vaapi-scaler-algorithm-to-fast.patch → debian/patches/0034-default-vaapi-scaler-algorithm-to-fast.patch
Vendored
View File
debian/patches/0038-dont-fail-on-vaapi-to-drm-export-vasync.patch → debian/patches/0035-dont-fail-on-vaapi-to-drm-export-vasync.patch
Vendored
View File
debian/patches/0039-skip-loading-plugins-on-vpl-runtime.patch → debian/patches/0036-skip-loading-plugins-on-vpl-runtime.patch
Vendored
View File
debian/patches/0040-tune-the-default-bf-for-qsv-enc.patch → debian/patches/0037-tune-the-default-bf-for-qsv-enc.patch
Vendored
View File
debian/patches/0041-allow-vaapi-import-drm-prime2-planar-formats.patch → debian/patches/0038-allow-vaapi-import-drm-prime2-planar-formats.patch
Vendored
View File
debian/patches/0042-add-vui-info-to-the-seq-header-of-hevc-vaapi-encoder.patch → debian/patches/0039-add-vui-info-to-the-seq-header-of-hevc-vaapi-encoder.patch
Vendored
View File
debian/patches/0043-sync-intel-d3d11va-textures-before-mapping-to-opencl.patch → debian/patches/0040-sync-intel-d3d11va-textures-before-mapping-to-opencl.patch
Vendored
View File
debian/patches/0044-add-icon-for-windows-version-ffmpeg.patch → debian/patches/0041-add-icon-for-windows-version-ffmpeg.patch
Vendored
View File
debian/patches/0046-add-full-hwa-pipeline-for-rockchip-rk3588-platform.patch → debian/patches/0042-add-full-hwa-pipeline-for-rockchip-rk3588-platform.patch
Vendored
+1048 -573
View File
File diff suppressed because it is too large Load Diff
debian/patches/0043-add-format-option-to-vt-scale-filter.patch
Vendored
+121
View File
@@ -0,0 +1,121 @@
Index: FFmpeg/libavfilter/vf_scale_vt.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_scale_vt.c
+++ FFmpeg/libavfilter/vf_scale_vt.c
@@ -41,11 +41,26 @@ typedef struct ScaleVtContext {
enum AVColorPrimaries colour_primaries;
enum AVColorTransferCharacteristic colour_transfer;
enum AVColorSpace colour_matrix;
+ enum AVPixelFormat format;
char *colour_primaries_string;
char *colour_transfer_string;
char *colour_matrix_string;
} ScaleVtContext;
+static const enum AVPixelFormat supported_formats[] = {
+ AV_PIX_FMT_NV12,
+ AV_PIX_FMT_P010,
+ AV_PIX_FMT_NONE,
+};
+
+static int format_is_supported(enum AVPixelFormat fmt)
+{
+ for (int i = 0; i < FF_ARRAY_ELEMS(supported_formats); i++)
+ if (supported_formats[i] == fmt)
+ return 1;
+ return 0;
+}
+
static av_cold int scale_vt_init(AVFilterContext *avctx)
{
ScaleVtContext *s = avctx->priv;
@@ -58,24 +73,6 @@ static av_cold int scale_vt_init(AVFilte
return AVERROR_EXTERNAL;
}
-#define STRING_OPTION(var_name, func_name, default_value) \
- do { \
- if (s->var_name##_string) { \
- int var = av_##func_name##_from_name(s->var_name##_string); \
- if (var < 0) { \
- av_log(avctx, AV_LOG_ERROR, "Invalid %s.\n", #var_name); \
- return AVERROR(EINVAL); \
- } \
- s->var_name = var; \
- } else { \
- s->var_name = default_value; \
- } \
- } while (0)
-
- STRING_OPTION(colour_primaries, color_primaries, AVCOL_PRI_UNSPECIFIED);
- STRING_OPTION(colour_transfer, color_transfer, AVCOL_TRC_UNSPECIFIED);
- STRING_OPTION(colour_matrix, color_space, AVCOL_SPC_UNSPECIFIED);
-
if (s->colour_primaries != AVCOL_PRI_UNSPECIFIED) {
value = av_map_videotoolbox_color_primaries_from_av(s->colour_primaries);
if (!value) {
@@ -241,6 +238,35 @@ static int scale_vt_config_output(AVFilt
FilterLink *inl = ff_filter_link(inlink);
AVHWFramesContext *hw_frame_ctx_in;
AVHWFramesContext *hw_frame_ctx_out;
+ enum AVPixelFormat out_format;
+
+#define STRING_OPTION(var_name, func_name, default_value) \
+ do { \
+ if (s->var_name##_string) { \
+ int var = av_##func_name##_from_name(s->var_name##_string); \
+ if (var < 0) { \
+ av_log(avctx, AV_LOG_ERROR, "Invalid %s.\n", #var_name); \
+ return AVERROR(EINVAL); \
+ } \
+ s->var_name = var; \
+ } else { \
+ s->var_name = default_value; \
+ } \
+ } while (0)
+
+ STRING_OPTION(colour_primaries, color_primaries, AVCOL_PRI_UNSPECIFIED);
+ STRING_OPTION(colour_transfer, color_transfer, AVCOL_TRC_UNSPECIFIED);
+ STRING_OPTION(colour_matrix, color_space, AVCOL_SPC_UNSPECIFIED);
+
+ if (!((s->colour_primaries == AVCOL_PRI_UNSPECIFIED &&
+ s->colour_transfer == AVCOL_TRC_UNSPECIFIED) ||
+ (s->colour_primaries == AVCOL_PRI_BT2020 &&
+ s->colour_transfer == AVCOL_TRC_SMPTE2084))) {
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
err = ff_scale_eval_dimensions(s, s->w_expr, s->h_expr, inlink, outlink,
&s->output_width,
@@ -263,11 +289,18 @@ static int scale_vt_config_output(AVFilt
hw_frame_ctx_in = (AVHWFramesContext *)inl->hw_frames_ctx->data;
+ out_format = (s->format == AV_PIX_FMT_NONE) ? hw_frame_ctx_in->sw_format : s->format;
+ if (!format_is_supported(s->format)) {
+ av_log(s, AV_LOG_ERROR, "Unsupported output format: %s\n",
+ av_get_pix_fmt_name(out_format));
+ return AVERROR(ENOSYS);
+ }
+
av_buffer_unref(&outl->hw_frames_ctx);
outl->hw_frames_ctx = av_hwframe_ctx_alloc(hw_frame_ctx_in->device_ref);
hw_frame_ctx_out = (AVHWFramesContext *)outl->hw_frames_ctx->data;
hw_frame_ctx_out->format = AV_PIX_FMT_VIDEOTOOLBOX;
- hw_frame_ctx_out->sw_format = hw_frame_ctx_in->sw_format;
+ hw_frame_ctx_out->sw_format = out_format;
hw_frame_ctx_out->width = outlink->w;
hw_frame_ctx_out->height = outlink->h;
((AVVTFramesContext *)hw_frame_ctx_out->hwctx)->color_range = ((AVVTFramesContext *)hw_frame_ctx_in->hwctx)->color_range;
@@ -300,6 +333,8 @@ static const AVOption scale_vt_options[]
OFFSET(colour_primaries_string), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS },
{ "color_transfer", "Output colour transfer characteristics",
OFFSET(colour_transfer_string), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS },
+ { "format", "Output pixel format",
+ OFFSET(format), AV_OPT_TYPE_PIXEL_FMT, { .i64 = AV_PIX_FMT_NONE }, AV_PIX_FMT_NONE, INT_MAX, .flags = FLAGS },
{ NULL },
};
debian/patches/0048-add-vf-overlay-videotoolbox-filter.patch → debian/patches/0044-add-vf-overlay-videotoolbox-filter.patch
Vendored
+26 -88
View File
@@ -2,118 +2,55 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3928,6 +3928,7 @@ overlay_opencl_filter_deps="opencl"
overlay_qsv_filter_deps="libmfx"
@@ -4191,6 +4191,7 @@ overlay_qsv_filter_deps="libmfx"
overlay_qsv_filter_select="qsvvpp"
overlay_rkrga_filter_deps="rkrga"
overlay_vaapi_filter_deps="vaapi VAProcPipelineCaps_blend_flags"
+overlay_videotoolbox_filter_deps="metal corevideo coreimage videotoolbox"
overlay_vulkan_filter_deps="vulkan spirv_compiler"
overlay_rkrga_filter_deps="rkrga"
overlay_vulkan_filter_deps="vulkan spirv_library"
owdenoise_filter_deps="gpl"
Index: FFmpeg/doc/filters.texi
===================================================================
--- FFmpeg.orig/doc/filters.texi
+++ FFmpeg/doc/filters.texi
@@ -19050,6 +19050,58 @@ See @ref{framesync}.
This filter also supports the @ref{framesync} options.
+@section overlay_videotoolbox
+
+Overlay one video on top of another.
+
+This is the VideoToolbox variant of the @ref{overlay} filter.
+It takes two inputs and has one output. The first input is the "main" video on which the second input is overlaid.
+It only accepts VideoToolbox frames. The underlying input pixel formats do not have to match.
+Different input pixel formats and color spaces will be automatically converted using hardware accelerated methods.
+The final output will have the same pixel format and color space as the "main" input.
+
+The filter accepts the following options:
+
+@table @option
+
+@item x
+Set the x coordinate of the overlaid video on the main video.
+Default value is @code{0}.
+
+@item y
+Set the y coordinate of the overlaid video on the main video.
+Default value is @code{0}.
+
+@item eof_action
+See @ref{framesync}.
+
+@item shortest
+See @ref{framesync}.
+
+@item repeatlast
+See @ref{framesync}.
+
+@end table
+
+@subsection Examples
+
+@itemize
+@item
+Overlay an image LOGO at the top-left corner of the INPUT video.
+The INPUT video is in nv12 format and the LOGO image is in rgba format.
+@example
+-hwaccel videotoolbox -i INPUT -i LOGO -codec:v:0 h264_videotoolbox -filter_complex "[0:v]format=nv12,hwupload[a], [1:v]format=rgba,hwupload[b], [a][b]overlay_videotoolbox" OUTPUT
+@end example
+@item
+Overlay an SDR video OVERLAY at the top-left corner of the HDR video MAIN.
+The INPUT video is in p010 format and the LOGO image is in nv12 format.
+The OUTPUT video will also be an HDR video with OVERLAY mapped to HDR.
+@example
+-hwaccel videotoolbox -i MAIN -i OVERLAY -codec:v:0 hevc_videotoolbox -tag:v hvc1 -filter_complex "[0:v]format=p010,hwupload[a], [1:v]format=nv12,hwupload[b], [a][b]overlay_videotoolbox" OUTPUT
+@end example
+
+@end itemize
+
@section owdenoise
Apply Overcomplete Wavelet denoiser.
pad_opencl_filter_deps="opencl"
Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -416,6 +416,9 @@ OBJS-$(CONFIG_OVERLAY_OPENCL_FILTER)
opencl/overlay.o framesync.o
@@ -429,6 +429,9 @@ OBJS-$(CONFIG_OVERLAY_OPENCL_FILTER)
OBJS-$(CONFIG_OVERLAY_QSV_FILTER) += vf_overlay_qsv.o framesync.o
OBJS-$(CONFIG_OVERLAY_RKRGA_FILTER) += vf_overlay_rkrga.o framesync.o
OBJS-$(CONFIG_OVERLAY_VAAPI_FILTER) += vf_overlay_vaapi.o framesync.o vaapi_vpp.o
+OBJS-$(CONFIG_OVERLAY_VIDEOTOOLBOX_FILTER) += vf_overlay_videotoolbox.o framesync.o \
+ metal/vf_overlay_videotoolbox.metallib.o \
+ metal/utils.o
OBJS-$(CONFIG_OVERLAY_VULKAN_FILTER) += vf_overlay_vulkan.o vulkan.o vulkan_filter.o
OBJS-$(CONFIG_OVERLAY_RKRGA_FILTER) += vf_overlay_rkrga.o framesync.o
OBJS-$(CONFIG_OWDENOISE_FILTER) += vf_owdenoise.o
OBJS-$(CONFIG_PAD_FILTER) += vf_pad.o
Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -389,6 +389,7 @@ extern const AVFilter ff_vf_overlay;
extern const AVFilter ff_vf_overlay_opencl;
extern const AVFilter ff_vf_overlay_qsv;
extern const AVFilter ff_vf_overlay_vaapi;
+extern const AVFilter ff_vf_overlay_videotoolbox;
extern const AVFilter ff_vf_overlay_vulkan;
extern const AVFilter ff_vf_overlay_cuda;
extern const AVFilter ff_vf_overlay_rkrga;
@@ -399,6 +399,7 @@ extern const FFFilter ff_vf_overlay_open
extern const FFFilter ff_vf_overlay_qsv;
extern const FFFilter ff_vf_overlay_rkrga;
extern const FFFilter ff_vf_overlay_vaapi;
+extern const FFFilter ff_vf_overlay_videotoolbox;
extern const FFFilter ff_vf_overlay_vulkan;
extern const FFFilter ff_vf_overlay_cuda;
extern const FFFilter ff_vf_owdenoise;
Index: FFmpeg/libavfilter/metal/utils.m
===================================================================
--- FFmpeg.orig/libavfilter/metal/utils.m
+++ FFmpeg/libavfilter/metal/utils.m
@@ -55,6 +55,9 @@ CVMetalTextureRef ff_metal_texture_from_
@@ -56,6 +56,9 @@ CVMetalTextureRef ff_metal_texture_from_
{
CVMetalTextureRef tex = NULL;
CVReturn ret;
+ bool is_planer = CVPixelBufferIsPlanar(pixbuf);
+ size_t width = is_planer ? CVPixelBufferGetWidthOfPlane(pixbuf, plane) : CVPixelBufferGetWidth(pixbuf);
+ size_t height = is_planer ? CVPixelBufferGetHeightOfPlane(pixbuf, plane) : CVPixelBufferGetHeight(pixbuf);
+ bool is_planar = CVPixelBufferIsPlanar(pixbuf);
+ size_t width = is_planar ? CVPixelBufferGetWidthOfPlane(pixbuf, plane) : CVPixelBufferGetWidth(pixbuf);
+ size_t height = is_planar ? CVPixelBufferGetHeightOfPlane(pixbuf, plane) : CVPixelBufferGetHeight(pixbuf);
ret = CVMetalTextureCacheCreateTextureFromImage(
NULL,
@@ -62,8 +65,8 @@ CVMetalTextureRef ff_metal_texture_from_
@@ -63,8 +66,8 @@ CVMetalTextureRef ff_metal_texture_from_
pixbuf,
NULL,
format,
@@ -232,7 +169,7 @@ Index: FFmpeg/libavfilter/vf_overlay_videotoolbox.m
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_overlay_videotoolbox.m
@@ -0,0 +1,758 @@
@@ -0,0 +1,759 @@
+/*
+ * Copyright (C) 2024 Gnattu OC <gnattuoc@me.com>
+ *
@@ -977,11 +914,11 @@ Index: FFmpeg/libavfilter/vf_overlay_videotoolbox.m
+ },
+};
+
+const AVFilter ff_vf_overlay_videotoolbox = {
+ .name = "overlay_videotoolbox",
+ .description = NULL_IF_CONFIG_SMALL("Overlay filter for VideoToolbox frames using Metal compute"),
+const FFFilter ff_vf_overlay_videotoolbox = {
+ .p.name = "overlay_videotoolbox",
+ .p.description = NULL_IF_CONFIG_SMALL("Overlay filter for VideoToolbox frames using Metal compute"),
+ .priv_size = OVERLAY_VT_CTX_SIZE,
+ .priv_class = &overlay_videotoolbox_class,
+ .p.priv_class = &overlay_videotoolbox_class,
+ .init = overlay_videotoolbox_init,
+ .uninit = overlay_videotoolbox_uninit,
+ .activate = overlay_videotoolbox_activate,
@@ -990,4 +927,5 @@ Index: FFmpeg/libavfilter/vf_overlay_videotoolbox.m
+ FILTER_INPUTS(overlay_videotoolbox_inputs),
+ FILTER_OUTPUTS(overlay_videotoolbox_outputs),
+ .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+ .p.flags = AVFILTER_FLAG_HWDEVICE,
+};
debian/patches/0049-add-coreimage-based-vf-transpose-vt-filter.patch → debian/patches/0045-add-coreimage-based-vf-transpose-vt-filter.patch
Vendored
+13 -12
View File
@@ -2,13 +2,13 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3979,7 +3979,7 @@ tonemap_vaapi_filter_deps="vaapi VAProcF
@@ -4243,7 +4243,7 @@ tonemap_vaapi_filter_deps="vaapi VAProcF
tonemap_opencl_filter_deps="opencl const_nan"
transpose_opencl_filter_deps="opencl"
transpose_vaapi_filter_deps="vaapi VAProcPipelineCaps_rotation_flags"
-transpose_vt_filter_deps="videotoolbox VTPixelRotationSessionCreate"
+transpose_vt_filter_deps="coreimage videotoolbox"
transpose_vulkan_filter_deps="vulkan spirv_compiler"
transpose_vulkan_filter_deps="vulkan spirv_library"
unsharp_opencl_filter_deps="opencl"
uspp_filter_deps="gpl avcodec"
Index: FFmpeg/libavfilter/vf_transpose_vt.c
@@ -294,24 +294,24 @@ Index: FFmpeg/libavfilter/vf_transpose_vt.c
- },
-};
-
-const AVFilter ff_vf_transpose_vt = {
- .name = "transpose_vt",
- .description = NULL_IF_CONFIG_SMALL("Transpose Videotoolbox frames"),
-const FFFilter ff_vf_transpose_vt = {
- .p.name = "transpose_vt",
- .p.description = NULL_IF_CONFIG_SMALL("Transpose Videotoolbox frames"),
- .p.priv_class = &transpose_vt_class,
- .p.flags = AVFILTER_FLAG_HWDEVICE,
- .priv_size = sizeof(TransposeVtContext),
- .init = transpose_vt_init,
- .uninit = transpose_vt_uninit,
- FILTER_INPUTS(transpose_vt_inputs),
- FILTER_OUTPUTS(transpose_vt_outputs),
- FILTER_SINGLE_PIXFMT(AV_PIX_FMT_VIDEOTOOLBOX),
- .priv_class = &transpose_vt_class,
- .flags = AVFILTER_FLAG_HWDEVICE,
- .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
-};
Index: FFmpeg/libavfilter/vf_transpose_vt.m
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_transpose_vt.m
@@ -0,0 +1,383 @@
@@ -0,0 +1,384 @@
+/*
+ * Copyright (c) 2023 Zhao Zhili <zhilizhao@tencent.com>
+ * Copyright (c) 2024 Gnattu OC <gnattuoc@me.com>
@@ -683,15 +683,16 @@ Index: FFmpeg/libavfilter/vf_transpose_vt.m
+ },
+};
+
+const AVFilter ff_vf_transpose_vt = {
+ .name = "transpose_vt",
+ .description = NULL_IF_CONFIG_SMALL("Transpose Videotoolbox frames"),
+const FFFilter ff_vf_transpose_vt = {
+ .p.name = "transpose_vt",
+ .p.description = NULL_IF_CONFIG_SMALL("Transpose Videotoolbox frames"),
+ .priv_size = sizeof(TransposeVtContext),
+ .init = transpose_vt_init,
+ .uninit = transpose_vt_uninit,
+ FILTER_INPUTS(transpose_vt_inputs),
+ FILTER_OUTPUTS(transpose_vt_outputs),
+ FILTER_SINGLE_PIXFMT(AV_PIX_FMT_VIDEOTOOLBOX),
+ .priv_class = &transpose_vt_class,
+ .p.priv_class = &transpose_vt_class,
+ .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+ .p.flags = AVFILTER_FLAG_HWDEVICE,
+};
debian/patches/0045-fix-libx265-encoded-fmp4-hls-playback-on-safari.patch
Vendored
-13
View File
@@ -1,13 +0,0 @@
Index: FFmpeg/libavformat/movenc.c
===================================================================
--- FFmpeg.orig/libavformat/movenc.c
+++ FFmpeg/libavformat/movenc.c
@@ -3202,7 +3202,7 @@ static int mov_write_stbl_tag(AVFormatCo
track->par->codec_tag == MKTAG('r','t','p',' ')) &&
track->has_keyframes && track->has_keyframes < track->entry)
mov_write_stss_tag(pb, track, MOV_SYNC_SAMPLE);
- if (track->par->codec_type == AVMEDIA_TYPE_VIDEO && track->has_disposable)
+ if (track->par->codec_type == AVMEDIA_TYPE_VIDEO && track->has_disposable && track->entry)
mov_write_sdtp_tag(pb, track);
if (track->mode == MODE_MOV && track->flags & MOV_TRACK_STPS)
mov_write_stss_tag(pb, track, MOV_PARTIAL_SYNC_SAMPLE);
debian/patches/0050-add-vf-tonemap-videotoolbox-filter.patch → debian/patches/0046-add-vf-tonemap-videotoolbox-filter.patch
Vendored
+32 -22
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3976,6 +3976,7 @@ tinterlace_merge_test_deps="tinterlace_f
@@ -4240,6 +4240,7 @@ tinterlace_merge_test_deps="tinterlace_f
tinterlace_pad_test_deps="tinterlace_filter"
tonemap_filter_deps="const_nan"
tonemap_vaapi_filter_deps="vaapi VAProcFilterParameterBufferHDRToneMapping"
@@ -14,7 +14,7 @@ Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -537,6 +537,9 @@ OBJS-$(CONFIG_TONEMAP_CUDA_FILTER)
@@ -554,6 +554,9 @@ OBJS-$(CONFIG_TONEMAP_CUDA_FILTER)
OBJS-$(CONFIG_TONEMAP_OPENCL_FILTER) += vf_tonemap_opencl.o opencl.o \
opencl/tonemap.o opencl/colorspace_common.o
OBJS-$(CONFIG_TONEMAP_VAAPI_FILTER) += vf_tonemap_vaapi.o vaapi_vpp.o
@@ -28,14 +28,14 @@ Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -502,6 +502,7 @@ extern const AVFilter ff_vf_tonemap;
extern const AVFilter ff_vf_tonemap_cuda;
extern const AVFilter ff_vf_tonemap_opencl;
extern const AVFilter ff_vf_tonemap_vaapi;
+extern const AVFilter ff_vf_tonemap_videotoolbox;
extern const AVFilter ff_vf_tpad;
extern const AVFilter ff_vf_transpose;
extern const AVFilter ff_vf_transpose_npp;
@@ -517,6 +517,7 @@ extern const FFFilter ff_vf_tonemap;
extern const FFFilter ff_vf_tonemap_cuda;
extern const FFFilter ff_vf_tonemap_opencl;
extern const FFFilter ff_vf_tonemap_vaapi;
+extern const FFFilter ff_vf_tonemap_videotoolbox;
extern const FFFilter ff_vf_tpad;
extern const FFFilter ff_vf_transpose;
extern const FFFilter ff_vf_transpose_npp;
Index: FFmpeg/libavfilter/metal/vf_tonemap_videotoolbox.metal
===================================================================
--- /dev/null
@@ -966,7 +966,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_videotoolbox.m
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_tonemap_videotoolbox.m
@@ -0,0 +1,1211 @@
@@ -0,0 +1,1221 @@
+/*
+ * Copyright (c) 2024 Gnattu OC <gnattuoc@me.com>
+ *
@@ -1061,9 +1061,11 @@ Index: FFmpeg/libavfilter/vf_tonemap_videotoolbox.m
+ enum AVPixelFormat in_fmt, out_fmt;
+ const AVPixFmtDescriptor *in_desc, *out_desc;
+ int in_planes, out_planes;
+ struct DoviMetadata *dovi;
+ enum TonemapAlgorithm tonemap;
+ enum TonemapMode tonemap_mode;
+ struct FFDOVIMetadataRemap *dovi;
+ /* enum TonemapAlgorithm */
+ int tonemap;
+ /* enum TonemapMode */
+ int tonemap_mode;
+ enum AVPixelFormat format;
+ int apply_dovi;
+ double peak;
@@ -1087,12 +1089,12 @@ Index: FFmpeg/libavfilter/vf_tonemap_videotoolbox.m
+ CVMetalTextureCacheRef texture_cache;
+} TonemapVideoToolboxContext;
+
+static const short linearize_funcs[AVCOL_TRC_NB] = {
+static const short linearize_funcs[] = {
+ [AVCOL_TRC_SMPTE2084] = 1, //"eotf_st2084",
+ [AVCOL_TRC_ARIB_STD_B67] = 2, //"eotf_arib_b67",
+};
+
+static const short delinearize_funcs[AVCOL_TRC_NB] = {
+static const short delinearize_funcs[] = {
+ [AVCOL_TRC_BT709] = 1, //"inverse_eotf_bt1886",
+ [AVCOL_TRC_BT2020_10] = 1, //"inverse_eotf_bt1886",
+};
@@ -1171,7 +1173,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_videotoolbox.m
+ for (c = 0; c < 3; c++) {
+ int has_poly = 0, has_mmr = 0, mmr_single = 1;
+ int mmr_idx = 0, min_order = 3, max_order = 1;
+ const struct ReshapeData *comp = &ctx->dovi->comp[c];
+ const struct FFDOVIReshapeData *comp = &ctx->dovi->comp[c];
+ if (!comp->num_pivots)
+ continue;
+ av_assert0(comp->num_pivots >= 2 && comp->num_pivots <= 9);
@@ -1771,6 +1773,13 @@ Index: FFmpeg/libavfilter/vf_tonemap_videotoolbox.m
+ return ret;
+ }
+
+ if (ctx->trc != AVCOL_TRC_SMPTE2084) {
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
+
+ return 0;
+}
+
@@ -1883,7 +1892,7 @@ Index: FFmpeg/libavfilter/vf_tonemap_videotoolbox.m
+ const AVDOVIRpuDataHeader *rpu = av_dovi_get_header(metadata);
+ // only map dovi rpus that don't require an EL
+ if (rpu->disable_residual_flag) {
+ struct DoviMetadata *dovi = av_malloc(sizeof(*dovi));
+ struct FFDOVIMetadataRemap *dovi = av_malloc(sizeof(*dovi));
+ ctx->dovi = dovi;
+ if (!ctx->dovi)
+ goto fail;
@@ -2166,15 +2175,16 @@ Index: FFmpeg/libavfilter/vf_tonemap_videotoolbox.m
+ },
+};
+
+const AVFilter ff_vf_tonemap_videotoolbox = {
+ .name = "tonemap_videotoolbox",
+ .description = NULL_IF_CONFIG_SMALL("Perform HDR to SDR conversion with Metal."),
+const FFFilter ff_vf_tonemap_videotoolbox = {
+ .p.name = "tonemap_videotoolbox",
+ .p.description = NULL_IF_CONFIG_SMALL("Perform HDR to SDR conversion with Metal."),
+ .priv_size = sizeof(TonemapVideoToolboxContext),
+ .priv_class = &tonemap_videotoolbox_class,
+ .p.priv_class = &tonemap_videotoolbox_class,
+ .preinit = tonemap_videotoolbox_preinit,
+ .uninit = tonemap_videotoolbox_uninit,
+ FILTER_INPUTS(tonemap_videotoolbox_inputs),
+ FILTER_OUTPUTS(tonemap_videotoolbox_outputs),
+ FILTER_QUERY_FUNC(tonemap_videotoolbox_query_formats),
+ .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+ .p.flags = AVFILTER_FLAG_HWDEVICE,
+};
debian/patches/0047-add-format-option-to-vt-scale-filter.patch
Vendored
-68
View File
@@ -1,68 +0,0 @@
Index: FFmpeg/libavfilter/vf_scale_vt.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_scale_vt.c
+++ FFmpeg/libavfilter/vf_scale_vt.c
@@ -41,11 +41,26 @@ typedef struct ScaleVtContext {
enum AVColorPrimaries colour_primaries;
enum AVColorTransferCharacteristic colour_transfer;
enum AVColorSpace colour_matrix;
+ enum AVPixelFormat format;
char *colour_primaries_string;
char *colour_transfer_string;
char *colour_matrix_string;
} ScaleVtContext;
+static const enum AVPixelFormat supported_formats[] = {
+ AV_PIX_FMT_NV12,
+ AV_PIX_FMT_P010,
+ AV_PIX_FMT_NONE,
+};
+
+static int format_is_supported(enum AVPixelFormat fmt)
+{
+ for (int i = 0; i < FF_ARRAY_ELEMS(supported_formats); i++)
+ if (supported_formats[i] == fmt)
+ return 1;
+ return 0;
+}
+
static av_cold int scale_vt_init(AVFilterContext *avctx)
{
ScaleVtContext *s = avctx->priv;
@@ -182,6 +197,7 @@ static int scale_vt_config_output(AVFilt
FilterLink *inl = ff_filter_link(inlink);
AVHWFramesContext *hw_frame_ctx_in;
AVHWFramesContext *hw_frame_ctx_out;
+ enum AVPixelFormat out_format;
err = ff_scale_eval_dimensions(s, s->w_expr, s->h_expr, inlink, outlink,
&s->output_width,
@@ -201,11 +217,18 @@ static int scale_vt_config_output(AVFilt
hw_frame_ctx_in = (AVHWFramesContext *)inl->hw_frames_ctx->data;
+ out_format = (s->format == AV_PIX_FMT_NONE) ? hw_frame_ctx_in->sw_format : s->format;
+ if (!format_is_supported(s->format)) {
+ av_log(s, AV_LOG_ERROR, "Unsupported output format: %s\n",
+ av_get_pix_fmt_name(out_format));
+ return AVERROR(ENOSYS);
+ }
+
av_buffer_unref(&outl->hw_frames_ctx);
outl->hw_frames_ctx = av_hwframe_ctx_alloc(hw_frame_ctx_in->device_ref);
hw_frame_ctx_out = (AVHWFramesContext *)outl->hw_frames_ctx->data;
hw_frame_ctx_out->format = AV_PIX_FMT_VIDEOTOOLBOX;
- hw_frame_ctx_out->sw_format = hw_frame_ctx_in->sw_format;
+ hw_frame_ctx_out->sw_format = out_format;
hw_frame_ctx_out->width = outlink->w;
hw_frame_ctx_out->height = outlink->h;
((AVVTFramesContext *)hw_frame_ctx_out->hwctx)->color_range = ((AVVTFramesContext *)hw_frame_ctx_in->hwctx)->color_range;
@@ -238,6 +261,8 @@ static const AVOption scale_vt_options[]
OFFSET(colour_primaries_string), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS },
{ "color_transfer", "Output colour transfer characteristics",
OFFSET(colour_transfer_string), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS },
+ { "format", "Output pixel format",
+ OFFSET(format), AV_OPT_TYPE_PIXEL_FMT, { .i64 = AV_PIX_FMT_NONE }, AV_PIX_FMT_NONE, INT_MAX, .flags = FLAGS },
{ NULL },
};
debian/patches/0051-add-mjpeg-videotoolbox-encoder.patch → debian/patches/0047-add-mjpeg-videotoolbox-encoder.patch
Vendored
+18 -18
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3525,6 +3525,8 @@ h264_videotoolbox_encoder_deps="pthreads
@@ -3767,6 +3767,8 @@ h264_videotoolbox_encoder_deps="pthreads
h264_videotoolbox_encoder_select="atsc_a53 videotoolbox_encoder"
hevc_videotoolbox_encoder_deps="pthreads"
hevc_videotoolbox_encoder_select="atsc_a53 videotoolbox_encoder"
@@ -15,31 +15,31 @@ Index: FFmpeg/libavcodec/Makefile
===================================================================
--- FFmpeg.orig/libavcodec/Makefile
+++ FFmpeg/libavcodec/Makefile
@@ -518,6 +518,7 @@ OBJS-$(CONFIG_MJPEGB_DECODER) +
OBJS-$(CONFIG_MJPEG_CUVID_DECODER) += cuviddec.o
OBJS-$(CONFIG_MJPEG_QSV_ENCODER) += qsvenc_jpeg.o
OBJS-$(CONFIG_MJPEG_VAAPI_ENCODER) += vaapi_encode_mjpeg.o
+OBJS-$(CONFIG_MJPEG_VIDEOTOOLBOX_ENCODER) += videotoolboxenc.o
@@ -544,6 +544,7 @@ OBJS-$(CONFIG_MJPEG_QSV_ENCODER) +
OBJS-$(CONFIG_MJPEG_RKMPP_DECODER) += rkmppdec.o
OBJS-$(CONFIG_MJPEG_RKMPP_ENCODER) += rkmppenc.o
OBJS-$(CONFIG_MJPEG_VAAPI_ENCODER) += vaapi_encode_mjpeg.o
+OBJS-$(CONFIG_MJPEG_VIDEOTOOLBOX_ENCODER) += videotoolboxenc.o
OBJS-$(CONFIG_MLP_DECODER) += mlpdec.o mlpdsp.o
OBJS-$(CONFIG_MLP_ENCODER) += mlpenc.o mlp.o
OBJS-$(CONFIG_MMVIDEO_DECODER) += mmvideo.o
Index: FFmpeg/libavcodec/allcodecs.c
===================================================================
--- FFmpeg.orig/libavcodec/allcodecs.c
+++ FFmpeg/libavcodec/allcodecs.c
@@ -875,6 +875,7 @@ extern const FFCodec ff_mjpeg_cuvid_deco
extern const FFCodec ff_mjpeg_qsv_encoder;
extern const FFCodec ff_mjpeg_qsv_decoder;
extern const FFCodec ff_mjpeg_vaapi_encoder;
+extern const FFCodec ff_mjpeg_videotoolbox_encoder;
@@ -916,6 +916,7 @@ extern const FFCodec ff_mjpeg_qsv_decode
extern const FFCodec ff_mjpeg_rkmpp_decoder;
extern const FFCodec ff_mjpeg_rkmpp_encoder;
extern const FFCodec ff_mjpeg_vaapi_encoder;
+extern const FFCodec ff_mjpeg_videotoolbox_encoder;
extern const FFCodec ff_mp3_mediacodec_decoder;
extern const FFCodec ff_mp3_mf_encoder;
extern const FFCodec ff_mpeg1_cuvid_decoder;
Index: FFmpeg/libavcodec/videotoolboxenc.c
===================================================================
--- FFmpeg.orig/libavcodec/videotoolboxenc.c
+++ FFmpeg/libavcodec/videotoolboxenc.c
@@ -546,6 +546,7 @@ static CMVideoCodecType get_cm_codec_typ
@@ -556,6 +556,7 @@ static CMVideoCodecType get_cm_codec_typ
else
return MKBETAG('a','p','c','n'); // kCMVideoCodecType_AppleProRes422
}
@@ -47,7 +47,7 @@ Index: FFmpeg/libavcodec/videotoolboxenc.c
default: return 0;
}
}
@@ -1243,7 +1244,7 @@ static int vtenc_create_encoder(AVCodecC
@@ -1258,7 +1259,7 @@ static int vtenc_create_encoder(AVCodecC
kVTCompressionPropertyKey_Quality,
quality_num);
CFRelease(quality_num);
@@ -56,7 +56,7 @@ Index: FFmpeg/libavcodec/videotoolboxenc.c
bit_rate_num = CFNumberCreate(kCFAllocatorDefault,
kCFNumberSInt32Type,
&bit_rate);
@@ -1357,7 +1358,7 @@ static int vtenc_create_encoder(AVCodecC
@@ -1374,7 +1375,7 @@ static int vtenc_create_encoder(AVCodecC
}
}
@@ -65,7 +65,7 @@ Index: FFmpeg/libavcodec/videotoolboxenc.c
CFNumberRef interval = CFNumberCreate(kCFAllocatorDefault,
kCFNumberIntType,
&avctx->gop_size);
@@ -1506,7 +1507,7 @@ static int vtenc_create_encoder(AVCodecC
@@ -1523,7 +1524,7 @@ static int vtenc_create_encoder(AVCodecC
}
}
@@ -74,7 +74,7 @@ Index: FFmpeg/libavcodec/videotoolboxenc.c
status = VTSessionSetProperty(vtctx->session,
kVTCompressionPropertyKey_AllowFrameReordering,
kCFBooleanFalse);
@@ -2875,6 +2876,13 @@ static const enum AVPixelFormat prores_p
@@ -2769,6 +2770,13 @@ static const enum AVPixelFormat prores_p
AV_PIX_FMT_NONE
};
@@ -88,7 +88,7 @@ Index: FFmpeg/libavcodec/videotoolboxenc.c
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
#define COMMON_OPTIONS \
{ "allow_sw", "Allow software encoding", OFFSET(allow_sw), AV_OPT_TYPE_BOOL, \
@@ -3046,4 +3054,36 @@ const FFCodec ff_prores_videotoolbox_enc
@@ -2944,4 +2952,36 @@ const FFCodec ff_prores_videotoolbox_enc
.caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
.p.wrapper_name = "videotoolbox",
.hw_configs = vt_encode_hw_configs,
@@ -114,7 +114,7 @@ Index: FFmpeg/libavcodec/videotoolboxenc.c
+ .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY |
+ AV_CODEC_CAP_HARDWARE,
+ .priv_data_size = sizeof(VTEncContext),
+ .p.pix_fmts = mjpeg_pix_fmts,
+ CODEC_PIXFMTS_ARRAY(mjpeg_pix_fmts),
+ .defaults = vt_defaults,
+ .color_ranges = AVCOL_RANGE_MPEG | AVCOL_RANGE_JPEG,
+ .init = vtenc_init,
debian/patches/0052-add-vt-low-priority-keyframe-decoding.patch → debian/patches/0048-add-vt-low-priority-keyframe-decoding.patch
Vendored
+9 -9
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/libavcodec/avcodec.h
===================================================================
--- FFmpeg.orig/libavcodec/avcodec.h
+++ FFmpeg/libavcodec/avcodec.h
@@ -2185,6 +2185,13 @@ typedef struct AVHWAccel {
@@ -2041,6 +2041,13 @@ typedef struct AVHWAccel {
#define AV_HWACCEL_FLAG_UNSAFE_OUTPUT (1 << 3)
/**
@@ -20,19 +20,19 @@ Index: FFmpeg/libavcodec/options_table.h
===================================================================
--- FFmpeg.orig/libavcodec/options_table.h
+++ FFmpeg/libavcodec/options_table.h
@@ -412,6 +412,7 @@ static const AVOption avcodec_options[]
{"mastering_display_metadata", .default_val.i64 = AV_PKT_DATA_MASTERING_DISPLAY_METADATA, .type = AV_OPT_TYPE_CONST, .flags = A|D, .unit = "side_data_pkt" },
{"content_light_level", .default_val.i64 = AV_PKT_DATA_CONTENT_LIGHT_LEVEL, .type = AV_OPT_TYPE_CONST, .flags = A|D, .unit = "side_data_pkt" },
{"icc_profile", .default_val.i64 = AV_PKT_DATA_ICC_PROFILE, .type = AV_OPT_TYPE_CONST, .flags = A|D, .unit = "side_data_pkt" },
@@ -401,6 +401,7 @@ static const AVOption avcodec_options[]
{"allow_high_depth", "allow to output YUV pixel formats with a different chroma sampling than 4:2:0 and/or other than 8 bits per component", 0, AV_OPT_TYPE_CONST, {.i64 = AV_HWACCEL_FLAG_ALLOW_HIGH_DEPTH }, INT_MIN, INT_MAX, V | D, .unit = "hwaccel_flags"},
{"allow_profile_mismatch", "attempt to decode anyway if HW accelerated decoder's supported profiles do not exactly match the stream", 0, AV_OPT_TYPE_CONST, {.i64 = AV_HWACCEL_FLAG_ALLOW_PROFILE_MISMATCH }, INT_MIN, INT_MAX, V | D, .unit = "hwaccel_flags"},
{"unsafe_output", "allow potentially unsafe hwaccel frame output that might require special care to process successfully", 0, AV_OPT_TYPE_CONST, {.i64 = AV_HWACCEL_FLAG_UNSAFE_OUTPUT }, INT_MIN, INT_MAX, V | D, .unit = "hwaccel_flags"},
+{"low_priority", "attempt to run decode pipeline at a lower priority than is used for realtime decoding", 0, AV_OPT_TYPE_CONST, {.i64 = AV_HWACCEL_FLAG_LOW_PRIORITY }, INT_MIN, INT_MAX, V | D, .unit = "hwaccel_flags"},
{NULL},
};
{"extra_hw_frames", "Number of extra hardware frames to allocate for the user", OFFSET(extra_hw_frames), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, V|D },
{"discard_damaged_percentage", "Percentage of damaged samples to discard a frame", OFFSET(discard_damaged_percentage), AV_OPT_TYPE_INT, {.i64 = 95 }, 0, 100, V|D },
{"side_data_prefer_packet", "Comma-separated list of side data types for which user-supplied (container) data is preferred over coded bytestream",
Index: FFmpeg/libavcodec/videotoolbox.c
===================================================================
--- FFmpeg.orig/libavcodec/videotoolbox.c
+++ FFmpeg/libavcodec/videotoolbox.c
@@ -985,6 +985,23 @@ static int videotoolbox_start(AVCodecCon
@@ -1025,6 +1025,23 @@ static int videotoolbox_start(AVCodecCon
av_log(avctx, AV_LOG_VERBOSE, "VideoToolbox reported invalid data.\n");
return AVERROR_INVALIDDATA;
case 0:
debian/patches/0053-add-dummy-device-derive-to-videotoolbox-hwcontext.patch → debian/patches/0049-add-dummy-device-derive-to-videotoolbox-hwcontext.patch
Vendored
View File
debian/patches/0054-add-cuda-transpose-filter-impl.patch → debian/patches/0050-add-cuda-transpose-filter-impl.patch
Vendored
+211 -106
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3321,6 +3321,8 @@ thumbnail_cuda_filter_deps="ffnvcodec"
@@ -3527,6 +3527,8 @@ thumbnail_cuda_filter_deps="ffnvcodec"
thumbnail_cuda_filter_deps_any="cuda_nvcc cuda_llvm"
tonemap_cuda_filter_deps="ffnvcodec const_nan"
tonemap_cuda_filter_deps_any="cuda_nvcc cuda_llvm"
@@ -15,7 +15,7 @@ Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -542,6 +542,8 @@ OBJS-$(CONFIG_TONEMAP_VIDEOTOOLBOX_FILTE
@@ -559,6 +559,8 @@ OBJS-$(CONFIG_TONEMAP_VIDEOTOOLBOX_FILTE
metal/utils.o
OBJS-$(CONFIG_TPAD_FILTER) += vf_tpad.o
OBJS-$(CONFIG_TRANSPOSE_FILTER) += vf_transpose.o
@@ -28,21 +28,21 @@ Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -505,6 +505,7 @@ extern const AVFilter ff_vf_tonemap_vaap
extern const AVFilter ff_vf_tonemap_videotoolbox;
extern const AVFilter ff_vf_tpad;
extern const AVFilter ff_vf_transpose;
+extern const AVFilter ff_vf_transpose_cuda;
extern const AVFilter ff_vf_transpose_npp;
extern const AVFilter ff_vf_transpose_opencl;
extern const AVFilter ff_vf_transpose_vaapi;
@@ -520,6 +520,7 @@ extern const FFFilter ff_vf_tonemap_vaap
extern const FFFilter ff_vf_tonemap_videotoolbox;
extern const FFFilter ff_vf_tpad;
extern const FFFilter ff_vf_transpose;
+extern const FFFilter ff_vf_transpose_cuda;
extern const FFFilter ff_vf_transpose_npp;
extern const FFFilter ff_vf_transpose_opencl;
extern const FFFilter ff_vf_transpose_vaapi;
Index: FFmpeg/libavfilter/vf_transpose_cuda.c
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_transpose_cuda.c
@@ -0,0 +1,483 @@
@@ -0,0 +1,564 @@
+/*
+ * Copyright (C) 2024 NyanMisaka
+ * Copyright (C) 2026 NyanMisaka
+ *
+ * This file is part of FFmpeg.
+ *
@@ -61,6 +61,7 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/avassert.h"
+#include "libavutil/common.h"
+#include "libavutil/hwcontext.h"
+#include "libavutil/hwcontext_cuda_internal.h"
@@ -76,19 +77,26 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+
+#include "cuda/load_helper.h"
+
+#define CHECK_CU(x) FF_CUDA_CHECK_DL(ctx, s->hwctx->internal->cuda_dl, x)
+#define DIV_UP(a, b) ( ((a) + (b) - 1) / (b) )
+
+#define BLOCK_X 32
+#define BLOCK_Y 16
+
+#define CHECK_CU(x) FF_CUDA_CHECK_DL(ctx, s->hwctx->internal->cuda_dl, x)
+
+static const enum AVPixelFormat supported_formats[] = {
+ AV_PIX_FMT_YUV420P,
+ AV_PIX_FMT_NV12,
+ AV_PIX_FMT_YUV422P,
+ AV_PIX_FMT_YUV444P,
+ AV_PIX_FMT_P010,
+ AV_PIX_FMT_P016,
+ AV_PIX_FMT_YUV420P10,
+ AV_PIX_FMT_YUV422P10,
+ AV_PIX_FMT_YUV444P10,
+ AV_PIX_FMT_YUV444P16,
+ AV_PIX_FMT_NV12,
+ AV_PIX_FMT_NV16,
+ AV_PIX_FMT_P010,
+ AV_PIX_FMT_P210,
+ AV_PIX_FMT_P016,
+ AV_PIX_FMT_P216,
+ AV_PIX_FMT_0RGB32,
+ AV_PIX_FMT_0BGR32,
+ AV_PIX_FMT_RGB32,
@@ -109,6 +117,9 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+ CUmodule cu_module;
+ CUfunction cu_func_uchar;
+ CUfunction cu_func_ushort;
+ CUfunction cu_func_uchar2;
+ CUfunction cu_func_ushort2;
+ CUfunction cu_func_uchar4;
+ CUstream cu_stream;
+
+ int flip_wh;
@@ -190,12 +201,9 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+
+static int format_is_supported(enum AVPixelFormat fmt)
+{
+ int i;
+
+ for (i = 0; i < FF_ARRAY_ELEMS(supported_formats); i++)
+ for (int i = 0; i < FF_ARRAY_ELEMS(supported_formats); i++)
+ if (supported_formats[i] == fmt)
+ return 1;
+
+ return 0;
+}
+
@@ -306,13 +314,23 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+ if (ret < 0)
+ goto exit;
+
+ ret = CHECK_CU(cu->cuModuleGetFunction(&s->cu_func_uchar,
+ s->cu_module, "Transpose_Cuda_uchar"));
+ ret = CHECK_CU(cu->cuModuleGetFunction(&s->cu_func_uchar, s->cu_module, "Transpose_Cuda_uchar"));
+ if (ret < 0)
+ goto exit;
+
+ ret = CHECK_CU(cu->cuModuleGetFunction(&s->cu_func_ushort,
+ s->cu_module, "Transpose_Cuda_ushort"));
+ ret = CHECK_CU(cu->cuModuleGetFunction(&s->cu_func_ushort, s->cu_module, "Transpose_Cuda_ushort"));
+ if (ret < 0)
+ goto exit;
+
+ ret = CHECK_CU(cu->cuModuleGetFunction(&s->cu_func_uchar2, s->cu_module, "Transpose_Cuda_uchar2"));
+ if (ret < 0)
+ goto exit;
+
+ ret = CHECK_CU(cu->cuModuleGetFunction(&s->cu_func_ushort2, s->cu_module, "Transpose_Cuda_ushort2"));
+ if (ret < 0)
+ goto exit;
+
+ ret = CHECK_CU(cu->cuModuleGetFunction(&s->cu_func_uchar4, s->cu_module, "Transpose_Cuda_uchar4"));
+ if (ret < 0)
+ goto exit;
+
@@ -325,61 +343,126 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+ return ret;
+}
+
+static int call_transpose_kernel(AVFilterContext *ctx,
+ int pix_size,
+ int pix_step,
+ int pix_offset,
+ CUdeviceptr dst_data,
+ int dst_width,
+ int dst_height,
+ int dst_pitch,
+ CUdeviceptr src_data,
+ int src_width,
+ int src_height,
+ int src_pitch)
+static CUresult call_kernel(AVFilterContext *ctx,
+ CUfunction cu_func,
+ CUarray_format cu_format,
+ int channels,
+ int is_422_uv, // Dst* & Src* are 4:2:2 UV planes
+ CUdeviceptr dst0,
+ CUdeviceptr dst1, // Dst1 is for fully planar V, optional
+ int dst_width, // Width is pixels per channel
+ int dst_height, // Height is pixels per channel
+ int dst_pitch, // Pitch is elements per channel
+ CUdeviceptr src0,
+ CUdeviceptr src1, // Src1 is for fully planar V, optional
+ int src_width, // Width is pixels per channel
+ int src_height, // Height is pixels per channel
+ int src_pitch)
+{
+ TransposeCUDAContext *s = ctx->priv;
+ CudaFunctions *cu = s->hwctx->internal->cuda_dl;
+ CUtexObject src0_tex = 0, src1_tex = 0;
+ int ret;
+
+ void* kernel_args[] = {
+ &dst_data, &dst_width, &dst_height, &dst_pitch,
+ &src_data, &src_width, &src_height, &src_pitch,
+ &pix_step, &pix_offset, &s->dir,
+ void *kernel_args[] = {
+ &dst0, &dst1, &dst_width, &dst_height, &dst_pitch,
+ &src0_tex, &src1_tex, &s->dir,
+ };
+
+ return CHECK_CU(cu->cuLaunchKernel(pix_size == 1 ? s->cu_func_uchar : s->cu_func_ushort,
+ DIV_UP(dst_width, BLOCK_X), DIV_UP(dst_height, BLOCK_Y), 1,
+ BLOCK_X, BLOCK_Y, 1, 0, s->cu_stream, kernel_args, NULL));
+ CUDA_TEXTURE_DESC tex_desc = {
+ .addressMode = { CU_TR_ADDRESS_MODE_CLAMP,
+ CU_TR_ADDRESS_MODE_CLAMP },
+ .filterMode = is_422_uv ? CU_TR_FILTER_MODE_LINEAR
+ : CU_TR_FILTER_MODE_POINT,
+ .flags = 2 /* CU_TRSF_NORMALIZED_COORDINATES */
+ };
+ CUDA_RESOURCE_DESC res_desc = {
+ .resType = CU_RESOURCE_TYPE_PITCH2D,
+ .res.pitch2D.format = cu_format,
+ .res.pitch2D.numChannels = channels,
+ .res.pitch2D.pitchInBytes = src_pitch,
+ .res.pitch2D.width = src_width,
+ .res.pitch2D.height = src_height
+ };
+
+ res_desc.res.pitch2D.devPtr = (CUdeviceptr)src0;
+ ret = CHECK_CU(cu->cuTexObjectCreate(&src0_tex, &res_desc, &tex_desc, NULL));
+ if (ret < 0)
+ goto exit;
+
+ if (src1) {
+ res_desc.res.pitch2D.devPtr = (CUdeviceptr)src1;
+ ret = CHECK_CU(cu->cuTexObjectCreate(&src1_tex, &res_desc, &tex_desc, NULL));
+ if (ret < 0)
+ goto exit;
+ }
+
+ ret = CHECK_CU(cu->cuLaunchKernel(cu_func,
+ DIV_UP(dst_width, BLOCK_X), DIV_UP(dst_height, BLOCK_Y), 1,
+ BLOCK_X, BLOCK_Y, 1, 0, s->cu_stream, kernel_args, NULL));
+exit:
+ if (src0_tex)
+ CHECK_CU(cu->cuTexObjectDestroy(src0_tex));
+ if (src1_tex)
+ CHECK_CU(cu->cuTexObjectDestroy(src1_tex));
+
+ return ret;
+}
+
+static int cudatranspose_rotate(AVFilterContext *ctx,
+ AVFrame *out, AVFrame *in)
+{
+ TransposeCUDAContext *s = ctx->priv;
+ int pix_size, pix_step, pix_offset;
+ int c, p, ret;
+ int ret;
+
+ for (c = 0; c < s->pix_desc->nb_components; c++) {
+ p = s->pix_desc->comp[c].plane;
+ if (!out->data[p] || !in->data[p])
+ return AVERROR(ENOMEM);
+ for (int c = 0; c < s->pix_desc->nb_components; c++) {
+ const AVComponentDescriptor *comp = &s->pix_desc->comp[c];
+ const int p = comp->plane;
+ int pix_size, channels;
+ int is_planar_u, is_planar_v, is_422_uv;
+ CUfunction func;
+ CUarray_format format;
+
+ pix_size = (s->pix_desc->comp[c].depth + s->pix_desc->comp[c].shift) / 8;
+ if (!(pix_size == 1 || pix_size == 2))
+ return AVERROR(ENOSYS);
+ pix_size = (comp->depth + 7) / 8;
+ channels = comp->step / pix_size;
+ if (pix_size > 2 || channels > 4)
+ av_unreachable("Unsupported pixel format!");
+
+ pix_step = s->pix_desc->comp[c].step / pix_size;
+ pix_offset = s->pix_desc->comp[c].offset / pix_size;
+ is_planar_u = p == 1 && channels == 1;
+ is_planar_v = p == 2 && channels == 1;
+ is_422_uv = p && s->pix_desc->log2_chroma_w == 1 && !s->pix_desc->log2_chroma_h;
+
+ ret = call_transpose_kernel(ctx, pix_size, pix_step, pix_offset,
+ (CUdeviceptr)out->data[p],
+ AV_CEIL_RSHIFT(out->width, p ? s->pix_desc->log2_chroma_w : 0),
+ AV_CEIL_RSHIFT(out->height, p ? s->pix_desc->log2_chroma_h : 0),
+ out->linesize[p] / pix_size,
+ (CUdeviceptr)in->data[p],
+ AV_CEIL_RSHIFT(in->width, p ? s->pix_desc->log2_chroma_w : 0),
+ AV_CEIL_RSHIFT(in->height, p ? s->pix_desc->log2_chroma_h : 0),
+ in->linesize[p] / pix_size);
+ if (comp->plane < c || is_planar_v) {
+ // We process planes as a whole, so don't reprocess
+ // them for additional components
+ continue;
+ }
+
+ switch (pix_size) {
+ case 1:
+ func = channels == 4 ? s->cu_func_uchar4 :
+ channels == 2 ? s->cu_func_uchar2 : s->cu_func_uchar;
+ format = CU_AD_FORMAT_UNSIGNED_INT8;
+ break;
+ case 2:
+ func = channels == 2 ? s->cu_func_ushort2 : s->cu_func_ushort;
+ format = CU_AD_FORMAT_UNSIGNED_INT16;
+ break;
+ default:
+ av_unreachable("Unsupported pixel format!");
+ }
+
+ ret = call_kernel(ctx, func, format, channels, is_422_uv,
+ (CUdeviceptr)out->data[p],
+ (CUdeviceptr)(is_planar_u ? out->data[p+1] : NULL),
+ AV_CEIL_RSHIFT(out->width, p ? s->pix_desc->log2_chroma_w : 0),
+ AV_CEIL_RSHIFT(out->height, p ? s->pix_desc->log2_chroma_h : 0),
+ out->linesize[p] / comp->step,
+ (CUdeviceptr)in->data[p],
+ (CUdeviceptr)(is_planar_u ? in->data[p+1] : NULL),
+ AV_CEIL_RSHIFT(in->width, p ? s->pix_desc->log2_chroma_w : 0),
+ AV_CEIL_RSHIFT(in->height, p ? s->pix_desc->log2_chroma_h : 0),
+ in->linesize[p]);
+ if (ret < 0)
+ return ret;
+ }
@@ -387,20 +470,18 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+ return 0;
+}
+
+static int cudatranspose_filter_frame_internal(AVFilterContext *ctx,
+ AVFrame *out, AVFrame *in)
+static int cudatranspose_transpose(AVFilterContext *ctx,
+ AVFrame *out, AVFrame *in)
+{
+ TransposeCUDAContext *s = ctx->priv;
+ AVFilterLink *outlink = ctx->outputs[0];
+ AVFrame *src = in;
+ int ret;
+
+ ret = cudatranspose_rotate(ctx, s->frame, src);
+ ret = cudatranspose_rotate(ctx, s->frame, in);
+ if (ret < 0)
+ return ret;
+
+ src = s->frame;
+ ret = av_hwframe_get_buffer(src->hw_frames_ctx, s->tmp_frame, 0);
+ ret = av_hwframe_get_buffer(s->frame->hw_frames_ctx, s->tmp_frame, 0);
+ if (ret < 0)
+ return ret;
+
@@ -447,7 +528,7 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+ if (ret < 0)
+ goto fail;
+
+ ret = cudatranspose_filter_frame_internal(ctx, out, in);
+ ret = cudatranspose_transpose(ctx, out, in);
+
+ CHECK_CU(cu->cuCtxPopCurrent(&dummy));
+ if (ret < 0)
@@ -497,9 +578,9 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+
+static const AVFilterPad cudatranspose_inputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .filter_frame = cudatranspose_filter_frame,
+ .name = "default",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .filter_frame = cudatranspose_filter_frame,
+ .get_buffer.video = cudatranspose_get_video_buffer,
+ },
+};
@@ -512,13 +593,13 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.c
+ },
+};
+
+const AVFilter ff_vf_transpose_cuda = {
+ .name = "transpose_cuda",
+ .description = NULL_IF_CONFIG_SMALL("Transpose input video using CUDA"),
+const FFFilter ff_vf_transpose_cuda = {
+ .p.name = "transpose_cuda",
+ .p.description = NULL_IF_CONFIG_SMALL("Transpose input video using CUDA"),
+ .p.priv_class = &cudatranspose_class,
+ .init = cudatranspose_init,
+ .uninit = cudatranspose_uninit,
+ .priv_size = sizeof(TransposeCUDAContext),
+ .priv_class = &cudatranspose_class,
+ FILTER_INPUTS(cudatranspose_inputs),
+ FILTER_OUTPUTS(cudatranspose_outputs),
+ FILTER_SINGLE_PIXFMT(AV_PIX_FMT_CUDA),
@@ -528,9 +609,9 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.cu
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_transpose_cuda.cu
@@ -0,0 +1,61 @@
@@ -0,0 +1,85 @@
+/*
+ * Copyright (C) 2024 NyanMisaka
+ * Copyright (C) 2026 NyanMisaka
+ *
+ * This file is part of FFmpeg.
+ *
@@ -549,44 +630,68 @@ Index: FFmpeg/libavfilter/vf_transpose_cuda.cu
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+template<typename T>
+__inline__ __device__ void transpose_func(
+ T* dst, int dst_width, int dst_height, int dst_pitch,
+ T* src, int src_width, int src_height, int src_pitch,
+ int pix_step, int pix_offset, int dir)
+#include "cuda/vector_helpers.cuh"
+
+__inline__ __device__ void map_input_ncoords(float *xi, float *yi,
+ int xo, int yo,
+ int dst_width, int dst_height,
+ int dir)
+{
+ int flip_wh = dir < 4;
+ *xi = flip_wh ? ((dir & 2) ? (dst_height - 1 - yo) : yo)
+ : ((dir == 6) ? xo : (dst_width - 1 - xo));
+ *yi = flip_wh ? ((dir & 1) ? (dst_width - 1 - xo) : xo)
+ : ((dir == 5) ? yo : (dst_height - 1 - yo));
+
+ *xi = (*xi + 0.5f) / (flip_wh ? dst_height : dst_width);
+ *yi = (*yi + 0.5f) / (flip_wh ? dst_width : dst_height);
+}
+
+template<typename T, int DST1, int FACTOR>
+__inline__ __device__ void Transpose_Cuda(
+ T *dst0, T *dst1, int dst_width, int dst_height, int dst_pitch,
+ cudaTextureObject_t src0_tex, cudaTextureObject_t src1_tex, int dir)
+{
+ int xo = blockIdx.x * blockDim.x + threadIdx.x;
+ int yo = blockIdx.y * blockDim.y + threadIdx.y;
+ if (xo >= dst_width || yo >= dst_height)
+ return;
+
+ int xi = (dir < 4) ? ((dir & 2) ? (dst_height - 1 - yo) : yo)
+ : ((dir == 6) ? xo : (dst_width - 1 - xo));
+ int yi = (dir < 4) ? ((dir & 1) ? (dst_width - 1 - xo) : xo)
+ : ((dir == 5) ? yo : (dst_height - 1 - yo));
+ if (xi >= src_width || yi >= src_height)
+ return;
+ float xi, yi;
+ map_input_ncoords(&xi, &yi, xo, yo,
+ dst_width, dst_height, dir);
+
+ int dst_pos = xo*pix_step + yo*dst_pitch + pix_offset;
+ int src_pos = xi*pix_step + yi*src_pitch + pix_offset;
+ dst[dst_pos] = src[src_pos];
+ dst0[yo*dst_pitch+xo] = from_floatN<T, floatT>(
+ saturate_rintf<floatT>(
+ tex2D<floatT>(src0_tex, xi, yi), FACTOR
+ )
+ );
+ if (DST1 && dst1 && src1_tex) {
+ dst1[yo*dst_pitch+xo] = from_floatN<T, floatT>(
+ saturate_rintf<floatT>(
+ tex2D<floatT>(src1_tex, xi, yi), FACTOR
+ )
+ );
+ }
+}
+
+extern "C" {
+
+#define TRANSPOSE_VARIANT(NAME, TYPE) \
+__global__ void Transpose_Cuda_ ## NAME( \
+ TYPE* dst, int dst_width, int dst_height, int dst_pitch, \
+ TYPE* src, int src_width, int src_height, int src_pitch, \
+ int pix_step, int pix_offset, int dir) \
+{ \
+ transpose_func( \
+ dst, dst_width, dst_height, dst_pitch, \
+ src, src_width, src_height, src_pitch, \
+ pix_step, pix_offset, dir); \
+#define TRANSPOSE_KERNEL(NAME, TYPE, DST1, FACTOR) \
+__global__ void Transpose_Cuda_ ## NAME( \
+ TYPE *dst0, TYPE *dst1, int dst_width, int dst_height, int dst_pitch, \
+ cudaTextureObject_t src0_tex, cudaTextureObject_t src1_tex, int dir) \
+{ \
+ Transpose_Cuda<TYPE, DST1, FACTOR>( \
+ dst0, dst1, dst_width, dst_height, dst_pitch, \
+ src0_tex, src1_tex, dir \
+ ); \
+}
+
+TRANSPOSE_VARIANT(uchar, unsigned char)
+TRANSPOSE_VARIANT(ushort, unsigned short)
+TRANSPOSE_KERNEL(uchar, uchar, 1, 0xFF)
+TRANSPOSE_KERNEL(ushort, ushort, 1, 0xFFFF)
+TRANSPOSE_KERNEL(uchar2, uchar2, 0, 0xFF)
+TRANSPOSE_KERNEL(ushort2, ushort2, 0, 0xFFFF)
+TRANSPOSE_KERNEL(uchar4, uchar4, 0, 0xFF)
+
+} /* extern "C" */
debian/patches/0055-add-flip-feat-to-opencl-transpose-filter.patch → debian/patches/0051-add-flip-feat-to-opencl-transpose-filter.patch
Vendored
+4 -4
View File
@@ -8,10 +8,10 @@ Index: FFmpeg/libavfilter/opencl/transpose.cl
- int xin = (dir & 2) ? (size.y - 1 - y) : y;
- int yin = (dir & 1) ? (size.x - 1 - x) : x;
+ int xin = (dir < 4) ? ((dir & 2) ? (size.y - 1 - y) : y)
+ : ((dir == 6) ? x : (size.x - 1 - x));
+ int yin = (dir < 4) ? ((dir & 1) ? (size.x - 1 - x) : x)
+ : ((dir == 5) ? y : (size.y - 1 - y));
+ int xin = dir < 4 ? ((dir & 2) ? (size.y - 1 - y) : y)
+ : ((dir == 6) ? x : (size.x - 1 - x));
+ int yin = dir < 4 ? ((dir & 1) ? (size.x - 1 - x) : x)
+ : ((dir == 5) ? y : (size.y - 1 - y));
float4 data = read_imagef(src, sampler, (int2)(xin, yin));
if (x < size.x && y < size.y)
debian/patches/0056-opus-allow-5point1-side-inputs.patch → debian/patches/0052-opus-allow-5point1-side-inputs.patch
Vendored
View File
debian/patches/0057-add-simd-optimized-tonemapx-filter.patch → debian/patches/0053-add-simd-optimized-tonemapx-filter.patch
Vendored
+86 -59
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -2319,6 +2319,9 @@ HEADERS_LIST="
@@ -2458,6 +2458,9 @@ HEADERS_LIST="
INTRINSICS_LIST="
intrinsics_neon
intrinsics_sse2
@@ -12,7 +12,7 @@ Index: FFmpeg/configure
"
MATH_FUNCS="
@@ -2753,6 +2756,9 @@ armv8_deps="aarch64"
@@ -2915,6 +2918,9 @@ armv8_deps="aarch64"
neon_deps_any="aarch64 arm"
intrinsics_neon_deps="neon"
intrinsics_sse2_deps="sse2"
@@ -22,7 +22,7 @@ Index: FFmpeg/configure
vfp_deps="arm"
vfpv3_deps="vfp"
setend_deps="arm"
@@ -3984,6 +3990,7 @@ tinterlace_filter_deps="gpl"
@@ -4243,6 +4249,7 @@ tinterlace_filter_deps="gpl"
tinterlace_merge_test_deps="tinterlace_filter"
tinterlace_pad_test_deps="tinterlace_filter"
tonemap_filter_deps="const_nan"
@@ -30,28 +30,28 @@ Index: FFmpeg/configure
tonemap_vaapi_filter_deps="vaapi VAProcFilterParameterBufferHDRToneMapping"
tonemap_videotoolbox_filter_deps="metal corevideo videotoolbox const_nan"
tonemap_opencl_filter_deps="opencl const_nan"
@@ -6529,6 +6536,19 @@ fi
check_cc intrinsics_neon arm_neon.h "int16x8_t test = vdupq_n_s16(0)"
check_cc intrinsics_sse2 emmintrin.h "__m128i test = _mm_setzero_si128()"
@@ -6835,6 +6842,19 @@ EOF
+disable intrinsics_sse42 && test_cc -msse4.2 <<EOF && enable intrinsics_sse42
check_cc intrinsics_sse2 emmintrin.h "__m128i test = _mm_setzero_si128()"
+ disable intrinsics_sse42 && test_cc -msse4.2 <<EOF && enable intrinsics_sse42
+#include <immintrin.h>
+int main(void) { __m128i t = _mm_cmpgt_epi64(_mm_setzero_si128(), _mm_setzero_si128()); return 0; }
+EOF
+disable intrinsics_fma3 && test_cc -mfma <<EOF && enable intrinsics_fma3
+ disable intrinsics_fma3 && test_cc -mfma <<EOF && enable intrinsics_fma3
+#include <immintrin.h>
+int main(void) { __m256 t = _mm256_fmadd_ps(_mm256_setzero_ps(), _mm256_setzero_ps(), _mm256_setzero_ps()); return 0; }
+EOF
+disable intrinsics_avx2 && test_cc -mavx2 <<EOF && enable intrinsics_avx2
+ disable intrinsics_avx2 && test_cc -mavx2 <<EOF && enable intrinsics_avx2
+#include <immintrin.h>
+int main(void) { __m256i t = _mm256_abs_epi32(_mm256_setzero_si256()); return 0; }
+EOF
+
check_ldflags -Wl,--as-needed
check_ldflags -Wl,-z,noexecstack
if [ $target_os = "darwin" ]; then
@@ -7661,6 +7681,16 @@ elif enabled gcc; then
check_cflags -mpreferred-stack-boundary=4
elif enabled loongarch; then
enabled lsx && check_inline_asm lsx '"vadd.b $vr0, $vr1, $vr2"' '-mlsx' && append LSXFLAGS '-mlsx'
enabled lasx && check_inline_asm lasx '"xvadd.b $xr0, $xr1, $xr2"' '-mlasx' && append LASXFLAGS '-mlasx'
@@ -8112,6 +8132,16 @@ elif enabled gcc; then
check_allcflags -mpreferred-stack-boundary=4
;;
esac
+ elif enabled x86_64; then
@@ -66,12 +66,12 @@ Index: FFmpeg/configure
+ esac
fi
elif enabled llvm_gcc; then
check_cflags -mllvm -stack-alignment=16
check_allcflags -mllvm -stack-alignment=16
Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -532,6 +532,7 @@ OBJS-$(CONFIG_TMEDIAN_FILTER)
@@ -549,6 +549,7 @@ OBJS-$(CONFIG_TMEDIAN_FILTER)
OBJS-$(CONFIG_TMIDEQUALIZER_FILTER) += vf_tmidequalizer.o
OBJS-$(CONFIG_TMIX_FILTER) += vf_mix.o framesync.o
OBJS-$(CONFIG_TONEMAP_FILTER) += vf_tonemap.o
@@ -83,13 +83,14 @@ Index: FFmpeg/libavfilter/aarch64/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/aarch64/Makefile
+++ FFmpeg/libavfilter/aarch64/Makefile
@@ -1,5 +1,6 @@
@@ -1,6 +1,7 @@
OBJS-$(CONFIG_BWDIF_FILTER) += aarch64/vf_bwdif_init_aarch64.o
OBJS-$(CONFIG_COLORDETECT_FILTER) += aarch64/vf_colordetect_init.o
OBJS-$(CONFIG_NLMEANS_FILTER) += aarch64/vf_nlmeans_init.o
+OBJS-$(CONFIG_TONEMAPX_FILTER) += aarch64/vf_tonemapx_intrin_neon.o
NEON-OBJS-$(CONFIG_BWDIF_FILTER) += aarch64/vf_bwdif_neon.o
NEON-OBJS-$(CONFIG_NLMEANS_FILTER) += aarch64/vf_nlmeans_neon.o
NEON-OBJS-$(CONFIG_COLORDETECT_FILTER) += aarch64/vf_colordetect_neon.o
Index: FFmpeg/libavfilter/aarch64/vf_tonemapx_intrin_neon.c
===================================================================
--- /dev/null
@@ -2539,14 +2540,14 @@ Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -499,6 +499,7 @@ extern const AVFilter ff_vf_tmedian;
extern const AVFilter ff_vf_tmidequalizer;
extern const AVFilter ff_vf_tmix;
extern const AVFilter ff_vf_tonemap;
+extern const AVFilter ff_vf_tonemapx;
extern const AVFilter ff_vf_tonemap_cuda;
extern const AVFilter ff_vf_tonemap_opencl;
extern const AVFilter ff_vf_tonemap_vaapi;
@@ -514,6 +514,7 @@ extern const FFFilter ff_vf_tmedian;
extern const FFFilter ff_vf_tmidequalizer;
extern const FFFilter ff_vf_tmix;
extern const FFFilter ff_vf_tonemap;
+extern const FFFilter ff_vf_tonemapx;
extern const FFFilter ff_vf_tonemap_cuda;
extern const FFFilter ff_vf_tonemap_opencl;
extern const FFFilter ff_vf_tonemap_vaapi;
Index: FFmpeg/libavfilter/colorspace.c
===================================================================
--- FFmpeg.orig/libavfilter/colorspace.c
@@ -2559,8 +2560,8 @@ Index: FFmpeg/libavfilter/colorspace.c
#include "libavutil/frame.h"
#include "libavutil/mastering_display_metadata.h"
#include "libavutil/pixdesc.h"
@@ -359,3 +360,51 @@ float inverse_eotf_arib_b67(float x) {
float inverse_eotf_bt1886(float x) {
@@ -359,3 +360,51 @@ float ff_inverse_eotf_arib_b67(float x)
float ff_inverse_eotf_bt1886(float x) {
return x > 0.0f ? powf(x, 1.0f / 2.4f) : 0.0f;
}
+
@@ -2615,20 +2616,21 @@ Index: FFmpeg/libavfilter/colorspace.h
===================================================================
--- FFmpeg.orig/libavfilter/colorspace.h
+++ FFmpeg/libavfilter/colorspace.h
@@ -110,4 +110,8 @@ float eotf_arib_b67(float x);
float inverse_eotf_arib_b67(float x);
float inverse_eotf_bt1886(float x);
@@ -110,4 +110,9 @@ float ff_eotf_arib_b67(float x);
float ff_inverse_eotf_arib_b67(float x);
float ff_inverse_eotf_bt1886(float x);
+int ff_get_range_off(int *off, int *y_rng, int *uv_rng,
+ enum AVColorRange rng, int depth);
+void ff_get_yuv_coeffs(int out[3][3][8], double (*table)[3],
+ int depth, int y_rng, int uv_rng, int yuv2rgb);
+
#endif
Index: FFmpeg/libavfilter/vf_tonemapx.c
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_tonemapx.c
@@ -0,0 +1,1948 @@
@@ -0,0 +1,1972 @@
+/*
+ * This file is part of FFmpeg.
+ *
@@ -2708,7 +2710,8 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+typedef struct TonemapxContext {
+ const AVClass *class;
+
+ enum TonemapAlgorithm tonemap;
+ /* enum TonemapAlgorithm */
+ int tonemap;
+ enum AVColorTransferCharacteristic trc;
+ enum AVColorSpace spc;
+ enum AVColorPrimaries pri;
@@ -2729,7 +2732,7 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+ uint16_t *delin_lut;
+ int in_yuv_off, out_yuv_off;
+
+ struct DoviMetadata *dovi;
+ struct FFDOVIMetadataRemap *dovi;
+
+ DECLARE_ALIGNED(16, float, dovi_pbuf)[3*(params_sz+pivots_sz+coeffs_sz+mmr_sz)];
+ DECLARE_ALIGNED(16, int, yuv2rgb_coeffs)[3][3][8];
@@ -2827,7 +2830,7 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+ for (c = 0; c < 3; c++) {
+ int has_poly = 0, has_mmr = 0, mmr_single = 1;
+ int mmr_idx = 0, min_order = 3, max_order = 1;
+ const struct ReshapeData *comp = &s->dovi->comp[c];
+ const struct FFDOVIReshapeData *comp = &s->dovi->comp[c];
+ if (!comp->num_pivots)
+ continue;
+ av_assert0(comp->num_pivots >= 2 && comp->num_pivots <= 9);
@@ -3099,13 +3102,13 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+
+static float bt2390(float s, float knee_offset, float peak)
+{
+ float peak_pq = inverse_eotf_st2084(peak, REFERENCE_WHITE_ALT);
+ float peak_pq = ff_inverse_eotf_st2084(peak, REFERENCE_WHITE_ALT);
+ float scale = peak_pq > 0.0f ? (1.0f / peak_pq) : 1.0f;
+
+ // SDR peak
+ float dst_peak = 1.0f;
+ float s_pq = inverse_eotf_st2084(s, REFERENCE_WHITE_ALT) * scale;
+ float max_lum = inverse_eotf_st2084(dst_peak, REFERENCE_WHITE_ALT) * scale;
+ float s_pq = ff_inverse_eotf_st2084(s, REFERENCE_WHITE_ALT) * scale;
+ float max_lum = ff_inverse_eotf_st2084(dst_peak, REFERENCE_WHITE_ALT) * scale;
+
+ float ks = (1.0f + knee_offset) * max_lum - knee_offset;
+ float tb = (s_pq - ks) / (1.0f - ks);
@@ -3116,7 +3119,7 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+ (-2.0f * tb3 + 3.0f * tb2) * max_lum;
+ float sig = MIX(pb, s_pq, s_pq < ks);
+
+ return eotf_st2084(sig * peak_pq, REFERENCE_WHITE_ALT);
+ return ff_eotf_st2084(sig * peak_pq, REFERENCE_WHITE_ALT);
+}
+
+static float mapsig(enum TonemapAlgorithm alg, float sig, double peak, double param)
@@ -3157,9 +3160,9 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+static float linearize(float x, enum AVColorTransferCharacteristic trc_src)
+{
+ if (trc_src == AVCOL_TRC_SMPTE2084)
+ return eotf_st2084(x, REFERENCE_WHITE_ALT);
+ return ff_eotf_st2084(x, REFERENCE_WHITE_ALT);
+ else if (trc_src == AVCOL_TRC_ARIB_STD_B67)
+ return eotf_arib_b67(x);
+ return ff_eotf_arib_b67(x);
+ else
+ return x;
+}
@@ -3167,7 +3170,7 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+static float delinearize(float x, enum AVColorTransferCharacteristic trc_dst)
+{
+ if (trc_dst == AVCOL_TRC_BT709 || trc_dst == AVCOL_TRC_BT2020_10)
+ return inverse_eotf_bt1886(x);
+ return ff_inverse_eotf_bt1886(x);
+ else
+ return x;
+}
@@ -4162,7 +4165,7 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+ // such videos usually have vdr_rpu_profile == 0, for example profile 5 videos
+ // this could be wrong as there is no public documentation on this field
+ if (rpu->disable_residual_flag && rpu->vdr_rpu_profile == 0) {
+ struct DoviMetadata *dovi = av_malloc(sizeof(*dovi));
+ struct FFDOVIMetadataRemap *dovi = av_malloc(sizeof(*dovi));
+ s->dovi = dovi;
+ if (!s->dovi)
+ goto fail;
@@ -4248,7 +4251,7 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+
+ av_frame_free(&in);
+
+ if (s->trc !=AVCOL_TRC_SMPTE2084) {
+ if (s->trc != AVCOL_TRC_SMPTE2084) {
+ av_frame_remove_side_data(out, AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_remove_side_data(out, AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
@@ -4516,6 +4519,21 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+ return 0;
+}
+
+static int config_props(AVFilterLink *outlink)
+{
+ AVFilterContext *ctx = outlink->src;
+ TonemapxContext *s = ctx->priv;
+
+ if (s->trc != AVCOL_TRC_SMPTE2084) {
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
+
+ return 0;
+}
+
+#define OFFSET(x) offsetof(TonemapxContext, x)
+#define FLAGS AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_FILTERING_PARAM
+static const AVOption tonemapx_options[] = {
@@ -4565,17 +4583,25 @@ Index: FFmpeg/libavfilter/vf_tonemapx.c
+ },
+};
+
+AVFilter ff_vf_tonemapx = {
+ .name = "tonemapx",
+ .description = NULL_IF_CONFIG_SMALL("SIMD optimized HDR to SDR tonemapping"),
+static const AVFilterPad tonemapx_outputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .config_props = config_props,
+ },
+};
+
+const FFFilter ff_vf_tonemapx = {
+ .p.name = "tonemapx",
+ .p.description = NULL_IF_CONFIG_SMALL("SIMD optimized HDR to SDR tonemapping"),
+ .init = init,
+ .uninit = uninit,
+ .priv_size = sizeof(TonemapxContext),
+ .priv_class = &tonemapx_class,
+ .p.priv_class = &tonemapx_class,
+ FILTER_INPUTS(tonemapx_inputs),
+ FILTER_OUTPUTS(ff_video_default_filterpad),
+ FILTER_OUTPUTS(tonemapx_outputs),
+ FILTER_QUERY_FUNC(query_formats),
+ .flags = AVFILTER_FLAG_SLICE_THREADS,
+ .p.flags = AVFILTER_FLAG_SLICE_THREADS,
+};
Index: FFmpeg/libavfilter/vf_tonemapx.h
===================================================================
@@ -4663,7 +4689,7 @@ Index: FFmpeg/libavfilter/vf_tonemapx.h
+ int rgb2rgb_passthrough;
+ const AVLumaCoefficients *coeffs, *ocoeffs;
+ double desat;
+ struct DoviMetadata *dovi;
+ struct FFDOVIMetadataRemap *dovi;
+ float *dovi_pbuf;
+ double (*lms2rgb_matrix)[3][3];
+ float (*ycc_offset)[3];
@@ -4730,15 +4756,14 @@ Index: FFmpeg/libavfilter/x86/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/x86/Makefile
+++ FFmpeg/libavfilter/x86/Makefile
@@ -36,6 +36,8 @@ OBJS-$(CONFIG_STEREO3D_FILTER)
OBJS-$(CONFIG_TBLEND_FILTER) += x86/vf_blend_init.o
OBJS-$(CONFIG_THRESHOLD_FILTER) += x86/vf_threshold_init.o
OBJS-$(CONFIG_TINTERLACE_FILTER) += x86/vf_tinterlace_init.o
@@ -1,5 +1,7 @@
OBJS-$(CONFIG_NOISE_FILTER) += x86/vf_noise.o
OBJS-$(CONFIG_SPP_FILTER) += x86/vf_spp.o
+OBJS-$(CONFIG_TONEMAPX_FILTER) += x86/vf_tonemapx_intrin_sse.o \
+ x86/vf_tonemapx_intrin_avx.o
OBJS-$(CONFIG_TRANSPOSE_FILTER) += x86/vf_transpose_init.o
OBJS-$(CONFIG_VOLUME_FILTER) += x86/af_volume_init.o
OBJS-$(CONFIG_V360_FILTER) += x86/vf_v360_init.o
X86ASM-OBJS-$(CONFIG_SCENE_SAD) += x86/scene_sad.o x86/scene_sad_init.o
Index: FFmpeg/libavfilter/x86/vf_tonemapx_intrin_avx.c
===================================================================
--- /dev/null
@@ -7405,7 +7430,7 @@ Index: FFmpeg/libavfilter/x86/vf_tonemapx_intrin_sse.c
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/x86/vf_tonemapx_intrin_sse.c
@@ -0,0 +1,2740 @@
@@ -0,0 +1,2742 @@
+/*
+ * Copyright (c) 2024 Gnattu OC <gnattuoc@me.com>
+ *
@@ -7470,6 +7495,7 @@ Index: FFmpeg/libavfilter/x86/vf_tonemapx_intrin_sse.c
+ return _mm_or_si128(_mm_and_si128(cmp, a), _mm_andnot_si128(cmp, xor_result));
+}
+
+/*
+X86_64_V2 inline static __m128 mix_float32x4(__m128 x, __m128 y, __m128 a)
+{
+ __m128 n = _mm_sub_ps(y, x);
@@ -7477,6 +7503,7 @@ Index: FFmpeg/libavfilter/x86/vf_tonemapx_intrin_sse.c
+ n = _mm_add_ps(n, x);
+ return n;
+}
+*/
+
+X86_64_V2 inline static float reduce_floatx4(__m128 x) {
+ x = _mm_hadd_ps(x, x);
debian/patches/0058-add-ac4-decoder-for-atsc-3-0.patch → debian/patches/0054-add-ac4-decoder-for-atsc-3-0.patch
Vendored
+66 -37
View File
@@ -1,38 +1,60 @@
Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3076,6 +3076,8 @@ ac3_decoder_select="ac3_parser ac3dsp bs
ac3_fixed_decoder_select="ac3_parser ac3dsp bswapdsp"
ac3_encoder_select="ac3dsp audiodsp me_cmp"
ac3_fixed_encoder_select="ac3dsp audiodsp me_cmp"
+ac4_decoder_select="ac4_parser"
+ac4_decoder_deps="gplv3"
acelp_kelvin_decoder_select="audiodsp celp_math"
adpcm_circus_decoder_deps="lgpl_gpl"
adpcm_g722_decoder_select="g722dsp"
@@ -3698,6 +3700,7 @@ vvc_qsv_decoder_select="vvc_mp4toannexb_
# parsers
aac_parser_select="adts_header mpeg4audio"
+ac4_parser_deps="gplv3"
ahx_parser_deps="lgpl_gpl"
apv_parser_select="cbs_apv"
av1_parser_select="cbs_av1"
Index: FFmpeg/libavcodec/Makefile
===================================================================
--- FFmpeg.orig/libavcodec/Makefile
+++ FFmpeg/libavcodec/Makefile
@@ -64,6 +64,8 @@ OBJS = ac3_parser.o
vorbis_parser.o \
@@ -65,6 +65,7 @@ OBJS = ac3_parser.o
xiph.o \
+SHLIBOBJS = float_fmul_reverse.o \
+
SHLIBOBJS = float_scalarproduct.o \
+ float_fmul_reverse.o \
timecode_internal.o \
# subsystems
include $(SRC_PATH)/libavcodec/aac/Makefile
include $(SRC_PATH)/libavcodec/hevc/Makefile
@@ -210,6 +212,7 @@ OBJS-$(CONFIG_AC3_ENCODER) +
@@ -219,6 +220,7 @@ OBJS-$(CONFIG_AC3_ENCODER) +
ac3.o kbdwin.o
OBJS-$(CONFIG_AC3_FIXED_ENCODER) += ac3enc_fixed.o ac3enc.o ac3tab.o ac3.o kbdwin.o
OBJS-$(CONFIG_AC3_MF_ENCODER) += mfenc.o mf_utils.o
+OBJS-$(CONFIG_AC4_DECODER) += ac4dec.o kbdwin.o
OBJS-$(CONFIG_ACELP_KELVIN_DECODER) += g729dec.o lsp.o celp_math.o celp_filters.o acelp_filters.o acelp_pitch_delay.o acelp_vectors.o g729postfilter.o
OBJS-$(CONFIG_ACELP_KELVIN_DECODER) += g729dec.o lsp.o celp_filters.o acelp_filters.o acelp_pitch_delay.o acelp_vectors.o g729postfilter.o
OBJS-$(CONFIG_AGM_DECODER) += agm.o jpegquanttables.o
OBJS-$(CONFIG_AIC_DECODER) += aic.o
@@ -1189,6 +1192,7 @@ OBJS-$(CONFIG_AAC_LATM_PARSER) +
OBJS-$(CONFIG_AHX_DECODER) += mpegaudiodec_float.o
@@ -1247,6 +1249,7 @@ OBJS-$(CONFIG_AAC_LATM_PARSER) +
OBJS-$(CONFIG_AAC_PARSER) += aac_parser.o aac_ac3_parser.o
OBJS-$(CONFIG_AC3_PARSER) += aac_ac3_parser.o ac3tab.o \
ac3_channel_layout_tab.o
+OBJS-$(CONFIG_AC4_PARSER) += ac4_parser.o
OBJS-$(CONFIG_ADX_PARSER) += adx_parser.o
OBJS-$(CONFIG_AHX_PARSER) += ahx_parser.o
OBJS-$(CONFIG_AMR_PARSER) += amr_parser.o
OBJS-$(CONFIG_AV1_PARSER) += av1_parser.o av1_parse.o
Index: FFmpeg/libavcodec/ac4_parser.c
===================================================================
--- /dev/null
+++ FFmpeg/libavcodec/ac4_parser.c
@@ -0,0 +1,96 @@
@@ -0,0 +1,99 @@
+/*
+ * This file is placed in the public domain.
+ *
+ * Copyright (c) 2025 Paul B Mahol
+ *
+ * This file is part of Librempeg
@@ -64,6 +86,7 @@ Index: FFmpeg/libavcodec/ac4_parser.c
+#include "ac4dec_data.h"
+#include "get_bits.h"
+#include "parser.h"
+#include "parser_internal.h"
+
+static int variable_bits(GetBitContext *gb, int bits)
+{
@@ -124,9 +147,9 @@ Index: FFmpeg/libavcodec/ac4_parser.c
+ return buf_size;
+}
+
+const AVCodecParser ff_ac4_parser = {
+ .codec_ids = { AV_CODEC_ID_AC4 },
+ .parser_parse = ac4_parse,
+const FFCodecParser ff_ac4_parser = {
+ PARSER_CODEC_LIST(AV_CODEC_ID_AC4),
+ .parse = ac4_parse,
+};
Index: FFmpeg/libavcodec/ac4dec.c
===================================================================
@@ -134,6 +157,8 @@ Index: FFmpeg/libavcodec/ac4dec.c
+++ FFmpeg/libavcodec/ac4dec.c
@@ -0,0 +1,6763 @@
+/*
+ * This file is placed in the public domain.
+ *
+ * AC-4 Audio Decoder
+ *
+ * Copyright (c) 2019 Paul B Mahol
@@ -4533,8 +4558,8 @@ Index: FFmpeg/libavcodec/ac4dec.c
+ return ret;
+ }
+
+ if (ss->im_codec_mode == IM_ASPX_AJCC)
+ ;
+ // if (ss->im_codec_mode == IM_ASPX_AJCC)
+ // ;
+
+ if (ss->im_codec_mode == IM_SCPL ||
+ ss->im_codec_mode == IM_ASPX_SCPL ||
@@ -6892,16 +6917,16 @@ Index: FFmpeg/libavcodec/ac4dec.c
+ FF_CODEC_DECODE_CB(ac4_decode_frame),
+ .flush = ac4_flush,
+ .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_CHANNEL_CONF,
+ .p.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
+ AV_SAMPLE_FMT_NONE },
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
+};
Index: FFmpeg/libavcodec/ac4dec_data.h
===================================================================
--- /dev/null
+++ FFmpeg/libavcodec/ac4dec_data.h
@@ -0,0 +1,1688 @@
@@ -0,0 +1,1690 @@
+/*
+ * This file is placed in the public domain.
+ *
+ * AC-4 Audio Decoder
+ *
+ * This file is part of Librempeg
@@ -8593,20 +8618,22 @@ Index: FFmpeg/libavcodec/allcodecs.c
===================================================================
--- FFmpeg.orig/libavcodec/allcodecs.c
+++ FFmpeg/libavcodec/allcodecs.c
@@ -430,6 +430,7 @@ extern const FFCodec ff_ac3_encoder;
@@ -442,6 +442,7 @@ extern const FFCodec ff_ac3_encoder;
extern const FFCodec ff_ac3_decoder;
extern const FFCodec ff_ac3_fixed_encoder;
extern const FFCodec ff_ac3_fixed_decoder;
+extern const FFCodec ff_ac4_decoder;
extern const FFCodec ff_acelp_kelvin_decoder;
extern const FFCodec ff_ahx_decoder;
extern const FFCodec ff_alac_encoder;
extern const FFCodec ff_alac_decoder;
Index: FFmpeg/libavcodec/float_fmul_reverse.c
===================================================================
--- /dev/null
+++ FFmpeg/libavcodec/float_fmul_reverse.c
@@ -0,0 +1,19 @@
@@ -0,0 +1,21 @@
+/*
+ * This file is placed in the public domain.
+ *
+ * This file is part of Librempeg
+ *
+ * Librempeg is free software; you can redistribute it and/or modify
@@ -8642,26 +8669,26 @@ Index: FFmpeg/libavcodec/parsers.c
===================================================================
--- FFmpeg.orig/libavcodec/parsers.c
+++ FFmpeg/libavcodec/parsers.c
@@ -23,6 +23,7 @@
extern const AVCodecParser ff_aac_parser;
extern const AVCodecParser ff_aac_latm_parser;
extern const AVCodecParser ff_ac3_parser;
@@ -40,6 +40,7 @@ FF_ENABLE_DEPRECATION_WARNINGS
extern const FFCodecParser ff_aac_parser;
extern const FFCodecParser ff_aac_latm_parser;
extern const FFCodecParser ff_ac3_parser;
+extern const AVCodecParser ff_ac4_parser;
extern const AVCodecParser ff_adx_parser;
extern const AVCodecParser ff_amr_parser;
extern const AVCodecParser ff_av1_parser;
extern const FFCodecParser ff_adx_parser;
extern const FFCodecParser ff_ahx_parser;
extern const FFCodecParser ff_amr_parser;
Index: FFmpeg/libavutil/Makefile
===================================================================
--- FFmpeg.orig/libavutil/Makefile
+++ FFmpeg/libavutil/Makefile
@@ -136,6 +136,7 @@ OBJS = adler32.o
@@ -142,6 +142,7 @@ OBJS = adler32.o
file.o \
file_open.o \
float_dsp.o \
+ float_fmul_reverse.o \
float_scalarproduct.o \
film_grain_params.o \
fixed_dsp.o \
frame.o \
hash.o \
Index: FFmpeg/libavutil/float_dsp.c
===================================================================
--- FFmpeg.orig/libavutil/float_dsp.c
@@ -8683,20 +8710,20 @@ Index: FFmpeg/libavutil/float_dsp.c
static void butterflies_float_c(float *restrict v1, float *restrict v2,
int len)
{
@@ -157,7 +147,7 @@ av_cold AVFloatDSPContext *avpriv_float_
@@ -146,7 +136,7 @@ av_cold AVFloatDSPContext *avpriv_float_
fdsp->vector_dmul_scalar = vector_dmul_scalar_c;
fdsp->vector_fmul_window = vector_fmul_window_c;
fdsp->vector_fmul_add = vector_fmul_add_c;
- fdsp->vector_fmul_reverse = vector_fmul_reverse_c;
+ fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_c;
fdsp->butterflies_float = butterflies_float_c;
fdsp->scalarproduct_float = avpriv_scalarproduct_float_c;
fdsp->scalarproduct_float = ff_scalarproduct_float_c;
fdsp->scalarproduct_double = ff_scalarproduct_double_c;
Index: FFmpeg/libavutil/float_dsp.h
===================================================================
--- FFmpeg.orig/libavutil/float_dsp.h
+++ FFmpeg/libavutil/float_dsp.h
@@ -229,6 +229,19 @@ float avpriv_scalarproduct_float_c(const
@@ -229,6 +229,19 @@ float ff_scalarproduct_float_c(const flo
double ff_scalarproduct_double_c(const double *v1, const double *v2,
size_t len);
@@ -8720,8 +8747,10 @@ Index: FFmpeg/libavutil/float_fmul_reverse.c
===================================================================
--- /dev/null
+++ FFmpeg/libavutil/float_fmul_reverse.c
@@ -0,0 +1,29 @@
@@ -0,0 +1,31 @@
+/*
+ * This file is placed in the public domain.
+ *
+ * This file is part of Librempeg
+ *
+ * Librempeg is free software; you can redistribute it and/or modify
debian/patches/0066-allow-vt-sw-decoder-for-every-codec.patch → debian/patches/0055-allow-vt-sw-decoder-for-every-codec.patch
Vendored
View File
debian/patches/0067-add-bwdif-videotoolbox-filter.patch → debian/patches/0056-add-bwdif-videotoolbox-filter.patch
Vendored
+34 -20
View File
@@ -2,19 +2,19 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3863,6 +3863,7 @@ boxblur_opencl_filter_deps="opencl gpl"
@@ -4120,6 +4120,7 @@ boxblur_opencl_filter_deps="opencl gpl"
bs2b_filter_deps="libbs2b"
bwdif_cuda_filter_deps="ffnvcodec"
bwdif_cuda_filter_deps_any="cuda_nvcc cuda_llvm"
+bwdif_videotoolbox_filter_deps="metal corevideo videotoolbox"
bwdif_vulkan_filter_deps="vulkan spirv_compiler"
chromaber_vulkan_filter_deps="vulkan spirv_compiler"
color_vulkan_filter_deps="vulkan spirv_compiler"
chromaber_vulkan_filter_deps="vulkan spirv_library"
color_vulkan_filter_deps="vulkan spirv_library"
Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -219,6 +219,10 @@ OBJS-$(CONFIG_BOXBLUR_OPENCL_FILTER)
@@ -224,6 +224,10 @@ OBJS-$(CONFIG_BOXBLUR_OPENCL_FILTER)
OBJS-$(CONFIG_BWDIF_FILTER) += vf_bwdif.o bwdifdsp.o yadif_common.o
OBJS-$(CONFIG_BWDIF_CUDA_FILTER) += vf_bwdif_cuda.o vf_bwdif_cuda.ptx.o \
yadif_common.o
@@ -29,19 +29,19 @@ Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -201,6 +201,7 @@ extern const AVFilter ff_vf_boxblur;
extern const AVFilter ff_vf_boxblur_opencl;
extern const AVFilter ff_vf_bwdif;
extern const AVFilter ff_vf_bwdif_cuda;
+extern const AVFilter ff_vf_bwdif_videotoolbox;
extern const AVFilter ff_vf_bwdif_vulkan;
extern const AVFilter ff_vf_cas;
extern const AVFilter ff_vf_ccrepack;
@@ -204,6 +204,7 @@ extern const FFFilter ff_vf_boxblur;
extern const FFFilter ff_vf_boxblur_opencl;
extern const FFFilter ff_vf_bwdif;
extern const FFFilter ff_vf_bwdif_cuda;
+extern const FFFilter ff_vf_bwdif_videotoolbox;
extern const FFFilter ff_vf_bwdif_vulkan;
extern const FFFilter ff_vf_cas;
extern const FFFilter ff_vf_ccrepack;
Index: FFmpeg/libavfilter/metal/vf_bwdif_videotoolbox.metal
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/metal/vf_bwdif_videotoolbox.metal
@@ -0,0 +1,271 @@
@@ -0,0 +1,270 @@
+/* bwdif.metal
+
+ Copyright (c) 2003-2024 HandBrake Team
@@ -312,12 +312,11 @@ Index: FFmpeg/libavfilter/metal/vf_bwdif_videotoolbox.metal
+ }
+ dst.write(pred.xyyy, pos);
+}
+
Index: FFmpeg/libavfilter/vf_bwdif_videotoolbox.m
===================================================================
--- /dev/null
+++ FFmpeg/libavfilter/vf_bwdif_videotoolbox.m
@@ -0,0 +1,448 @@
@@ -0,0 +1,449 @@
+/*
+ * Copyright (C) 2018 Philip Langdale <philipl@overt.org>
+ * 2020 Aman Karmani <aman@tmm1.net>
@@ -753,16 +752,31 @@ Index: FFmpeg/libavfilter/vf_bwdif_videotoolbox.m
+ },
+};
+
+const AVFilter ff_vf_bwdif_videotoolbox = {
+ .name = "bwdif_videotoolbox",
+ .description = NULL_IF_CONFIG_SMALL("BWDIF for VideoToolbox frames using Metal compute"),
+const FFFilter ff_vf_bwdif_videotoolbox = {
+ .p.name = "bwdif_videotoolbox",
+ .p.description = NULL_IF_CONFIG_SMALL("BWDIF for VideoToolbox frames using Metal compute"),
+ .priv_size = BWDIF_VT_CTX_SIZE,
+ .priv_class = &bwdif_videotoolbox_class,
+ .p.priv_class = &bwdif_videotoolbox_class,
+ .init = bwdif_videotoolbox_init,
+ .uninit = bwdif_videotoolbox_uninit,
+ FILTER_SINGLE_PIXFMT(AV_PIX_FMT_VIDEOTOOLBOX),
+ FILTER_INPUTS(bwdif_videotoolbox_inputs),
+ FILTER_OUTPUTS(bwdif_videotoolbox_outputs),
+ .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
+ .p.flags = AVFILTER_FLAG_HWDEVICE |
+ AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
+ .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+};
Index: FFmpeg/libavfilter/vf_yadif_videotoolbox.m
===================================================================
--- FFmpeg.orig/libavfilter/vf_yadif_videotoolbox.m
+++ FFmpeg/libavfilter/vf_yadif_videotoolbox.m
@@ -436,7 +436,8 @@ const FFFilter ff_vf_yadif_videotoolbox
.p.name = "yadif_videotoolbox",
.p.description = NULL_IF_CONFIG_SMALL("YADIF for VideoToolbox frames using Metal compute"),
.p.priv_class = &yadif_videotoolbox_class,
- .p.flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
+ .p.flags = AVFILTER_FLAG_HWDEVICE |
+ AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
.priv_size = YADIF_VT_CTX_SIZE,
.init = yadif_videotoolbox_init,
.uninit = yadif_videotoolbox_uninit,
debian/patches/0068-add-12bit-decoding-on-videotoolbox.patch → debian/patches/0057-add-12bit-decoding-on-videotoolbox.patch
Vendored
+4 -4
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/libavcodec/hevc/hevcdec.c
===================================================================
--- FFmpeg.orig/libavcodec/hevc/hevcdec.c
+++ FFmpeg/libavcodec/hevc/hevcdec.c
@@ -660,6 +660,9 @@ static enum AVPixelFormat get_format(HEV
@@ -719,6 +719,9 @@ static enum AVPixelFormat get_format(HEV
#if CONFIG_HEVC_NVDEC_HWACCEL
*fmt++ = AV_PIX_FMT_CUDA;
#endif
@@ -12,9 +12,9 @@ Index: FFmpeg/libavcodec/hevc/hevcdec.c
break;
case AV_PIX_FMT_YUV422P12:
#if CONFIG_HEVC_DXVA2_HWACCEL
@@ -675,6 +678,9 @@ static enum AVPixelFormat get_format(HEV
#if CONFIG_HEVC_VULKAN_HWACCEL
*fmt++ = AV_PIX_FMT_VULKAN;
@@ -737,6 +740,9 @@ static enum AVPixelFormat get_format(HEV
#if CONFIG_HEVC_NVDEC_HWACCEL
*fmt++ = AV_PIX_FMT_CUDA;
#endif
+#if CONFIG_HEVC_VIDEOTOOLBOX_HWACCEL
+ *fmt++ = AV_PIX_FMT_VIDEOTOOLBOX;
debian/patches/0069-fix-the-sub2video-perf-regressions.patch → debian/patches/0058-fix-the-sub2video-perf-regressions.patch
Vendored
+3 -17
View File
@@ -2,13 +2,13 @@ Index: FFmpeg/fftools/ffmpeg_filter.c
===================================================================
--- FFmpeg.orig/fftools/ffmpeg_filter.c
+++ FFmpeg/fftools/ffmpeg_filter.c
@@ -2651,7 +2651,8 @@ static void sub2video_heartbeat(InputFil
@@ -2948,7 +2948,8 @@ static void sub2video_heartbeat(InputFil
or if we need to initialize the system, update the
overlayed subpicture and its start/end times */
overlaid subpicture and its start/end times */
sub2video_update(ifp, pts2 + 1, NULL);
- else
+
+ if (av_buffersrc_get_nb_failed_requests(ifp->filter))
+ if (av_buffersrc_get_nb_failed_requests(ifp->ifilter.filter))
sub2video_push_ref(ifp, pts2);
}
@@ -31,17 +31,3 @@ Index: FFmpeg/tests/ref/fate/filter-overlay-dvdsub-2397
1, 3952, 3952, 10, 2013, 0x0e5db67e
1, 3962, 3962, 10, 2013, 0xfc9baf97
0, 119, 119, 1, 518400, 0x588534fc
Index: FFmpeg/tests/ref/fate/sub2video
===================================================================
--- FFmpeg.orig/tests/ref/fate/sub2video
+++ FFmpeg/tests/ref/fate/sub2video
@@ -68,8 +68,7 @@
0, 258, 258, 1, 518400, 0x34cdddee
0, 269, 269, 1, 518400, 0xbab197ea
1, 53910000, 53910000, 2696000, 2095, 0x61bb15ed
-0, 270, 270, 1, 518400, 0xbab197ea
-0, 271, 271, 1, 518400, 0x4db4ce51
+0, 270, 270, 1, 518400, 0x4db4ce51
0, 283, 283, 1, 518400, 0xbab197ea
1, 56663000, 56663000, 1262000, 1013, 0xc9ae89b7
0, 284, 284, 1, 518400, 0xe6bc0ea9
debian/patches/0070-allow-vpl-qsv-to-init-with-the-legacy-msdk-path.patch → debian/patches/0059-allow-vpl-qsv-to-init-with-the-legacy-msdk-path.patch
Vendored
+8 -8
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -2533,6 +2533,7 @@ HAVE_LIST="
@@ -2687,6 +2687,7 @@ HAVE_LIST="
xmllint
zlib_gzip
openvino2
@@ -10,9 +10,9 @@ Index: FFmpeg/configure
"
# options emitted with CONFIG_ prefix but not available on the command line
@@ -6989,6 +6990,9 @@ elif enabled libvpl; then
@@ -7369,6 +7370,9 @@ elif enabled libvpl; then
die "ERROR: libvpl >= 2.6 not found"
add_cflags -DMFX_DEPRECATED_OFF
add_cppflags -DMFX_DEPRECATED_OFF
check_type "vpl/mfxdefs.h vpl/mfxvideo.h" "struct mfxConfigInterface"
+# Ancient Intel driver on Windows still require the legacy init functions.
+# No compiler option is changed if the test fails, just needs to add back the disabled libmfx.
@@ -84,7 +84,7 @@ Index: FFmpeg/libavcodec/qsvenc.c
===================================================================
--- FFmpeg.orig/libavcodec/qsvenc.c
+++ FFmpeg/libavcodec/qsvenc.c
@@ -1192,7 +1192,7 @@ static int init_video_param(AVCodecConte
@@ -1190,7 +1190,7 @@ static int init_video_param(AVCodecConte
#if QSV_HAVE_EXT_AV1_PARAM
if (avctx->codec_id == AV_CODEC_ID_AV1) {
@@ -93,7 +93,7 @@ Index: FFmpeg/libavcodec/qsvenc.c
q->extav1tileparam.Header.BufferId = MFX_EXTBUFF_AV1_TILE_PARAM;
q->extav1tileparam.Header.BufferSz = sizeof(q->extav1tileparam);
q->extav1tileparam.NumTileColumns = q->tile_cols;
@@ -1416,7 +1416,7 @@ static int qsv_retrieve_enc_av1_params(A
@@ -1414,7 +1414,7 @@ static int qsv_retrieve_enc_av1_params(A
(mfxExtBuffer*)&co3,
};
@@ -106,7 +106,7 @@ Index: FFmpeg/libavfilter/qsvvpp.c
===================================================================
--- FFmpeg.orig/libavfilter/qsvvpp.c
+++ FFmpeg/libavfilter/qsvvpp.c
@@ -1065,6 +1065,16 @@ int ff_qsvvpp_filter_frame(QSVVPPContext
@@ -1066,6 +1066,16 @@ int ff_qsvvpp_filter_frame(QSVVPPContext
return 0;
}
@@ -123,7 +123,7 @@ Index: FFmpeg/libavfilter/qsvvpp.c
#if QSV_ONEVPL
int ff_qsvvpp_create_mfx_session(void *ctx,
@@ -1108,6 +1118,15 @@ int ff_qsvvpp_create_mfx_session(void *c
@@ -1109,6 +1119,15 @@ int ff_qsvvpp_create_mfx_session(void *c
impl_idx++;
}
@@ -139,7 +139,7 @@ Index: FFmpeg/libavfilter/qsvvpp.c
if (sts < 0)
return ff_qsvvpp_print_error(ctx, sts,
"Error creating a MFX session");
@@ -1125,6 +1144,19 @@ int ff_qsvvpp_create_mfx_session(void *c
@@ -1126,6 +1145,19 @@ int ff_qsvvpp_create_mfx_session(void *c
mfxVersion *pver,
mfxSession *psession)
{
debian/patches/0059-tune-the-default-inter-thread-queue-sizes.patch
Vendored
-29
View File
@@ -1,29 +0,0 @@
Index: FFmpeg/fftools/ffmpeg_sched.h
===================================================================
--- FFmpeg.orig/fftools/ffmpeg_sched.h
+++ FFmpeg/fftools/ffmpeg_sched.h
@@ -257,7 +257,10 @@ int sch_add_mux(Scheduler *sch, SchThrea
/**
* Default size of a frame thread queue.
*/
-#define DEFAULT_FRAME_THREAD_QUEUE_SIZE 8
+#define DEFAULT_FRAME_THREAD_QUEUE_SIZE 1
+
+// The new default value of 8 does little to help hwaccel, but instead increases
+// extra_hw_frames, which causes video memory on dGPU to be exhausted more easily.
/**
* Add a muxed stream for a previously added muxer.
Index: FFmpeg/tests/ref/fate/ffmpeg-fix_sub_duration_heartbeat
===================================================================
--- FFmpeg.orig/tests/ref/fate/ffmpeg-fix_sub_duration_heartbeat
+++ FFmpeg/tests/ref/fate/ffmpeg-fix_sub_duration_heartbeat
@@ -33,3 +33,8 @@
<font face="Monospace">{\an7}(<i> inaudible radio chatter</i> )
>> Safety remains our numb</font>
+9
+00:00:03,704 --> 00:00:04,004
+<font face="Monospace">{\an7}(<i> inaudible radio chatter</i> )
+>> Safety remains our number one</font>
+
debian/patches/0071-alway-set-videotoolboxenc-pixel-buffer-info.patch → debian/patches/0060-alway-set-videotoolboxenc-pixel-buffer-info.patch
Vendored
View File
debian/patches/0060-backport-fixes-for-vulkan-from-upstream.patch
Vendored
-185
View File
@@ -1,185 +0,0 @@
Index: FFmpeg/libavutil/hwcontext_vulkan.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_vulkan.c
+++ FFmpeg/libavutil/hwcontext_vulkan.c
@@ -90,6 +90,9 @@ typedef struct VulkanDevicePriv {
VkPhysicalDeviceMemoryProperties mprops;
VkPhysicalDeviceExternalMemoryHostPropertiesEXT hprops;
+ /* Opaque FD external semaphore properties */
+ VkExternalSemaphoreProperties ext_sem_props_opaque;
+
/* Features */
VkPhysicalDeviceVulkan11Features device_features_1_1;
VkPhysicalDeviceVulkan12Features device_features_1_2;
@@ -1212,7 +1215,7 @@ static int setup_queue_families(AVHWDevi
};
qf[i] = (VkQueueFamilyProperties2) {
.sType = VK_STRUCTURE_TYPE_QUEUE_FAMILY_PROPERTIES_2,
- .pNext = &qf_vid[i],
+ .pNext = p->vkctx.extensions & FF_VK_EXT_VIDEO_QUEUE ? &qf_vid[i] : NULL,
};
}
@@ -1642,6 +1645,7 @@ static int vulkan_device_init(AVHWDevice
FFVulkanFunctions *vk = &p->vkctx.vkfn;
VkQueueFamilyProperties2 *qf;
VkQueueFamilyVideoPropertiesKHR *qf_vid;
+ VkPhysicalDeviceExternalSemaphoreInfo ext_sem_props_info;
int graph_index, comp_index, tx_index, enc_index, dec_index;
/* Set device extension flags */
@@ -1687,6 +1691,24 @@ static int vulkan_device_init(AVHWDevice
return AVERROR_EXTERNAL;
}
+ ext_sem_props_info = (VkPhysicalDeviceExternalSemaphoreInfo) {
+ .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTERNAL_SEMAPHORE_INFO,
+ };
+
+ /* Opaque FD semaphore properties */
+ ext_sem_props_info.handleType =
+#ifdef _WIN32
+ IsWindows8OrGreater()
+ ? VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32_BIT
+ : VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32_KMT_BIT;
+#else
+ VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT;
+#endif
+ p->ext_sem_props_opaque.sType = VK_STRUCTURE_TYPE_EXTERNAL_SEMAPHORE_PROPERTIES;
+ vk->GetPhysicalDeviceExternalSemaphoreProperties(hwctx->phys_dev,
+ &ext_sem_props_info,
+ &p->ext_sem_props_opaque);
+
qf = av_malloc_array(qf_num, sizeof(VkQueueFamilyProperties2));
if (!qf)
return AVERROR(ENOMEM);
@@ -1703,7 +1725,7 @@ static int vulkan_device_init(AVHWDevice
};
qf[i] = (VkQueueFamilyProperties2) {
.sType = VK_STRUCTURE_TYPE_QUEUE_FAMILY_PROPERTIES_2,
- .pNext = &qf_vid[i],
+ .pNext = p->vkctx.extensions & FF_VK_EXT_VIDEO_QUEUE ? &qf_vid[i] : NULL,
};
}
@@ -1980,7 +2002,7 @@ static int vulkan_frames_get_constraints
count += vkfmt_from_pixfmt2(ctx, vk_formats_list[i].pixfmt,
p->use_linear_images ? VK_IMAGE_TILING_LINEAR :
VK_IMAGE_TILING_OPTIMAL,
- NULL, NULL, NULL, NULL, 0, 0) >= 0;
+ NULL, NULL, NULL, NULL, p->disable_multiplane, 1) >= 0;
}
constraints->valid_sw_formats = av_malloc_array(count + 1,
@@ -1993,7 +2015,7 @@ static int vulkan_frames_get_constraints
if (vkfmt_from_pixfmt2(ctx, vk_formats_list[i].pixfmt,
p->use_linear_images ? VK_IMAGE_TILING_LINEAR :
VK_IMAGE_TILING_OPTIMAL,
- NULL, NULL, NULL, NULL, 0, 0) >= 0) {
+ NULL, NULL, NULL, NULL, p->disable_multiplane, 1) >= 0) {
constraints->valid_sw_formats[count++] = vk_formats_list[i].pixfmt;
}
}
@@ -2346,8 +2368,19 @@ static int create_frame(AVHWFramesContex
VulkanDevicePriv *p = ctx->hwctx;
AVVulkanDeviceContext *hwctx = &p->p;
FFVulkanFunctions *vk = &p->vkctx.vkfn;
+ AVVkFrame *f;
- VkExportSemaphoreCreateInfo ext_sem_info = {
+ VkSemaphoreTypeCreateInfo sem_type_info = {
+ .sType = VK_STRUCTURE_TYPE_SEMAPHORE_TYPE_CREATE_INFO,
+ .semaphoreType = VK_SEMAPHORE_TYPE_TIMELINE,
+ .initialValue = 0,
+ };
+ VkSemaphoreCreateInfo sem_spawn = {
+ .sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO,
+ .pNext = &sem_type_info,
+ };
+
+ VkExportSemaphoreCreateInfo ext_sem_info_opaque = {
.sType = VK_STRUCTURE_TYPE_EXPORT_SEMAPHORE_CREATE_INFO,
#ifdef _WIN32
.handleTypes = IsWindows8OrGreater()
@@ -2358,23 +2391,13 @@ static int create_frame(AVHWFramesContex
#endif
};
- VkSemaphoreTypeCreateInfo sem_type_info = {
- .sType = VK_STRUCTURE_TYPE_SEMAPHORE_TYPE_CREATE_INFO,
-#ifdef _WIN32
- .pNext = p->vkctx.extensions & FF_VK_EXT_EXTERNAL_WIN32_SEM ? &ext_sem_info : NULL,
-#else
- .pNext = p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_SEM ? &ext_sem_info : NULL,
-#endif
- .semaphoreType = VK_SEMAPHORE_TYPE_TIMELINE,
- .initialValue = 0,
- };
-
- VkSemaphoreCreateInfo sem_spawn = {
- .sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO,
- .pNext = &sem_type_info,
- };
+ /* Check if exporting is supported before chaining any structs */
+ if (p->ext_sem_props_opaque.externalSemaphoreFeatures & VK_EXTERNAL_SEMAPHORE_FEATURE_EXPORTABLE_BIT) {
+ if (p->vkctx.extensions & (FF_VK_EXT_EXTERNAL_WIN32_SEM | FF_VK_EXT_EXTERNAL_FD_SEM))
+ ff_vk_link_struct(&sem_type_info, &ext_sem_info_opaque);
+ }
- AVVkFrame *f = av_vk_frame_alloc();
+ f = av_vk_frame_alloc();
if (!f) {
av_log(ctx, AV_LOG_ERROR, "Unable to allocate memory for AVVkFrame!\n");
return AVERROR(ENOMEM);
@@ -2689,11 +2712,11 @@ static int vulkan_frames_init(AVHWFrames
!(hwctx->usage & VK_IMAGE_USAGE_VIDEO_DECODE_DST_BIT_KHR)));
int sampleable = hwctx->usage & (VK_IMAGE_USAGE_SAMPLED_BIT |
VK_IMAGE_USAGE_STORAGE_BIT);
+ hwctx->img_flags = VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT;
if (sampleable && !is_lone_dpb) {
- hwctx->img_flags = VK_IMAGE_CREATE_ALIAS_BIT;
+ hwctx->img_flags |= VK_IMAGE_CREATE_ALIAS_BIT;
if ((fmt->vk_planes > 1) && (hwctx->format[0] == fmt->vkf))
- hwctx->img_flags |= VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT |
- VK_IMAGE_CREATE_EXTENDED_USAGE_BIT;
+ hwctx->img_flags |= VK_IMAGE_CREATE_EXTENDED_USAGE_BIT;
}
}
Index: FFmpeg/libavutil/vulkan.c
===================================================================
--- FFmpeg.orig/libavutil/vulkan.c
+++ FFmpeg/libavutil/vulkan.c
@@ -185,7 +185,7 @@ int ff_vk_load_props(FFVulkanContext *s)
};
s->qf_props[i] = (VkQueueFamilyProperties2) {
.sType = VK_STRUCTURE_TYPE_QUEUE_FAMILY_PROPERTIES_2,
- .pNext = &s->video_props[i],
+ .pNext = s->extensions & FF_VK_EXT_VIDEO_QUEUE ? &s->video_props[i] : NULL,
};
}
Index: FFmpeg/libavutil/vulkan_functions.h
===================================================================
--- FFmpeg.orig/libavutil/vulkan_functions.h
+++ FFmpeg/libavutil/vulkan_functions.h
@@ -33,10 +33,8 @@ typedef enum FFVulkanExtensions {
FF_VK_EXT_EXTERNAL_FD_SEM = 1ULL << 3, /* VK_KHR_external_semaphore_fd */
FF_VK_EXT_EXTERNAL_HOST_MEMORY = 1ULL << 4, /* VK_EXT_external_memory_host */
FF_VK_EXT_DEBUG_UTILS = 1ULL << 5, /* VK_EXT_debug_utils */
-#ifdef _WIN32
FF_VK_EXT_EXTERNAL_WIN32_MEMORY = 1ULL << 6, /* VK_KHR_external_memory_win32 */
FF_VK_EXT_EXTERNAL_WIN32_SEM = 1ULL << 7, /* VK_KHR_external_semaphore_win32 */
-#endif
FF_VK_EXT_DESCRIPTOR_BUFFER = 1ULL << 8, /* VK_EXT_descriptor_buffer */
FF_VK_EXT_DEVICE_DRM = 1ULL << 9, /* VK_EXT_physical_device_drm */
FF_VK_EXT_VIDEO_QUEUE = 1ULL << 10, /* VK_KHR_video_queue */
@@ -75,6 +73,7 @@ typedef enum FFVulkanExtensions {
MACRO(1, 0, FF_VK_EXT_NO_FLAG, CreateDevice) \
MACRO(1, 0, FF_VK_EXT_NO_FLAG, GetPhysicalDeviceFeatures2) \
MACRO(1, 0, FF_VK_EXT_NO_FLAG, GetPhysicalDeviceProperties) \
+ MACRO(1, 0, FF_VK_EXT_NO_FLAG, GetPhysicalDeviceExternalSemaphoreProperties) \
MACRO(1, 0, FF_VK_EXT_VIDEO_QUEUE, GetPhysicalDeviceVideoCapabilitiesKHR) \
MACRO(1, 0, FF_VK_EXT_VIDEO_QUEUE, GetPhysicalDeviceVideoFormatPropertiesKHR) \
MACRO(1, 0, FF_VK_EXT_NO_FLAG, DeviceWaitIdle) \
debian/patches/0072-add-remove-dovi-hdr10plus-bsf.patch → debian/patches/0061-add-remove-dovi-hdr10plus-bsf.patch
Vendored
+2 -2
View File
@@ -55,7 +55,7 @@ Index: FFmpeg/libavcodec/bsf/av1_metadata.c
+ }
+ }
+
+ if (ctx->remove_hdr10plus && provider_code == ITU_T_T35_PROVIDER_CODE_SMTPE) {
+ if (ctx->remove_hdr10plus && provider_code == ITU_T_T35_PROVIDER_CODE_SAMSUNG) {
+ provider_oriented_code = AV_RB16(t35->payload + 2);
+ application_identifier = AV_RB8(t35->payload + 4);
+ // HDR10+ Metadata
@@ -131,7 +131,7 @@ Index: FFmpeg/libavcodec/bsf/h265_metadata.c
+ uint8_t *nal_sei = au->units[i].data;
+ // This Matches ITU-T T.35 SMPTE ST 2094-40
+ if (nal_sei[0] == 0x4E && nal_sei[1] == 0x01 && nal_sei[2] == 0x04) {
+ if (nal_sei[4] == ITU_T_T35_COUNTRY_CODE_US && nal_sei[6] == ITU_T_T35_PROVIDER_CODE_SMTPE) {
+ if (nal_sei[4] == ITU_T_T35_COUNTRY_CODE_US && nal_sei[6] == ITU_T_T35_PROVIDER_CODE_SAMSUNG) {
+ // identifier for HDR10+
+ const uint8_t smpte2094_40_provider_oriented_code = 0x01;
+ const uint8_t smpte2094_40_application_identifier = 0x04;
debian/patches/0061-revert-the-broken-vulkan-hwupload-rewrite.patch
Vendored
-624
View File
@@ -1,624 +0,0 @@
Index: FFmpeg/libavutil/hwcontext_vulkan.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_vulkan.c
+++ FFmpeg/libavutil/hwcontext_vulkan.c
@@ -139,9 +139,6 @@ typedef struct VulkanFramesPriv {
FFVkExecPool upload_exec;
FFVkExecPool download_exec;
- /* Temporary buffer pools */
- AVBufferPool *tmp;
-
/* Modifier info list to free at uninit */
VkImageDrmFormatModifierListCreateInfoEXT *modifier_info;
} VulkanFramesPriv;
@@ -2622,8 +2619,6 @@ static void vulkan_frames_uninit(AVHWFra
ff_vk_exec_pool_free(&p->vkctx, &fp->compute_exec);
ff_vk_exec_pool_free(&p->vkctx, &fp->upload_exec);
ff_vk_exec_pool_free(&p->vkctx, &fp->download_exec);
-
- av_buffer_pool_uninit(&fp->tmp);
}
static int vulkan_frames_init(AVHWFramesContext *hwfc)
@@ -3809,290 +3804,128 @@ static int vulkan_map_from(AVHWFramesCon
return AVERROR(ENOSYS);
}
-static int copy_buffer_data(AVHWFramesContext *hwfc, AVBufferRef *buf,
- AVFrame *swf, VkBufferImageCopy *region,
- int planes, int upload)
+static size_t get_req_buffer_size(VulkanDevicePriv *p, int *stride, int height)
{
- VkResult ret;
- VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
- FFVulkanFunctions *vk = &p->vkctx.vkfn;
- AVVulkanDeviceContext *hwctx = &p->p;
-
- FFVkBuffer *vkbuf = (FFVkBuffer *)buf->data;
-
- const VkMappedMemoryRange flush_info = {
- .sType = VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE,
- .memory = vkbuf->mem,
- .size = VK_WHOLE_SIZE,
- };
-
- if (!(vkbuf->flags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) && !upload) {
- ret = vk->InvalidateMappedMemoryRanges(hwctx->act_dev, 1,
- &flush_info);
- if (ret != VK_SUCCESS) {
- av_log(hwfc, AV_LOG_ERROR, "Failed to invalidate buffer data: %s\n",
- ff_vk_ret2str(ret));
- return AVERROR_EXTERNAL;
- }
- }
-
- for (int i = 0; i < planes; i++)
- av_image_copy_plane(vkbuf->mapped_mem + region[i].bufferOffset,
- region[i].bufferRowLength,
- swf->data[i],
- swf->linesize[i],
- swf->linesize[i],
- region[i].imageExtent.height);
-
- if (!(vkbuf->flags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) && upload) {
- ret = vk->FlushMappedMemoryRanges(hwctx->act_dev, 1,
- &flush_info);
- if (ret != VK_SUCCESS) {
- av_log(hwfc, AV_LOG_ERROR, "Failed to flush buffer data: %s\n",
- ff_vk_ret2str(ret));
- return AVERROR_EXTERNAL;
- }
- }
-
- return 0;
+ size_t size;
+ *stride = FFALIGN(*stride, p->props.properties.limits.optimalBufferCopyRowPitchAlignment);
+ size = height*(*stride);
+ size = FFALIGN(size, p->props.properties.limits.minMemoryMapAlignment);
+ return size;
}
-static int get_plane_buf(AVHWFramesContext *hwfc, AVBufferRef **dst,
- AVFrame *swf, VkBufferImageCopy *region, int upload)
+static int transfer_image_buf(AVHWFramesContext *hwfc, AVFrame *f,
+ AVBufferRef **bufs, size_t *buf_offsets,
+ const int *buf_stride, int w,
+ int h, enum AVPixelFormat pix_fmt, int to_buf)
{
int err;
+ AVVkFrame *frame = (AVVkFrame *)f->data[0];
VulkanFramesPriv *fp = hwfc->hwctx;
VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
- const int planes = av_pix_fmt_count_planes(swf->format);
-
- size_t buf_offset = 0;
- for (int i = 0; i < planes; i++) {
- size_t size;
- ptrdiff_t linesize = swf->linesize[i];
-
- uint32_t p_w, p_h;
- get_plane_wh(&p_w, &p_h, swf->format, swf->width, swf->height, i);
+ FFVulkanFunctions *vk = &p->vkctx.vkfn;
+ VkImageMemoryBarrier2 img_bar[AV_NUM_DATA_POINTERS];
+ int nb_img_bar = 0;
- linesize = FFALIGN(linesize,
- p->props.properties.limits.optimalBufferCopyRowPitchAlignment);
- size = p_h*linesize;
-
- region[i] = (VkBufferImageCopy) {
- .bufferOffset = buf_offset,
- .bufferRowLength = linesize,
- .bufferImageHeight = p_h,
- .imageSubresource.layerCount = 1,
- .imageExtent = (VkExtent3D){ p_w, p_h, 1 },
- /* Rest of the fields adjusted/filled in later */
- };
+ const int nb_images = ff_vk_count_images(frame);
+ int pixfmt_planes = av_pix_fmt_count_planes(pix_fmt);
+ const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
- buf_offset = FFALIGN(buf_offset + size,
- p->props.properties.limits.optimalBufferCopyOffsetAlignment);
- }
+ VkCommandBuffer cmd_buf;
+ FFVkExecContext *exec = ff_vk_exec_get(to_buf ? &fp->download_exec :
+ &fp->upload_exec);
+ cmd_buf = exec->buf;
+ ff_vk_exec_start(&p->vkctx, exec);
- err = ff_vk_get_pooled_buffer(&p->vkctx, &fp->tmp, dst,
- VK_BUFFER_USAGE_TRANSFER_SRC_BIT |
- VK_BUFFER_USAGE_TRANSFER_DST_BIT,
- NULL, buf_offset,
- VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
- VK_MEMORY_PROPERTY_HOST_CACHED_BIT);
+ err = ff_vk_exec_add_dep_buf(&p->vkctx, exec, bufs, pixfmt_planes, 1);
if (err < 0)
return err;
- return 0;
-}
-
-static int create_mapped_buffer(AVHWFramesContext *hwfc,
- FFVkBuffer *vkb, VkBufferUsageFlags usage,
- size_t size,
- VkExternalMemoryBufferCreateInfo *create_desc,
- VkImportMemoryHostPointerInfoEXT *import_desc,
- VkMemoryHostPointerPropertiesEXT props)
-{
- int err;
- VkResult ret;
- VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
- FFVulkanFunctions *vk = &p->vkctx.vkfn;
- AVVulkanDeviceContext *hwctx = &p->p;
-
- VkBufferCreateInfo buf_spawn = {
- .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
- .pNext = create_desc,
- .usage = usage,
- .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
- .size = size,
- };
- VkMemoryRequirements req = {
- .size = size,
- .alignment = p->hprops.minImportedHostPointerAlignment,
- .memoryTypeBits = props.memoryTypeBits,
- };
-
- err = ff_vk_alloc_mem(&p->vkctx, &req,
- VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT,
- import_desc, &vkb->flags, &vkb->mem);
+ err = ff_vk_exec_add_dep_frame(&p->vkctx, exec, f,
+ VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT,
+ VK_PIPELINE_STAGE_2_TRANSFER_BIT);
if (err < 0)
return err;
- ret = vk->CreateBuffer(hwctx->act_dev, &buf_spawn, hwctx->alloc, &vkb->buf);
- if (ret != VK_SUCCESS) {
- vk->FreeMemory(hwctx->act_dev, vkb->mem, hwctx->alloc);
- return AVERROR_EXTERNAL;
- }
-
- ret = vk->BindBufferMemory(hwctx->act_dev, vkb->buf, vkb->mem, 0);
- if (ret != VK_SUCCESS) {
- vk->FreeMemory(hwctx->act_dev, vkb->mem, hwctx->alloc);
- vk->DestroyBuffer(hwctx->act_dev, vkb->buf, hwctx->alloc);
- return AVERROR_EXTERNAL;
- }
-
- return 0;
-}
-
-static void destroy_avvkbuf(void *opaque, uint8_t *data)
-{
- FFVulkanContext *s = opaque;
- FFVkBuffer *buf = (FFVkBuffer *)data;
- ff_vk_free_buf(s, buf);
- av_free(buf);
-}
-
-static int host_map_frame(AVHWFramesContext *hwfc, AVBufferRef **dst, int *nb_bufs,
- AVFrame *swf, VkBufferImageCopy *region, int upload)
-{
- int err;
- VkResult ret;
- VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
- FFVulkanFunctions *vk = &p->vkctx.vkfn;
- AVVulkanDeviceContext *hwctx = &p->p;
-
- const int planes = av_pix_fmt_count_planes(swf->format);
-
- VkExternalMemoryBufferCreateInfo create_desc = {
- .sType = VK_STRUCTURE_TYPE_EXTERNAL_MEMORY_BUFFER_CREATE_INFO,
- .handleTypes = VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT,
- };
- VkImportMemoryHostPointerInfoEXT import_desc = {
- .sType = VK_STRUCTURE_TYPE_IMPORT_MEMORY_HOST_POINTER_INFO_EXT,
- .handleType = VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT,
- };
- VkMemoryHostPointerPropertiesEXT props;
-
- for (int i = 0; i < planes; i++) {
- FFVkBuffer *vkb;
- uint32_t p_w, p_h;
- size_t offs;
- size_t buffer_size;
-
- /* We can't host map images with negative strides */
- if (swf->linesize[i] < 0) {
- err = AVERROR(EINVAL);
- goto fail;
- }
-
- get_plane_wh(&p_w, &p_h, swf->format, swf->width, swf->height, i);
-
- /* Get the previous point at which mapping was possible and use it */
- offs = (uintptr_t)swf->data[i] % p->hprops.minImportedHostPointerAlignment;
- import_desc.pHostPointer = swf->data[i] - offs;
+ ff_vk_frame_barrier(&p->vkctx, exec, f, img_bar, &nb_img_bar,
+ VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT,
+ VK_PIPELINE_STAGE_2_TRANSFER_BIT_KHR,
+ to_buf ? VK_ACCESS_TRANSFER_READ_BIT :
+ VK_ACCESS_TRANSFER_WRITE_BIT,
+ to_buf ? VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL :
+ VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ VK_QUEUE_FAMILY_IGNORED);
- props = (VkMemoryHostPointerPropertiesEXT) {
- VK_STRUCTURE_TYPE_MEMORY_HOST_POINTER_PROPERTIES_EXT,
- };
- ret = vk->GetMemoryHostPointerPropertiesEXT(hwctx->act_dev,
- import_desc.handleType,
- import_desc.pHostPointer,
- &props);
- if (!(ret == VK_SUCCESS && props.memoryTypeBits)) {
- err = AVERROR(EINVAL);
- goto fail;
- }
+ vk->CmdPipelineBarrier2(cmd_buf, &(VkDependencyInfo) {
+ .sType = VK_STRUCTURE_TYPE_DEPENDENCY_INFO,
+ .pImageMemoryBarriers = img_bar,
+ .imageMemoryBarrierCount = nb_img_bar,
+ });
- /* Buffer region for this plane */
- region[i] = (VkBufferImageCopy) {
- .bufferOffset = offs,
- .bufferRowLength = swf->linesize[i],
- .bufferImageHeight = p_h,
+ /* Schedule a copy for each plane */
+ for (int i = 0; i < pixfmt_planes; i++) {
+ int idx = FFMIN(i, nb_images - 1);
+ VkImageAspectFlags plane_aspect[] = { VK_IMAGE_ASPECT_COLOR_BIT,
+ VK_IMAGE_ASPECT_PLANE_0_BIT,
+ VK_IMAGE_ASPECT_PLANE_1_BIT,
+ VK_IMAGE_ASPECT_PLANE_2_BIT, };
+
+ FFVkBuffer *vkbuf = (FFVkBuffer *)bufs[i]->data;
+ VkBufferImageCopy buf_reg = {
+ .bufferOffset = buf_offsets[i],
+ .bufferRowLength = buf_stride[i] / desc->comp[i].step,
.imageSubresource.layerCount = 1,
- .imageExtent = (VkExtent3D){ p_w, p_h, 1 },
- /* Rest of the fields adjusted/filled in later */
+ .imageSubresource.aspectMask = plane_aspect[(pixfmt_planes != nb_images) +
+ i*(pixfmt_planes != nb_images)],
+ .imageOffset = { 0, 0, 0, },
};
- /* Add the offset at the start, which gets ignored */
- buffer_size = offs + swf->linesize[i]*p_h;
- buffer_size = FFALIGN(buffer_size, p->props.properties.limits.minMemoryMapAlignment);
- buffer_size = FFALIGN(buffer_size, p->hprops.minImportedHostPointerAlignment);
-
- /* Create a buffer */
- vkb = av_mallocz(sizeof(*vkb));
- if (!vkb) {
- err = AVERROR(ENOMEM);
- goto fail;
- }
-
- err = create_mapped_buffer(hwfc, vkb,
- upload ? VK_BUFFER_USAGE_TRANSFER_SRC_BIT :
- VK_BUFFER_USAGE_TRANSFER_DST_BIT,
- buffer_size, &create_desc, &import_desc,
- props);
- if (err < 0) {
- av_free(vkb);
- goto fail;
- }
+ uint32_t p_w, p_h;
+ get_plane_wh(&p_w, &p_h, pix_fmt, w, h, i);
- /* Create a ref */
- dst[*nb_bufs] = av_buffer_create((uint8_t *)vkb, sizeof(*vkb),
- destroy_avvkbuf, &p->vkctx, 0);
- if (!dst[*nb_bufs]) {
- destroy_avvkbuf(&p->vkctx, (uint8_t *)vkb);
- err = AVERROR(ENOMEM);
- goto fail;
- }
+ buf_reg.bufferImageHeight = p_h;
+ buf_reg.imageExtent = (VkExtent3D){ p_w, p_h, 1, };
- (*nb_bufs)++;
+ if (to_buf)
+ vk->CmdCopyImageToBuffer(cmd_buf, frame->img[idx],
+ img_bar[0].newLayout,
+ vkbuf->buf,
+ 1, &buf_reg);
+ else
+ vk->CmdCopyBufferToImage(cmd_buf, vkbuf->buf, frame->img[idx],
+ img_bar[0].newLayout,
+ 1, &buf_reg);
}
- return 0;
+ err = ff_vk_exec_submit(&p->vkctx, exec);
+ if (err < 0)
+ return err;
-fail:
- for (int i = 0; i < (*nb_bufs); i++)
- av_buffer_unref(&dst[i]);
- return err;
+ ff_vk_exec_wait(&p->vkctx, exec);
+
+ return 0;
}
-static int vulkan_transfer_frame(AVHWFramesContext *hwfc,
- AVFrame *swf, AVFrame *hwf,
- int upload)
+static int vulkan_transfer_data(AVHWFramesContext *hwfc, const AVFrame *vkf,
+ const AVFrame *swf, int from)
{
- int err;
- VulkanFramesPriv *fp = hwfc->hwctx;
- VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
+ int err = 0;
+ VkResult ret;
+ AVHWDeviceContext *dev_ctx = hwfc->device_ctx;
+ VulkanDevicePriv *p = dev_ctx->hwctx;
+ AVVulkanDeviceContext *hwctx = &p->p;
FFVulkanFunctions *vk = &p->vkctx.vkfn;
- int host_mapped = 0;
-
- AVVkFrame *hwf_vk = (AVVkFrame *)hwf->data[0];
- VkBufferImageCopy region[AV_NUM_DATA_POINTERS]; // always one per plane
+ AVFrame tmp;
+ FFVkBuffer *vkbufs[AV_NUM_DATA_POINTERS];
+ AVBufferRef *bufs[AV_NUM_DATA_POINTERS] = { 0 };
+ size_t buf_offsets[AV_NUM_DATA_POINTERS] = { 0 };
+ uint32_t p_w, p_h;
const int planes = av_pix_fmt_count_planes(swf->format);
- const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(swf->format);
- const int nb_images = ff_vk_count_images(hwf_vk);
- static const VkImageAspectFlags plane_aspect[] = { VK_IMAGE_ASPECT_COLOR_BIT,
- VK_IMAGE_ASPECT_PLANE_0_BIT,
- VK_IMAGE_ASPECT_PLANE_1_BIT,
- VK_IMAGE_ASPECT_PLANE_2_BIT, };
- VkImageMemoryBarrier2 img_bar[AV_NUM_DATA_POINTERS];
- int nb_img_bar = 0;
-
- AVBufferRef *bufs[AV_NUM_DATA_POINTERS];
- int nb_bufs = 0;
+ int host_mapped[AV_NUM_DATA_POINTERS] = { 0 };
+ const int map_host = !!(p->vkctx.extensions & FF_VK_EXT_EXTERNAL_HOST_MEMORY);
- VkCommandBuffer cmd_buf;
- FFVkExecContext *exec;
-
- /* Sanity checking */
if ((swf->format != AV_PIX_FMT_NONE && !av_vkfmt_from_pixfmt(swf->format))) {
av_log(hwfc, AV_LOG_ERROR, "Unsupported software frame pixel format!\n");
return AVERROR(EINVAL);
@@ -4101,97 +3934,115 @@ static int vulkan_transfer_frame(AVHWFra
if (swf->width > hwfc->width || swf->height > hwfc->height)
return AVERROR(EINVAL);
- /* Setup buffers first */
- if (p->vkctx.extensions & FF_VK_EXT_EXTERNAL_HOST_MEMORY) {
- err = host_map_frame(hwfc, bufs, &nb_bufs, swf, region, upload);
- if (err >= 0)
- host_mapped = 1;
- }
+ /* Create buffers */
+ for (int i = 0; i < planes; i++) {
+ size_t req_size;
+
+ VkExternalMemoryBufferCreateInfo create_desc = {
+ .sType = VK_STRUCTURE_TYPE_EXTERNAL_MEMORY_BUFFER_CREATE_INFO,
+ .handleTypes = VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT,
+ };
+
+ VkImportMemoryHostPointerInfoEXT import_desc = {
+ .sType = VK_STRUCTURE_TYPE_IMPORT_MEMORY_HOST_POINTER_INFO_EXT,
+ .handleType = VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT,
+ };
+
+ VkMemoryHostPointerPropertiesEXT p_props = {
+ .sType = VK_STRUCTURE_TYPE_MEMORY_HOST_POINTER_PROPERTIES_EXT,
+ };
+
+ get_plane_wh(&p_w, &p_h, swf->format, swf->width, swf->height, i);
- if (!host_mapped) {
- err = get_plane_buf(hwfc, &bufs[0], swf, region, upload);
+ tmp.linesize[i] = FFABS(swf->linesize[i]);
+
+ /* Do not map images with a negative stride */
+ if (map_host && swf->linesize[i] > 0) {
+ size_t offs;
+ offs = (uintptr_t)swf->data[i] % p->hprops.minImportedHostPointerAlignment;
+ import_desc.pHostPointer = swf->data[i] - offs;
+
+ /* We have to compensate for the few extra bytes of padding we
+ * completely ignore at the start */
+ req_size = FFALIGN(offs + tmp.linesize[i] * p_h,
+ p->hprops.minImportedHostPointerAlignment);
+
+ ret = vk->GetMemoryHostPointerPropertiesEXT(hwctx->act_dev,
+ import_desc.handleType,
+ import_desc.pHostPointer,
+ &p_props);
+ if (ret == VK_SUCCESS && p_props.memoryTypeBits) {
+ host_mapped[i] = 1;
+ buf_offsets[i] = offs;
+ }
+ }
+
+ if (!host_mapped[i])
+ req_size = get_req_buffer_size(p, &tmp.linesize[i], p_h);
+
+ err = ff_vk_create_avbuf(&p->vkctx, &bufs[i], req_size,
+ host_mapped[i] ? &create_desc : NULL,
+ host_mapped[i] ? &import_desc : NULL,
+ from ? VK_BUFFER_USAGE_TRANSFER_DST_BIT :
+ VK_BUFFER_USAGE_TRANSFER_SRC_BIT,
+ VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
+ (host_mapped[i] ?
+ VK_MEMORY_PROPERTY_HOST_COHERENT_BIT : 0x0));
if (err < 0)
goto end;
- nb_bufs = 1;
- if (upload) {
- err = copy_buffer_data(hwfc, bufs[0], swf, region, planes, 1);
- if (err < 0)
- goto end;
- }
+ vkbufs[i] = (FFVkBuffer *)bufs[i]->data;
}
- exec = ff_vk_exec_get(&fp->upload_exec);
- cmd_buf = exec->buf;
+ if (!from) {
+ /* Map, copy image TO buffer (which then goes to the VkImage), unmap */
+ if ((err = ff_vk_map_buffers(&p->vkctx, vkbufs, tmp.data, planes, 0)))
+ goto end;
- ff_vk_exec_start(&p->vkctx, exec);
+ for (int i = 0; i < planes; i++) {
+ if (host_mapped[i])
+ continue;
- /* Prep destination Vulkan frame */
- err = ff_vk_exec_add_dep_frame(&p->vkctx, exec, hwf,
- VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT,
- VK_PIPELINE_STAGE_2_TRANSFER_BIT);
- if (err < 0)
- goto end;
+ get_plane_wh(&p_w, &p_h, swf->format, swf->width, swf->height, i);
- /* No need to declare buf deps for synchronous transfers */
- if (upload) {
- err = ff_vk_exec_add_dep_buf(&p->vkctx, exec, bufs, nb_bufs, 1);
- if (err < 0) {
- ff_vk_exec_discard_deps(&p->vkctx, exec);
- goto end;
+ av_image_copy_plane(tmp.data[i], tmp.linesize[i],
+ (const uint8_t *)swf->data[i], swf->linesize[i],
+ FFMIN(tmp.linesize[i], FFABS(swf->linesize[i])),
+ p_h);
}
+
+ if ((err = ff_vk_unmap_buffers(&p->vkctx, vkbufs, planes, 1)))
+ goto end;
}
- ff_vk_frame_barrier(&p->vkctx, exec, hwf, img_bar, &nb_img_bar,
- VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT,
- VK_PIPELINE_STAGE_2_TRANSFER_BIT_KHR,
- upload ? VK_ACCESS_TRANSFER_WRITE_BIT :
- VK_ACCESS_TRANSFER_READ_BIT,
- upload ? VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL :
- VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
- VK_QUEUE_FAMILY_IGNORED);
+ /* Copy buffers into/from image */
+ err = transfer_image_buf(hwfc, (AVFrame *)vkf, bufs, buf_offsets,
+ tmp.linesize, swf->width, swf->height, swf->format,
+ from);
+
+ if (from) {
+ /* Map, copy buffer (which came FROM the VkImage) to the frame, unmap */
+ if ((err = ff_vk_map_buffers(&p->vkctx, vkbufs, tmp.data, planes, 0)))
+ goto end;
- vk->CmdPipelineBarrier2(cmd_buf, &(VkDependencyInfo) {
- .sType = VK_STRUCTURE_TYPE_DEPENDENCY_INFO,
- .pImageMemoryBarriers = img_bar,
- .imageMemoryBarrierCount = nb_img_bar,
- });
+ for (int i = 0; i < planes; i++) {
+ if (host_mapped[i])
+ continue;
- for (int i = 0; i < planes; i++) {
- int buf_idx = FFMIN(i, (nb_bufs - 1));
- int img_idx = FFMIN(i, (nb_images - 1));
- FFVkBuffer *vkbuf = (FFVkBuffer *)bufs[buf_idx]->data;
-
- uint32_t orig_stride = region[i].bufferRowLength;
- region[i].bufferRowLength /= desc->comp[i].step;
- region[i].imageSubresource.aspectMask = plane_aspect[(planes != nb_images) +
- i*(planes != nb_images)];
-
- if (upload)
- vk->CmdCopyBufferToImage(cmd_buf, vkbuf->buf,
- hwf_vk->img[img_idx],
- img_bar[img_idx].newLayout,
- 1, &region[i]);
- else
- vk->CmdCopyImageToBuffer(cmd_buf, hwf_vk->img[img_idx],
- img_bar[img_idx].newLayout,
- vkbuf->buf,
- 1, &region[i]);
+ get_plane_wh(&p_w, &p_h, swf->format, swf->width, swf->height, i);
- region[i].bufferRowLength = orig_stride;
- }
+ av_image_copy_plane_uc_from(swf->data[i], swf->linesize[i],
+ (const uint8_t *)tmp.data[i], tmp.linesize[i],
+ FFMIN(tmp.linesize[i], FFABS(swf->linesize[i])),
+ p_h);
+ }
- err = ff_vk_exec_submit(&p->vkctx, exec);
- if (err < 0) {
- ff_vk_exec_discard_deps(&p->vkctx, exec);
- } else if (!upload) {
- ff_vk_exec_wait(&p->vkctx, exec);
- if (!host_mapped)
- err = copy_buffer_data(hwfc, bufs[0], swf, region, planes, 0);
+ if ((err = ff_vk_unmap_buffers(&p->vkctx, vkbufs, planes, 1)))
+ goto end;
}
end:
- for (int i = 0; i < nb_bufs; i++)
+ for (int i = 0; i < planes; i++)
av_buffer_unref(&bufs[i]);
return err;
@@ -4218,7 +4069,7 @@ static int vulkan_transfer_data_to(AVHWF
if (src->hw_frames_ctx)
return AVERROR(ENOSYS);
else
- return vulkan_transfer_frame(hwfc, (AVFrame *)src, dst, 1);
+ return vulkan_transfer_data(hwfc, dst, src, 0);
}
}
@@ -4335,7 +4186,7 @@ static int vulkan_transfer_data_from(AVH
if (dst->hw_frames_ctx)
return AVERROR(ENOSYS);
else
- return vulkan_transfer_frame(hwfc, dst, (AVFrame *)src, 0);
+ return vulkan_transfer_data(hwfc, src, dst, 1);
}
}
Index: FFmpeg/libavutil/vulkan.c
===================================================================
--- FFmpeg.orig/libavutil/vulkan.c
+++ FFmpeg/libavutil/vulkan.c
@@ -887,8 +887,11 @@ int ff_vk_alloc_mem(FFVulkanContext *s,
ret = vk->AllocateMemory(s->hwctx->act_dev, &alloc_info,
s->hwctx->alloc, mem);
- if (ret != VK_SUCCESS)
+ if (ret != VK_SUCCESS) {
+ av_log(s, AV_LOG_ERROR, "Failed to allocate memory: %s\n",
+ ff_vk_ret2str(ret));
return AVERROR(ENOMEM);
+ }
if (mem_flags)
*mem_flags |= s->mprops.memoryTypes[index].propertyFlags;
debian/patches/0073-fix-atenc-layout-samplerate.patch → debian/patches/0062-fix-atenc-layout-samplerate.patch
Vendored
+12 -14
View File
@@ -22,19 +22,17 @@ Index: FFmpeg/libavcodec/audiotoolboxenc.c
FFAT_ENC_CLASS(NAME) \
const FFCodec ff_##NAME##_at_encoder = { \
.p.name = #NAME "_at", \
@@ -652,9 +652,10 @@ static const AVOption options[] = {
AV_CODEC_CAP_ENCODER_FLUSH CAPS, \
.p.ch_layouts = CH_LAYOUTS, \
.p.sample_fmts = (const enum AVSampleFormat[]) { \
- AV_SAMPLE_FMT_S16, \
+ AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S32,\
AV_SAMPLE_FMT_U8, AV_SAMPLE_FMT_NONE \
}, \
+ .p.supported_samplerates = SAMPLE_RATES, \
@@ -653,7 +653,8 @@ static const AVOption options[] = {
.p.profiles = PROFILES, \
.p.wrapper_name = "at", \
CODEC_CH_LAYOUTS_ARRAY(CH_LAYOUTS), \
- CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_U8), \
+ CODEC_SAMPLERATES_ARRAY(SAMPLE_RATES), \
+ CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_U8), \
};
@@ -669,15 +670,25 @@ static const AVChannelLayout aac_at_ch_l
static const AVChannelLayout aac_at_ch_layouts[] = {
@@ -666,15 +667,25 @@ static const AVChannelLayout aac_at_ch_l
AV_CHANNEL_LAYOUT_6POINT0,
AV_CHANNEL_LAYOUT_6POINT1,
AV_CHANNEL_LAYOUT_7POINT0,
@@ -46,13 +44,13 @@ Index: FFmpeg/libavcodec/audiotoolboxenc.c
};
-FFAT_ENC(aac, AV_CODEC_ID_AAC, aac_profiles, , aac_at_channel_layouts, aac_at_ch_layouts)
+static const int aac_at_sample_rates[9] = {
+ 48000, 44100, 32000,24000, 22050, 16000, 12000, 11025, 8000
+static const int aac_at_sample_rates[] = {
+ 48000, 44100, 32000,24000, 22050, 16000, 12000, 11025, 8000, 0
+};
+
+static const int alac_at_sample_rates[16] = {
+static const int alac_at_sample_rates[] = {
+ 384000, 352800, 192000, 176400, 96000, 88200, 64000, 48000,
+ 44100, 32000, 24000, 22050, 16000, 12000, 11025, 8000
+ 44100, 32000, 24000, 22050, 16000, 12000, 11025, 8000, 0
+};
+
+FFAT_ENC(aac, AV_CODEC_ID_AAC, aac_profiles, , aac_at_channel_layouts, aac_at_ch_layouts, aac_at_sample_rates)
debian/patches/0062-use-linear-contiguous-vulkan-images-for-amd-encoder.patch
Vendored
-320
View File
@@ -1,320 +0,0 @@
Index: FFmpeg/libavutil/hwcontext_vulkan.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_vulkan.c
+++ FFmpeg/libavutil/hwcontext_vulkan.c
@@ -64,6 +64,13 @@
#if HAVE_LINUX_DMA_BUF_H
#include <sys/ioctl.h>
#include <linux/dma-buf.h>
+#if 0 // ndef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+#define DMA_BUF_IOCTL_EXPORT_SYNC_FILE _IOWR(DMA_BUF_BASE, 2, struct dma_buf_export_sync_file)
+struct dma_buf_export_sync_file {
+ uint32_t flags;
+ int32_t fd;
+};
+#endif
#endif
#if CONFIG_CUDA
@@ -1400,6 +1407,7 @@ static void vulkan_device_uninit(AVHWDev
static int vulkan_device_create_internal(AVHWDeviceContext *ctx,
VulkanDeviceSelection *dev_select,
+ int use_linear_images,
int disable_multiplane,
AVDictionary *opts, int flags)
{
@@ -1599,10 +1607,14 @@ static int vulkan_device_create_internal
goto end;
}
- /* Tiled images setting, use them by default */
- opt_d = av_dict_get(opts, "linear_images", NULL, 0);
- if (opt_d)
- p->use_linear_images = strtol(opt_d->value, NULL, 10);
+ /* Tiled images setting, use them by default.
+ * The use_linear_images argument takes precedent over the option */
+ p->use_linear_images = use_linear_images;
+ if (!p->use_linear_images) {
+ opt_d = av_dict_get(opts, "linear_images", NULL, 0);
+ if (opt_d)
+ p->use_linear_images = strtol(opt_d->value, NULL, 10);
+ }
/*
* The disable_multiplane argument takes precedent over the option.
@@ -1882,7 +1894,7 @@ static int vulkan_device_create(AVHWDevi
}
}
- return vulkan_device_create_internal(ctx, &dev_select, 0, opts, flags);
+ return vulkan_device_create_internal(ctx, &dev_select, 0, 0, opts, flags);
}
static int vulkan_device_derive(AVHWDeviceContext *ctx,
@@ -1906,6 +1918,7 @@ static int vulkan_device_derive(AVHWDevi
};
#endif
const char *vendor;
+ int use_linear_images = 0;
#if VA_CHECK_VERSION(1, 15, 0)
vas = vaGetDisplayAttributes(dpy, &attr, 1);
@@ -1920,11 +1933,14 @@ static int vulkan_device_derive(AVHWDevi
return AVERROR_EXTERNAL;
}
- if (strstr(vendor, "AMD"))
+ if (strstr(vendor, "AMD")) {
dev_select.vendor_id = 0x1002;
+ use_linear_images = 1;
+ }
}
- return vulkan_device_create_internal(ctx, &dev_select, 0, opts, flags);
+ /* AMD VA-API encoders require linear contiguous (multiplane) surface */
+ return vulkan_device_create_internal(ctx, &dev_select, use_linear_images, 0, opts, flags);
}
#endif
#if CONFIG_LIBDRM
@@ -1933,6 +1949,7 @@ static int vulkan_device_derive(AVHWDevi
struct stat drm_node_info;
drmDevice *drm_dev_info;
AVDRMDeviceContext *src_hwctx = src_ctx->hwctx;
+ int use_linear_images = 0;
err = fstat(src_hwctx->fd, &drm_node_info);
if (err) {
@@ -1952,12 +1969,15 @@ static int vulkan_device_derive(AVHWDevi
return AVERROR_EXTERNAL;
}
- if (drm_dev_info->bustype == DRM_BUS_PCI)
+ if (drm_dev_info->bustype == DRM_BUS_PCI) {
dev_select.pci_device = drm_dev_info->deviceinfo.pci->device_id;
+ use_linear_images = drm_dev_info->deviceinfo.pci->vendor_id == 0x1002;
+ }
drmFreeDevice(&drm_dev_info);
- return vulkan_device_create_internal(ctx, &dev_select, 0, opts, flags);
+ /* AMD VA-API encoders require linear contiguous (multiplane) surface */
+ return vulkan_device_create_internal(ctx, &dev_select, use_linear_images, 0, opts, flags);
}
#endif
#if CONFIG_CUDA
@@ -1980,7 +2000,7 @@ static int vulkan_device_derive(AVHWDevi
* CUDA is not able to import multiplane images, so always derive a
* Vulkan device with multiplane disabled.
*/
- return vulkan_device_create_internal(ctx, &dev_select, 1, opts, flags);
+ return vulkan_device_create_internal(ctx, &dev_select, 0, 1, opts, flags);
}
#endif
default:
@@ -2876,6 +2896,7 @@ static int vulkan_map_from_drm_frame_des
const AVDRMFrameDescriptor *desc = (AVDRMFrameDescriptor *)src->data[0];
VkBindImageMemoryInfo bind_info[AV_DRM_MAX_PLANES];
VkBindImagePlaneMemoryInfo plane_info[AV_DRM_MAX_PLANES];
+ const int has_modifiers = !!(p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS);
for (int i = 0; i < desc->nb_layers; i++) {
if (drm_to_vulkan_fmt(desc->layers[i].format) == VK_FORMAT_UNDEFINED) {
@@ -2885,13 +2906,21 @@ static int vulkan_map_from_drm_frame_des
}
}
+ if (!has_modifiers &&
+ desc->objects[0].format_modifier != DRM_FORMAT_MOD_INVALID &&
+ desc->objects[0].format_modifier != DRM_FORMAT_MOD_LINEAR) {
+ av_log(ctx, AV_LOG_ERROR, "The driver can only import DRM frame with invalid/linear modifier!\n");
+ err = AVERROR_EXTERNAL;
+ goto fail;
+ }
+
if (!(f = av_vk_frame_alloc())) {
av_log(ctx, AV_LOG_ERROR, "Unable to allocate memory for AVVkFrame!\n");
err = AVERROR(ENOMEM);
goto fail;
}
- f->tiling = VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT;
+ f->tiling = has_modifiers ? VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT : VK_IMAGE_TILING_LINEAR;
for (int i = 0; i < desc->nb_layers; i++) {
const int planes = desc->layers[i].nb_planes;
@@ -2929,7 +2958,7 @@ static int vulkan_map_from_drm_frame_des
.mipLevels = 1,
.arrayLayers = 1,
.flags = 0x0,
- .tiling = VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT,
+ .tiling = f->tiling,
.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED, /* specs say so */
.usage = 0x0, /* filled in below */
.samples = VK_SAMPLE_COUNT_1_BIT,
@@ -2970,7 +2999,7 @@ static int vulkan_map_from_drm_frame_des
fmt_props = (VkPhysicalDeviceImageFormatInfo2) {
.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_IMAGE_FORMAT_INFO_2,
- .pNext = &props_ext,
+ .pNext = has_modifiers ? &props_ext : NULL,
.format = create_info.format,
.type = create_info.imageType,
.tiling = create_info.tiling,
@@ -3145,7 +3174,7 @@ static int vulkan_map_from_drm_frame_syn
const AVDRMFrameDescriptor *desc = (AVDRMFrameDescriptor *)src->data[0];
-#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+#if 0 // def DMA_BUF_IOCTL_EXPORT_SYNC_FILE
if (p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_SEM) {
VkCommandBuffer cmd_buf;
FFVkExecContext *exec;
@@ -3171,8 +3200,10 @@ static int vulkan_map_from_drm_frame_syn
if (ioctl(desc->objects[i].fd, DMA_BUF_IOCTL_EXPORT_SYNC_FILE,
&implicit_fd_info)) {
err = AVERROR(errno);
- av_log(hwctx, AV_LOG_ERROR, "Failed to retrieve implicit DRM sync file: %s\n",
+ av_log(hwctx, i ? AV_LOG_ERROR : AV_LOG_DEBUG, "Failed to retrieve implicit DRM sync file: %s\n",
av_err2str(err));
+ if (i == 0) /* Bail out from the plane0 if ioctl failed */
+ goto fallback;
for (; i >= 0; i--)
vk->DestroySemaphore(hwctx->act_dev, drm_sync_sem[i], hwctx->alloc);
return err;
@@ -3248,10 +3279,12 @@ static int vulkan_map_from_drm_frame_syn
} else
#endif
{
- AVVkFrame *f = (AVVkFrame *)dst->data[0];
- av_log(hwctx, AV_LOG_WARNING, "No support for synchronization when importing DMA-BUFs, "
- "image may be corrupted.\n");
- err = prepare_frame(hwfc, &fp->compute_exec, f, PREP_MODE_EXTERNAL_IMPORT);
+#if 0 // def DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+fallback:
+#endif
+ av_log(hwctx, AV_LOG_DEBUG, "No support for synchronization when importing DMA-BUFs, "
+ "image may be corrupted.\n");
+ err = prepare_frame(hwfc, &fp->compute_exec, (AVVkFrame *)dst->data[0], PREP_MODE_EXTERNAL_IMPORT);
if (err)
return err;
}
@@ -3594,22 +3627,14 @@ fail:
static int vulkan_map_to(AVHWFramesContext *hwfc, AVFrame *dst,
const AVFrame *src, int flags)
{
- av_unused VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
-
switch (src->format) {
#if CONFIG_LIBDRM
#if CONFIG_VAAPI
case AV_PIX_FMT_VAAPI:
- if (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS)
- return vulkan_map_from_vaapi(hwfc, dst, src, flags);
- else
- return AVERROR(ENOSYS);
+ return vulkan_map_from_vaapi(hwfc, dst, src, flags);
#endif
case AV_PIX_FMT_DRM_PRIME:
- if (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS)
- return vulkan_map_from_drm(hwfc, dst, src, flags);
- else
- return AVERROR(ENOSYS);
+ return vulkan_map_from_drm(hwfc, dst, src, flags);
#endif
default:
return AVERROR(ENOSYS);
@@ -3652,13 +3677,14 @@ static int vulkan_map_to_drm(AVHWFramesC
VulkanFramesPriv *fp = hwfc->hwctx;
AVVulkanFramesContext *hwfctx = &fp->p;
const int planes = av_pix_fmt_count_planes(hwfc->sw_format);
+ const int has_modifiers = !!(p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS);
VkImageDrmFormatModifierPropertiesEXT drm_mod = {
.sType = VK_STRUCTURE_TYPE_IMAGE_DRM_FORMAT_MODIFIER_PROPERTIES_EXT,
};
VkSemaphoreWaitInfo wait_info = {
.sType = VK_STRUCTURE_TYPE_SEMAPHORE_WAIT_INFO,
.flags = 0x0,
- .semaphoreCount = planes,
+ .semaphoreCount = p->disable_multiplane ? planes : 1,
};
AVDRMFrameDescriptor *drm_desc = av_mallocz(sizeof(*drm_desc));
@@ -3679,10 +3705,16 @@ static int vulkan_map_to_drm(AVHWFramesC
if (err < 0)
goto end;
- ret = vk->GetImageDrmFormatModifierPropertiesEXT(hwctx->act_dev, f->img[0],
- &drm_mod);
- if (ret != VK_SUCCESS) {
- av_log(hwfc, AV_LOG_ERROR, "Failed to retrieve DRM format modifier!\n");
+ if (has_modifiers) {
+ ret = vk->GetImageDrmFormatModifierPropertiesEXT(hwctx->act_dev, f->img[0],
+ &drm_mod);
+ if (ret != VK_SUCCESS) {
+ av_log(hwfc, AV_LOG_ERROR, "Failed to retrieve DRM format modifier!\n");
+ err = AVERROR_EXTERNAL;
+ goto end;
+ }
+ } else if (f->tiling != VK_IMAGE_TILING_LINEAR) {
+ av_log(hwfc, AV_LOG_ERROR, "The driver can only export linear images to DRM frame!\n");
err = AVERROR_EXTERNAL;
goto end;
}
@@ -3704,7 +3736,7 @@ static int vulkan_map_to_drm(AVHWFramesC
drm_desc->nb_objects++;
drm_desc->objects[i].size = f->size[i];
- drm_desc->objects[i].format_modifier = drm_mod.drmFormatModifier;
+ drm_desc->objects[i].format_modifier = has_modifiers ? drm_mod.drmFormatModifier : 0x0;
}
drm_desc->nb_layers = planes;
@@ -3715,6 +3747,14 @@ static int vulkan_map_to_drm(AVHWFramesC
};
VkFormat plane_vkfmt = av_vkfmt_from_pixfmt(hwfc->sw_format)[i];
+ if (drm_desc->nb_layers > 1 && !p->disable_multiplane) {
+ switch (i) {
+ case 0: sub.aspectMask = VK_IMAGE_ASPECT_PLANE_0_BIT; break;
+ case 1: sub.aspectMask = VK_IMAGE_ASPECT_PLANE_1_BIT; break;
+ case 2: sub.aspectMask = VK_IMAGE_ASPECT_PLANE_2_BIT; break;
+ }
+ }
+
drm_desc->layers[i].format = vulkan_fmt_to_drm(plane_vkfmt);
drm_desc->layers[i].nb_planes = 1;
@@ -3729,7 +3769,7 @@ static int vulkan_map_to_drm(AVHWFramesC
if (f->tiling == VK_IMAGE_TILING_OPTIMAL)
continue;
- vk->GetImageSubresourceLayout(hwctx->act_dev, f->img[i], &sub, &layout);
+ vk->GetImageSubresourceLayout(hwctx->act_dev, f->img[p->disable_multiplane ? i : 0], &sub, &layout);
drm_desc->layers[i].planes[0].offset = layout.offset;
drm_desc->layers[i].planes[0].pitch = layout.rowPitch;
@@ -3781,21 +3821,13 @@ fail:
static int vulkan_map_from(AVHWFramesContext *hwfc, AVFrame *dst,
const AVFrame *src, int flags)
{
- av_unused VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
-
switch (dst->format) {
#if CONFIG_LIBDRM
case AV_PIX_FMT_DRM_PRIME:
- if (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS)
- return vulkan_map_to_drm(hwfc, dst, src, flags);
- else
- return AVERROR(ENOSYS);
+ return vulkan_map_to_drm(hwfc, dst, src, flags);
#if CONFIG_VAAPI
case AV_PIX_FMT_VAAPI:
- if (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS)
- return vulkan_map_to_vaapi(hwfc, dst, src, flags);
- else
- return AVERROR(ENOSYS);
+ return vulkan_map_to_vaapi(hwfc, dst, src, flags);
#endif
#endif
default:
debian/patches/0074-videotoolbox-remove-opengl-compatability.patch → debian/patches/0063-videotoolbox-remove-opengl-compatability.patch
Vendored
View File
debian/patches/0064-add-pgs-support-to-vulkan-overlay.patch
Vendored
-365
View File
@@ -1,365 +0,0 @@
Index: FFmpeg/libavfilter/vf_overlay_vulkan.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_overlay_vulkan.c
+++ FFmpeg/libavfilter/vf_overlay_vulkan.c
@@ -33,9 +33,11 @@ typedef struct OverlayVulkanContext {
int initialized;
FFVulkanPipeline pl;
+ FFVulkanPipeline pl_pass;
FFVkExecPool e;
FFVkQueueFamilyCtx qf;
FFVkSPIRVShader shd;
+ FFVkSPIRVShader shd_pass;
VkSampler sampler;
/* Push constants / options */
@@ -48,6 +50,10 @@ typedef struct OverlayVulkanContext {
int overlay_y;
int overlay_w;
int overlay_h;
+
+ int opt_repeatlast;
+ int opt_shortest;
+ int opt_eof_action;
} OverlayVulkanContext;
static const char overlay_noalpha[] = {
@@ -84,15 +90,16 @@ static const char overlay_alpha[] = {
static av_cold int init_filter(AVFilterContext *ctx)
{
int err;
- uint8_t *spv_data;
+ uint8_t *spv_data, *spv_data_pass;
size_t spv_len;
void *spv_opaque = NULL;
+ void *spv_opaque_pass = NULL;
OverlayVulkanContext *s = ctx->priv;
FFVulkanContext *vkctx = &s->vkctx;
const int planes = av_pix_fmt_count_planes(s->vkctx.output_format);
const int ialpha = av_pix_fmt_desc_get(s->vkctx.input_format)->flags & AV_PIX_FMT_FLAG_ALPHA;
const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(s->vkctx.output_format);
- FFVkSPIRVShader *shd = &s->shd;
+ FFVkSPIRVShader *shd;
FFVkSPIRVCompiler *spv;
FFVulkanDescriptorSetBinding *desc;
@@ -105,90 +112,143 @@ static av_cold int init_filter(AVFilterC
ff_vk_qf_init(vkctx, &s->qf, VK_QUEUE_COMPUTE_BIT);
RET(ff_vk_exec_pool_init(vkctx, &s->qf, &s->e, s->qf.nb_queues*4, 0, 0, 0, NULL));
RET(ff_vk_init_sampler(vkctx, &s->sampler, 1, VK_FILTER_NEAREST));
+
+ /* overlay_compute */
RET(ff_vk_shader_init(&s->pl, &s->shd, "overlay_compute",
VK_SHADER_STAGE_COMPUTE_BIT, 0));
+ {
+ shd = &s->shd;
+ ff_vk_shader_set_compute_sizes(shd, 32, 32, 1);
+
+ GLSLC(0, layout(push_constant, std430) uniform pushConstants { );
+ GLSLC(1, ivec2 o_offset[3]; );
+ GLSLC(1, ivec2 o_size[3]; );
+ GLSLC(0, }; );
+ GLSLC(0, );
+
+ ff_vk_add_push_constant(&s->pl, 0, sizeof(s->opts),
+ VK_SHADER_STAGE_COMPUTE_BIT);
+
+ desc = (FFVulkanDescriptorSetBinding []) {
+ {
+ .name = "main_img",
+ .type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .dimensions = 2,
+ .elems = planes,
+ .stages = VK_SHADER_STAGE_COMPUTE_BIT,
+ .samplers = DUP_SAMPLER(s->sampler),
+ },
+ {
+ .name = "overlay_img",
+ .type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .dimensions = 2,
+ .elems = planes,
+ .stages = VK_SHADER_STAGE_COMPUTE_BIT,
+ .samplers = DUP_SAMPLER(s->sampler),
+ },
+ {
+ .name = "output_img",
+ .type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
+ .mem_layout = ff_vk_shader_rep_fmt(s->vkctx.output_format),
+ .mem_quali = "writeonly",
+ .dimensions = 2,
+ .elems = planes,
+ .stages = VK_SHADER_STAGE_COMPUTE_BIT,
+ },
+ };
+
+ RET(ff_vk_pipeline_descriptor_set_add(vkctx, &s->pl, shd, desc, 3, 0, 0));
+
+ GLSLD( overlay_noalpha );
+ GLSLD( overlay_alpha );
+ GLSLC(0, void main() );
+ GLSLC(0, { );
+ GLSLC(1, ivec2 pos = ivec2(gl_GlobalInvocationID.xy); );
+ GLSLF(1, int planes = %i; ,planes);
+ GLSLC(1, for (int i = 0; i < planes; i++) { );
+ if (ialpha)
+ GLSLC(2, overlay_alpha_opaque(i, pos); );
+ else
+ GLSLC(2, overlay_noalpha(i, pos); );
+ GLSLC(1, } );
+ GLSLC(0, } );
+
+ RET(spv->compile_shader(spv, ctx, shd, &spv_data, &spv_len, "main",
+ &spv_opaque));
+ RET(ff_vk_shader_create(vkctx, shd, spv_data, spv_len, "main"));
+
+ RET(ff_vk_init_compute_pipeline(vkctx, &s->pl, shd));
+ RET(ff_vk_exec_pipeline_register(vkctx, &s->e, &s->pl));
+
+ s->opts.o_offset[0] = s->overlay_x;
+ s->opts.o_offset[1] = s->overlay_y;
+ s->opts.o_offset[2] = s->opts.o_offset[0] >> pix_desc->log2_chroma_w;
+ s->opts.o_offset[3] = s->opts.o_offset[1] >> pix_desc->log2_chroma_h;
+ s->opts.o_offset[4] = s->opts.o_offset[0] >> pix_desc->log2_chroma_w;
+ s->opts.o_offset[5] = s->opts.o_offset[1] >> pix_desc->log2_chroma_h;
+
+ s->opts.o_size[0] = s->overlay_w;
+ s->opts.o_size[1] = s->overlay_h;
+ s->opts.o_size[2] = s->opts.o_size[0] >> pix_desc->log2_chroma_w;
+ s->opts.o_size[3] = s->opts.o_size[1] >> pix_desc->log2_chroma_h;
+ s->opts.o_size[4] = s->opts.o_size[0] >> pix_desc->log2_chroma_w;
+ s->opts.o_size[5] = s->opts.o_size[1] >> pix_desc->log2_chroma_h;
+ }
- ff_vk_shader_set_compute_sizes(&s->shd, 32, 32, 1);
+ /* overlay_compute_pass */
+ RET(ff_vk_shader_init(&s->pl_pass, &s->shd_pass, "overlay_compute_pass",
+ VK_SHADER_STAGE_COMPUTE_BIT, 0));
+ {
+ shd = &s->shd_pass;
+ ff_vk_shader_set_compute_sizes(shd, 32, 32, 1);
+
+ desc = (FFVulkanDescriptorSetBinding []) {
+ {
+ .name = "main_img",
+ .type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .dimensions = 2,
+ .elems = planes,
+ .stages = VK_SHADER_STAGE_COMPUTE_BIT,
+ .samplers = DUP_SAMPLER(s->sampler),
+ },
+ {
+ .name = "output_img",
+ .type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
+ .mem_layout = ff_vk_shader_rep_fmt(s->vkctx.output_format),
+ .mem_quali = "writeonly",
+ .dimensions = 2,
+ .elems = planes,
+ .stages = VK_SHADER_STAGE_COMPUTE_BIT,
+ },
+ };
+
+ RET(ff_vk_pipeline_descriptor_set_add(vkctx, &s->pl_pass, shd, desc, 2, 0, 0));
+
+ GLSLC(0, void main() );
+ GLSLC(0, { );
+ GLSLC(1, ivec2 pos = ivec2(gl_GlobalInvocationID.xy); );
+ GLSLF(1, int planes = %i; ,planes);
+ GLSLC(1, for (int i = 0; i < planes; i++) { );
+ GLSLC(2, vec4 res = texture(main_img[i], pos); );
+ GLSLC(2, imageStore(output_img[i], pos, res); );
+ GLSLC(1, } );
+ GLSLC(0, } );
+
+ RET(spv->compile_shader(spv, ctx, shd, &spv_data_pass, &spv_len, "main",
+ &spv_opaque));
+ RET(ff_vk_shader_create(vkctx, shd, spv_data_pass, spv_len, "main"));
- GLSLC(0, layout(push_constant, std430) uniform pushConstants { );
- GLSLC(1, ivec2 o_offset[3]; );
- GLSLC(1, ivec2 o_size[3]; );
- GLSLC(0, }; );
- GLSLC(0, );
-
- ff_vk_add_push_constant(&s->pl, 0, sizeof(s->opts),
- VK_SHADER_STAGE_COMPUTE_BIT);
-
- desc = (FFVulkanDescriptorSetBinding []) {
- {
- .name = "main_img",
- .type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
- .dimensions = 2,
- .elems = planes,
- .stages = VK_SHADER_STAGE_COMPUTE_BIT,
- .samplers = DUP_SAMPLER(s->sampler),
- },
- {
- .name = "overlay_img",
- .type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
- .dimensions = 2,
- .elems = planes,
- .stages = VK_SHADER_STAGE_COMPUTE_BIT,
- .samplers = DUP_SAMPLER(s->sampler),
- },
- {
- .name = "output_img",
- .type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
- .mem_layout = ff_vk_shader_rep_fmt(s->vkctx.output_format),
- .mem_quali = "writeonly",
- .dimensions = 2,
- .elems = planes,
- .stages = VK_SHADER_STAGE_COMPUTE_BIT,
- },
- };
-
- RET(ff_vk_pipeline_descriptor_set_add(vkctx, &s->pl, shd, desc, 3, 0, 0));
-
- GLSLD( overlay_noalpha );
- GLSLD( overlay_alpha );
- GLSLC(0, void main() );
- GLSLC(0, { );
- GLSLC(1, ivec2 pos = ivec2(gl_GlobalInvocationID.xy); );
- GLSLF(1, int planes = %i; ,planes);
- GLSLC(1, for (int i = 0; i < planes; i++) { );
- if (ialpha)
- GLSLC(2, overlay_alpha_opaque(i, pos); );
- else
- GLSLC(2, overlay_noalpha(i, pos); );
- GLSLC(1, } );
- GLSLC(0, } );
-
- RET(spv->compile_shader(spv, ctx, shd, &spv_data, &spv_len, "main",
- &spv_opaque));
- RET(ff_vk_shader_create(vkctx, shd, spv_data, spv_len, "main"));
-
- RET(ff_vk_init_compute_pipeline(vkctx, &s->pl, shd));
- RET(ff_vk_exec_pipeline_register(vkctx, &s->e, &s->pl));
-
- s->opts.o_offset[0] = s->overlay_x;
- s->opts.o_offset[1] = s->overlay_y;
- s->opts.o_offset[2] = s->opts.o_offset[0] >> pix_desc->log2_chroma_w;
- s->opts.o_offset[3] = s->opts.o_offset[1] >> pix_desc->log2_chroma_h;
- s->opts.o_offset[4] = s->opts.o_offset[0] >> pix_desc->log2_chroma_w;
- s->opts.o_offset[5] = s->opts.o_offset[1] >> pix_desc->log2_chroma_h;
-
- s->opts.o_size[0] = s->overlay_w;
- s->opts.o_size[1] = s->overlay_h;
- s->opts.o_size[2] = s->opts.o_size[0] >> pix_desc->log2_chroma_w;
- s->opts.o_size[3] = s->opts.o_size[1] >> pix_desc->log2_chroma_h;
- s->opts.o_size[4] = s->opts.o_size[0] >> pix_desc->log2_chroma_w;
- s->opts.o_size[5] = s->opts.o_size[1] >> pix_desc->log2_chroma_h;
+ RET(ff_vk_init_compute_pipeline(vkctx, &s->pl_pass, shd));
+ RET(ff_vk_exec_pipeline_register(vkctx, &s->e, &s->pl_pass));
+ }
s->initialized = 1;
fail:
if (spv_opaque)
spv->free_shader(spv, &spv_opaque);
+ if (spv_opaque_pass)
+ spv->free_shader(spv, &spv_opaque_pass);
if (spv)
spv->uninit(&spv);
@@ -210,22 +270,11 @@ static int overlay_vulkan_blend(FFFrameS
if (err < 0)
goto fail;
- if (!input_main || !input_overlay)
- return 0;
-
- if (!s->initialized) {
- AVHWFramesContext *main_fc = (AVHWFramesContext*)input_main->hw_frames_ctx->data;
- AVHWFramesContext *overlay_fc = (AVHWFramesContext*)input_overlay->hw_frames_ctx->data;
- if (main_fc->sw_format != overlay_fc->sw_format) {
- av_log(ctx, AV_LOG_ERROR, "Mismatching sw formats!\n");
- return AVERROR(EINVAL);
- }
-
- s->overlay_w = input_overlay->width;
- s->overlay_h = input_overlay->height;
+ if (!input_main)
+ return AVERROR_BUG;
+ if (!s->initialized)
RET(init_filter(ctx));
- }
out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
if (!out) {
@@ -233,9 +282,13 @@ static int overlay_vulkan_blend(FFFrameS
goto fail;
}
- RET(ff_vk_filter_process_Nin(&s->vkctx, &s->e, &s->pl,
- out, (AVFrame *[]){ input_main, input_overlay }, 2,
- s->sampler, &s->opts, sizeof(s->opts)));
+ if (input_overlay)
+ RET(ff_vk_filter_process_Nin(&s->vkctx, &s->e, &s->pl,
+ out, (AVFrame *[]){ input_main, input_overlay }, 2,
+ s->sampler, &s->opts, sizeof(s->opts)));
+ else /* passthrough */
+ RET(ff_vk_filter_process_simple(&s->vkctx, &s->e, &s->pl_pass, out, input_main,
+ s->sampler, NULL, 0));
err = av_frame_copy_props(out, input_main);
if (err < 0)
@@ -253,6 +306,20 @@ static int overlay_vulkan_config_output(
int err;
AVFilterContext *avctx = outlink->src;
OverlayVulkanContext *s = avctx->priv;
+ AVFilterLink *inlink = avctx->inputs[0];
+ AVFilterLink *inlink_overlay = avctx->inputs[1];
+ FilterLink *inl = ff_filter_link(inlink);
+ FilterLink *inl_overlay = ff_filter_link(inlink_overlay);
+ AVHWFramesContext *main_fc = (AVHWFramesContext*)inl->hw_frames_ctx->data;
+ AVHWFramesContext *overlay_fc = (AVHWFramesContext*)inl_overlay->hw_frames_ctx->data;
+
+ if (main_fc->sw_format != overlay_fc->sw_format) {
+ av_log(avctx, AV_LOG_ERROR, "Mismatching sw formats!\n");
+ return AVERROR(EINVAL);
+ }
+
+ s->overlay_w = inlink_overlay->w;
+ s->overlay_h = inlink_overlay->h;
err = ff_vk_filter_config_output(outlink);
if (err < 0)
@@ -262,6 +329,11 @@ static int overlay_vulkan_config_output(
if (err < 0)
return err;
+ s->fs.opt_repeatlast = s->opt_repeatlast;
+ s->fs.opt_shortest = s->opt_shortest;
+ s->fs.opt_eof_action = s->opt_eof_action;
+ s->fs.time_base = outlink->time_base = inlink->time_base;
+
return ff_framesync_configure(&s->fs);
}
@@ -289,7 +361,9 @@ static void overlay_vulkan_uninit(AVFilt
ff_vk_exec_pool_free(vkctx, &s->e);
ff_vk_pipeline_free(vkctx, &s->pl);
+ ff_vk_pipeline_free(vkctx, &s->pl_pass);
ff_vk_shader_free(vkctx, &s->shd);
+ ff_vk_shader_free(vkctx, &s->shd_pass);
if (s->sampler)
vk->DestroySampler(vkctx->hwctx->act_dev, s->sampler,
@@ -306,6 +380,14 @@ static void overlay_vulkan_uninit(AVFilt
static const AVOption overlay_vulkan_options[] = {
{ "x", "Set horizontal offset", OFFSET(overlay_x), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, .flags = FLAGS },
{ "y", "Set vertical offset", OFFSET(overlay_y), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, .flags = FLAGS },
+ { "eof_action", "Action to take when encountering EOF from secondary input ",
+ OFFSET(opt_eof_action), AV_OPT_TYPE_INT, { .i64 = EOF_ACTION_REPEAT },
+ EOF_ACTION_REPEAT, EOF_ACTION_PASS, .flags = FLAGS, .unit = "eof_action" },
+ { "repeat", "Repeat the previous frame.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_REPEAT }, .flags = FLAGS, .unit = "eof_action" },
+ { "endall", "End both streams.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_ENDALL }, .flags = FLAGS, .unit = "eof_action" },
+ { "pass", "Pass through the main input.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_PASS }, .flags = FLAGS, .unit = "eof_action" },
+ { "shortest", "force termination when the shortest input terminates", OFFSET(opt_shortest), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, FLAGS },
+ { "repeatlast", "repeat overlay of the last overlay frame", OFFSET(opt_repeatlast), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS },
{ NULL },
};
debian/patches/0075-use-dynamic-pool-for-vpl-qsv-hwupload.patch → debian/patches/0064-use-dynamic-pool-for-vpl-qsv-hwupload.patch
Vendored
View File
debian/patches/0077-fix-ass-incorrect-null-copy.patch → debian/patches/0065-fix-ass-incorrect-null-copy.patch
Vendored
View File
debian/patches/0065-fix-yuv420p-to-p01x-unscaled-conversion.patch
Vendored
-217
View File
@@ -1,217 +0,0 @@
Index: FFmpeg/libswscale/swscale_unscaled.c
===================================================================
--- FFmpeg.orig/libswscale/swscale_unscaled.c
+++ FFmpeg/libswscale/swscale_unscaled.c
@@ -396,7 +396,7 @@ static int planar8ToP01xleWrapper(SwsCon
const uint8_t *tsrc0 = src[0];
for (x = c->srcW; x > 0; x--) {
t = *tsrc0++;
- output_pixel(tdstY++, t | (t << 8));
+ output_pixel(tdstY++, (t << 8));
}
src[0] += srcStride[0];
dstY += dstStride[0] / 2;
@@ -407,9 +407,9 @@ static int planar8ToP01xleWrapper(SwsCon
const uint8_t *tsrc2 = src[2];
for (x = c->srcW / 2; x > 0; x--) {
t = *tsrc1++;
- output_pixel(tdstUV++, t | (t << 8));
+ output_pixel(tdstUV++, (t << 8));
t = *tsrc2++;
- output_pixel(tdstUV++, t | (t << 8));
+ output_pixel(tdstUV++, (t << 8));
}
src[1] += srcStride[1];
src[2] += srcStride[2];
Index: FFmpeg/tests/ref/fate/filter-pixdesc-p010le
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixdesc-p010le
+++ FFmpeg/tests/ref/fate/filter-pixdesc-p010le
@@ -1 +1 @@
-pixdesc-p010le 7b4a503997eb4e14cba80ee52db85e39
+pixdesc-p010le 0268fd44f63022e21ada69704534fc85
Index: FFmpeg/tests/ref/fate/filter-pixdesc-p016le
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixdesc-p016le
+++ FFmpeg/tests/ref/fate/filter-pixdesc-p016le
@@ -1 +1 @@
-pixdesc-p016le ed04897de0a6788bb3458e7365f10d36
+pixdesc-p016le 0268fd44f63022e21ada69704534fc85
Index: FFmpeg/tests/ref/fate/filter-pixfmts-copy
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-copy
+++ FFmpeg/tests/ref/fate/filter-pixfmts-copy
@@ -63,11 +63,11 @@ nv21 335d85c9af6110f26ae9
nv24 f30fc8d0ac40af69e119ea919a314572
nv42 29a212f70f8780fe0eb99abcae81894d
p010be 7f9842d6015026136bad60d03c035cc3
-p010le c453421b9f726bdaf2bacf59a492c43b
+p010le 1929db89609c4b8c6d9c9030a9e7843d
p012be 7f9842d6015026136bad60d03c035cc3
p012le 1929db89609c4b8c6d9c9030a9e7843d
p016be 7f9842d6015026136bad60d03c035cc3
-p016le c453421b9f726bdaf2bacf59a492c43b
+p016le 1929db89609c4b8c6d9c9030a9e7843d
p210be 847e9c6e292b17349e69570829252b3e
p210le c06e4b76cf504e908128081f92b60ce2
p212be 4df641ed058718ad27a01889f923b04f
Index: FFmpeg/tests/ref/fate/filter-pixfmts-crop
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-crop
+++ FFmpeg/tests/ref/fate/filter-pixfmts-crop
@@ -61,11 +61,11 @@ nv21 1bcfc197f4fb95de85ba
nv24 514c8f12082f0737e558778cbe7de258
nv42 ece9baae1c5de579dac2c66a89e08ef3
p010be 8b2de2eb6b099bbf355bfc55a0694ddc
-p010le 373b50c766dfd0a8e79c9a73246d803a
+p010le a1e4f713e145dfc465bfe0cc77096a03
p012be 8b2de2eb6b099bbf355bfc55a0694ddc
p012le a1e4f713e145dfc465bfe0cc77096a03
p016be 8b2de2eb6b099bbf355bfc55a0694ddc
-p016le 373b50c766dfd0a8e79c9a73246d803a
+p016le a1e4f713e145dfc465bfe0cc77096a03
p210be 2947f43774352ef61f9e83777548c7c5
p210le 74fcd5a32eee687eebe002c884103963
p212be c983aa869bae2c70e7b01810902ffc05
Index: FFmpeg/tests/ref/fate/filter-pixfmts-field
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-field
+++ FFmpeg/tests/ref/fate/filter-pixfmts-field
@@ -63,11 +63,11 @@ nv21 7294574037cc7f9373ef
nv24 3b100fb527b64ee2b2d7120da573faf5
nv42 1841ce853152d86b27c130f319ea0db2
p010be a0311a09bba7383553267d2b3b9c075e
-p010le ee09a18aefa3ebe97715b3a7312cb8ff
+p010le f1cc90d292046109a626db2da9f0f9b6
p012be a0311a09bba7383553267d2b3b9c075e
p012le f1cc90d292046109a626db2da9f0f9b6
p016be a0311a09bba7383553267d2b3b9c075e
-p016le ee09a18aefa3ebe97715b3a7312cb8ff
+p016le f1cc90d292046109a626db2da9f0f9b6
p210be 58d46f566ab28e3bcfb715c7aa53cf58
p210le 8d68f7655a3d76f2f8436bd25beb3973
p212be a8901966c5bc111e9e62d3989b0b666b
Index: FFmpeg/tests/ref/fate/filter-pixfmts-hflip
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-hflip
+++ FFmpeg/tests/ref/fate/filter-pixfmts-hflip
@@ -61,11 +61,11 @@ nv21 9f10dfff8963dc327d33
nv24 f0c5b2f42970f8d4003621d8857a872f
nv42 4dcf9aec82b110712b396a8b365dcb13
p010be 744b13e44d39e1ff7588983fa03e0101
-p010le a50b160346ab94f55a425065b57006f0
+p010le aeb31f50c66f376b0530c7bb6287212b
p012be 744b13e44d39e1ff7588983fa03e0101
p012le aeb31f50c66f376b0530c7bb6287212b
p016be 744b13e44d39e1ff7588983fa03e0101
-p016le a50b160346ab94f55a425065b57006f0
+p016le aeb31f50c66f376b0530c7bb6287212b
p210be 6f5a76d6467b86d55fe5589d3af8a7ea
p210le b6982912b2376371edea4fccf99fe40c
p212be 9ffa4664543233ec7c9b99a627cb7003
Index: FFmpeg/tests/ref/fate/filter-pixfmts-il
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-il
+++ FFmpeg/tests/ref/fate/filter-pixfmts-il
@@ -63,11 +63,11 @@ nv21 ab586d8781246b5a32d8
nv24 554153c71d142e3fd8e40b7dcaaec229
nv42 d699724c8deaeb4f87faf2766512eec3
p010be 3df51286ef66b53e3e283dbbab582263
-p010le eadcd8241e97e35b2b47d5eb2eaea6cd
+p010le 38945445b360fa737e9e37257393e823
p012be 3df51286ef66b53e3e283dbbab582263
p012le 38945445b360fa737e9e37257393e823
p016be 3df51286ef66b53e3e283dbbab582263
-p016le eadcd8241e97e35b2b47d5eb2eaea6cd
+p016le 38945445b360fa737e9e37257393e823
p210be 29ec4e8912d456cd15203a96487c42e8
p210le c695064fb9f2cc4e35957d4d649cc281
p212be ee6f88801823da3d617fb9e073e88068
Index: FFmpeg/tests/ref/fate/filter-pixfmts-null
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-null
+++ FFmpeg/tests/ref/fate/filter-pixfmts-null
@@ -63,11 +63,11 @@ nv21 335d85c9af6110f26ae9
nv24 f30fc8d0ac40af69e119ea919a314572
nv42 29a212f70f8780fe0eb99abcae81894d
p010be 7f9842d6015026136bad60d03c035cc3
-p010le c453421b9f726bdaf2bacf59a492c43b
+p010le 1929db89609c4b8c6d9c9030a9e7843d
p012be 7f9842d6015026136bad60d03c035cc3
p012le 1929db89609c4b8c6d9c9030a9e7843d
p016be 7f9842d6015026136bad60d03c035cc3
-p016le c453421b9f726bdaf2bacf59a492c43b
+p016le 1929db89609c4b8c6d9c9030a9e7843d
p210be 847e9c6e292b17349e69570829252b3e
p210le c06e4b76cf504e908128081f92b60ce2
p212be 4df641ed058718ad27a01889f923b04f
Index: FFmpeg/tests/ref/fate/filter-pixfmts-pad
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-pad
+++ FFmpeg/tests/ref/fate/filter-pixfmts-pad
@@ -28,9 +28,9 @@ nv16 d3a50501d2ea8535489f
nv21 0fdeb2cdd56cf5a7147dc273456fa217
nv24 193b9eadcc06ad5081609f76249b3e47
nv42 1738ad3c31c6c16e17679f5b09ce4677
-p010le fbbc23cc1d764a5e6fb71883d985f3ed
+p010le 3a92c1bd3e9de050bf6abcc3fd911ab7
p012le 3a92c1bd3e9de050bf6abcc3fd911ab7
-p016le fbbc23cc1d764a5e6fb71883d985f3ed
+p016le 3a92c1bd3e9de050bf6abcc3fd911ab7
p210le 680912c059de39c3401cac856bd1b0c1
p212le a2f88017bcce2383ba60bc4872e639ba
p216le 8718662e226a4581561e7bb532af2d83
Index: FFmpeg/tests/ref/fate/filter-pixfmts-scale
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-scale
+++ FFmpeg/tests/ref/fate/filter-pixfmts-scale
@@ -63,11 +63,11 @@ nv21 c74bb1c10dbbdee8a1f6
nv24 2aa6e805bf6d4179ed8d7dea37d75db3
nv42 80714d1eb2d8bcaeab3abc3124df1abd
p010be 1d6726d94bf1385996a9a9840dd0e878
-p010le 4b316f2b9e18972299beb73511278fa8
+p010le 5d436e6b35292a0e356d81f37f989b66
p012be e4dc7ccd654c2d74fde9c7b2711d960b
p012le cd4b6bdcd8967fc0e869ce3b8a014133
p016be 31e204018cbb53f8988c4e1174ea8ce9
-p016le d5afe557f492a09317e525d7cb782f5b
+p016le 6832661b5fe5f9a7a882f482a881b679
p210be 2cc6dfcf5e006c8ed5238988a06fd45e
p210le 04efb8f14a9d98417af40954a06aa187
p212be 611c6e267e7a694ce89467779e44060b
Index: FFmpeg/tests/ref/fate/filter-pixfmts-transpose
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-transpose
+++ FFmpeg/tests/ref/fate/filter-pixfmts-transpose
@@ -60,11 +60,11 @@ nv21 292adaf5271c5c8516b7
nv24 ea9de8b47faed722ee40182f89489beb
nv42 636af6cd6a4f3ac5edc0fc3ce3c56d63
p010be ad0de2cc9bff81688b182a870fcf7000
-p010le e7ff5143595021246733ce6bd0a769e8
+p010le 024ef1cf56a4872f202b96a6a4bbf10a
p012be ad0de2cc9bff81688b182a870fcf7000
p012le 024ef1cf56a4872f202b96a6a4bbf10a
p016be ad0de2cc9bff81688b182a870fcf7000
-p016le e7ff5143595021246733ce6bd0a769e8
+p016le 024ef1cf56a4872f202b96a6a4bbf10a
p410be 8b3e0ccb31b6a20ff00a29253fb2dec3
p410le 4e5f78dfccda9a6387e81354a56a033a
p412be 88e4578d2c6d99399a6cf1db9e4c0553
Index: FFmpeg/tests/ref/fate/filter-pixfmts-vflip
===================================================================
--- FFmpeg.orig/tests/ref/fate/filter-pixfmts-vflip
+++ FFmpeg/tests/ref/fate/filter-pixfmts-vflip
@@ -63,11 +63,11 @@ nv21 2909feacd27bebb080c8
nv24 334420b9d3df84499d2ca16bb66eed2b
nv42 ba4063e2795c17fea3c8a646b01fd1f5
p010be 06e9354b6e0e38ba41736352cedc0bd5
-p010le fd18d322bffbf5816902c13102872e22
+p010le cdf6a3c38d9d4e3f079fa369e1dda662
p012be 06e9354b6e0e38ba41736352cedc0bd5
p012le cdf6a3c38d9d4e3f079fa369e1dda662
p016be 06e9354b6e0e38ba41736352cedc0bd5
-p016le fd18d322bffbf5816902c13102872e22
+p016le cdf6a3c38d9d4e3f079fa369e1dda662
p210be ca886ab2b3ea5c153f1954b3709f7249
p210le d71c2d4e483030ffd87fa6a68c83fce0
p212be 1734e5840d4e75defe7a28683c3f8856
debian/patches/0079-add-first-vframe-only-to-ffprobe.patch → debian/patches/0066-add-first-vframe-only-to-ffprobe.patch
Vendored
+16 -16
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/fftools/ffprobe.c
===================================================================
--- FFmpeg.orig/fftools/ffprobe.c
+++ FFmpeg/fftools/ffprobe.c
@@ -147,6 +147,12 @@ static int show_private_data
@@ -141,6 +141,12 @@ static const char *video_codec_name = NU
#define SHOW_OPTIONAL_FIELDS_ALWAYS 1
static int show_optional_fields = SHOW_OPTIONAL_FIELDS_AUTO;
@@ -15,7 +15,7 @@ Index: FFmpeg/fftools/ffprobe.c
static char *output_format;
static char *stream_specifier;
static char *show_data_hash;
@@ -3108,9 +3114,13 @@ static int read_interval_packets(WriterC
@@ -1596,9 +1602,13 @@ static int read_interval_packets(AVTextF
AVFormatContext *fmt_ctx = ifile->fmt_ctx;
AVPacket *pkt = NULL;
AVFrame *frame = NULL;
@@ -30,7 +30,7 @@ Index: FFmpeg/fftools/ffprobe.c
av_log(NULL, AV_LOG_VERBOSE, "Processing read interval ");
log_read_interval(interval, NULL, AV_LOG_VERBOSE);
@@ -3149,6 +3159,35 @@ static int read_interval_packets(WriterC
@@ -1637,6 +1647,35 @@ static int read_interval_packets(AVTextF
ret = AVERROR(ENOMEM);
goto end;
}
@@ -66,8 +66,8 @@ Index: FFmpeg/fftools/ffprobe.c
while (!av_read_frame(fmt_ctx, pkt)) {
if (fmt_ctx->nb_streams > nb_streams) {
REALLOCZ_ARRAY_STREAM(nb_streams_frames, nb_streams, fmt_ctx->nb_streams);
@@ -3156,6 +3195,14 @@ static int read_interval_packets(WriterC
REALLOCZ_ARRAY_STREAM(selected_streams, nb_streams, fmt_ctx->nb_streams);
@@ -1646,6 +1685,14 @@ static int read_interval_packets(AVTextF
REALLOCZ_ARRAY_STREAM(streams_with_film_grain, nb_streams, fmt_ctx->nb_streams);
nb_streams = fmt_ctx->nb_streams;
}
+ if (only_show_first_video_frame && nb_finished_video_streams < nb_video_streams &&
@@ -81,7 +81,7 @@ Index: FFmpeg/fftools/ffprobe.c
if (selected_streams[pkt->stream_index]) {
AVRational tb = ifile->streams[pkt->stream_index].st->time_base;
int64_t pts = pkt->pts != AV_NOPTS_VALUE ? pkt->pts : pkt->dts;
@@ -3181,6 +3228,28 @@ static int read_interval_packets(WriterC
@@ -1671,6 +1718,28 @@ static int read_interval_packets(AVTextF
}
frame_count++;
@@ -109,8 +109,8 @@ Index: FFmpeg/fftools/ffprobe.c
+
if (do_read_packets) {
if (do_show_packets)
show_packet(w, ifile, pkt, i++);
@@ -3188,6 +3257,8 @@ static int read_interval_packets(WriterC
show_packet(tfc, ifile, pkt, i++);
@@ -1678,6 +1747,8 @@ static int read_interval_packets(AVTextF
}
if (do_read_frames) {
int packet_new = 1;
@@ -119,13 +119,13 @@ Index: FFmpeg/fftools/ffprobe.c
FrameData *fd;
pkt->opaque_ref = av_buffer_allocz(sizeof(*fd));
@@ -3199,25 +3270,51 @@ static int read_interval_packets(WriterC
@@ -1689,25 +1760,51 @@ static int read_interval_packets(AVTextF
fd->pkt_pos = pkt->pos;
fd->pkt_size = pkt->size;
- while (process_frame(w, ifile, frame, pkt, &packet_new) > 0);
- while (process_frame(tfc, ifile, frame, pkt, &packet_new) > 0);
+ while (1) {
+ int frame_ret = process_frame(w, ifile, frame, pkt, &packet_new);
+ int frame_ret = process_frame(tfc, ifile, frame, pkt, &packet_new);
+
+ if (frame_ret <= 0)
+ break;
@@ -155,7 +155,7 @@ Index: FFmpeg/fftools/ffprobe.c
- for (i = 0; i < ifile->nb_streams; i++) {
- pkt->stream_index = i;
- if (do_read_frames) {
- while (process_frame(w, ifile, frame, pkt, &(int){1}) > 0);
- while (process_frame(tfc, ifile, frame, pkt, &(int){1}) > 0);
- if (ifile->streams[i].dec_ctx)
- avcodec_flush_buffers(ifile->streams[i].dec_ctx);
+ if (!only_show_first_video_frame) {
@@ -163,7 +163,7 @@ Index: FFmpeg/fftools/ffprobe.c
+ for (i = 0; i < ifile->nb_streams; i++) {
+ pkt->stream_index = i;
+ if (do_read_frames) {
+ while (process_frame(w, ifile, frame, pkt, &(int){1}) > 0);
+ while (process_frame(tfc, ifile, frame, pkt, &(int){1}) > 0);
+ if (ifile->streams[i].dec_ctx)
+ avcodec_flush_buffers(ifile->streams[i].dec_ctx);
+ }
@@ -179,10 +179,10 @@ Index: FFmpeg/fftools/ffprobe.c
if (ret < 0) {
av_log(NULL, AV_LOG_ERROR, "Could not read packets in interval ");
log_read_interval(interval, NULL, AV_LOG_ERROR);
@@ -4609,6 +4706,7 @@ static const OptionDef real_options[] =
{ "print_filename", OPT_TYPE_FUNC, OPT_FUNC_ARG, {.func_arg = opt_print_filename}, "override the printed input filename", "print_file"},
{ "find_stream_info", OPT_TYPE_BOOL, OPT_INPUT | OPT_EXPERT, { &find_stream_info },
@@ -3208,6 +3305,7 @@ static const OptionDef real_options[] =
"read and decode the streams to fill missing information with heuristics" },
{ "c", OPT_TYPE_FUNC, OPT_FUNC_ARG, { .func_arg = opt_codec}, "force decoder", "decoder_name" },
{ "codec", OPT_TYPE_FUNC, OPT_FUNC_ARG, { .func_arg = opt_codec}, "alias for -c (force decoder)", "decoder_name" },
+ { "only_first_vframe", OPT_TYPE_BOOL, 0, { &only_show_first_video_frame }, "only show first video frame when show_frames is used" },
{ NULL, },
};
debian/patches/0080-fix-auto-inserting-sw-color-conv-filters-between-hw-fmts.patch → debian/patches/0067-fix-auto-inserting-sw-color-conv-filters-between-hw-fmts.patch
Vendored
+4 -2
View File
@@ -2,12 +2,13 @@ Index: FFmpeg/fftools/ffmpeg_filter.c
===================================================================
--- FFmpeg.orig/fftools/ffmpeg_filter.c
+++ FFmpeg/fftools/ffmpeg_filter.c
@@ -1543,8 +1543,21 @@ static int configure_output_video_filter
ofp->format != AV_PIX_FMT_NONE || !ofp->formats);
@@ -1708,9 +1708,22 @@ static int configure_output_video_filter
ofp->format != AV_PIX_FMT_NONE || !ofp->pix_fmts);
av_bprint_init(&bprint, 0, AV_BPRINT_SIZE_UNLIMITED);
choose_pix_fmts(ofp, &bprint);
- choose_color_spaces(ofp, &bprint);
- choose_color_ranges(ofp, &bprint);
- choose_alpha_modes(ofp, &bprint);
+ /* SW filter cannot handle color conversions between HW pixel formats. */
+ {
+ int bprint_color_options = 1;
@@ -21,6 +22,7 @@ Index: FFmpeg/fftools/ffmpeg_filter.c
+ if (bprint_color_options) {
+ choose_color_spaces(ofp, &bprint);
+ choose_color_ranges(ofp, &bprint);
+ choose_alpha_modes(ofp, &bprint);
+ }
+ }
if (!av_bprint_is_complete(&bprint))
debian/patches/0081-fix-dummy-hw-frame-missing-hw-frames-ctx.patch → debian/patches/0068-fix-dummy-hw-frame-missing-hw-frames-ctx.patch
Vendored
+1 -1
View File
@@ -14,7 +14,7 @@ Index: FFmpeg/fftools/ffmpeg_filter.c
+ ofp_fmt_desc = av_pix_fmt_desc_get(ofp->format);
+ // a dummy frame with HW pixel format requires a valid hw_frames_ctx
+ if (ofp_fmt_desc->flags & AV_PIX_FMT_FLAG_HWACCEL) {
+ AVFilterContext *sink = ofp->filter;
+ AVFilterContext *sink = ofp->ofilter.filter;
+ AVBufferRef *hwfc = av_buffersink_get_hw_frames_ctx(sink);
+ ret = av_buffer_replace(&frame->hw_frames_ctx, hwfc);
+ if (ret < 0)
debian/patches/0082-add-yadif-and-bwdif-opencl-filter-impl.patch → debian/patches/0069-add-yadif-and-bwdif-opencl-filter-impl.patch
Vendored
+34 -34
View File
@@ -2,16 +2,16 @@ Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -3867,6 +3867,7 @@ boxblur_opencl_filter_deps="opencl gpl"
@@ -4121,6 +4121,7 @@ boxblur_opencl_filter_deps="opencl gpl"
bs2b_filter_deps="libbs2b"
bwdif_cuda_filter_deps="ffnvcodec"
bwdif_cuda_filter_deps_any="cuda_nvcc cuda_llvm"
+bwdif_opencl_filter_deps="opencl"
bwdif_videotoolbox_filter_deps="metal corevideo videotoolbox"
bwdif_vulkan_filter_deps="vulkan spirv_compiler"
chromaber_vulkan_filter_deps="vulkan spirv_compiler"
@@ -4019,6 +4020,7 @@ xfade_opencl_filter_deps="opencl"
xfade_vulkan_filter_deps="vulkan spirv_compiler"
chromaber_vulkan_filter_deps="vulkan spirv_library"
@@ -4280,6 +4281,7 @@ xfade_opencl_filter_deps="opencl"
xfade_vulkan_filter_deps="vulkan spirv_library"
yadif_cuda_filter_deps="ffnvcodec"
yadif_cuda_filter_deps_any="cuda_nvcc cuda_llvm"
+yadif_opencl_filter_deps="opencl"
@@ -22,7 +22,7 @@ Index: FFmpeg/libavfilter/Makefile
===================================================================
--- FFmpeg.orig/libavfilter/Makefile
+++ FFmpeg/libavfilter/Makefile
@@ -219,6 +219,8 @@ OBJS-$(CONFIG_BOXBLUR_OPENCL_FILTER)
@@ -224,6 +224,8 @@ OBJS-$(CONFIG_BOXBLUR_OPENCL_FILTER)
OBJS-$(CONFIG_BWDIF_FILTER) += vf_bwdif.o bwdifdsp.o yadif_common.o
OBJS-$(CONFIG_BWDIF_CUDA_FILTER) += vf_bwdif_cuda.o vf_bwdif_cuda.ptx.o \
yadif_common.o
@@ -31,7 +31,7 @@ Index: FFmpeg/libavfilter/Makefile
OBJS-$(CONFIG_BWDIF_VIDEOTOOLBOX_FILTER) += vf_bwdif_videotoolbox.o \
metal/vf_bwdif_videotoolbox.metallib.o \
metal/utils.o \
@@ -591,6 +593,8 @@ OBJS-$(CONFIG_XSTACK_FILTER)
@@ -609,6 +611,8 @@ OBJS-$(CONFIG_XSTACK_FILTER)
OBJS-$(CONFIG_YADIF_FILTER) += vf_yadif.o yadif_common.o
OBJS-$(CONFIG_YADIF_CUDA_FILTER) += vf_yadif_cuda.o vf_yadif_cuda.ptx.o \
yadif_common.o cuda/load_helper.o
@@ -44,22 +44,22 @@ Index: FFmpeg/libavfilter/allfilters.c
===================================================================
--- FFmpeg.orig/libavfilter/allfilters.c
+++ FFmpeg/libavfilter/allfilters.c
@@ -201,6 +201,7 @@ extern const AVFilter ff_vf_boxblur;
extern const AVFilter ff_vf_boxblur_opencl;
extern const AVFilter ff_vf_bwdif;
extern const AVFilter ff_vf_bwdif_cuda;
+extern const AVFilter ff_vf_bwdif_opencl;
extern const AVFilter ff_vf_bwdif_videotoolbox;
extern const AVFilter ff_vf_bwdif_vulkan;
extern const AVFilter ff_vf_cas;
@@ -548,6 +549,7 @@ extern const AVFilter ff_vf_xpsnr;
extern const AVFilter ff_vf_xstack;
extern const AVFilter ff_vf_yadif;
extern const AVFilter ff_vf_yadif_cuda;
+extern const AVFilter ff_vf_yadif_opencl;
extern const AVFilter ff_vf_yadif_videotoolbox;
extern const AVFilter ff_vf_yaepblur;
extern const AVFilter ff_vf_zmq;
@@ -204,6 +204,7 @@ extern const FFFilter ff_vf_boxblur;
extern const FFFilter ff_vf_boxblur_opencl;
extern const FFFilter ff_vf_bwdif;
extern const FFFilter ff_vf_bwdif_cuda;
+extern const FFFilter ff_vf_bwdif_opencl;
extern const FFFilter ff_vf_bwdif_videotoolbox;
extern const FFFilter ff_vf_bwdif_vulkan;
extern const FFFilter ff_vf_cas;
@@ -562,6 +563,7 @@ extern const FFFilter ff_vf_xpsnr;
extern const FFFilter ff_vf_xstack;
extern const FFFilter ff_vf_yadif;
extern const FFFilter ff_vf_yadif_cuda;
+extern const FFFilter ff_vf_yadif_opencl;
extern const FFFilter ff_vf_yadif_videotoolbox;
extern const FFFilter ff_vf_yaepblur;
extern const FFFilter ff_vf_zmq;
Index: FFmpeg/libavfilter/opencl.c
===================================================================
--- FFmpeg.orig/libavfilter/opencl.c
@@ -583,7 +583,7 @@ Index: FFmpeg/libavfilter/vf_bwdif_opencl.c
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+
+#include "libavutil/avassert.h"
+#include "libavutil/common.h"
+#include "libavutil/imgutils.h"
@@ -878,17 +878,17 @@ Index: FFmpeg/libavfilter/vf_bwdif_opencl.c
+ },
+};
+
+const AVFilter ff_vf_bwdif_opencl = {
+ .name = "bwdif_opencl",
+ .description = NULL_IF_CONFIG_SMALL("Deinterlace (BWDIF) the video through OpenCL."),
+const FFFilter ff_vf_bwdif_opencl = {
+ .p.name = "bwdif_opencl",
+ .p.description = NULL_IF_CONFIG_SMALL("Deinterlace (BWDIF) the video through OpenCL."),
+ .priv_size = sizeof(DeintOpenCLContext),
+ .priv_class = &bwdif_opencl_class,
+ .p.priv_class = &bwdif_opencl_class,
+ .init = &deint_opencl_init,
+ .uninit = &deint_opencl_uninit,
+ FILTER_INPUTS(deint_opencl_inputs),
+ FILTER_OUTPUTS(deint_opencl_outputs),
+ FILTER_SINGLE_PIXFMT(AV_PIX_FMT_OPENCL),
+ .flags = AVFILTER_FLAG_HWDEVICE |
+ .p.flags = AVFILTER_FLAG_HWDEVICE |
+ AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
+ .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+};
@@ -916,7 +916,7 @@ Index: FFmpeg/libavfilter/vf_yadif_opencl.c
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+
+#include "libavutil/avassert.h"
+#include "libavutil/common.h"
+#include "libavutil/imgutils.h"
@@ -1203,17 +1203,17 @@ Index: FFmpeg/libavfilter/vf_yadif_opencl.c
+ },
+};
+
+const AVFilter ff_vf_yadif_opencl = {
+ .name = "yadif_opencl",
+ .description = NULL_IF_CONFIG_SMALL("Deinterlace (YADIF) the video through OpenCL."),
+const FFFilter ff_vf_yadif_opencl = {
+ .p.name = "yadif_opencl",
+ .p.description = NULL_IF_CONFIG_SMALL("Deinterlace (YADIF) the video through OpenCL."),
+ .priv_size = sizeof(DeintOpenCLContext),
+ .priv_class = &yadif_opencl_class,
+ .p.priv_class = &yadif_opencl_class,
+ .init = &deint_opencl_init,
+ .uninit = &deint_opencl_uninit,
+ FILTER_INPUTS(deint_opencl_inputs),
+ FILTER_OUTPUTS(deint_opencl_outputs),
+ FILTER_SINGLE_PIXFMT(AV_PIX_FMT_OPENCL),
+ .flags = AVFILTER_FLAG_HWDEVICE |
+ .p.flags = AVFILTER_FLAG_HWDEVICE |
+ AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
+ .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+};
debian/patches/0070-allow-vpp-tonemap-without-hdr10-metadata.patch
Vendored
+159
View File
@@ -0,0 +1,159 @@
Index: FFmpeg/libavfilter/vf_tonemap_vaapi.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_tonemap_vaapi.c
+++ FFmpeg/libavfilter/vf_tonemap_vaapi.c
@@ -56,7 +56,7 @@ static int tonemap_vaapi_save_metadata(A
AVContentLightMetadata *light_meta;
if (input_frame->color_trc != AVCOL_TRC_SMPTE2084) {
- av_log(avctx, AV_LOG_WARNING, "Only support HDR10 as input for vaapi tone-mapping\n");
+ av_log(avctx, AV_LOG_DEBUG, "Only support HDR10 as input for vaapi tone-mapping\n");
}
ctx->src_display = av_frame_get_side_data(input_frame,
@@ -122,8 +122,7 @@ static int tonemap_vaapi_save_metadata(A
ctx->in_metadata.white_point_y);
}
} else {
- av_log(avctx, AV_LOG_ERROR, "No mastering display data from input\n");
- return AVERROR(EINVAL);
+ av_log(avctx, AV_LOG_DEBUG, "No mastering display data from input\n");
}
ctx->src_light = av_frame_get_side_data(input_frame,
@@ -150,23 +149,20 @@ static int tonemap_vaapi_save_metadata(A
return 0;
}
-static int tonemap_vaapi_update_sidedata(AVFilterContext *avctx, AVFrame *output_frame)
+static int tonemap_vaapi_update_sidedata(AVFilterContext *avctx,
+ AVFrameSideData *metadata,
+ AVFrameSideData *metadata_lt)
{
HDRVAAPIContext *ctx = avctx->priv;
- AVFrameSideData *metadata;
AVMasteringDisplayMetadata *hdr_meta;
- AVFrameSideData *metadata_lt;
AVContentLightMetadata *hdr_meta_lt;
int i;
const int mapping[3] = {1, 2, 0}; //green, blue, red
const int chroma_den = 50000;
const int luma_den = 10000;
- metadata = av_frame_new_side_data(output_frame,
- AV_FRAME_DATA_MASTERING_DISPLAY_METADATA,
- sizeof(AVMasteringDisplayMetadata));
- if (!metadata)
- return AVERROR(ENOMEM);
+ if (!metadata || !metadata_lt)
+ return AVERROR(EINVAL);
hdr_meta = (AVMasteringDisplayMetadata *)metadata->data;
@@ -210,12 +206,6 @@ static int tonemap_vaapi_update_sidedata
ctx->out_metadata.max_display_mastering_luminance,
ctx->out_metadata.min_display_mastering_luminance);
- metadata_lt = av_frame_new_side_data(output_frame,
- AV_FRAME_DATA_CONTENT_LIGHT_LEVEL,
- sizeof(AVContentLightMetadata));
- if (!metadata_lt)
- return AVERROR(ENOMEM);
-
hdr_meta_lt = (AVContentLightMetadata *)metadata_lt->data;
hdr_meta_lt->MaxCLL = FFMIN(ctx->out_metadata.max_content_light_level, 65535);
@@ -396,7 +386,14 @@ static int tonemap_vaapi_filter_frame(AV
output_frame->colorspace = ctx->color_matrix;
if (ctx->mastering_display) {
- err = tonemap_vaapi_update_sidedata(avctx, output_frame);
+ AVFrameSideData *sd, *sd_lt;
+ sd = av_frame_new_side_data(output_frame,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA,
+ sizeof(AVMasteringDisplayMetadata));
+ sd_lt = av_frame_new_side_data(output_frame,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL,
+ sizeof(AVContentLightMetadata));
+ err = tonemap_vaapi_update_sidedata(avctx, sd, sd_lt);
if (err < 0)
goto fail;
}
@@ -457,6 +454,20 @@ static av_cold int tonemap_vaapi_init(AV
}
}
+ return 0;
+}
+
+static int tonemap_vaapi_config_output(AVFilterLink *outlink)
+{
+ AVFilterContext *avctx = outlink->src;
+ HDRVAAPIContext *ctx = avctx->priv;
+ int update_sd = 0, drop_sd = 0;
+ int err;
+
+ err = ff_vaapi_vpp_config_output(outlink);
+ if (err < 0)
+ return err;
+
#define STRING_OPTION(var_name, func_name, default_value) do { \
if (ctx->var_name ## _string) { \
int var = av_ ## func_name ## _from_name(ctx->var_name ## _string); \
@@ -505,6 +516,47 @@ static av_cold int tonemap_vaapi_init(AV
}
}
+ if (ctx->output_format_string) {
+ enum AVPixelFormat fmt = av_get_pix_fmt(ctx->output_format_string);
+ const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(fmt);
+
+ if (!pix_desc || (pix_desc->flags & AV_PIX_FMT_FLAG_HWACCEL))
+ return AVERROR(EINVAL);
+
+ if (pix_desc->comp[0].depth < 10)
+ drop_sd = 1;
+ else if (ctx->mastering_display)
+ update_sd = 1;
+ else
+ drop_sd = 1;
+ } else {
+ if (ctx->mastering_display)
+ update_sd = 1;
+ else
+ drop_sd = 1;
+ }
+
+ if (drop_sd) {
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
+ if (update_sd) {
+ AVFrameSideData *sd, *sd_lt;
+ sd = av_frame_side_data_new(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA,
+ sizeof(AVMasteringDisplayMetadata),
+ AV_FRAME_SIDE_DATA_FLAG_UNIQUE);
+ sd_lt = av_frame_side_data_new(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL,
+ sizeof(AVContentLightMetadata),
+ AV_FRAME_SIDE_DATA_FLAG_UNIQUE);
+ err = tonemap_vaapi_update_sidedata(avctx, sd, sd_lt);
+ if (err < 0)
+ return err;
+ }
+
return 0;
}
@@ -555,7 +607,7 @@ static const AVFilterPad tonemap_vaapi_o
{
.name = "default",
.type = AVMEDIA_TYPE_VIDEO,
- .config_props = &ff_vaapi_vpp_config_output,
+ .config_props = &tonemap_vaapi_config_output,
},
};
debian/patches/0084-allow-qsvenc-to-bail-out-from-unsupported-low-power-mode.patch → debian/patches/0071-allow-qsvenc-to-bail-out-from-unsupported-low-power-mode.patch
Vendored
View File
debian/patches/0086-add-array-of-textures-support-for-d3d11va-hwaccel.patch → debian/patches/0072-add-array-of-textures-support-for-d3d11va-hwaccel.patch
Vendored
+3 -3
View File
@@ -232,7 +232,7 @@ Index: FFmpeg/libavcodec/dxva2.c
av_log((void *)avctx, AV_LOG_ERROR, "get_buffer frame is invalid!\n");
return NULL;
}
@@ -853,8 +932,10 @@ unsigned ff_dxva2_get_surface_index(cons
@@ -852,8 +931,10 @@ unsigned ff_dxva2_get_surface_index(cons
}
#endif
#if CONFIG_D3D11VA
@@ -261,7 +261,7 @@ Index: FFmpeg/libavutil/hwcontext_d3d11va.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_d3d11va.c
+++ FFmpeg/libavutil/hwcontext_d3d11va.c
@@ -318,7 +318,8 @@ static int d3d11va_frames_init(AVHWFrame
@@ -329,7 +329,8 @@ static int d3d11va_frames_init(AVHWFrame
ctx->initial_pool_size = texDesc2.ArraySize;
hwctx->BindFlags = texDesc2.BindFlags;
hwctx->MiscFlags = texDesc2.MiscFlags;
@@ -275,7 +275,7 @@ Index: FFmpeg/libavutil/hwcontext_d3d11va.h
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_d3d11va.h
+++ FFmpeg/libavutil/hwcontext_d3d11va.h
@@ -188,6 +188,11 @@ typedef struct AVD3D11VAFramesContext {
@@ -202,6 +202,11 @@ typedef struct AVD3D11VAFramesContext {
* Whether the frames require extra sync when exporting as external memory.
*/
int require_sync;
debian/patches/0090-dont-return-ext-error-for-invalid-vt-frame.patch → debian/patches/0073-dont-return-ext-error-for-invalid-vt-frame.patch
Vendored
View File
debian/patches/0074-fix-mapped-hwframe-to-swframe-swscale-conversion.patch
Vendored
+18
View File
@@ -0,0 +1,18 @@
Index: FFmpeg/libswscale/swscale.c
===================================================================
--- FFmpeg.orig/libswscale/swscale.c
+++ FFmpeg/libswscale/swscale.c
@@ -1430,7 +1430,12 @@ int sws_frame_setup(SwsContext *ctx, con
/* For now, if a single frame has a context, then both need a context */
if (!!src->hw_frames_ctx != !!dst->hw_frames_ctx) {
- return AVERROR(ENOTSUP);
+ const AVPixFmtDescriptor *src_desc = av_pix_fmt_desc_get(src->format);
+ const AVPixFmtDescriptor *dst_desc = av_pix_fmt_desc_get(dst->format);
+ const int is_mapped_hwframe_to_sw = !(src_desc->flags & AV_PIX_FMT_FLAG_HWACCEL) &&
+ !(dst_desc->flags & AV_PIX_FMT_FLAG_HWACCEL);
+ if (!is_mapped_hwframe_to_sw)
+ return AVERROR(ENOTSUP);
} else if (!!src->hw_frames_ctx) {
/* Both hardware frames must already be allocated */
if (!src->data[0] || !dst->data[0])
debian/patches/0075-fix-hwupload-filter-cannot-use-devices-created-by-hwaccel.patch
Vendored
+101
View File
@@ -0,0 +1,101 @@
Index: FFmpeg/libavfilter/vf_hwupload.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_hwupload.c
+++ FFmpeg/libavfilter/vf_hwupload.c
@@ -47,43 +47,35 @@ typedef struct HWUploadContext {
char *device_type;
} HWUploadContext;
-static int hwupload_init(AVFilterContext *avctx)
+static int hwupload_query_formats(AVFilterContext *avctx)
{
HWUploadContext *ctx = avctx->priv;
- int err;
+ AVHWFramesConstraints *constraints = NULL;
+ const enum AVPixelFormat *input_pix_fmts, *output_pix_fmts;
+ AVFilterFormats *input_formats = NULL;
+ int err, i;
- if (!avctx->hw_device_ctx) {
+ if (ctx->hwdevice_ref) {
+ /* We already have a specified device. */
+ } else if (avctx->hw_device_ctx) {
+ if (ctx->device_type) {
+ err = av_hwdevice_ctx_create_derived(
+ &ctx->hwdevice_ref,
+ av_hwdevice_find_type_by_name(ctx->device_type),
+ avctx->hw_device_ctx, 0);
+ if (err < 0)
+ return err;
+ } else {
+ ctx->hwdevice_ref = av_buffer_ref(avctx->hw_device_ctx);
+ if (!ctx->hwdevice_ref)
+ return AVERROR(ENOMEM);
+ }
+ } else {
av_log(ctx, AV_LOG_ERROR, "A hardware device reference is required "
"to upload frames to.\n");
return AVERROR(EINVAL);
}
- if (ctx->device_type) {
- err = av_hwdevice_ctx_create_derived(
- &ctx->hwdevice_ref,
- av_hwdevice_find_type_by_name(ctx->device_type),
- avctx->hw_device_ctx, 0);
- if (err < 0)
- return err;
- } else {
- ctx->hwdevice_ref = av_buffer_ref(avctx->hw_device_ctx);
- if (!ctx->hwdevice_ref)
- return AVERROR(ENOMEM);
- }
-
- return 0;
-}
-
-static int hwupload_query_formats(const AVFilterContext *avctx,
- AVFilterFormatsConfig **cfg_in,
- AVFilterFormatsConfig **cfg_out)
-{
- const HWUploadContext *ctx = avctx->priv;
- AVHWFramesConstraints *constraints = NULL;
- const enum AVPixelFormat *input_pix_fmts, *output_pix_fmts;
- AVFilterFormats *input_formats = NULL;
- int err, i;
-
constraints = av_hwdevice_get_hwframe_constraints(ctx->hwdevice_ref, NULL);
if (!constraints) {
err = AVERROR(EINVAL);
@@ -106,15 +98,16 @@ static int hwupload_query_formats(const
}
}
- if ((err = ff_formats_ref(input_formats, &cfg_in[0]->formats)) < 0 ||
+ if ((err = ff_formats_ref(input_formats, &avctx->inputs[0]->outcfg.formats)) < 0 ||
(err = ff_formats_ref(ff_make_pixel_format_list(output_pix_fmts),
- &cfg_out[0]->formats)) < 0)
+ &avctx->outputs[0]->incfg.formats)) < 0)
goto fail;
av_hwframe_constraints_free(&constraints);
return 0;
fail:
+ av_buffer_unref(&ctx->hwdevice_ref);
av_hwframe_constraints_free(&constraints);
return err;
}
@@ -287,11 +280,10 @@ const FFFilter ff_vf_hwupload = {
.p.description = NULL_IF_CONFIG_SMALL("Upload a normal frame to a hardware frame"),
.p.priv_class = &hwupload_class,
.p.flags = AVFILTER_FLAG_HWDEVICE,
- .init = hwupload_init,
.uninit = hwupload_uninit,
.priv_size = sizeof(HWUploadContext),
FILTER_INPUTS(hwupload_inputs),
FILTER_OUTPUTS(hwupload_outputs),
- FILTER_QUERY_FUNC2(hwupload_query_formats),
+ FILTER_QUERY_FUNC(hwupload_query_formats),
.flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
};
debian/patches/0076-backport-av1-videotoolbox.patch
Vendored
-354
View File
@@ -1,354 +0,0 @@
Index: FFmpeg/configure
===================================================================
--- FFmpeg.orig/configure
+++ FFmpeg/configure
@@ -2467,6 +2467,7 @@ TYPES_LIST="
kCMVideoCodecType_HEVC
kCMVideoCodecType_HEVCWithAlpha
kCMVideoCodecType_VP9
+ kCMVideoCodecType_AV1
kCVPixelFormatType_420YpCbCr10BiPlanarVideoRange
kCVPixelFormatType_422YpCbCr8BiPlanarVideoRange
kCVPixelFormatType_422YpCbCr10BiPlanarVideoRange
@@ -3176,6 +3177,8 @@ av1_vaapi_hwaccel_deps="vaapi VADecPictu
av1_vaapi_hwaccel_select="av1_decoder"
av1_vdpau_hwaccel_deps="vdpau VdpPictureInfoAV1"
av1_vdpau_hwaccel_select="av1_decoder"
+av1_videotoolbox_hwaccel_deps="videotoolbox"
+av1_videotoolbox_hwaccel_select="av1_decoder"
av1_vulkan_hwaccel_deps="vulkan"
av1_vulkan_hwaccel_select="av1_decoder"
h263_vaapi_hwaccel_deps="vaapi"
@@ -6744,6 +6747,7 @@ enabled videotoolbox && {
check_func_headers CoreMedia/CMFormatDescription.h kCMVideoCodecType_HEVC "-framework CoreMedia"
check_func_headers CoreMedia/CMFormatDescription.h kCMVideoCodecType_HEVCWithAlpha "-framework CoreMedia"
check_func_headers CoreMedia/CMFormatDescription.h kCMVideoCodecType_VP9 "-framework CoreMedia"
+ check_func_headers CoreMedia/CMFormatDescription.h kCMVideoCodecType_AV1 "-framework CoreMedia"
check_func_headers CoreVideo/CVPixelBuffer.h kCVPixelFormatType_420YpCbCr10BiPlanarVideoRange "-framework CoreVideo"
check_func_headers CoreVideo/CVPixelBuffer.h kCVPixelFormatType_422YpCbCr8BiPlanarVideoRange "-framework CoreVideo"
check_func_headers CoreVideo/CVPixelBuffer.h kCVPixelFormatType_422YpCbCr10BiPlanarVideoRange "-framework CoreVideo"
Index: FFmpeg/libavcodec/Makefile
===================================================================
--- FFmpeg.orig/libavcodec/Makefile
+++ FFmpeg/libavcodec/Makefile
@@ -1018,6 +1018,7 @@ OBJS-$(CONFIG_AV1_D3D12VA_HWACCEL)
OBJS-$(CONFIG_AV1_NVDEC_HWACCEL) += nvdec_av1.o
OBJS-$(CONFIG_AV1_VAAPI_HWACCEL) += vaapi_av1.o
OBJS-$(CONFIG_AV1_VDPAU_HWACCEL) += vdpau_av1.o
+OBJS-$(CONFIG_AV1_VIDEOTOOLBOX_HWACCEL) += videotoolbox_av1.o
OBJS-$(CONFIG_AV1_VULKAN_HWACCEL) += vulkan_decode.o vulkan_av1.o
OBJS-$(CONFIG_H263_VAAPI_HWACCEL) += vaapi_mpeg4.o
OBJS-$(CONFIG_H263_VIDEOTOOLBOX_HWACCEL) += videotoolbox.o
Index: FFmpeg/libavcodec/av1dec.c
===================================================================
--- FFmpeg.orig/libavcodec/av1dec.c
+++ FFmpeg/libavcodec/av1dec.c
@@ -541,6 +541,7 @@ static int get_pixel_format(AVCodecConte
CONFIG_AV1_NVDEC_HWACCEL + \
CONFIG_AV1_VAAPI_HWACCEL + \
CONFIG_AV1_VDPAU_HWACCEL + \
+ CONFIG_AV1_VIDEOTOOLBOX_HWACCEL + \
CONFIG_AV1_VULKAN_HWACCEL)
enum AVPixelFormat pix_fmts[HWACCEL_MAX + 2], *fmtp = pix_fmts;
@@ -568,6 +569,9 @@ static int get_pixel_format(AVCodecConte
#if CONFIG_AV1_VDPAU_HWACCEL
*fmtp++ = AV_PIX_FMT_VDPAU;
#endif
+#if CONFIG_AV1_VIDEOTOOLBOX_HWACCEL
+ *fmtp++ = AV_PIX_FMT_VIDEOTOOLBOX;
+#endif
#if CONFIG_AV1_VULKAN_HWACCEL
*fmtp++ = AV_PIX_FMT_VULKAN;
#endif
@@ -592,6 +596,9 @@ static int get_pixel_format(AVCodecConte
#if CONFIG_AV1_VDPAU_HWACCEL
*fmtp++ = AV_PIX_FMT_VDPAU;
#endif
+#if CONFIG_AV1_VIDEOTOOLBOX_HWACCEL
+ *fmtp++ = AV_PIX_FMT_VIDEOTOOLBOX;
+#endif
#if CONFIG_AV1_VULKAN_HWACCEL
*fmtp++ = AV_PIX_FMT_VULKAN;
#endif
@@ -1439,6 +1446,10 @@ static int av1_receive_frame_internal(AV
if (raw_tile_group && (s->tile_num == raw_tile_group->tg_end + 1)) {
int show_frame = s->raw_frame_header->show_frame;
+ // Set nb_unit to point at the next OBU, to indicate which
+ // OBUs have been processed for this current frame. (If this
+ // frame gets output, we set nb_unit to this value later too.)
+ s->nb_unit = i + 1;
if (avctx->hwaccel && s->cur_frame.f) {
ret = FF_HW_SIMPLE_CALL(avctx, end_frame);
if (ret < 0) {
@@ -1449,6 +1460,8 @@ static int av1_receive_frame_internal(AV
update_reference_list(avctx);
+ // Set start_unit to indicate the first OBU of the next frame.
+ s->start_unit = s->nb_unit;
raw_tile_group = NULL;
s->raw_frame_header = NULL;
@@ -1478,7 +1491,7 @@ end:
s->raw_frame_header = NULL;
av_packet_unref(s->pkt);
ff_cbs_fragment_reset(&s->current_obu);
- s->nb_unit = 0;
+ s->nb_unit = s->start_unit = 0;
}
if (!ret && !frame->buf[0])
ret = AVERROR(EAGAIN);
@@ -1505,7 +1518,7 @@ static int av1_receive_frame(AVCodecCont
return ret;
}
- s->nb_unit = 0;
+ s->nb_unit = s->start_unit = 0;
av_log(avctx, AV_LOG_DEBUG, "Total OBUs on this packet: %d.\n",
s->current_obu.nb_units);
}
@@ -1526,7 +1539,7 @@ static void av1_decode_flush(AVCodecCont
av1_frame_unref(&s->cur_frame);
s->operating_point_idc = 0;
- s->nb_unit = 0;
+ s->nb_unit = s->start_unit = 0;
s->raw_frame_header = NULL;
s->raw_seq = NULL;
s->cll = NULL;
@@ -1594,6 +1607,9 @@ const FFCodec ff_av1_decoder = {
#if CONFIG_AV1_VDPAU_HWACCEL
HWACCEL_VDPAU(av1),
#endif
+#if CONFIG_AV1_VIDEOTOOLBOX_HWACCEL
+ HWACCEL_VIDEOTOOLBOX(av1),
+#endif
#if CONFIG_AV1_VULKAN_HWACCEL
HWACCEL_VULKAN(av1),
#endif
Index: FFmpeg/libavcodec/av1dec.h
===================================================================
--- FFmpeg.orig/libavcodec/av1dec.h
+++ FFmpeg/libavcodec/av1dec.h
@@ -114,7 +114,8 @@ typedef struct AV1DecContext {
AV1Frame ref[AV1_NUM_REF_FRAMES];
AV1Frame cur_frame;
- int nb_unit;
+ int nb_unit; ///< The index of the next OBU to be processed.
+ int start_unit; ///< The index of the first OBU of the current frame.
// AVOptions
int operating_point;
Index: FFmpeg/libavcodec/hwaccels.h
===================================================================
--- FFmpeg.orig/libavcodec/hwaccels.h
+++ FFmpeg/libavcodec/hwaccels.h
@@ -26,6 +26,7 @@ extern const struct FFHWAccel ff_av1_dxv
extern const struct FFHWAccel ff_av1_nvdec_hwaccel;
extern const struct FFHWAccel ff_av1_vaapi_hwaccel;
extern const struct FFHWAccel ff_av1_vdpau_hwaccel;
+extern const struct FFHWAccel ff_av1_videotoolbox_hwaccel;
extern const struct FFHWAccel ff_av1_vulkan_hwaccel;
extern const struct FFHWAccel ff_h263_vaapi_hwaccel;
extern const struct FFHWAccel ff_h263_videotoolbox_hwaccel;
Index: FFmpeg/libavcodec/videotoolbox.c
===================================================================
--- FFmpeg.orig/libavcodec/videotoolbox.c
+++ FFmpeg/libavcodec/videotoolbox.c
@@ -56,6 +56,10 @@ enum { kCMVideoCodecType_HEVC = 'hvc1' }
enum { kCMVideoCodecType_VP9 = 'vp09' };
#endif
+#if !HAVE_KCMVIDEOCODECTYPE_AV1
+enum { kCMVideoCodecType_AV1 = 'av01' };
+#endif
+
#define VIDEOTOOLBOX_ESDS_EXTRADATA_PADDING 12
typedef struct VTHWFrame {
@@ -92,6 +96,26 @@ int ff_videotoolbox_buffer_copy(VTContex
return 0;
}
+int ff_videotoolbox_buffer_append(VTContext *vtctx,
+ const uint8_t *buffer,
+ uint32_t size)
+{
+ void *tmp;
+
+ tmp = av_fast_realloc(vtctx->bitstream,
+ &vtctx->allocated_size,
+ vtctx->bitstream_size + size);
+
+ if (!tmp)
+ return AVERROR(ENOMEM);
+
+ vtctx->bitstream = tmp;
+ memcpy(vtctx->bitstream + vtctx->bitstream_size, buffer, size);
+ vtctx->bitstream_size += size;
+
+ return 0;
+}
+
static int videotoolbox_postproc_frame(void *avctx, AVFrame *frame)
{
int ret;
@@ -840,6 +864,13 @@ static CFDictionaryRef videotoolbox_deco
CFDictionarySetValue(avc_info, CFSTR("vpcC"), data);
break;
#endif
+#if CONFIG_AV1_VIDEOTOOLBOX_HWACCEL
+ case kCMVideoCodecType_AV1 :
+ data = ff_videotoolbox_av1c_extradata_create(avctx);
+ if (data)
+ CFDictionarySetValue(avc_info, CFSTR("av1C"), data);
+ break;
+#endif
default:
break;
}
@@ -905,6 +936,9 @@ static int videotoolbox_start(AVCodecCon
case AV_CODEC_ID_VP9 :
videotoolbox->cm_codec_type = kCMVideoCodecType_VP9;
break;
+ case AV_CODEC_ID_AV1 :
+ videotoolbox->cm_codec_type = kCMVideoCodecType_AV1;
+ break;
default :
break;
}
Index: FFmpeg/libavcodec/videotoolbox_av1.c
===================================================================
--- /dev/null
+++ FFmpeg/libavcodec/videotoolbox_av1.c
@@ -0,0 +1,105 @@
+/*
+ * Videotoolbox hardware acceleration for AV1
+ * Copyright (c) 2023 Jan Ekström
+ * Copyright (c) 2024 Ruslan Chernenko
+ * Copyright (c) 2024 Martin Storsjö
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/mem.h"
+
+#include "av1dec.h"
+#include "hwaccel_internal.h"
+#include "internal.h"
+#include "vt_internal.h"
+
+CFDataRef ff_videotoolbox_av1c_extradata_create(AVCodecContext *avctx)
+{
+ AV1DecContext *s = avctx->priv_data;
+ uint8_t *buf;
+ CFDataRef data;
+ if (!s->raw_seq)
+ return NULL;
+
+ buf = av_malloc(s->seq_data_ref->size + 4);
+ if (!buf)
+ return NULL;
+ buf[0] = 0x81; // version and marker (constant)
+ buf[1] = s->raw_seq->seq_profile << 5 | s->raw_seq->seq_level_idx[0];
+ buf[2] = s->raw_seq->seq_tier[0] << 7 |
+ s->raw_seq->color_config.high_bitdepth << 6 |
+ s->raw_seq->color_config.twelve_bit << 5 |
+ s->raw_seq->color_config.mono_chrome << 4 |
+ s->raw_seq->color_config.subsampling_x << 3 |
+ s->raw_seq->color_config.subsampling_y << 2 |
+ s->raw_seq->color_config.chroma_sample_position;
+
+ if (s->raw_seq->initial_display_delay_present_flag)
+ buf[3] = 0 << 5 |
+ s->raw_seq->initial_display_delay_present_flag << 4 |
+ s->raw_seq->initial_display_delay_minus_1[0];
+ else
+ buf[3] = 0x00;
+ memcpy(buf + 4, s->seq_data_ref->data, s->seq_data_ref->size);
+ data = CFDataCreate(kCFAllocatorDefault, buf, s->seq_data_ref->size + 4);
+ av_free(buf);
+ return data;
+};
+
+
+static int videotoolbox_av1_start_frame(AVCodecContext *avctx,
+ const uint8_t *buffer,
+ uint32_t size)
+{
+ return 0;
+}
+
+static int videotoolbox_av1_decode_slice(AVCodecContext *avctx,
+ const uint8_t *buffer,
+ uint32_t size)
+{
+ return 0;
+}
+
+static int videotoolbox_av1_end_frame(AVCodecContext *avctx)
+{
+ const AV1DecContext *s = avctx->priv_data;
+ VTContext *vtctx = avctx->internal->hwaccel_priv_data;
+ AVFrame *frame = s->cur_frame.f;
+
+ vtctx->bitstream_size = 0;
+ for (int i = s->start_unit; i < s->nb_unit; i++)
+ ff_videotoolbox_buffer_append(vtctx, s->current_obu.units[i].data,
+ s->current_obu.units[i].data_size);
+ return ff_videotoolbox_common_end_frame(avctx, frame);
+}
+
+const FFHWAccel ff_av1_videotoolbox_hwaccel = {
+ .p.name = "av1_videotoolbox",
+ .p.type = AVMEDIA_TYPE_VIDEO,
+ .p.id = AV_CODEC_ID_AV1,
+ .p.pix_fmt = AV_PIX_FMT_VIDEOTOOLBOX,
+ .alloc_frame = ff_videotoolbox_alloc_frame,
+ .start_frame = videotoolbox_av1_start_frame,
+ .decode_slice = videotoolbox_av1_decode_slice,
+ .end_frame = videotoolbox_av1_end_frame,
+ .frame_params = ff_videotoolbox_frame_params,
+ .init = ff_videotoolbox_common_init,
+ .uninit = ff_videotoolbox_uninit,
+ .priv_data_size = sizeof(VTContext),
+};
Index: FFmpeg/libavcodec/vt_internal.h
===================================================================
--- FFmpeg.orig/libavcodec/vt_internal.h
+++ FFmpeg/libavcodec/vt_internal.h
@@ -56,6 +56,9 @@ int ff_videotoolbox_frame_params(AVCodec
int ff_videotoolbox_buffer_copy(VTContext *vtctx,
const uint8_t *buffer,
uint32_t size);
+int ff_videotoolbox_buffer_append(VTContext *vtctx,
+ const uint8_t *buffer,
+ uint32_t size);
int ff_videotoolbox_uninit(AVCodecContext *avctx);
int ff_videotoolbox_h264_start_frame(AVCodecContext *avctx,
const uint8_t *buffer,
@@ -64,6 +67,7 @@ int ff_videotoolbox_h264_decode_slice(AV
const uint8_t *buffer,
uint32_t size);
int ff_videotoolbox_common_end_frame(AVCodecContext *avctx, AVFrame *frame);
+CFDataRef ff_videotoolbox_av1c_extradata_create(AVCodecContext *avctx);
CFDataRef ff_videotoolbox_avcc_extradata_create(AVCodecContext *avctx);
CFDataRef ff_videotoolbox_hvcc_extradata_create(AVCodecContext *avctx);
CFDataRef ff_videotoolbox_vpcc_extradata_create(AVCodecContext *avctx);
debian/patches/0076-fix-seeking-h264-open-gop-videos-with-d3d11va-on-amd.patch
Vendored
+13
View File
@@ -0,0 +1,13 @@
Index: FFmpeg/libavcodec/h264_slice.c
===================================================================
--- FFmpeg.orig/libavcodec/h264_slice.c
+++ FFmpeg/libavcodec/h264_slice.c
@@ -1150,7 +1150,7 @@ static int h264_init_ps(H264Context *h,
if (flush_changes)
ff_h264_flush_change(h);
- if ((ret = get_pixel_format(h, must_reinit || needs_reinit)) < 0)
+ if ((ret = get_pixel_format(h, 1)) < 0)
return ret;
h->avctx->pix_fmt = ret;
debian/patches/0077-fix-low-default-readrate-catchup-caused-remuxing-to-stall.patch
Vendored
+22
View File
@@ -0,0 +1,22 @@
Index: FFmpeg/fftools/ffmpeg_demux.c
===================================================================
--- FFmpeg.orig/fftools/ffmpeg_demux.c
+++ FFmpeg/fftools/ffmpeg_demux.c
@@ -534,7 +534,7 @@ static void readrate_sleep(Demuxer *d)
ds->lag = lag;
ds->resume_wc = now;
ds->resume_pts = pts;
- av_log_once(ds, AV_LOG_WARNING, AV_LOG_DEBUG, &resume_warn,
+ av_log_once(ds, AV_LOG_VERBOSE, AV_LOG_DEBUG, &resume_warn,
"Resumed reading at pts %0.3f with rate %0.3f after a lag of %0.3fs\n",
(float)pts/AV_TIME_BASE, d->readrate_catchup, (float)lag/AV_TIME_BASE);
}
@@ -2152,7 +2152,7 @@ int ifile_open(const OptionsContext *o,
d->readrate_initial_burst);
return AVERROR(EINVAL);
}
- d->readrate_catchup = o->readrate_catchup ? o->readrate_catchup : d->readrate * 1.05;
+ d->readrate_catchup = o->readrate_catchup ? o->readrate_catchup : d->readrate * 100;
if (d->readrate_catchup < d->readrate) {
av_log(d, AV_LOG_ERROR,
"Option -readrate_catchup is %0.3f; it must be at least equal to %0.3f.\n",
debian/patches/0078-backport-fixes-for-vulkan-from-upstream.patch
Vendored
+509
View File
@@ -0,0 +1,509 @@
Index: FFmpeg/libavcodec/vulkan_encode.c
===================================================================
--- FFmpeg.orig/libavcodec/vulkan_encode.c
+++ FFmpeg/libavcodec/vulkan_encode.c
@@ -772,14 +772,6 @@ av_cold int ff_vulkan_encode_init(AVCode
return AVERROR(EINVAL);
}
- if ((ctx->enc_caps.supportedEncodeFeedbackFlags & feedback_flags) !=
- feedback_flags) {
- av_log (avctx, AV_LOG_ERROR,
- "Driver does not support required encode feedback flags "
- "(BUFFER_OFFSET and BYTES_WRITTEN).\n");
- return AVERROR(ENOTSUP);
- }
-
ctx->base.op = &vulkan_base_encode_ops;
ctx->codec = codec;
@@ -881,6 +873,14 @@ av_cold int ff_vulkan_encode_init(AVCode
return AVERROR_EXTERNAL;
}
+ if ((ctx->enc_caps.supportedEncodeFeedbackFlags & feedback_flags) !=
+ feedback_flags) {
+ av_log(avctx, AV_LOG_ERROR,
+ "Driver does not support required encode feedback flags "
+ "(BUFFER_OFFSET and BYTES_WRITTEN).\n");
+ return AVERROR(ENOTSUP);
+ }
+
err = init_rc(avctx, ctx);
if (err < 0)
return err;
Index: FFmpeg/libavcodec/vulkan_encode.h
===================================================================
--- FFmpeg.orig/libavcodec/vulkan_encode.h
+++ FFmpeg/libavcodec/vulkan_encode.h
@@ -203,7 +203,7 @@ typedef struct FFVulkanEncodeContext {
#define VULKAN_ENCODE_COMMON_OPTIONS \
{ "qp", "Use an explicit constant quantizer for the whole stream", OFFSET(common.opts.qp), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 255, FLAGS }, \
- { "quality", "Set encode quality (trades off against speed, higher is faster)", OFFSET(common.opts.quality), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FLAGS }, \
+ { "quality", "Set encode quality (trades off against speed, higher is slower)", OFFSET(common.opts.quality), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FLAGS }, \
{ "rc_mode", "Select rate control type", OFFSET(common.opts.rc_mode), AV_OPT_TYPE_INT, { .i64 = FF_VK_RC_MODE_AUTO }, 0, FF_VK_RC_MODE_AUTO, FLAGS, "rc_mode" }, \
{ "auto", "Choose mode automatically based on parameters", 0, AV_OPT_TYPE_CONST, { .i64 = FF_VK_RC_MODE_AUTO }, INT_MIN, INT_MAX, FLAGS, "rc_mode" }, \
{ "driver", "Driver-specific rate control", 0, AV_OPT_TYPE_CONST, { .i64 = VK_VIDEO_ENCODE_RATE_CONTROL_MODE_DEFAULT_KHR }, INT_MIN, INT_MAX, FLAGS, "rc_mode" }, \
Index: FFmpeg/libavcodec/vulkan_encode_h265.c
===================================================================
--- FFmpeg.orig/libavcodec/vulkan_encode_h265.c
+++ FFmpeg/libavcodec/vulkan_encode_h265.c
@@ -1595,23 +1595,21 @@ static av_cold int vulkan_encode_h265_in
av_log(avctx, AV_LOG_VERBOSE, " Capability flags:\n");
av_log(avctx, AV_LOG_VERBOSE, " hdr_compliance: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_HRD_COMPLIANCE_BIT_KHR));
+ !!(enc->caps.flags & VK_VIDEO_ENCODE_H265_CAPABILITY_HRD_COMPLIANCE_BIT_KHR));
av_log(avctx, AV_LOG_VERBOSE, " pred_weight_table_generated: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_PREDICTION_WEIGHT_TABLE_GENERATED_BIT_KHR));
+ !!(enc->caps.flags & VK_VIDEO_ENCODE_H265_CAPABILITY_PREDICTION_WEIGHT_TABLE_GENERATED_BIT_KHR));
av_log(avctx, AV_LOG_VERBOSE, " row_unaligned_slice: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_ROW_UNALIGNED_SLICE_BIT_KHR));
+ !!(enc->caps.flags & VK_VIDEO_ENCODE_H265_CAPABILITY_ROW_UNALIGNED_SLICE_SEGMENT_BIT_KHR));
av_log(avctx, AV_LOG_VERBOSE, " different_slice_type: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_DIFFERENT_SLICE_TYPE_BIT_KHR));
+ !!(enc->caps.flags & VK_VIDEO_ENCODE_H265_CAPABILITY_DIFFERENT_SLICE_SEGMENT_TYPE_BIT_KHR));
av_log(avctx, AV_LOG_VERBOSE, " b_frame_in_l0_list: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_B_FRAME_IN_L0_LIST_BIT_KHR));
+ !!(enc->caps.flags & VK_VIDEO_ENCODE_H265_CAPABILITY_B_FRAME_IN_L0_LIST_BIT_KHR));
av_log(avctx, AV_LOG_VERBOSE, " b_frame_in_l1_list: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_B_FRAME_IN_L1_LIST_BIT_KHR));
+ !!(enc->caps.flags & VK_VIDEO_ENCODE_H265_CAPABILITY_B_FRAME_IN_L1_LIST_BIT_KHR));
av_log(avctx, AV_LOG_VERBOSE, " per_pict_type_min_max_qp: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_PER_PICTURE_TYPE_MIN_MAX_QP_BIT_KHR));
+ !!(enc->caps.flags & VK_VIDEO_ENCODE_H265_CAPABILITY_PER_PICTURE_TYPE_MIN_MAX_QP_BIT_KHR));
av_log(avctx, AV_LOG_VERBOSE, " per_slice_constant_qp: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_PER_SLICE_CONSTANT_QP_BIT_KHR));
- av_log(avctx, AV_LOG_VERBOSE, " generate_prefix_nalu: %i\n",
- !!(enc->caps.flags & VK_VIDEO_ENCODE_H264_CAPABILITY_GENERATE_PREFIX_NALU_BIT_KHR));
+ !!(enc->caps.flags & VK_VIDEO_ENCODE_H265_CAPABILITY_PER_SLICE_SEGMENT_CONSTANT_QP_BIT_KHR));
av_log(avctx, AV_LOG_VERBOSE, " Capabilities:\n");
av_log(avctx, AV_LOG_VERBOSE, " maxLevelIdc: %i\n",
Index: FFmpeg/libavutil/hwcontext_vulkan.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_vulkan.c
+++ FFmpeg/libavutil/hwcontext_vulkan.c
@@ -190,11 +190,18 @@ typedef struct VulkanFramesPriv {
/* Properties for DRM modifier for each plane in the image */
VkDrmFormatModifierPropertiesEXT drm_format_modifier_properties[5];
+
+ /* Set when physical device reports DEDICATED_ONLY for DMA-BUF export (try_export_flags) */
+ int export_requires_dedicated;
} VulkanFramesPriv;
typedef struct AVVkFrameInternal {
pthread_mutex_t update_mutex;
+ /* Binary semaphore for SYNC_FD export at DRM map time. Created once lazily,
+ * re-signaled each time via a submit in vulkan_map_to_drm. */
+ VkSemaphore drm_sync_sem;
+
#if CONFIG_CUDA
/* Importing external memory into cuda is really expensive so we keep the
* memory imported all the time */
@@ -2370,7 +2377,7 @@ static int alloc_mem(AVHWDeviceContext *
return 0;
}
-static void vulkan_free_internal(AVVkFrame *f)
+static void vulkan_free_internal(VulkanDevicePriv *p, AVVkFrame *f)
{
av_unused AVVkFrameInternal *internal = f->internal;
@@ -2402,6 +2409,10 @@ static void vulkan_free_internal(AVVkFra
}
#endif
+ if (internal->drm_sync_sem != VK_NULL_HANDLE)
+ p->vkctx.vkfn.DestroySemaphore(p->p.act_dev, internal->drm_sync_sem,
+ p->p.alloc);
+
pthread_mutex_destroy(&internal->update_mutex);
av_freep(&f->internal);
}
@@ -2429,7 +2440,7 @@ static void vulkan_frame_free(AVHWFrames
vk->WaitSemaphores(hwctx->act_dev, &sem_wait, UINT64_MAX);
}
- vulkan_free_internal(f);
+ vulkan_free_internal(p, f);
for (int i = 0; i < nb_images; i++) {
vk->DestroyImage(hwctx->act_dev, f->img[i], hwctx->alloc);
@@ -2476,6 +2487,10 @@ static int alloc_bind_mem(AVHWFramesCont
vk->GetImageMemoryRequirements2(hwctx->act_dev, &req_desc, &req);
+ av_log(hwfc, AV_LOG_TRACE,
+ "plane %d: driver reports prefersDedicatedAllocation=%i requiresDedicatedAllocation=%i\n",
+ img_cnt, ded_req.prefersDedicatedAllocation, ded_req.requiresDedicatedAllocation);
+
if (f->tiling == VK_IMAGE_TILING_LINEAR)
req.memoryRequirements.size = FFALIGN(req.memoryRequirements.size,
p->props.properties.limits.minMemoryMapAlignment);
@@ -2483,6 +2498,8 @@ static int alloc_bind_mem(AVHWFramesCont
/* In case the implementation prefers/requires dedicated allocation */
use_ded_mem = ded_req.prefersDedicatedAllocation |
ded_req.requiresDedicatedAllocation;
+ if (((VulkanFramesPriv *)hwfc->hwctx)->export_requires_dedicated)
+ use_ded_mem = 1;
if (use_ded_mem)
ded_alloc.image = f->img[img_cnt];
@@ -2854,7 +2871,8 @@ static void try_export_flags(AVHWFramesC
.type = VK_IMAGE_TYPE_2D,
.tiling = hwctx->tiling,
.usage = hwctx->usage,
- .flags = VK_IMAGE_CREATE_ALIAS_BIT,
+ .flags = (hwctx->tiling == VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT && has_mods) ?
+ (hwctx->img_flags) : (VkImageCreateFlags)(VK_IMAGE_CREATE_ALIAS_BIT),
};
nb_mods = has_mods ? drm_mod_info->drmFormatModifierCount : 1;
@@ -2865,9 +2883,18 @@ static void try_export_flags(AVHWFramesC
ret = vk->GetPhysicalDeviceImageFormatProperties2(dev_hwctx->phys_dev,
&pinfo, &props);
+ if (has_mods)
+ av_log(hwfc, AV_LOG_VERBOSE, "GetPhysicalDeviceImageFormatProperties2: mod[%d]=0x%llx -> %s\n",
+ i, (unsigned long long)phy_dev_mod_info.drmFormatModifier,
+ ret == VK_SUCCESS ? "OK" : "FAIL");
if (ret == VK_SUCCESS) {
*iexp |= exp;
*comp_handle_types |= eprops.externalMemoryProperties.compatibleHandleTypes;
+ if (exp == VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT) {
+ VulkanFramesPriv *fp = hwfc->hwctx;
+ fp->export_requires_dedicated = !!(eprops.externalMemoryProperties.externalMemoryFeatures &
+ VK_EXTERNAL_MEMORY_FEATURE_DEDICATED_ONLY_BIT);
+ }
}
}
}
@@ -2895,7 +2922,8 @@ static AVBufferRef *vulkan_pool_alloc(vo
? VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_BIT
: VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT_BIT);
#else
- if (p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_MEMORY)
+ if ((p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_MEMORY) &&
+ (hwctx->tiling != VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT))
try_export_flags(hwfc, &eiinfo.handleTypes, &e,
VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT);
@@ -2914,8 +2942,10 @@ static AVBufferRef *vulkan_pool_alloc(vo
err = create_frame(hwfc, &f, hwctx->tiling, hwctx->usage, hwctx->img_flags,
hwctx->nb_layers,
eiinfo.handleTypes ? &eiinfo : hwctx->create_pnext);
- if (err)
+ if (err) {
+ av_log(hwfc, AV_LOG_ERROR, "vulkan_pool_alloc failed: create_frame failed: %d\n", err);
return NULL;
+ }
err = alloc_bind_mem(hwfc, f, eminfo, sizeof(*eminfo));
if (err)
@@ -2943,6 +2973,7 @@ static AVBufferRef *vulkan_pool_alloc(vo
return avbuf;
fail:
+ av_log(hwfc, AV_LOG_ERROR, "vulkan_pool_alloc failed with error %d\n", err);
vulkan_frame_free(hwfc, f);
return NULL;
}
@@ -3044,7 +3075,14 @@ static int vulkan_frames_init(AVHWFrames
}
/* Lone DPB images do not need additional flags. */
- if (!is_lone_dpb) {
+ /* With DRM modifier + video profile the caller has already chosen a valid
+ * usage/img_flags/chain; do not add usage or img_flags (supported_usage does
+ * not consider the actual modifier or video profile). */
+ int drm_mod_with_video = (hwctx->tiling == VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT &&
+ ff_vk_find_struct(hwctx->create_pnext,
+ VK_STRUCTURE_TYPE_VIDEO_PROFILE_LIST_INFO_KHR));
+
+ if (!is_lone_dpb && !drm_mod_with_video) {
/* Image usage flags */
hwctx->usage |= supported_usage & (VK_IMAGE_USAGE_TRANSFER_DST_BIT |
VK_IMAGE_USAGE_TRANSFER_SRC_BIT |
@@ -3163,8 +3201,8 @@ static int vulkan_frames_init(AVHWFrames
}
vk->GetPhysicalDeviceFormatProperties2(dev_hwctx->phys_dev, fmt->fallback[i], &fmtp);
- for (uint32_t i = 0; i < modp.drmFormatModifierCount; ++i) {
- VkDrmFormatModifierPropertiesEXT *m = &modp.pDrmFormatModifierProperties[i];
+ for (uint32_t j = 0; j < modp.drmFormatModifierCount; ++j) {
+ VkDrmFormatModifierPropertiesEXT *m = &modp.pDrmFormatModifierProperties[j];
if (m->drmFormatModifier == drm_mod.drmFormatModifier) {
mod_props = m;
break;
@@ -3502,7 +3540,7 @@ static int vulkan_map_from_drm_frame_des
&f->flags, &f->mem[i]);
if (err) {
close(idesc.fd);
- return err;
+ goto fail;
}
f->size[i] = req2.memoryRequirements.size;
@@ -3969,7 +4007,7 @@ static int vulkan_export_to_cuda(AVHWFra
return 0;
fail:
- vulkan_free_internal(dst_f);
+ vulkan_free_internal(p, dst_f);
return err;
}
@@ -3980,6 +4018,7 @@ static int vulkan_transfer_data_from_cud
CUcontext dummy;
AVVkFrame *dst_f;
AVVkFrameInternal *dst_int;
+ VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
VulkanFramesPriv *fp = hwfc->hwctx;
const int planes = av_pix_fmt_count_planes(hwfc->sw_format);
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(hwfc->sw_format);
@@ -4058,7 +4097,7 @@ static int vulkan_transfer_data_from_cud
fail:
CHECK_CU(cu->cuCtxPopCurrent(&dummy));
- vulkan_free_internal(dst_f);
+ vulkan_free_internal(p, dst_f);
av_buffer_unref(&dst->buf[0]);
return err;
}
@@ -4128,6 +4167,72 @@ static VkImageAspectFlags plane_index_to
return VK_IMAGE_ASPECT_MEMORY_PLANE_0_BIT_EXT;
}
+#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+static int vulkan_drm_export_sync_fd(AVHWFramesContext *hwfc, AVVkFrame *f,
+ VulkanFramesPriv *fp, int nb_sems)
+{
+ int sync_fd = -1;
+ VkResult ret;
+ VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
+ AVVulkanDeviceContext *hwctx = &p->p;
+ FFVulkanFunctions *vk = &p->vkctx.vkfn;
+
+ if (f->internal->drm_sync_sem == VK_NULL_HANDLE) {
+ VkExportSemaphoreCreateInfo exp_info = {
+ .sType = VK_STRUCTURE_TYPE_EXPORT_SEMAPHORE_CREATE_INFO,
+ .handleTypes = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT,
+ };
+ VkSemaphoreTypeCreateInfo type_info = {
+ .sType = VK_STRUCTURE_TYPE_SEMAPHORE_TYPE_CREATE_INFO,
+ .pNext = &exp_info,
+ .semaphoreType = VK_SEMAPHORE_TYPE_BINARY,
+ };
+ VkSemaphoreCreateInfo sem_create = {
+ .sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO,
+ .pNext = &type_info,
+ };
+ ret = vk->CreateSemaphore(hwctx->act_dev, &sem_create, hwctx->alloc,
+ &f->internal->drm_sync_sem);
+ if (ret != VK_SUCCESS) {
+ av_log(hwctx, AV_LOG_ERROR, "Failed to create DRM export semaphore: %s\n",
+ ff_vk_ret2str(ret));
+ return AVERROR_EXTERNAL;
+ }
+ }
+
+ /* Submit a lightweight exec that waits on the timeline semaphore
+ * (true last operation on the frame) and signals the binary semaphore,
+ * so any Vulkan frame can get a SYNC_FD regardless of origin. */
+ FFVkExecContext *exec = ff_vk_exec_get(&p->vkctx, &fp->compute_exec);
+ if (ff_vk_exec_start(&p->vkctx, exec) >= 0) {
+ for (int i = 0; i < nb_sems; i++)
+ ff_vk_exec_add_dep_wait_sem(&p->vkctx, exec, f->sem[i],
+ f->sem_value[i],
+ VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT);
+ ff_vk_exec_add_dep_bool_sem(&p->vkctx, exec, &f->internal->drm_sync_sem, 1,
+ VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT, 0);
+ if (ff_vk_exec_submit(&p->vkctx, exec) >= 0) {
+ VkSemaphoreGetFdInfoKHR get_fd_info = {
+ .sType = VK_STRUCTURE_TYPE_SEMAPHORE_GET_FD_INFO_KHR,
+ .semaphore = f->internal->drm_sync_sem,
+ .handleType = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT,
+ };
+ ret = vk->GetSemaphoreFdKHR(hwctx->act_dev, &get_fd_info, &sync_fd);
+ if (ret != VK_SUCCESS) {
+ av_log(hwctx, AV_LOG_WARNING,
+ "Failed to get sync fd from DRM map export semaphore: %s\n",
+ ff_vk_ret2str(ret));
+ sync_fd = -1;
+ }
+ } else {
+ ff_vk_exec_discard_deps(&p->vkctx, exec);
+ }
+ }
+
+ return sync_fd;
+}
+#endif
+
static int vulkan_map_to_drm(AVHWFramesContext *hwfc, AVFrame *dst,
const AVFrame *src, int flags)
{
@@ -4138,15 +4243,14 @@ static int vulkan_map_to_drm(AVHWFramesC
AVVulkanDeviceContext *hwctx = &p->p;
FFVulkanFunctions *vk = &p->vkctx.vkfn;
VulkanFramesPriv *fp = hwfc->hwctx;
+ const int planes = av_pix_fmt_count_planes(hwfc->sw_format);
const int nb_images = ff_vk_count_images(f);
VkImageDrmFormatModifierPropertiesEXT drm_mod = {
.sType = VK_STRUCTURE_TYPE_IMAGE_DRM_FORMAT_MODIFIER_PROPERTIES_EXT,
};
- VkSemaphoreWaitInfo wait_info = {
- .sType = VK_STRUCTURE_TYPE_SEMAPHORE_WAIT_INFO,
- .flags = 0x0,
- .semaphoreCount = nb_images,
- };
+ const int nb_sems = nb_images;
+ int free_drm_desc_on_err = 1;
+ int sync_fd = -1;
AVDRMFrameDescriptor *drm_desc = av_mallocz(sizeof(*drm_desc));
if (!drm_desc)
@@ -4156,16 +4260,36 @@ static int vulkan_map_to_drm(AVHWFramesC
if (err < 0)
goto end;
- /* Wait for the operation to finish so we can cleanly export it. */
- wait_info.pSemaphores = f->sem;
- wait_info.pValues = f->sem_value;
+#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+ if ((p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_SEM) &&
+ f->tiling == VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT &&
+ vk->GetSemaphoreFdKHR && vk->CreateSemaphore) {
+ err = vulkan_drm_export_sync_fd(hwfc, f, fp, nb_sems);
+ if (err < 0)
+ goto end;
+ sync_fd = err;
+ err = 0;
+ }
+#endif
- vk->WaitSemaphores(hwctx->act_dev, &wait_info, UINT64_MAX);
+ if (sync_fd < 0) {
+ VkSemaphoreWaitInfo wait_info = {
+ .sType = VK_STRUCTURE_TYPE_SEMAPHORE_WAIT_INFO,
+ .flags = 0x0,
+ .semaphoreCount = nb_sems,
+ .pSemaphores = f->sem,
+ .pValues = f->sem_value,
+ };
+ vk->WaitSemaphores(hwctx->act_dev, &wait_info, UINT64_MAX);
+ }
err = ff_hwframe_map_create(src->hw_frames_ctx, dst, src, &vulkan_unmap_to_drm, drm_desc);
if (err < 0)
goto end;
+ /* It will be freed in ff_hwframe_map_create callback */
+ free_drm_desc_on_err = 0;
+
ret = vk->GetImageDrmFormatModifierPropertiesEXT(hwctx->act_dev, f->img[0],
&drm_mod);
if (ret != VK_SUCCESS) {
@@ -4174,7 +4298,7 @@ static int vulkan_map_to_drm(AVHWFramesC
goto end;
}
- for (int i = 0; (i < nb_images) && (f->mem[i]); i++) {
+ for (int i = 0; (i < planes) && (f->mem[i]); i++) {
VkMemoryGetFdInfoKHR export_info = {
.sType = VK_STRUCTURE_TYPE_MEMORY_GET_FD_INFO_KHR,
.memory = f->mem[i],
@@ -4189,12 +4313,30 @@ static int vulkan_map_to_drm(AVHWFramesC
goto end;
}
+#if HAVE_LINUX_DMA_BUF_H && defined(DMA_BUF_IOCTL_IMPORT_SYNC_FILE)
+ if (sync_fd >= 0) {
+ int dup_fd = dup(sync_fd);
+ if (dup_fd >= 0) {
+ struct dma_buf_import_sync_file import_info = {
+ .flags = DMA_BUF_SYNC_WRITE,
+ .fd = dup_fd,
+ };
+ if (ioctl(drm_desc->objects[i].fd, DMA_BUF_IOCTL_IMPORT_SYNC_FILE, &import_info) < 0)
+ av_log(hwfc, AV_LOG_WARNING, "DMA_BUF_IOCTL_IMPORT_SYNC_FILE failed: %s\n", av_err2str(AVERROR(errno)));
+ close(dup_fd);
+ } else {
+ av_log(hwfc, AV_LOG_WARNING, "dup(sync_fd) failed: %s\n", av_err2str(AVERROR(errno)));
+ }
+ }
+#endif
+
drm_desc->nb_objects++;
drm_desc->objects[i].size = f->size[i];
drm_desc->objects[i].format_modifier = drm_mod.drmFormatModifier;
}
- drm_desc->nb_layers = nb_images;
+ /* NV12 has 2 planes but 1 image/semaphore */
+ drm_desc->nb_layers = FFMAX(planes, nb_images);
for (int i = 0; i < drm_desc->nb_layers; i++) {
VkFormat plane_vkfmt = av_vkfmt_from_pixfmt(hwfc->sw_format)[i];
@@ -4209,13 +4351,14 @@ static int vulkan_map_to_drm(AVHWFramesC
for (int j = 0; j < drm_desc->layers[i].nb_planes; j++) {
VkSubresourceLayout layout;
+ int aspect_plane = (nb_images == 1) ? i : j;
VkImageSubresource sub = {
- .aspectMask = plane_index_to_aspect(j),
+ .aspectMask = plane_index_to_aspect(aspect_plane),
};
drm_desc->layers[i].planes[j].object_index = FFMIN(i, drm_desc->nb_objects - 1);
- vk->GetImageSubresourceLayout(hwctx->act_dev, f->img[i], &sub, &layout);
+ vk->GetImageSubresourceLayout(hwctx->act_dev, f->img[FFMIN(i, nb_images - 1)], &sub, &layout);
drm_desc->layers[i].planes[j].offset = layout.offset;
drm_desc->layers[i].planes[j].pitch = layout.rowPitch;
}
@@ -4235,13 +4378,22 @@ static int vulkan_map_to_drm(AVHWFramesC
dst->width = src->width;
dst->height = src->height;
dst->data[0] = (uint8_t *)drm_desc;
+ dst->hw_frames_ctx = av_buffer_ref(src->hw_frames_ctx);
+
+ if (sync_fd >= 0)
+ close(sync_fd);
av_log(hwfc, AV_LOG_VERBOSE, "Mapped AVVkFrame to a DRM object!\n");
return 0;
end:
- av_free(drm_desc);
+ for (int i = 0; i < drm_desc->nb_objects; i++)
+ close(drm_desc->objects[i].fd);
+ if (free_drm_desc_on_err)
+ av_free(drm_desc);
+ if (sync_fd >= 0)
+ close(sync_fd);
return err;
}
@@ -4751,6 +4903,7 @@ static int vulkan_transfer_data_to_cuda(
CUcontext dummy;
AVVkFrame *dst_f;
AVVkFrameInternal *dst_int;
+ VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
VulkanFramesPriv *fp = hwfc->hwctx;
const int planes = av_pix_fmt_count_planes(hwfc->sw_format);
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(hwfc->sw_format);
@@ -4831,7 +4984,7 @@ static int vulkan_transfer_data_to_cuda(
fail:
CHECK_CU(cu->cuCtxPopCurrent(&dummy));
- vulkan_free_internal(dst_f);
+ vulkan_free_internal(p, dst_f);
av_buffer_unref(&dst->buf[0]);
return err;
}
debian/patches/0078-fix-a-race-condition-in-vaapi-csc-to-yuv420p.patch
Vendored
-58
View File
@@ -1,58 +0,0 @@
Index: FFmpeg/libavutil/hwcontext_vaapi.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_vaapi.c
+++ FFmpeg/libavutil/hwcontext_vaapi.c
@@ -72,6 +72,7 @@ typedef struct VAAPIDevicePriv {
typedef struct VAAPISurfaceFormat {
enum AVPixelFormat pix_fmt;
VAImageFormat image_format;
+ unsigned int fourcc;
} VAAPISurfaceFormat;
typedef struct VAAPIDeviceContext {
@@ -221,15 +222,21 @@ static int vaapi_get_image_format(AVHWDe
VAImageFormat **image_format)
{
VAAPIDeviceContext *ctx = hwdev->hwctx;
+ const VAAPIFormatDescriptor *desc;
int i;
+ desc = vaapi_format_from_pix_fmt(pix_fmt);
+ if (!desc || !image_format)
+ goto fail;
+
for (i = 0; i < ctx->nb_formats; i++) {
- if (ctx->formats[i].pix_fmt == pix_fmt) {
- if (image_format)
- *image_format = &ctx->formats[i].image_format;
+ if (ctx->formats[i].fourcc == desc->fourcc) {
+ *image_format = &ctx->formats[i].image_format;
return 0;
}
}
+
+fail:
return AVERROR(ENOSYS);
}
@@ -438,6 +445,7 @@ static int vaapi_device_init(AVHWDeviceC
av_log(hwdev, AV_LOG_DEBUG, "Format %#x -> %s.\n",
fourcc, av_get_pix_fmt_name(pix_fmt));
ctx->formats[ctx->nb_formats].pix_fmt = pix_fmt;
+ ctx->formats[ctx->nb_formats].fourcc = fourcc;
ctx->formats[ctx->nb_formats].image_format = image_list[i];
++ctx->nb_formats;
}
@@ -1014,12 +1022,6 @@ static int vaapi_map_to_memory(AVHWFrame
{
int err;
- if (dst->format != AV_PIX_FMT_NONE) {
- err = vaapi_get_image_format(hwfc->device_ctx, dst->format, NULL);
- if (err < 0)
- return err;
- }
-
err = vaapi_map_frame(hwfc, dst, src, flags);
if (err)
return err;
debian/patches/0079-add-fixes-for-vaapi-drm-prime-vulkan-interop.patch
Vendored
+769
View File
@@ -0,0 +1,769 @@
Index: FFmpeg/libavutil/hwcontext_vulkan.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_vulkan.c
+++ FFmpeg/libavutil/hwcontext_vulkan.c
@@ -59,11 +59,47 @@
#include <xf86drm.h>
#include <drm_fourcc.h>
#include "hwcontext_drm.h"
+#ifndef DRM_FORMAT_P010
+#define DRM_FORMAT_P010 fourcc_code('P', '0', '1', '0')
+#define DRM_FORMAT_P012 fourcc_code('P', '0', '1', '2')
+#define DRM_FORMAT_P016 fourcc_code('P', '0', '1', '6')
+#endif
+#ifndef DRM_FORMAT_P210
+#define DRM_FORMAT_P210 fourcc_code('P', '2', '1', '0')
+#endif
+#ifndef DRM_FORMAT_S010
+#define DRM_FORMAT_S010 fourcc_code('S', '0', '1', '0')
+#define DRM_FORMAT_S210 fourcc_code('S', '2', '1', '0')
+#define DRM_FORMAT_S410 fourcc_code('S', '4', '1', '0')
+#define DRM_FORMAT_S012 fourcc_code('S', '0', '1', '2')
+#define DRM_FORMAT_S212 fourcc_code('S', '2', '1', '2')
+#define DRM_FORMAT_S412 fourcc_code('S', '4', '1', '2')
+#define DRM_FORMAT_S016 fourcc_code('S', '0', '1', '6')
+#define DRM_FORMAT_S216 fourcc_code('S', '2', '1', '6')
+#define DRM_FORMAT_S416 fourcc_code('S', '4', '1', '6')
+#endif
#endif
#if HAVE_LINUX_DMA_BUF_H
#include <sys/ioctl.h>
#include <linux/dma-buf.h>
+/* dma-buf export/import sync is only available in v6.0+ */
+#ifndef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+struct dma_buf_export_sync_file {
+ uint32_t flags;
+ int32_t fd;
+};
+#define DMA_BUF_IOCTL_EXPORT_SYNC_FILE \
+ _IOWR(DMA_BUF_BASE, 2, struct dma_buf_export_sync_file)
+#endif
+#ifndef DMA_BUF_IOCTL_IMPORT_SYNC_FILE
+struct dma_buf_import_sync_file {
+ uint32_t flags;
+ int32_t fd;
+};
+#define DMA_BUF_IOCTL_IMPORT_SYNC_FILE \
+ _IOW(DMA_BUF_BASE, 3, struct dma_buf_import_sync_file)
+#endif
#endif
#if CONFIG_CUDA
@@ -167,6 +203,11 @@ typedef struct VulkanDevicePriv {
/* Maximum queues */
int limit_queues;
+
+ /* DRM dma-buf sync file ioctl ret code, tested on first use.
+ * 0: success, other: fail (use fallback) */
+ int drm_export_sync_file_ret;
+ int drm_import_sync_file_ret;
} VulkanDevicePriv;
typedef struct VulkanFramesPriv {
@@ -1794,6 +1835,7 @@ static void vulkan_device_uninit(AVHWDev
static int vulkan_device_create_internal(AVHWDeviceContext *ctx,
VulkanDeviceSelection *dev_select,
+ int use_linear_images,
int disable_multiplane,
AVDictionary *opts, int flags)
{
@@ -1866,10 +1908,14 @@ static int vulkan_device_create_internal
goto end;
}
- /* Tiled images setting, use them by default */
- opt_d = av_dict_get(opts, "linear_images", NULL, 0);
- if (opt_d)
- p->use_linear_images = strtol(opt_d->value, NULL, 10);
+ /* Tiled images setting, use them by default.
+ * The use_linear_images argument takes precedent over the option */
+ p->use_linear_images = use_linear_images;
+ if (!p->use_linear_images) {
+ opt_d = av_dict_get(opts, "linear_images", NULL, 0);
+ if (opt_d)
+ p->use_linear_images = strtol(opt_d->value, NULL, 10);
+ }
/* The disable_multiplane argument takes precedent over the option */
p->disable_multiplane = disable_multiplane;
@@ -2167,7 +2213,7 @@ static int vulkan_device_create(AVHWDevi
}
}
- return vulkan_device_create_internal(ctx, &dev_select, 0, opts, flags);
+ return vulkan_device_create_internal(ctx, &dev_select, 0, 0, opts, flags);
}
static int vulkan_device_derive(AVHWDeviceContext *ctx,
@@ -2209,7 +2255,8 @@ static int vulkan_device_derive(AVHWDevi
dev_select.vendor_id = 0x1002;
}
- return vulkan_device_create_internal(ctx, &dev_select, 0, opts, flags);
+ /* Linear multiplane image is used to ensure compatibility with VAAPI */
+ return vulkan_device_create_internal(ctx, &dev_select, 1, 0, opts, flags);
}
#endif
#if CONFIG_LIBDRM
@@ -2242,7 +2289,14 @@ static int vulkan_device_derive(AVHWDevi
drmFreeDevice(&drm_dev_info);
- return vulkan_device_create_internal(ctx, &dev_select, 0, opts, flags);
+ /* Linear multiplane image is used to ensure compatibility with DRM */
+ return vulkan_device_create_internal(ctx, &dev_select, 1, 0, opts, flags);
+ }
+#endif
+#if CONFIG_RKMPP
+ case AV_HWDEVICE_TYPE_RKMPP: {
+ /* Linear multiplane image is used to ensure compatibility with RKMPP */
+ return vulkan_device_create_internal(ctx, &dev_select, 1, 0, opts, flags);
}
#endif
#if CONFIG_CUDA
@@ -2265,7 +2319,7 @@ static int vulkan_device_derive(AVHWDevi
* CUDA is not able to import multiplane images, so always derive a
* Vulkan device with multiplane disabled.
*/
- return vulkan_device_create_internal(ctx, &dev_select, 1, opts, flags);
+ return vulkan_device_create_internal(ctx, &dev_select, 0, 1, opts, flags);
}
#endif
default:
@@ -2874,6 +2928,8 @@ static void try_export_flags(AVHWFramesC
.flags = (hwctx->tiling == VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT && has_mods) ?
(hwctx->img_flags) : (VkImageCreateFlags)(VK_IMAGE_CREATE_ALIAS_BIT),
};
+ VkPhysicalDeviceImageFormatInfo2 pinfo_fallback = pinfo;
+ pinfo_fallback.format = av_vkfmt_from_pixfmt(hwfc->sw_format)[0];
nb_mods = has_mods ? drm_mod_info->drmFormatModifierCount : 1;
for (int i = 0; i < nb_mods; i++) {
@@ -2881,7 +2937,10 @@ static void try_export_flags(AVHWFramesC
phy_dev_mod_info.drmFormatModifier = drm_mod_info->pDrmFormatModifiers[i];
ret = vk->GetPhysicalDeviceImageFormatProperties2(dev_hwctx->phys_dev,
- &pinfo, &props);
+ &pinfo, &props);
+ if (ret != VK_SUCCESS)
+ ret = vk->GetPhysicalDeviceImageFormatProperties2(dev_hwctx->phys_dev,
+ &pinfo_fallback, &props);
if (has_mods)
av_log(hwfc, AV_LOG_VERBOSE, "GetPhysicalDeviceImageFormatProperties2: mod[%d]=0x%llx -> %s\n",
@@ -3285,28 +3344,65 @@ static void vulkan_unmap_from_drm(AVHWFr
static const struct {
uint32_t drm_fourcc;
VkFormat vk_format;
+ int nb_planes;
+ uint32_t fallbacks[AV_DRM_MAX_PLANES];
} vulkan_drm_format_map[] = {
- { DRM_FORMAT_R8, VK_FORMAT_R8_UNORM },
- { DRM_FORMAT_R16, VK_FORMAT_R16_UNORM },
- { DRM_FORMAT_GR88, VK_FORMAT_R8G8_UNORM },
- { DRM_FORMAT_RG88, VK_FORMAT_R8G8_UNORM },
- { DRM_FORMAT_GR1616, VK_FORMAT_R16G16_UNORM },
- { DRM_FORMAT_RG1616, VK_FORMAT_R16G16_UNORM },
- { DRM_FORMAT_ARGB8888, VK_FORMAT_B8G8R8A8_UNORM },
- { DRM_FORMAT_XRGB8888, VK_FORMAT_B8G8R8A8_UNORM },
- { DRM_FORMAT_ABGR8888, VK_FORMAT_R8G8B8A8_UNORM },
- { DRM_FORMAT_XBGR8888, VK_FORMAT_R8G8B8A8_UNORM },
- { DRM_FORMAT_ARGB2101010, VK_FORMAT_A2B10G10R10_UNORM_PACK32 },
- { DRM_FORMAT_ABGR2101010, VK_FORMAT_A2R10G10B10_UNORM_PACK32 },
- { DRM_FORMAT_XRGB2101010, VK_FORMAT_A2B10G10R10_UNORM_PACK32 },
- { DRM_FORMAT_XBGR2101010, VK_FORMAT_A2R10G10B10_UNORM_PACK32 },
+ { DRM_FORMAT_R8, VK_FORMAT_R8_UNORM, 1, { DRM_FORMAT_R8 } },
+ { DRM_FORMAT_R16, VK_FORMAT_R16_UNORM, 1, { DRM_FORMAT_R16 } },
+ { DRM_FORMAT_GR88, VK_FORMAT_R8G8_UNORM, 1, { DRM_FORMAT_GR88 } },
+ { DRM_FORMAT_RG88, VK_FORMAT_R8G8_UNORM, 1, { DRM_FORMAT_RG88 } },
+ { DRM_FORMAT_GR1616, VK_FORMAT_R16G16_UNORM, 1, { DRM_FORMAT_GR1616 } },
+ { DRM_FORMAT_RG1616, VK_FORMAT_R16G16_UNORM, 1, { DRM_FORMAT_RG1616 } },
+ { DRM_FORMAT_ARGB8888, VK_FORMAT_B8G8R8A8_UNORM, 1, { DRM_FORMAT_ARGB8888 } },
+ { DRM_FORMAT_XRGB8888, VK_FORMAT_B8G8R8A8_UNORM, 1, { DRM_FORMAT_XRGB8888 } },
+ { DRM_FORMAT_ABGR8888, VK_FORMAT_R8G8B8A8_UNORM, 1, { DRM_FORMAT_ABGR8888 } },
+ { DRM_FORMAT_XBGR8888, VK_FORMAT_R8G8B8A8_UNORM, 1, { DRM_FORMAT_XBGR8888 } },
+ { DRM_FORMAT_ARGB2101010, VK_FORMAT_A2B10G10R10_UNORM_PACK32, 1, { DRM_FORMAT_ARGB2101010 } },
+ { DRM_FORMAT_ABGR2101010, VK_FORMAT_A2R10G10B10_UNORM_PACK32, 1, { DRM_FORMAT_ABGR2101010 } },
+ { DRM_FORMAT_XRGB2101010, VK_FORMAT_A2B10G10R10_UNORM_PACK32, 1, { DRM_FORMAT_XRGB2101010 } },
+ { DRM_FORMAT_XBGR2101010, VK_FORMAT_A2R10G10B10_UNORM_PACK32, 1, { DRM_FORMAT_XBGR2101010 } },
// All these DRM_FORMATs were added in the same libdrm commit.
#ifdef DRM_FORMAT_XYUV8888
- { DRM_FORMAT_XYUV8888, VK_FORMAT_R8G8B8A8_UNORM },
- { DRM_FORMAT_XVYU2101010, VK_FORMAT_A2R10G10B10_UNORM_PACK32 } ,
- { DRM_FORMAT_XVYU12_16161616, VK_FORMAT_R12X4G12X4B12X4A12X4_UNORM_4PACK16 } ,
- { DRM_FORMAT_XVYU16161616, VK_FORMAT_R16G16B16A16_UNORM } ,
+ { DRM_FORMAT_XYUV8888, VK_FORMAT_R8G8B8A8_UNORM, 1, { DRM_FORMAT_XYUV8888 } },
+ { DRM_FORMAT_XVYU2101010, VK_FORMAT_A2R10G10B10_UNORM_PACK32, 1, { DRM_FORMAT_XVYU2101010 } },
+ { DRM_FORMAT_XVYU12_16161616, VK_FORMAT_R12X4G12X4B12X4A12X4_UNORM_4PACK16, 1, { DRM_FORMAT_XVYU12_16161616 } },
+ { DRM_FORMAT_XVYU16161616, VK_FORMAT_R16G16B16A16_UNORM, 1, { DRM_FORMAT_XVYU16161616 } },
+#endif
+
+ { DRM_FORMAT_RGB888, VK_FORMAT_R8G8B8_UNORM, 1, { DRM_FORMAT_RGB888 } },
+ { DRM_FORMAT_BGR888, VK_FORMAT_B8G8R8_UNORM, 1, { DRM_FORMAT_BGR888 } },
+
+ { DRM_FORMAT_NV12, VK_FORMAT_G8_B8R8_2PLANE_420_UNORM, 2, { DRM_FORMAT_R8, DRM_FORMAT_GR88 } },
+ { DRM_FORMAT_NV16, VK_FORMAT_G8_B8R8_2PLANE_422_UNORM, 2, { DRM_FORMAT_R8, DRM_FORMAT_GR88 } },
+ { DRM_FORMAT_NV24, VK_FORMAT_G8_B8R8_2PLANE_444_UNORM, 2, { DRM_FORMAT_R8, DRM_FORMAT_GR88 } },
+
+ // All these DRM_FORMATs were added in the same libdrm commit.
+#ifdef DRM_FORMAT_P010
+ { DRM_FORMAT_P010, VK_FORMAT_G10X6_B10X6R10X6_2PLANE_420_UNORM_3PACK16, 2, { DRM_FORMAT_R16, DRM_FORMAT_GR1616 } },
+ { DRM_FORMAT_P012, VK_FORMAT_G12X4_B12X4R12X4_2PLANE_420_UNORM_3PACK16, 2, { DRM_FORMAT_R16, DRM_FORMAT_GR1616 } },
+ { DRM_FORMAT_P016, VK_FORMAT_G16_B16R16_2PLANE_420_UNORM, 2, { DRM_FORMAT_R16, DRM_FORMAT_GR1616 } },
+#endif
+
+#ifdef DRM_FORMAT_P210
+ { DRM_FORMAT_P210, VK_FORMAT_G10X6_B10X6R10X6_2PLANE_422_UNORM_3PACK16, 2, { DRM_FORMAT_R16, DRM_FORMAT_GR1616 } },
+#endif
+
+ { DRM_FORMAT_YUV420, VK_FORMAT_G8_B8_R8_3PLANE_420_UNORM, 3, { DRM_FORMAT_R8, DRM_FORMAT_R8, DRM_FORMAT_R8 } },
+ { DRM_FORMAT_YUV422, VK_FORMAT_G8_B8_R8_3PLANE_422_UNORM, 3, { DRM_FORMAT_R8, DRM_FORMAT_R8, DRM_FORMAT_R8 } },
+ { DRM_FORMAT_YUV444, VK_FORMAT_G8_B8_R8_3PLANE_444_UNORM, 3, { DRM_FORMAT_R8, DRM_FORMAT_R8, DRM_FORMAT_R8 } },
+
+ // All these DRM_FORMATs were added in the same libdrm commit.
+#ifdef DRM_FORMAT_S010
+ { DRM_FORMAT_S010, VK_FORMAT_G16_B16_R16_3PLANE_420_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
+ { DRM_FORMAT_S012, VK_FORMAT_G16_B16_R16_3PLANE_420_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
+ { DRM_FORMAT_S016, VK_FORMAT_G16_B16_R16_3PLANE_420_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
+ { DRM_FORMAT_S210, VK_FORMAT_G16_B16_R16_3PLANE_422_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
+ { DRM_FORMAT_S212, VK_FORMAT_G16_B16_R16_3PLANE_422_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
+ { DRM_FORMAT_S216, VK_FORMAT_G16_B16_R16_3PLANE_422_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
+ { DRM_FORMAT_S410, VK_FORMAT_G16_B16_R16_3PLANE_444_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
+ { DRM_FORMAT_S412, VK_FORMAT_G16_B16_R16_3PLANE_444_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
+ { DRM_FORMAT_S416, VK_FORMAT_G16_B16_R16_3PLANE_444_UNORM, 3, { DRM_FORMAT_R16, DRM_FORMAT_R16, DRM_FORMAT_R16 } },
#endif
};
@@ -3318,6 +3414,51 @@ static inline VkFormat drm_to_vulkan_fmt
return VK_FORMAT_UNDEFINED;
}
+static inline int decompose_drm_frame_desc(AVDRMFrameDescriptor *dst,
+ const AVDRMFrameDescriptor *src)
+{
+ int map_idx = -1;
+ int p = 0;
+
+ if (!src || !dst)
+ return AVERROR(EINVAL);
+
+ for (int i = 0; i < FF_ARRAY_ELEMS(vulkan_drm_format_map); i++) {
+ if (vulkan_drm_format_map[i].drm_fourcc == src->layers[0].format) {
+ map_idx = i;
+ break;
+ }
+ }
+ if (map_idx == -1)
+ return AVERROR(ENOSYS);
+ if (vulkan_drm_format_map[map_idx].nb_planes == 1 &&
+ vulkan_drm_format_map[map_idx].fallbacks[0] == src->layers[0].format) {
+ return AVERROR(ENOSYS);
+ }
+
+ dst->nb_objects = src->nb_objects;
+ memcpy(dst->objects, src->objects, sizeof(src->objects));
+
+ for (int i = 0; i < src->nb_layers; i++) {
+ for (int j = 0; j < src->layers[i].nb_planes; j++) {
+ if (p >= vulkan_drm_format_map[map_idx].nb_planes)
+ goto end;
+
+ dst->layers[p].format = vulkan_drm_format_map[map_idx].fallbacks[p];
+ dst->layers[p].nb_planes = 1;
+ dst->layers[p].planes[0] = src->layers[i].planes[j];
+ p++;
+ }
+ }
+
+end:
+ dst->nb_layers = p;
+ return 0;
+}
+
+static VkImageAspectFlags plane_index_to_aspect(int plane);
+static VkImageAspectFlags plane_index_to_aspect_linear(int plane, int nb_images, int nb_planes);
+
static int vulkan_map_from_drm_frame_desc(AVHWFramesContext *hwfc, AVVkFrame **frame,
const AVFrame *src, int flags)
{
@@ -3329,10 +3470,15 @@ static int vulkan_map_from_drm_frame_des
VulkanDevicePriv *p = ctx->hwctx;
AVVulkanDeviceContext *hwctx = &p->p;
FFVulkanFunctions *vk = &p->vkctx.vkfn;
- const AVDRMFrameDescriptor *desc = (AVDRMFrameDescriptor *)src->data[0];
+ AVDRMFrameDescriptor desc_decomposed = { 0 };
+ AVDRMFrameDescriptor *desc_src = (AVDRMFrameDescriptor *)src->data[0];
+ AVDRMFrameDescriptor *desc = desc_src;
VkBindImageMemoryInfo bind_info[AV_DRM_MAX_PLANES];
VkBindImagePlaneMemoryInfo plane_info[AV_DRM_MAX_PLANES];
+ const int has_modifiers = !!(p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS);
+ int retry = 0;
+start:
for (int i = 0; i < desc->nb_layers; i++) {
if (drm_to_vulkan_fmt(desc->layers[i].format) == VK_FORMAT_UNDEFINED) {
av_log(ctx, AV_LOG_ERROR, "Unsupported DMABUF layer format %#08x!\n",
@@ -3341,13 +3487,21 @@ static int vulkan_map_from_drm_frame_des
}
}
+ if (!has_modifiers &&
+ desc->objects[0].format_modifier != DRM_FORMAT_MOD_LINEAR &&
+ desc->objects[0].format_modifier != DRM_FORMAT_MOD_INVALID) {
+ av_log(ctx, AV_LOG_ERROR, "Cannot map non-linear tiling image from DRM!\n");
+ err = AVERROR_EXTERNAL;
+ goto fail;
+ }
+
if (!(f = av_vk_frame_alloc())) {
av_log(ctx, AV_LOG_ERROR, "Unable to allocate memory for AVVkFrame!\n");
err = AVERROR(ENOMEM);
goto fail;
}
- f->tiling = VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT;
+ f->tiling = has_modifiers ? VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT : VK_IMAGE_TILING_LINEAR;
for (int i = 0; i < desc->nb_layers; i++) {
const int planes = desc->layers[i].nb_planes;
@@ -3385,7 +3539,7 @@ static int vulkan_map_from_drm_frame_des
.mipLevels = 1,
.arrayLayers = 1,
.flags = 0x0,
- .tiling = VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT,
+ .tiling = f->tiling,
.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED, /* specs say so */
.usage = 0x0, /* filled in below */
.samples = VK_SAMPLE_COUNT_1_BIT,
@@ -3426,7 +3580,7 @@ static int vulkan_map_from_drm_frame_des
fmt_props = (VkPhysicalDeviceImageFormatInfo2) {
.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_IMAGE_FORMAT_INFO_2,
- .pNext = &props_ext,
+ .pNext = has_modifiers ? &props_ext : NULL,
.format = create_info.format,
.type = create_info.imageType,
.tiling = create_info.tiling,
@@ -3438,6 +3592,13 @@ static int vulkan_map_from_drm_frame_des
ret = vk->GetPhysicalDeviceImageFormatProperties2(hwctx->phys_dev,
&fmt_props, &props_ret);
if (ret != VK_SUCCESS) {
+ /* Decompose the AVDRMFrameDescriptor and retry */
+ if (!retry && !decompose_drm_frame_desc(&desc_decomposed, desc_src)) {
+ retry++;
+ desc = &desc_decomposed;
+ vulkan_frame_free(hwfc, f);
+ goto start;
+ }
av_log(ctx, AV_LOG_ERROR, "Cannot map DRM frame to Vulkan: %s\n",
ff_vk_ret2str(ret));
err = AVERROR_EXTERNAL;
@@ -3549,9 +3710,12 @@ static int vulkan_map_from_drm_frame_des
for (int i = 0; i < desc->nb_layers; i++) {
const int planes = desc->layers[i].nb_planes;
for (int j = 0; j < planes; j++) {
- VkImageAspectFlagBits aspect = j == 0 ? VK_IMAGE_ASPECT_MEMORY_PLANE_0_BIT_EXT :
- j == 1 ? VK_IMAGE_ASPECT_MEMORY_PLANE_1_BIT_EXT :
- VK_IMAGE_ASPECT_MEMORY_PLANE_2_BIT_EXT;
+ VkImageAspectFlagBits aspect;
+
+ if (has_modifiers)
+ aspect = plane_index_to_aspect(j);
+ else
+ aspect = plane_index_to_aspect_linear(j, desc->nb_layers, planes);
plane_info[bind_counts].sType = VK_STRUCTURE_TYPE_BIND_IMAGE_PLANE_MEMORY_INFO;
plane_info[bind_counts].pNext = NULL;
@@ -3592,15 +3756,20 @@ static int vulkan_map_from_drm_frame_syn
const AVDRMFrameDescriptor *desc, int flags)
{
int err;
+#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
VkResult ret;
+#endif
AVHWDeviceContext *ctx = hwfc->device_ctx;
VulkanDevicePriv *p = ctx->hwctx;
VulkanFramesPriv *fp = hwfc->hwctx;
AVVulkanDeviceContext *hwctx = &p->p;
+#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
FFVulkanFunctions *vk = &p->vkctx.vkfn;
+#endif
#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
- if (p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_SEM) {
+ if ((p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_SEM) &&
+ !p->drm_export_sync_file_ret) {
VkCommandBuffer cmd_buf;
FFVkExecContext *exec;
VkImageMemoryBarrier2 img_bar[AV_NUM_DATA_POINTERS];
@@ -3618,18 +3787,19 @@ static int vulkan_map_from_drm_frame_syn
};
VkImportSemaphoreFdInfoKHR import_info;
struct dma_buf_export_sync_file implicit_fd_info = {
- .flags = DMA_BUF_SYNC_READ,
+ .flags = DMA_BUF_SYNC_RW,
.fd = -1,
};
- if (ioctl(desc->objects[i].fd, DMA_BUF_IOCTL_EXPORT_SYNC_FILE,
- &implicit_fd_info)) {
+ if (p->drm_export_sync_file_ret = ioctl(desc->objects[i].fd,
+ DMA_BUF_IOCTL_EXPORT_SYNC_FILE,
+ &implicit_fd_info)) {
err = AVERROR(errno);
- av_log(hwctx, AV_LOG_ERROR, "Failed to retrieve implicit DRM sync file: %s\n",
- av_err2str(err));
+ av_log(hwctx, AV_LOG_WARNING, "Failed to retrieve implicit DRM sync file: %s, "
+ "disabling for future frames\n", av_err2str(err));
for (; i >= 0; i--)
vk->DestroySemaphore(hwctx->act_dev, drm_sync_sem[i], hwctx->alloc);
- return err;
+ goto fallback;
}
ret = vk->CreateSemaphore(hwctx->act_dev, &sem_spawn,
@@ -3638,6 +3808,7 @@ static int vulkan_map_from_drm_frame_syn
av_log(hwctx, AV_LOG_ERROR, "Failed to create semaphore: %s\n",
ff_vk_ret2str(ret));
err = AVERROR_EXTERNAL;
+ close(implicit_fd_info.fd);
for (; i >= 0; i--)
vk->DestroySemaphore(hwctx->act_dev, drm_sync_sem[i], hwctx->alloc);
return err;
@@ -3656,6 +3827,7 @@ static int vulkan_map_from_drm_frame_syn
av_log(hwctx, AV_LOG_ERROR, "Failed to import semaphore: %s\n",
ff_vk_ret2str(ret));
err = AVERROR_EXTERNAL;
+ close(implicit_fd_info.fd);
for (; i >= 0; i--)
vk->DestroySemaphore(hwctx->act_dev, drm_sync_sem[i], hwctx->alloc);
return err;
@@ -3672,13 +3844,13 @@ static int vulkan_map_from_drm_frame_syn
drm_sync_sem, desc->nb_objects,
VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT, 1);
if (err < 0)
- return err;
+ goto fail;
err = ff_vk_exec_add_dep_frame(&p->vkctx, exec, dst,
VK_PIPELINE_STAGE_2_NONE,
VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT);
if (err < 0)
- return err;
+ goto fail;
ff_vk_frame_barrier(&p->vkctx, exec, dst, img_bar, &nb_img_bar,
VK_PIPELINE_STAGE_2_NONE,
@@ -3698,14 +3870,22 @@ static int vulkan_map_from_drm_frame_syn
err = ff_vk_exec_submit(&p->vkctx, exec);
if (err < 0)
- return err;
+ goto fail;
+
+ return 0;
+fail:
+ ff_vk_exec_discard_deps(&p->vkctx, exec);
+ return err;
} else
#endif
{
- AVVkFrame *f = (AVVkFrame *)dst->data[0];
- av_log(hwctx, AV_LOG_WARNING, "No support for synchronization when importing DMA-BUFs, "
- "image may be corrupted.\n");
- err = prepare_frame(hwfc, &fp->compute_exec, f, PREP_MODE_EXTERNAL_IMPORT);
+#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+fallback:
+#endif
+ av_log(hwctx, AV_LOG_DEBUG, "No support for synchronization when importing DMA-BUFs, "
+ "image may be corrupted.\n");
+ err = prepare_frame(hwfc, &fp->compute_exec,
+ (AVVkFrame *)dst->data[0], PREP_MODE_EXTERNAL_IMPORT);
if (err)
return err;
}
@@ -3742,7 +3922,7 @@ static int vulkan_map_from_drm(AVHWFrame
return 0;
fail:
- vulkan_frame_free(hwfc->device_ctx->hwctx, f);
+ vulkan_frame_free(hwfc, f);
dst->data[0] = NULL;
return err;
}
@@ -4112,16 +4292,10 @@ static int vulkan_map_to(AVHWFramesConte
#if CONFIG_LIBDRM
#if CONFIG_VAAPI
case AV_PIX_FMT_VAAPI:
- if (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS)
- return vulkan_map_from_vaapi(hwfc, dst, src, flags);
- else
- return AVERROR(ENOSYS);
+ return vulkan_map_from_vaapi(hwfc, dst, src, flags);
#endif
case AV_PIX_FMT_DRM_PRIME:
- if (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS)
- return vulkan_map_from_drm(hwfc, dst, src, flags);
- else
- return AVERROR(ENOSYS);
+ return vulkan_map_from_drm(hwfc, dst, src, flags);
#endif
default:
return AVERROR(ENOSYS);
@@ -4137,10 +4311,51 @@ typedef struct VulkanDRMMapping {
static void vulkan_unmap_to_drm(AVHWFramesContext *hwfc, HWMapDescriptor *hwmap)
{
AVDRMFrameDescriptor *drm_desc = hwmap->priv;
+#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+ AVVkFrame *f = (AVVkFrame *)hwmap->source->data[0];
+ AVHWDeviceContext *ctx = hwfc->device_ctx;
+ VulkanDevicePriv *p = ctx->hwctx;
+ AVVulkanDeviceContext *hwctx = &p->p;
+ FFVulkanFunctions *vk = &p->vkctx.vkfn;
+#endif
+
+ /* On unmap from DRM, import any fences the external consumer attached
+ * back into the original Vulkan frame's semaphore, so Vulkan waits on
+ * them before reusing the frame. We don't know if the consumer did reads
+ * or writes, so export both. */
+#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
+ if ((p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_SEM) &&
+ !p->drm_export_sync_file_ret) {
+ for (int i = 0; i < drm_desc->nb_objects; i++) {
+ struct dma_buf_export_sync_file implicit_fd_info = {
+ .flags = DMA_BUF_SYNC_RW,
+ .fd = -1,
+ };
- /* on unmap from DRM, make sure to import sync objects so that we are sync'd with any work that was
- * done on the buffer while exported. We don't know if who used the dmabuf did reads or writes, so protect against both */
- vulkan_map_from_drm_frame_sync(hwfc, hwmap->source, drm_desc, AV_HWFRAME_MAP_READ | AV_HWFRAME_MAP_WRITE);
+ if (drm_desc->objects[i].fd < 0)
+ break;
+
+ if (p->drm_export_sync_file_ret = ioctl(drm_desc->objects[i].fd,
+ DMA_BUF_IOCTL_EXPORT_SYNC_FILE,
+ &implicit_fd_info)) {
+ av_log(hwfc, AV_LOG_WARNING, "DMA_BUF_IOCTL_EXPORT_SYNC_FILE failed: %s, "
+ "disabling for future frames\n", av_err2str(AVERROR(errno)));
+ break;
+ }
+
+ VkImportSemaphoreFdInfoKHR import_info = {
+ .sType = VK_STRUCTURE_TYPE_IMPORT_SEMAPHORE_FD_INFO_KHR,
+ .semaphore = f->sem[i],
+ .handleType = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT,
+ .flags = VK_SEMAPHORE_IMPORT_TEMPORARY_BIT,
+ .fd = implicit_fd_info.fd,
+ };
+
+ if (vk->ImportSemaphoreFdKHR(hwctx->act_dev, &import_info) != VK_SUCCESS)
+ close(implicit_fd_info.fd);
+ }
+ }
+#endif
for (int i = 0; i < drm_desc->nb_objects; i++)
close(drm_desc->objects[i].fd);
@@ -4167,6 +4382,20 @@ static VkImageAspectFlags plane_index_to
return VK_IMAGE_ASPECT_MEMORY_PLANE_0_BIT_EXT;
}
+static VkImageAspectFlags plane_index_to_aspect_linear(int plane, int nb_images, int nb_planes)
+{
+ if (nb_images == 1 && nb_planes > 1) {
+ if (plane == 0) return VK_IMAGE_ASPECT_PLANE_0_BIT;
+ if (plane == 1) return VK_IMAGE_ASPECT_PLANE_1_BIT;
+ if (plane == 2) return VK_IMAGE_ASPECT_PLANE_2_BIT;
+
+ av_assert2(0 && "Invalid plane index");
+ return VK_IMAGE_ASPECT_PLANE_0_BIT;
+ }
+
+ return VK_IMAGE_ASPECT_COLOR_BIT;
+}
+
#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
static int vulkan_drm_export_sync_fd(AVHWFramesContext *hwfc, AVVkFrame *f,
VulkanFramesPriv *fp, int nb_sems)
@@ -4245,14 +4474,22 @@ static int vulkan_map_to_drm(AVHWFramesC
VulkanFramesPriv *fp = hwfc->hwctx;
const int planes = av_pix_fmt_count_planes(hwfc->sw_format);
const int nb_images = ff_vk_count_images(f);
+ const int is_modifier_tiling = (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS) &&
+ f->tiling == VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT;
VkImageDrmFormatModifierPropertiesEXT drm_mod = {
.sType = VK_STRUCTURE_TYPE_IMAGE_DRM_FORMAT_MODIFIER_PROPERTIES_EXT,
};
const int nb_sems = nb_images;
int free_drm_desc_on_err = 1;
int sync_fd = -1;
+ AVDRMFrameDescriptor *drm_desc = NULL;
- AVDRMFrameDescriptor *drm_desc = av_mallocz(sizeof(*drm_desc));
+ if (f->tiling == VK_IMAGE_TILING_OPTIMAL) {
+ av_log(hwfc, AV_LOG_ERROR, "Cannot map optimal tiling image to DRM!\n");
+ return AVERROR(EINVAL);
+ }
+
+ drm_desc = av_mallocz(sizeof(*drm_desc));
if (!drm_desc)
return AVERROR(ENOMEM);
@@ -4261,8 +4498,9 @@ static int vulkan_map_to_drm(AVHWFramesC
goto end;
#ifdef DMA_BUF_IOCTL_EXPORT_SYNC_FILE
- if ((p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_SEM) &&
- f->tiling == VK_IMAGE_TILING_DRM_FORMAT_MODIFIER_EXT &&
+ // Known to be not working on AMD GPU with VAAPI, disable it for now
+ if (0 && (p->vkctx.extensions & FF_VK_EXT_EXTERNAL_FD_SEM) &&
+ !p->drm_import_sync_file_ret &&
vk->GetSemaphoreFdKHR && vk->CreateSemaphore) {
err = vulkan_drm_export_sync_fd(hwfc, f, fp, nb_sems);
if (err < 0)
@@ -4290,13 +4528,16 @@ static int vulkan_map_to_drm(AVHWFramesC
/* It will be freed in ff_hwframe_map_create callback */
free_drm_desc_on_err = 0;
- ret = vk->GetImageDrmFormatModifierPropertiesEXT(hwctx->act_dev, f->img[0],
- &drm_mod);
- if (ret != VK_SUCCESS) {
- av_log(hwfc, AV_LOG_ERROR, "Failed to retrieve DRM format modifier!\n");
- err = AVERROR_EXTERNAL;
- goto end;
- }
+ if (is_modifier_tiling) {
+ ret = vk->GetImageDrmFormatModifierPropertiesEXT(hwctx->act_dev, f->img[0],
+ &drm_mod);
+ if (ret != VK_SUCCESS) {
+ av_log(hwfc, AV_LOG_ERROR, "Failed to retrieve DRM format modifier!\n");
+ err = AVERROR_EXTERNAL;
+ goto end;
+ }
+ } else
+ drm_mod.drmFormatModifier = DRM_FORMAT_MOD_LINEAR;
for (int i = 0; (i < planes) && (f->mem[i]); i++) {
VkMemoryGetFdInfoKHR export_info = {
@@ -4313,16 +4554,20 @@ static int vulkan_map_to_drm(AVHWFramesC
goto end;
}
-#if HAVE_LINUX_DMA_BUF_H && defined(DMA_BUF_IOCTL_IMPORT_SYNC_FILE)
- if (sync_fd >= 0) {
+#ifdef DMA_BUF_IOCTL_IMPORT_SYNC_FILE
+ // Known to be not working on AMD GPU with VAAPI, disable it for now
+ if (0 && sync_fd >= 0 && !p->drm_import_sync_file_ret) {
int dup_fd = dup(sync_fd);
if (dup_fd >= 0) {
struct dma_buf_import_sync_file import_info = {
- .flags = DMA_BUF_SYNC_WRITE,
+ .flags = DMA_BUF_SYNC_RW,
.fd = dup_fd,
};
- if (ioctl(drm_desc->objects[i].fd, DMA_BUF_IOCTL_IMPORT_SYNC_FILE, &import_info) < 0)
- av_log(hwfc, AV_LOG_WARNING, "DMA_BUF_IOCTL_IMPORT_SYNC_FILE failed: %s\n", av_err2str(AVERROR(errno)));
+ if (p->drm_import_sync_file_ret = ioctl(drm_desc->objects[i].fd,
+ DMA_BUF_IOCTL_IMPORT_SYNC_FILE,
+ &import_info))
+ av_log(hwfc, AV_LOG_WARNING, "DMA_BUF_IOCTL_IMPORT_SYNC_FILE failed: %s, "
+ "disabling for future frames\n", av_err2str(AVERROR(errno)));
close(dup_fd);
} else {
av_log(hwfc, AV_LOG_WARNING, "dup(sync_fd) failed: %s\n", av_err2str(AVERROR(errno)));
@@ -4341,8 +4586,13 @@ static int vulkan_map_to_drm(AVHWFramesC
VkFormat plane_vkfmt = av_vkfmt_from_pixfmt(hwfc->sw_format)[i];
drm_desc->layers[i].format = vulkan_fmt_to_drm(plane_vkfmt);
- drm_desc->layers[i].nb_planes = fp->drm_format_modifier_properties[i].drmFormatModifierPlaneCount;
+ drm_desc->layers[i].nb_planes = is_modifier_tiling ? fp->drm_format_modifier_properties[i].drmFormatModifierPlaneCount : 1;
+ if (drm_desc->layers[i].format == DRM_FORMAT_INVALID) {
+ av_log(hwfc, AV_LOG_ERROR, "Cannot map to DRM layer, unsupported!\n");
+ err = AVERROR_PATCHWELCOME;
+ goto end;
+ }
if (drm_desc->layers[i].nb_planes > MAX_MEMORY_PLANES) {
av_log(hwfc, AV_LOG_ERROR, "Too many memory planes for DRM format!\n");
err = AVERROR_EXTERNAL;
@@ -4351,10 +4601,13 @@ static int vulkan_map_to_drm(AVHWFramesC
for (int j = 0; j < drm_desc->layers[i].nb_planes; j++) {
VkSubresourceLayout layout;
- int aspect_plane = (nb_images == 1) ? i : j;
- VkImageSubresource sub = {
- .aspectMask = plane_index_to_aspect(aspect_plane),
- };
+ VkImageSubresource sub = { 0 };
+
+ if (is_modifier_tiling) {
+ int aspect_plane = (nb_images == 1) ? i : j;
+ sub.aspectMask = plane_index_to_aspect(aspect_plane);
+ } else
+ sub.aspectMask = plane_index_to_aspect_linear(i, nb_images, planes);
drm_desc->layers[i].planes[j].object_index = FFMIN(i, drm_desc->nb_objects - 1);
@@ -4362,34 +4615,24 @@ static int vulkan_map_to_drm(AVHWFramesC
drm_desc->layers[i].planes[j].offset = layout.offset;
drm_desc->layers[i].planes[j].pitch = layout.rowPitch;
}
-
- if (drm_desc->layers[i].format == DRM_FORMAT_INVALID) {
- av_log(hwfc, AV_LOG_ERROR, "Cannot map to DRM layer, unsupported!\n");
- err = AVERROR_PATCHWELCOME;
- goto end;
- }
-
-
- if (f->tiling == VK_IMAGE_TILING_OPTIMAL)
- continue;
-
}
dst->width = src->width;
dst->height = src->height;
dst->data[0] = (uint8_t *)drm_desc;
- dst->hw_frames_ctx = av_buffer_ref(src->hw_frames_ctx);
if (sync_fd >= 0)
close(sync_fd);
- av_log(hwfc, AV_LOG_VERBOSE, "Mapped AVVkFrame to a DRM object!\n");
+ av_log(hwfc, AV_LOG_DEBUG, "Mapped AVVkFrame to a DRM object!\n");
return 0;
end:
- for (int i = 0; i < drm_desc->nb_objects; i++)
+ for (int i = 0; i < drm_desc->nb_objects; i++) {
close(drm_desc->objects[i].fd);
+ drm_desc->objects[i].fd = -1;
+ }
if (free_drm_desc_on_err)
av_free(drm_desc);
if (sync_fd >= 0)
@@ -4433,16 +4676,10 @@ static int vulkan_map_from(AVHWFramesCon
switch (dst->format) {
#if CONFIG_LIBDRM
case AV_PIX_FMT_DRM_PRIME:
- if (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS)
- return vulkan_map_to_drm(hwfc, dst, src, flags);
- else
- return AVERROR(ENOSYS);
+ return vulkan_map_to_drm(hwfc, dst, src, flags);
#if CONFIG_VAAPI
case AV_PIX_FMT_VAAPI:
- if (p->vkctx.extensions & FF_VK_EXT_DRM_MODIFIER_FLAGS)
- return vulkan_map_to_vaapi(hwfc, dst, src, flags);
- else
- return AVERROR(ENOSYS);
+ return vulkan_map_to_vaapi(hwfc, dst, src, flags);
#endif
#endif
default:
debian/patches/0063-prefer-vulkan-device-with-higher-api-version.patch → debian/patches/0080-prefer-vulkan-device-with-higher-api-version.patch
Vendored
+2 -2
View File
@@ -2,7 +2,7 @@ Index: FFmpeg/libavutil/hwcontext_vulkan.c
===================================================================
--- FFmpeg.orig/libavutil/hwcontext_vulkan.c
+++ FFmpeg/libavutil/hwcontext_vulkan.c
@@ -968,7 +968,7 @@ static const char *vk_dev_type(enum VkPh
@@ -1355,7 +1355,7 @@ static const char *vk_dev_type(enum VkPh
static int find_device(AVHWDeviceContext *ctx, VulkanDeviceSelection *select)
{
int err = 0, choice = -1;
@@ -11,7 +11,7 @@ Index: FFmpeg/libavutil/hwcontext_vulkan.c
VkResult ret;
VulkanDevicePriv *p = ctx->hwctx;
AVVulkanDeviceContext *hwctx = &p->p;
@@ -1035,63 +1035,78 @@ static int find_device(AVHWDeviceContext
@@ -1431,63 +1431,78 @@ static int find_device(AVHWDeviceContext
if (select->has_uuid) {
for (int i = 0; i < num; i++) {
debian/patches/0081-signal-vulkan-semaphores-in-drm-prime-reverse-mapping.patch
Vendored
+94
View File
@@ -0,0 +1,94 @@
Index: FFmpeg/libavfilter/vf_hwmap.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_hwmap.c
+++ FFmpeg/libavfilter/vf_hwmap.c
@@ -26,6 +26,12 @@
#include "libavutil/hwcontext_d3d11va.h"
#endif
+#if CONFIG_VULKAN
+#include "libavutil/hwcontext_vulkan.h"
+#include "libavutil/vulkan_loader.h"
+#include "libavutil/vulkan.h"
+#endif
+
#include "avfilter.h"
#include "filters.h"
#include "formats.h"
@@ -39,6 +45,11 @@ typedef struct HWMapContext {
int mode;
char *derive_device_type;
int reverse;
+
+#if CONFIG_VULKAN
+ FFVulkanFunctions vkfn;
+ int vkfn_loaded;
+#endif
} HWMapContext;
static int hwmap_query_formats(const AVFilterContext *avctx,
@@ -193,6 +204,18 @@ static int hwmap_config_output(AVFilterL
goto fail;
}
+#if CONFIG_VULKAN
+ if (inl->hw_frames_ctx && inlink->format == AV_PIX_FMT_VULKAN) {
+ AVHWFramesContext *hwfc_src = (AVHWFramesContext *)inl->hw_frames_ctx->data;
+ AVVulkanFramesContext *vkfc_src = hwfc_src->hwctx;
+ AVHWFramesContext *hwfc_dst = (AVHWFramesContext *)source->data;
+ AVVulkanFramesContext *vkfc_dst = hwfc_dst->hwctx;
+
+ // Passthrough the VK_IMAGE_USAGE_*_BIT
+ vkfc_dst->usage = vkfc_src->usage;
+ }
+#endif
+
// 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
@@ -375,6 +398,45 @@ static int hwmap_filter_frame(AVFilterLi
if (err < 0)
goto fail;
+#if CONFIG_VULKAN
+ if (ctx->reverse &&
+ input->hw_frames_ctx && map->hw_frames_ctx &&
+ input->format == AV_PIX_FMT_VULKAN &&
+ (map->format == AV_PIX_FMT_VAAPI ||
+ map->format == AV_PIX_FMT_DRM_PRIME)) {
+ // If we mapped backwards from vulkan to drm_prime, we need
+ // to wait for the AVVkFrame semaphores to be signaled.
+ AVHWFramesContext *hwfc = (AVHWFramesContext *)input->hw_frames_ctx->data;
+ AVVulkanDeviceContext *vk_dev = hwfc->device_ctx->hwctx;
+ AVVkFrame *vkf = (AVVkFrame *)input->data[0];
+ const int nb_sems = vkf ? ff_vk_count_images(vkf) : 0;
+
+ if (hwfc->device_ctx->type != AV_HWDEVICE_TYPE_VULKAN || !vkf || !nb_sems)
+ goto exit;
+
+ if (!ctx->vkfn_loaded) {
+ uint64_t exts = ff_vk_extensions_to_mask(vk_dev->enabled_dev_extensions,
+ vk_dev->nb_enabled_dev_extensions);
+ err = ff_vk_load_functions(hwfc->device_ctx, &ctx->vkfn, exts, 1, 1);
+ if (err < 0)
+ goto fail;
+ ctx->vkfn_loaded = 1;
+ }
+ if (ctx->vkfn.WaitSemaphores) {
+ VkSemaphoreWaitInfo wait_info = {
+ .sType = VK_STRUCTURE_TYPE_SEMAPHORE_WAIT_INFO,
+ .flags = 0x0,
+ .semaphoreCount = nb_sems,
+ .pSemaphores = vkf->sem,
+ .pValues = vkf->sem_value,
+ };
+ ctx->vkfn.WaitSemaphores(vk_dev->act_dev, &wait_info, UINT64_MAX);
+ av_log(avctx, AV_LOG_DEBUG, "Vulkan sems for reverse-mapped DRM objects signaled!\n");
+ }
+ }
+exit:
+#endif
+
av_frame_free(&input);
av_log(ctx, AV_LOG_DEBUG, "Filter output: %s, %ux%u (%"PRId64").\n",
debian/patches/0082-add-fixes-and-enhancements-for-vulkan-filters.patch
Vendored
+506
View File
@@ -0,0 +1,506 @@
Index: FFmpeg/libavfilter/vf_flip_vulkan.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_flip_vulkan.c
+++ FFmpeg/libavfilter/vf_flip_vulkan.c
@@ -109,13 +109,13 @@ static av_cold int init_filter(AVFilterC
switch (type)
{
case FLIP_HORIZONTAL:
- GLSLF(2, vec4 res = imageLoad(input_image[%i], ivec2(size.x - pos.x, pos.y)); ,i);
+ GLSLF(2, vec4 res = imageLoad(input_image[%i], ivec2(size.x - 1 - pos.x, pos.y)); ,i);
break;
case FLIP_VERTICAL:
- GLSLF(2, vec4 res = imageLoad(input_image[%i], ivec2(pos.x, size.y - pos.y)); ,i);
+ GLSLF(2, vec4 res = imageLoad(input_image[%i], ivec2(pos.x, size.y - 1 - pos.y)); ,i);
break;
case FLIP_BOTH:
- GLSLF(2, vec4 res = imageLoad(input_image[%i], ivec2(size.xy - pos.xy));, i);
+ GLSLF(2, vec4 res = imageLoad(input_image[%i], ivec2(size.xy - 1 - pos.xy));, i);
break;
default:
GLSLF(2, vec4 res = imageLoad(input_image[%i], pos); ,i);
Index: FFmpeg/libavfilter/vf_overlay_vulkan.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_overlay_vulkan.c
+++ FFmpeg/libavfilter/vf_overlay_vulkan.c
@@ -35,6 +35,7 @@ typedef struct OverlayVulkanContext {
FFVkExecPool e;
AVVulkanDeviceQueueFamily *qf;
FFVulkanShader shd;
+ FFVulkanShader shd_pass;
/* Push constants / options */
struct {
@@ -46,6 +47,10 @@ typedef struct OverlayVulkanContext {
int overlay_y;
int overlay_w;
int overlay_h;
+
+ int opt_repeatlast;
+ int opt_shortest;
+ int opt_eof_action;
} OverlayVulkanContext;
static const char overlay_noalpha[] = {
@@ -85,12 +90,13 @@ static av_cold int init_filter(AVFilterC
uint8_t *spv_data;
size_t spv_len;
void *spv_opaque = NULL;
+ void *spv_opaque_pass = NULL;
OverlayVulkanContext *s = ctx->priv;
FFVulkanContext *vkctx = &s->vkctx;
const int planes = av_pix_fmt_count_planes(s->vkctx.output_format);
const int ialpha = av_pix_fmt_desc_get(s->vkctx.input_format)->flags & AV_PIX_FMT_FLAG_ALPHA;
const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(s->vkctx.output_format);
- FFVulkanShader *shd = &s->shd;
+ FFVulkanShader *shd;
FFVkSPIRVCompiler *spv;
FFVulkanDescriptorSetBinding *desc;
@@ -108,12 +114,15 @@ static av_cold int init_filter(AVFilterC
}
RET(ff_vk_exec_pool_init(vkctx, s->qf, &s->e, s->qf->num*4, 0, 0, 0, NULL));
+
+ /* overlay */
RET(ff_vk_shader_init(vkctx, &s->shd, "overlay",
VK_SHADER_STAGE_COMPUTE_BIT,
NULL, 0,
32, 32, 1,
0));
+ shd = &s->shd;
GLSLC(0, layout(push_constant, std430) uniform pushConstants { );
GLSLC(1, ivec2 o_offset[3]; );
GLSLC(1, ivec2 o_size[3]; );
@@ -169,9 +178,9 @@ static av_cold int init_filter(AVFilterC
GLSLC(1, } );
GLSLC(0, } );
- RET(spv->compile_shader(vkctx, spv, shd, &spv_data, &spv_len, "main",
+ RET(spv->compile_shader(vkctx, spv, &s->shd, &spv_data, &spv_len, "main",
&spv_opaque));
- RET(ff_vk_shader_link(vkctx, shd, spv_data, spv_len, "main"));
+ RET(ff_vk_shader_link(vkctx, &s->shd, spv_data, spv_len, "main"));
RET(ff_vk_shader_register_exec(vkctx, &s->e, &s->shd));
@@ -189,11 +198,59 @@ static av_cold int init_filter(AVFilterC
s->opts.o_size[4] = s->opts.o_size[0] >> pix_desc->log2_chroma_w;
s->opts.o_size[5] = s->opts.o_size[1] >> pix_desc->log2_chroma_h;
+ /* overlay_pass */
+ RET(ff_vk_shader_init(vkctx, &s->shd_pass, "overlay_pass",
+ VK_SHADER_STAGE_COMPUTE_BIT,
+ NULL, 0,
+ 32, 32, 1,
+ 0));
+
+ desc = (FFVulkanDescriptorSetBinding []) {
+ {
+ .name = "main_img",
+ .type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
+ .mem_layout = ff_vk_shader_rep_fmt(s->vkctx.input_format, FF_VK_REP_FLOAT),
+ .mem_quali = "readonly",
+ .dimensions = 2,
+ .elems = planes,
+ .stages = VK_SHADER_STAGE_COMPUTE_BIT,
+ },
+ {
+ .name = "output_img",
+ .type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
+ .mem_layout = ff_vk_shader_rep_fmt(s->vkctx.output_format, FF_VK_REP_FLOAT),
+ .mem_quali = "writeonly",
+ .dimensions = 2,
+ .elems = planes,
+ .stages = VK_SHADER_STAGE_COMPUTE_BIT,
+ },
+ };
+
+ RET(ff_vk_shader_add_descriptor_set(vkctx, &s->shd_pass, desc, 2, 0, 0));
+
+ shd = &s->shd_pass;
+ GLSLC(0, void main() );
+ GLSLC(0, { );
+ GLSLC(1, ivec2 pos = ivec2(gl_GlobalInvocationID.xy); );
+ GLSLF(1, int planes = %i; ,planes);
+ GLSLC(1, for (int i = 0; i < planes; i++) { );
+ GLSLC(2, vec4 res = imageLoad(main_img[i], pos); );
+ GLSLC(2, imageStore(output_img[i], pos, res); );
+ GLSLC(1, } );
+ GLSLC(0, } );
+
+ RET(spv->compile_shader(vkctx, spv, &s->shd_pass, &spv_data, &spv_len, "main",
+ &spv_opaque_pass));
+ RET(ff_vk_shader_link(vkctx, &s->shd_pass, spv_data, spv_len, "main"));
+ RET(ff_vk_shader_register_exec(vkctx, &s->e, &s->shd_pass));
+
s->initialized = 1;
fail:
if (spv_opaque)
spv->free_shader(spv, &spv_opaque);
+ if (spv_opaque_pass)
+ spv->free_shader(spv, &spv_opaque_pass);
if (spv)
spv->uninit(&spv);
@@ -215,22 +272,11 @@ static int overlay_vulkan_blend(FFFrameS
if (err < 0)
goto fail;
- if (!input_main || !input_overlay)
- return 0;
-
- if (!s->initialized) {
- AVHWFramesContext *main_fc = (AVHWFramesContext*)input_main->hw_frames_ctx->data;
- AVHWFramesContext *overlay_fc = (AVHWFramesContext*)input_overlay->hw_frames_ctx->data;
- if (main_fc->sw_format != overlay_fc->sw_format) {
- av_log(ctx, AV_LOG_ERROR, "Mismatching sw formats!\n");
- return AVERROR(EINVAL);
- }
-
- s->overlay_w = input_overlay->width;
- s->overlay_h = input_overlay->height;
+ if (!input_main)
+ return AVERROR_BUG;
+ if (!s->initialized)
RET(init_filter(ctx));
- }
out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
if (!out) {
@@ -238,9 +284,14 @@ static int overlay_vulkan_blend(FFFrameS
goto fail;
}
- RET(ff_vk_filter_process_Nin(&s->vkctx, &s->e, &s->shd,
- out, (AVFrame *[]){ input_main, input_overlay }, 2,
- VK_NULL_HANDLE, &s->opts, sizeof(s->opts)));
+ if (input_overlay)
+ RET(ff_vk_filter_process_Nin(&s->vkctx, &s->e, &s->shd,
+ out, (AVFrame *[]){ input_main, input_overlay }, 2,
+ VK_NULL_HANDLE, &s->opts, sizeof(s->opts)));
+ else /* passthrough */
+ RET(ff_vk_filter_process_simple(&s->vkctx, &s->e, &s->shd_pass,
+ out, input_main, VK_NULL_HANDLE,
+ &s->opts, sizeof(s->opts)));
err = av_frame_copy_props(out, input_main);
if (err < 0)
@@ -258,6 +309,20 @@ static int overlay_vulkan_config_output(
int err;
AVFilterContext *avctx = outlink->src;
OverlayVulkanContext *s = avctx->priv;
+ AVFilterLink *inlink = avctx->inputs[0];
+ AVFilterLink *inlink_overlay = avctx->inputs[1];
+ FilterLink *inl = ff_filter_link(inlink);
+ FilterLink *inl_overlay = ff_filter_link(inlink_overlay);
+ AVHWFramesContext *main_fc = (AVHWFramesContext*)inl->hw_frames_ctx->data;
+ AVHWFramesContext *overlay_fc = (AVHWFramesContext*)inl_overlay->hw_frames_ctx->data;
+
+ if (main_fc->sw_format != overlay_fc->sw_format) {
+ av_log(avctx, AV_LOG_ERROR, "Mismatching sw formats!\n");
+ return AVERROR(EINVAL);
+ }
+
+ s->overlay_w = inlink_overlay->w;
+ s->overlay_h = inlink_overlay->h;
err = ff_vk_filter_config_output(outlink);
if (err < 0)
@@ -267,6 +332,11 @@ static int overlay_vulkan_config_output(
if (err < 0)
return err;
+ s->fs.opt_repeatlast = s->opt_repeatlast;
+ s->fs.opt_shortest = s->opt_shortest;
+ s->fs.opt_eof_action = s->opt_eof_action;
+ s->fs.time_base = outlink->time_base = inlink->time_base;
+
return ff_framesync_configure(&s->fs);
}
@@ -293,6 +363,7 @@ static void overlay_vulkan_uninit(AVFilt
ff_vk_exec_pool_free(vkctx, &s->e);
ff_vk_shader_free(vkctx, &s->shd);
+ ff_vk_shader_free(vkctx, &s->shd_pass);
ff_vk_uninit(&s->vkctx);
ff_framesync_uninit(&s->fs);
@@ -305,6 +376,14 @@ static void overlay_vulkan_uninit(AVFilt
static const AVOption overlay_vulkan_options[] = {
{ "x", "Set horizontal offset", OFFSET(overlay_x), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, .flags = FLAGS },
{ "y", "Set vertical offset", OFFSET(overlay_y), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, .flags = FLAGS },
+ { "eof_action", "Action to take when encountering EOF from secondary input ",
+ OFFSET(opt_eof_action), AV_OPT_TYPE_INT, { .i64 = EOF_ACTION_REPEAT },
+ EOF_ACTION_REPEAT, EOF_ACTION_PASS, .flags = FLAGS, .unit = "eof_action" },
+ { "repeat", "Repeat the previous frame.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_REPEAT }, .flags = FLAGS, .unit = "eof_action" },
+ { "endall", "End both streams.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_ENDALL }, .flags = FLAGS, .unit = "eof_action" },
+ { "pass", "Pass through the main input.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_PASS }, .flags = FLAGS, .unit = "eof_action" },
+ { "shortest", "force termination when the shortest input terminates", OFFSET(opt_shortest), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, FLAGS },
+ { "repeatlast", "repeat overlay of the last overlay frame", OFFSET(opt_repeatlast), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS },
{ NULL },
};
Index: FFmpeg/libavfilter/vf_scale_vulkan.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_scale_vulkan.c
+++ FFmpeg/libavfilter/vf_scale_vulkan.c
@@ -104,6 +104,23 @@ static const char write_nv12[] = {
C(0, } )
};
+static const char write_nv16[] = {
+ C(0, void write_nv16(vec4 src, ivec2 pos) )
+ C(0, { )
+ C(1, imageStore(output_img[0], pos, vec4(src.r, 0.0, 0.0, 0.0)); )
+ C(1, pos.x >>= 1; )
+ C(1, imageStore(output_img[1], pos, vec4(src.g, src.b, 0.0, 0.0)); )
+ C(0, } )
+};
+
+static const char write_nv24[] = {
+ C(0, void write_nv24(vec4 src, ivec2 pos) )
+ C(0, { )
+ C(1, imageStore(output_img[0], pos, vec4(src.r, 0.0, 0.0, 0.0)); )
+ C(1, imageStore(output_img[1], pos, vec4(src.g, src.b, 0.0, 0.0)); )
+ C(0, } )
+};
+
static const char write_420[] = {
C(0, void write_420(vec4 src, ivec2 pos) )
C(0, { )
@@ -114,6 +131,16 @@ static const char write_420[] = {
C(0, } )
};
+static const char write_422[] = {
+ C(0, void write_422(vec4 src, ivec2 pos) )
+ C(0, { )
+ C(1, imageStore(output_img[0], pos, vec4(src.r, 0.0, 0.0, 0.0)); )
+ C(1, pos.x >>= 1; )
+ C(1, imageStore(output_img[1], pos, vec4(src.g, 0.0, 0.0, 0.0)); )
+ C(1, imageStore(output_img[2], pos, vec4(src.b, 0.0, 0.0, 0.0)); )
+ C(0, } )
+};
+
static const char write_444[] = {
C(0, void write_444(vec4 src, ivec2 pos) )
C(0, { )
@@ -134,9 +161,30 @@ static int init_scale_shader(AVFilterCon
}
switch (s->vkctx.output_format) {
- case AV_PIX_FMT_NV12: GLSLD(write_nv12); break;
- case AV_PIX_FMT_YUV420P: GLSLD( write_420); break;
- case AV_PIX_FMT_YUV444P: GLSLD( write_444); break;
+ case AV_PIX_FMT_NV12:
+ case AV_PIX_FMT_P010:
+ case AV_PIX_FMT_P012:
+ case AV_PIX_FMT_P016: GLSLD(write_nv12); break;
+ case AV_PIX_FMT_NV16:
+ case AV_PIX_FMT_P210:
+ case AV_PIX_FMT_P212:
+ case AV_PIX_FMT_P216: GLSLD(write_nv16); break;
+ case AV_PIX_FMT_NV24:
+ case AV_PIX_FMT_P410:
+ case AV_PIX_FMT_P412:
+ case AV_PIX_FMT_P416: GLSLD(write_nv24); break;
+ case AV_PIX_FMT_YUV420P:
+ case AV_PIX_FMT_YUV420P10:
+ case AV_PIX_FMT_YUV420P12:
+ case AV_PIX_FMT_YUV420P16: GLSLD( write_420); break;
+ case AV_PIX_FMT_YUV422P:
+ case AV_PIX_FMT_YUV422P10:
+ case AV_PIX_FMT_YUV422P12:
+ case AV_PIX_FMT_YUV422P16: GLSLD( write_422); break;
+ case AV_PIX_FMT_YUV444P:
+ case AV_PIX_FMT_YUV444P10:
+ case AV_PIX_FMT_YUV444P12:
+ case AV_PIX_FMT_YUV444P16: GLSLD( write_444); break;
default: break;
}
@@ -149,7 +197,8 @@ static int init_scale_shader(AVFilterCon
GLSLC(1, vec2 c_o = vec2(crop_x, crop_y) / in_d; );
GLSLC(0, );
- if (s->vkctx.output_format == s->vkctx.input_format) {
+ if (s->vkctx.output_format == s->vkctx.input_format ||
+ !ff_vk_mt_is_np_rgb(s->vkctx.input_format)) {
for (int i = 0; i < desc[1].elems; i++) {
GLSLF(1, size = imageSize(output_img[%i]); ,i);
GLSLC(1, if (IS_WITHIN(pos, size)) { );
@@ -166,16 +215,38 @@ static int init_scale_shader(AVFilterCon
GLSLC(1, vec4 res = scale_bilinear(0, pos, c_r, c_o); );
GLSLF(1, res = rgb2yuv(res, %i); ,s->out_range == AVCOL_RANGE_JPEG);
switch (s->vkctx.output_format) {
- case AV_PIX_FMT_NV12: GLSLC(1, write_nv12(res, pos); ); break;
- case AV_PIX_FMT_YUV420P: GLSLC(1, write_420(res, pos); ); break;
- case AV_PIX_FMT_YUV444P: GLSLC(1, write_444(res, pos); ); break;
+ case AV_PIX_FMT_NV12:
+ case AV_PIX_FMT_P010:
+ case AV_PIX_FMT_P012:
+ case AV_PIX_FMT_P016: GLSLC(1, write_nv12(res, pos); ); break;
+ case AV_PIX_FMT_NV16:
+ case AV_PIX_FMT_P210:
+ case AV_PIX_FMT_P212:
+ case AV_PIX_FMT_P216: GLSLC(1, write_nv16(res, pos); ); break;
+ case AV_PIX_FMT_NV24:
+ case AV_PIX_FMT_P410:
+ case AV_PIX_FMT_P412:
+ case AV_PIX_FMT_P416: GLSLC(1, write_nv24(res, pos); ); break;
+ case AV_PIX_FMT_YUV420P:
+ case AV_PIX_FMT_YUV420P10:
+ case AV_PIX_FMT_YUV420P12:
+ case AV_PIX_FMT_YUV420P16: GLSLC(1, write_420(res, pos); ); break;
+ case AV_PIX_FMT_YUV422P:
+ case AV_PIX_FMT_YUV422P10:
+ case AV_PIX_FMT_YUV422P12:
+ case AV_PIX_FMT_YUV422P16: GLSLC(1, write_422(res, pos); ); break;
+ case AV_PIX_FMT_YUV444P:
+ case AV_PIX_FMT_YUV444P10:
+ case AV_PIX_FMT_YUV444P12:
+ case AV_PIX_FMT_YUV444P16: GLSLC(1, write_444(res, pos); ); break;
default: return AVERROR(EINVAL);
}
}
GLSLC(0, } );
- if (s->vkctx.output_format != s->vkctx.input_format) {
+ if (s->vkctx.output_format != s->vkctx.input_format &&
+ ff_vk_mt_is_np_rgb(s->vkctx.input_format)) {
const AVLumaCoefficients *lcoeffs;
double tmp_mat[3][3];
@@ -385,7 +456,8 @@ static int scale_vulkan_filter_frame(AVF
if (s->out_range != AVCOL_RANGE_UNSPECIFIED)
out->color_range = s->out_range;
- if (s->vkctx.output_format != s->vkctx.input_format)
+ if (s->vkctx.output_format != s->vkctx.input_format &&
+ ff_vk_mt_is_np_rgb(s->vkctx.input_format))
out->chroma_location = AVCHROMA_LOC_TOPLEFT;
av_frame_free(&in);
@@ -440,13 +512,58 @@ static int scale_vulkan_config_output(AV
return AVERROR_PATCHWELCOME;
}
} else if (s->vkctx.output_format != s->vkctx.input_format) {
- if (!ff_vk_mt_is_np_rgb(s->vkctx.input_format)) {
+ const AVPixFmtDescriptor *idesc = av_pix_fmt_desc_get(s->vkctx.input_format);
+ const AVPixFmtDescriptor *odesc = av_pix_fmt_desc_get(s->vkctx.output_format);
+ const int iplanes = av_pix_fmt_count_planes(s->vkctx.input_format);
+ const int oplanes = av_pix_fmt_count_planes(s->vkctx.output_format);
+
+ const int irgb = idesc->flags & AV_PIX_FMT_FLAG_RGB;
+ const int iyuv = !irgb && idesc->nb_components >= 2;
+ const int iplanar = idesc->flags & AV_PIX_FMT_FLAG_PLANAR;
+ const int orgb = odesc->flags & AV_PIX_FMT_FLAG_RGB;
+ const int oyuv = !orgb && idesc->nb_components >= 2;
+ const int oplanar = odesc->flags & AV_PIX_FMT_FLAG_PLANAR;
+
+ if (iyuv && oyuv && iplanar && oplanar && iplanes == oplanes && iplanes > 1) {
+ if (idesc->log2_chroma_w != odesc->log2_chroma_w ||
+ idesc->log2_chroma_h != odesc->log2_chroma_h) {
+ av_log(avctx, AV_LOG_ERROR, "Unsupported input format for conversion\n");
+ return AVERROR(EINVAL);
+ }
+ if (s->out_range != AVCOL_RANGE_UNSPECIFIED) {
+ av_log(avctx, AV_LOG_ERROR, "Cannot change range in yuv2yuv conversion\n");
+ return AVERROR(EINVAL);
+ }
+ if (inlink->w == outlink->w && inlink->h == outlink->h)
+ s->scaler = F_NEAREST;
+ } else if (!ff_vk_mt_is_np_rgb(s->vkctx.input_format)) {
av_log(avctx, AV_LOG_ERROR, "Unsupported input format for conversion\n");
return AVERROR(EINVAL);
}
if (s->vkctx.output_format != AV_PIX_FMT_NV12 &&
+ s->vkctx.output_format != AV_PIX_FMT_P010 &&
+ s->vkctx.output_format != AV_PIX_FMT_P012 &&
+ s->vkctx.output_format != AV_PIX_FMT_P016 &&
+ s->vkctx.output_format != AV_PIX_FMT_NV16 &&
+ s->vkctx.output_format != AV_PIX_FMT_P210 &&
+ s->vkctx.output_format != AV_PIX_FMT_P212 &&
+ s->vkctx.output_format != AV_PIX_FMT_P216 &&
+ s->vkctx.output_format != AV_PIX_FMT_NV24 &&
+ s->vkctx.output_format != AV_PIX_FMT_P410 &&
+ s->vkctx.output_format != AV_PIX_FMT_P412 &&
+ s->vkctx.output_format != AV_PIX_FMT_P416 &&
s->vkctx.output_format != AV_PIX_FMT_YUV420P &&
- s->vkctx.output_format != AV_PIX_FMT_YUV444P) {
+ s->vkctx.output_format != AV_PIX_FMT_YUV420P10 &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV420P12 &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV420P16 &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV422P &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV422P10 &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV422P12 &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV422P16 &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV444P &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV444P10 &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV444P12 &&
+ s->vkctx.output_format != AV_PIX_FMT_YUV444P16) {
av_log(avctx, AV_LOG_ERROR, "Unsupported output format\n");
return AVERROR(EINVAL);
}
Index: FFmpeg/libavfilter/vf_transpose_vulkan.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_transpose_vulkan.c
+++ FFmpeg/libavfilter/vf_transpose_vulkan.c
@@ -49,6 +49,8 @@ static av_cold int init_filter(AVFilterC
TransposeVulkanContext *s = ctx->priv;
FFVulkanContext *vkctx = &s->vkctx;
+ const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(s->vkctx.output_format);
+ const int is_422 = pix_desc->log2_chroma_w == 1 && !pix_desc->log2_chroma_h;
const int planes = av_pix_fmt_count_planes(s->vkctx.output_format);
FFVulkanShader *shd = &s->shd;
FFVkSPIRVCompiler *spv;
@@ -100,19 +102,24 @@ static av_cold int init_filter(AVFilterC
GLSLC(0, void main() );
GLSLC(0, { );
GLSLC(1, ivec2 size; );
+ GLSLC(1, ivec2 ipos; );
GLSLC(1, ivec2 pos = ivec2(gl_GlobalInvocationID.xy); );
for (int i = 0; i < planes; i++) {
+ int is_422_uv = i && is_422;
GLSLC(0, );
GLSLF(1, size = imageSize(output_images[%i]); ,i);
GLSLC(1, if (IS_WITHIN(pos, size)) { );
if (s->dir == TRANSPOSE_CCLOCK)
- GLSLF(2, vec4 res = imageLoad(input_images[%i], ivec2(size.y - pos.y, pos.x)); ,i);
+ GLSLC(2, ipos = ivec2(size.y - 1 - pos.y, pos.x); );
else if (s->dir == TRANSPOSE_CLOCK_FLIP || s->dir == TRANSPOSE_CLOCK) {
- GLSLF(2, vec4 res = imageLoad(input_images[%i], ivec2(size.yx - pos.yx)); ,i);
+ GLSLC(2, ipos = ivec2(size.yx - 1 - pos.yx); );
if (s->dir == TRANSPOSE_CLOCK)
- GLSLC(2, pos = ivec2(pos.x, size.y - pos.y); );
+ GLSLC(2, pos = ivec2(pos.x, size.y - 1 - pos.y); );
} else
- GLSLF(2, vec4 res = imageLoad(input_images[%i], pos.yx); ,i);
+ GLSLC(2, ipos = pos.yx; );
+ if (is_422_uv)
+ GLSLC(2, ipos = ivec2(ipos.x >> 1, ipos.y << 1); );
+ GLSLF(2, vec4 res = imageLoad(input_images[%i], ipos); ,i);
GLSLF(2, imageStore(output_images[%i], pos, res); ,i);
GLSLC(1, } );
}
@@ -161,11 +168,9 @@ static int filter_frame(AVFilterLink *in
RET(av_frame_copy_props(out, in));
if (in->sample_aspect_ratio.num)
+ out->sample_aspect_ratio = av_inv_q(in->sample_aspect_ratio);
+ else
out->sample_aspect_ratio = in->sample_aspect_ratio;
- else {
- out->sample_aspect_ratio.num = in->sample_aspect_ratio.den;
- out->sample_aspect_ratio.den = in->sample_aspect_ratio.num;
- }
av_frame_free(&in);
debian/patches/0083-allow-vpp-tonemap-without-hdr10-metadata.patch
Vendored
-43
View File
@@ -1,43 +0,0 @@
Index: FFmpeg/libavfilter/vf_tonemap_vaapi.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_tonemap_vaapi.c
+++ FFmpeg/libavfilter/vf_tonemap_vaapi.c
@@ -56,7 +56,7 @@ static int tonemap_vaapi_save_metadata(A
AVContentLightMetadata *light_meta;
if (input_frame->color_trc != AVCOL_TRC_SMPTE2084) {
- av_log(avctx, AV_LOG_WARNING, "Only support HDR10 as input for vaapi tone-mapping\n");
+ av_log(avctx, AV_LOG_DEBUG, "Only support HDR10 as input for vaapi tone-mapping\n");
}
ctx->src_display = av_frame_get_side_data(input_frame,
@@ -64,8 +64,7 @@ static int tonemap_vaapi_save_metadata(A
if (ctx->src_display) {
hdr_meta = (AVMasteringDisplayMetadata *)ctx->src_display->data;
if (!hdr_meta) {
- av_log(avctx, AV_LOG_ERROR, "No mastering display data\n");
- return AVERROR(EINVAL);
+ av_log(avctx, AV_LOG_DEBUG, "No mastering display data\n");
}
if (hdr_meta->has_luminance) {
@@ -122,8 +121,7 @@ static int tonemap_vaapi_save_metadata(A
ctx->in_metadata.white_point_y);
}
} else {
- av_log(avctx, AV_LOG_ERROR, "No mastering display data from input\n");
- return AVERROR(EINVAL);
+ av_log(avctx, AV_LOG_DEBUG, "No mastering display data from input\n");
}
ctx->src_light = av_frame_get_side_data(input_frame,
@@ -131,8 +129,7 @@ static int tonemap_vaapi_save_metadata(A
if (ctx->src_light) {
light_meta = (AVContentLightMetadata *)ctx->src_light->data;
if (!light_meta) {
- av_log(avctx, AV_LOG_ERROR, "No light metadata\n");
- return AVERROR(EINVAL);
+ av_log(avctx, AV_LOG_DEBUG, "No light metadata\n");
}
ctx->in_metadata.max_content_light_level = light_meta->MaxCLL;
debian/patches/0083-backport-fixes-for-nvenc-and-cuvid-from-upstream.patch
Vendored
+95
View File
@@ -0,0 +1,95 @@
Index: FFmpeg/libavcodec/cuviddec.c
===================================================================
--- FFmpeg.orig/libavcodec/cuviddec.c
+++ FFmpeg/libavcodec/cuviddec.c
@@ -164,9 +164,17 @@ static int CUDAAPI cuvid_handle_video_se
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;
+ // NVDEC target dimensions must be even-aligned for internal surface allocation.
+ // For chroma-subsampled formats (420/422), the output dimensions must also be
+ // even. For monochrome/444, keep the original output dimensions and only
+ // even-align the NVDEC target — the frame copy will crop to avctx dimensions.
+ cuinfo.ulTargetWidth = (avctx->width + 1) & ~1;
+ cuinfo.ulTargetHeight = (avctx->height + 1) & ~1;
+ if (format->chroma_format == cudaVideoChromaFormat_420 ||
+ format->chroma_format == cudaVideoChromaFormat_422) {
+ avctx->width = cuinfo.ulTargetWidth;
+ avctx->height = cuinfo.ulTargetHeight;
+ }
// aspect ratio conversion, 1:1, depends on scaled resolution
cuinfo.target_rect.left = 0;
@@ -916,7 +924,7 @@ static av_cold int cuvid_decode_init(AVC
if (probe_desc && probe_desc->nb_components)
probed_bit_depth = probe_desc->comp[0].depth;
- if (probe_desc && !probe_desc->log2_chroma_w && !probe_desc->log2_chroma_h)
+ if (probe_desc && probe_desc->nb_components > 1 && !probe_desc->log2_chroma_w && !probe_desc->log2_chroma_h)
is_yuv444 = 1;
#ifdef NVDEC_HAVE_422_SUPPORT
@@ -1103,7 +1111,7 @@ static av_cold int cuvid_decode_init(AVC
// Skip first 4 bytes of AV1CodecConfigurationRecord to keep configOBUs
// only, otherwise cuvidParseVideoData report unknown error.
if (avctx->codec->id == AV_CODEC_ID_AV1 &&
- extradata_size > 4 &&
+ extradata_size >= 4 &&
extradata[0] & 0x80) {
extradata += 4;
extradata_size -= 4;
Index: FFmpeg/libavcodec/nvdec.c
===================================================================
--- FFmpeg.orig/libavcodec/nvdec.c
+++ FFmpeg/libavcodec/nvdec.c
@@ -732,8 +732,18 @@ int ff_nvdec_frame_params(AVCodecContext
chroma_444 = supports_444 && cuvid_chroma_format == cudaVideoChromaFormat_444;
frames_ctx->format = AV_PIX_FMT_CUDA;
- frames_ctx->width = (avctx->coded_width + 1) & ~1;
- frames_ctx->height = (avctx->coded_height + 1) & ~1;
+ // NVDEC target dimensions must be even-aligned for internal surface allocation.
+ // For chroma-subsampled formats (420/422), the output dimensions must also be
+ // even. For monochrome/444, keep the original output dimensions and only
+ // even-align the NVDEC target — the frame copy will crop to avctx dimensions.
+ if (cuvid_chroma_format == cudaVideoChromaFormat_420 ||
+ cuvid_chroma_format == cudaVideoChromaFormat_422) {
+ frames_ctx->width = (avctx->coded_width + 1) & ~1;
+ frames_ctx->height = (avctx->coded_height + 1) & ~1;
+ } else {
+ frames_ctx->width = avctx->coded_width;
+ frames_ctx->height = avctx->coded_height;
+ }
/*
* We add two extra frames to the pool to account for deinterlacing filters
* holding onto their frames.
Index: FFmpeg/libavcodec/nvenc.c
===================================================================
--- FFmpeg.orig/libavcodec/nvenc.c
+++ FFmpeg/libavcodec/nvenc.c
@@ -683,7 +683,9 @@ static int nvenc_check_capabilities(AVCo
#ifdef NVENC_HAVE_MVHEVC
ctx->multiview_supported = nvenc_check_cap(avctx, NV_ENC_CAPS_SUPPORT_MVHEVC_ENCODE) > 0;
- if(ctx->profile == NV_ENC_HEVC_PROFILE_MULTIVIEW_MAIN && !ctx->multiview_supported) {
+ if (avctx->codec_id == AV_CODEC_ID_HEVC &&
+ ctx->profile == NV_ENC_HEVC_PROFILE_MULTIVIEW_MAIN &&
+ !ctx->multiview_supported) {
av_log(avctx, AV_LOG_WARNING, "Multiview not supported by the device\n");
return AVERROR(ENOSYS);
}
@@ -1359,6 +1361,11 @@ static av_cold int nvenc_setup_h264_conf
case NV_ENC_H264_PROFILE_BASELINE:
cc->profileGUID = NV_ENC_H264_PROFILE_BASELINE_GUID;
avctx->profile = AV_PROFILE_H264_BASELINE;
+ if (cc->frameIntervalP > 1) {
+ av_log(avctx, AV_LOG_WARNING,
+ "B-frames are not supported by H.264 Baseline profile, disabling.\n");
+ cc->frameIntervalP = 1;
+ }
break;
case NV_ENC_H264_PROFILE_MAIN:
cc->profileGUID = NV_ENC_H264_PROFILE_MAIN_GUID;
debian/patches/0084-backport-fixes-for-detecting-ac4-streams-in-mpegts.patch
Vendored
+52
View File
@@ -0,0 +1,52 @@
Index: FFmpeg/libavformat/mpegts.c
===================================================================
--- FFmpeg.orig/libavformat/mpegts.c
+++ FFmpeg/libavformat/mpegts.c
@@ -2263,6 +2263,7 @@ int ff_parse_mpeg2_descriptor(AVFormatCo
sti->need_parsing = AVSTREAM_PARSE_FULL;
sti->need_context_update = 1;
}
+ break;
}
if (ext_desc_tag == SUPPLEMENTARY_AUDIO_DESCRIPTOR) {
int flags;
@@ -2300,22 +2301,14 @@ int ff_parse_mpeg2_descriptor(AVFormatCo
if (language[0])
av_dict_set(&st->metadata, "language", language, 0);
}
+ break;
}
- break;
- case AC3_DESCRIPTOR:
- {
- int component_type_flag = get8(pp, desc_end) & (1 << 7);
- if (component_type_flag) {
- int component_type = get8(pp, desc_end);
- int service_type_mask = 0x38; // 0b00111000
- int service_type = ((component_type & service_type_mask) >> 3);
- if (service_type == 0x02 /* 0b010 */) {
- st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
- av_log(ts ? ts->stream : fc, AV_LOG_DEBUG, "New track disposition for id %u: %u\n", st->id, st->disposition);
- }
- }
+ if (ext_desc_tag == AC4_DESCRIPTOR) {
+ st->codecpar->codec_id = AV_CODEC_ID_AC4;
+ st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
}
break;
+ case AC3_DESCRIPTOR:
case ENHANCED_AC3_DESCRIPTOR:
{
int component_type_flag = get8(pp, desc_end) & (1 << 7);
Index: FFmpeg/libavformat/mpegts.h
===================================================================
--- FFmpeg.orig/libavformat/mpegts.h
+++ FFmpeg/libavformat/mpegts.h
@@ -228,6 +228,7 @@ https://developer.apple.com/library/arch
/* DVB descriptor_tag_extension values from
ETSI EN 300 468 Table 109: Possible locations of extended descriptors */
#define SUPPLEMENTARY_AUDIO_DESCRIPTOR 0x06
+#define AC4_DESCRIPTOR 0x15
/** see "Dolby Vision Streams Within the MPEG-2 Transport Stream Format"
https://professional.dolby.com/siteassets/content-creation/dolby-vision-for-content-creators/dolby-vision-bitstreams-in-mpeg-2-transport-stream-multiplex-v1.2.pdf */
debian/patches/0085-fix-hdr10-sidedata-not-removed-in-libplacebo-outlink.patch
Vendored
+21
View File
@@ -0,0 +1,21 @@
Index: FFmpeg/libavfilter/vf_libplacebo.c
===================================================================
--- FFmpeg.orig/libavfilter/vf_libplacebo.c
+++ FFmpeg/libavfilter/vf_libplacebo.c
@@ -1478,6 +1478,16 @@ static int libplacebo_config_output(AVFi
AVHWFramesContext *hwfc;
AVVulkanFramesContext *vkfc;
+ if (!((s->color_primaries == -1 &&
+ s->color_trc == -1) ||
+ (s->color_primaries == AVCOL_PRI_BT2020 &&
+ s->color_trc == AVCOL_TRC_SMPTE2084))) {
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_MASTERING_DISPLAY_METADATA);
+ av_frame_side_data_remove(&outlink->side_data, &outlink->nb_side_data,
+ AV_FRAME_DATA_CONTENT_LIGHT_LEVEL);
+ }
+
/* Frame dimensions */
RET(ff_scale_eval_dimensions(s, s->w_expr, s->h_expr, inlink, outlink,
&outlink->w, &outlink->h));
debian/patches/0085-pass-eac3-addbsi-to-movenc-for-dolby-atmos-remuxing.patch
Vendored
-278
View File
@@ -1,278 +0,0 @@
Index: FFmpeg/libavcodec/ac3_parser.c
===================================================================
--- FFmpeg.orig/libavcodec/ac3_parser.c
+++ FFmpeg/libavcodec/ac3_parser.c
@@ -73,6 +73,201 @@ int ff_ac3_find_syncword(const uint8_t *
return i;
}
+#ifndef EAC3_SR_CODE_REDUCED
+ #define EAC3_SR_CODE_REDUCED 3
+#endif
+
+// see also ff_eac3_parse_header()
+static int eac3_parse_addbsi_from_header(GetBitContext *gbc,
+ AC3HeaderInfo *hdr,
+ uint8_t *flag_ec3_extension_type_a,
+ uint8_t *complexity_index_type_a)
+{
+ int i, blk;
+
+ if (!gbc || !hdr || !flag_ec3_extension_type_a || !complexity_index_type_a)
+ return AVERROR(EINVAL);
+
+ if (hdr->frame_type == EAC3_FRAME_TYPE_RESERVED)
+ return AC3_PARSE_ERROR_FRAME_TYPE;
+ if (hdr->substreamid)
+ return AC3_PARSE_ERROR_FRAME_TYPE;
+ if (hdr->sr_code == EAC3_SR_CODE_REDUCED)
+ return AVERROR_PATCHWELCOME;
+
+ skip_bits(gbc, 5); // skip bitstream id
+
+ /* volume control params */
+ for (i = 0; i < (hdr->channel_mode ? 1 : 2); i++) {
+ skip_bits(gbc, 5); // skip dialog_normalization[i]
+ if (get_bits1(gbc)) {
+ skip_bits(gbc, 8); // skip heavy_dynamic_range[i]
+ }
+ }
+
+ /* dependent stream channel map */
+ if (hdr->frame_type == EAC3_FRAME_TYPE_DEPENDENT) {
+ if (get_bits1(gbc)) {
+ skip_bits(gbc, 16); // skip channel_map
+ }
+ }
+
+ /* mixing metadata */
+ if (get_bits1(gbc)) {
+ /* center and surround mix levels */
+ if (hdr->channel_mode > AC3_CHMODE_STEREO) {
+ skip_bits(gbc, 2); // skip preferred_downmix
+ if (hdr->channel_mode & 1) {
+ /* if three front channels exist */
+ skip_bits(gbc, 3); // skip center_mix_level_ltrt
+ skip_bits(gbc, 3); // skip center_mix_level
+ }
+ if (hdr->channel_mode & 4) {
+ /* if a surround channel exists */
+ skip_bits(gbc, 3); // skip surround_mix_level_ltrt
+ skip_bits(gbc, 3); // skip surround_mix_level
+ }
+ }
+
+ /* lfe mix level */
+ if (hdr->lfe_on && get_bits1(gbc)) {
+ skip_bits(gbc, 5); // skip lfe_mix_level
+ }
+
+ /* info for mixing with other streams and substreams */
+ if (hdr->frame_type == EAC3_FRAME_TYPE_INDEPENDENT) {
+ for (i = 0; i < (hdr->channel_mode ? 1 : 2); i++) {
+ // TODO: apply program scale factor
+ if (get_bits1(gbc)) {
+ skip_bits(gbc, 6); // skip program scale factor
+ }
+ }
+ if (get_bits1(gbc)) {
+ skip_bits(gbc, 6); // skip external program scale factor
+ }
+ /* skip mixing parameter data */
+ switch(get_bits(gbc, 2)) {
+ case 1: skip_bits(gbc, 5); break;
+ case 2: skip_bits(gbc, 12); break;
+ case 3: {
+ int mix_data_size = (get_bits(gbc, 5) + 2) << 3;
+ skip_bits_long(gbc, mix_data_size);
+ break;
+ }
+ }
+ /* skip pan information for mono or dual mono source */
+ if (hdr->channel_mode < AC3_CHMODE_STEREO) {
+ for (i = 0; i < (hdr->channel_mode ? 1 : 2); i++) {
+ if (get_bits1(gbc)) {
+ /* note: this is not in the ATSC A/52B specification
+ reference: ETSI TS 102 366 V1.1.1
+ section: E.1.3.1.25 */
+ skip_bits(gbc, 8); // skip pan mean direction index
+ skip_bits(gbc, 6); // skip reserved paninfo bits
+ }
+ }
+ }
+ /* skip mixing configuration information */
+ if (get_bits1(gbc)) {
+ for (blk = 0; blk < hdr->num_blocks; blk++) {
+ if (hdr->num_blocks == 1 || get_bits1(gbc)) {
+ skip_bits(gbc, 5);
+ }
+ }
+ }
+ }
+ }
+
+ /* informational metadata */
+ if (get_bits1(gbc)) {
+ skip_bits(gbc, 3); // skip bitstream_mode
+ skip_bits(gbc, 2); // skip copyright bit and original bitstream bit
+ if (hdr->channel_mode == AC3_CHMODE_STEREO) {
+ skip_bits(gbc, 2); // skip dolby_surround_mode
+ skip_bits(gbc, 2); // skip dolby_headphone_mode
+ }
+ if (hdr->channel_mode >= AC3_CHMODE_2F2R) {
+ skip_bits(gbc, 2); // skip dolby_surround_ex_mode
+ }
+ for (i = 0; i < (hdr->channel_mode ? 1 : 2); i++) {
+ if (get_bits1(gbc)) {
+ skip_bits(gbc, 8); // skip mix level, room type, and A/D converter type
+ }
+ }
+ if (hdr->sr_code != EAC3_SR_CODE_REDUCED) {
+ skip_bits1(gbc); // skip source sample rate code
+ }
+ }
+
+ /* converter synchronization flag
+ If frames are less than six blocks, this bit should be turned on
+ once every 6 blocks to indicate the start of a frame set.
+ reference: RFC 4598, Section 2.1.3 Frame Sets */
+ if (hdr->frame_type == EAC3_FRAME_TYPE_INDEPENDENT && hdr->num_blocks != 6) {
+ skip_bits1(gbc); // skip converter synchronization flag
+ }
+
+ /* original frame size code if this stream was converted from AC-3 */
+ if (hdr->frame_type == EAC3_FRAME_TYPE_AC3_CONVERT &&
+ (hdr->num_blocks == 6 || get_bits1(gbc))) {
+ skip_bits(gbc, 6); // skip frame size code
+ }
+
+ /* additional bitstream info */
+ if (get_bits1(gbc)) {
+ int addbsil = get_bits(gbc, 6);
+ for (i = 0; i < addbsil + 1; i++) {
+ if (i == 0) {
+ /* In this 8 bit chunk, the LSB is equal to flag_ec3_extension_type_a
+ which can be used to detect Atmos presence */
+ skip_bits(gbc, 7);
+ if (get_bits1(gbc)) {
+ *flag_ec3_extension_type_a = 1;
+ *complexity_index_type_a = get_bits(gbc, 8);
+ /* The upper limit of complexity_index_type_a is 16 */
+ *complexity_index_type_a = FFMIN(*complexity_index_type_a, 16);
+ ++i;
+ }
+ } else {
+ skip_bits(gbc, 8); // skip additional bit stream info
+ }
+ }
+ }
+ return 0;
+}
+
+int avpriv_ac3_parse_header2(AC3HeaderInfo **phdr,
+ const uint8_t *buf, size_t size,
+ uint8_t *flag_ec3_extension_type_a,
+ uint8_t *complexity_index_type_a)
+{
+ GetBitContext gb;
+ AC3HeaderInfo *hdr;
+ int err;
+
+ if (!*phdr)
+ *phdr = av_mallocz(sizeof(AC3HeaderInfo));
+ if (!*phdr)
+ return AVERROR(ENOMEM);
+ hdr = *phdr;
+
+ err = init_get_bits8(&gb, buf, size);
+ if (err < 0)
+ return AVERROR_INVALIDDATA;
+ err = ff_ac3_parse_header(&gb, hdr);
+ if (err < 0)
+ return AVERROR_INVALIDDATA;
+
+ // eac3 addbsi
+ if (!(hdr->bitstream_id <= 10)) {
+ eac3_parse_addbsi_from_header(&gb, hdr,
+ flag_ec3_extension_type_a,
+ complexity_index_type_a);
+ }
+
+ return get_bits_count(&gb);
+}
+
int ff_ac3_parse_header(GetBitContext *gbc, AC3HeaderInfo *hdr)
{
int frame_size_code;
Index: FFmpeg/libavcodec/ac3_parser_internal.h
===================================================================
--- FFmpeg.orig/libavcodec/ac3_parser_internal.h
+++ FFmpeg/libavcodec/ac3_parser_internal.h
@@ -86,6 +86,11 @@ int ff_ac3_parse_header(GetBitContext *g
int avpriv_ac3_parse_header(AC3HeaderInfo **hdr, const uint8_t *buf,
size_t size);
+int avpriv_ac3_parse_header2(AC3HeaderInfo **phdr,
+ const uint8_t *buf, size_t size,
+ uint8_t *flag_ec3_extension_type_a,
+ uint8_t *complexity_index_type_a);
+
int ff_ac3_find_syncword(const uint8_t *buf, int buf_size);
#endif /* AVCODEC_AC3_PARSER_INTERNAL_H */
Index: FFmpeg/libavformat/movenc.c
===================================================================
--- FFmpeg.orig/libavformat/movenc.c
+++ FFmpeg/libavformat/movenc.c
@@ -393,6 +393,9 @@ struct eac3_info {
uint16_t chan_loc;
/* if there is no dependent substream, then one bit reserved instead */
} substream[1]; /* TODO: support 8 independent substreams */
+ /* addbsi */
+ uint8_t flag_ec3_extension_type_a; /* indicates the enhanced AC-3 extension, 1 bit */
+ uint8_t complexity_index_type_a; /* indicates the decoding complexity, 8 bits */
};
static int mov_write_ac3_tag(AVFormatContext *s, AVIOContext *pb, MOVTrack *track)
@@ -458,7 +461,9 @@ static int handle_eac3(MOVMuxContext *mo
if (!info->pkt && !(info->pkt = av_packet_alloc()))
return AVERROR(ENOMEM);
- if ((ret = avpriv_ac3_parse_header(&hdr, pkt->data, pkt->size)) < 0) {
+ if ((ret = avpriv_ac3_parse_header2(&hdr, pkt->data, pkt->size,
+ &info->flag_ec3_extension_type_a,
+ &info->complexity_index_type_a)) < 0) {
if (ret == AVERROR(ENOMEM))
goto end;
@@ -534,7 +539,8 @@ static int handle_eac3(MOVMuxContext *mo
while (cumul_size != pkt->size) {
GetBitContext gbc;
int i;
- ret = avpriv_ac3_parse_header(&hdr, pkt->data + cumul_size, pkt->size - cumul_size);
+ ret = avpriv_ac3_parse_header2(&hdr, pkt->data + cumul_size, pkt->size - cumul_size,
+ &info->flag_ec3_extension_type_a, &info->complexity_index_type_a);
if (ret < 0)
goto end;
if (hdr->frame_type != EAC3_FRAME_TYPE_DEPENDENT) {
@@ -614,7 +620,7 @@ static int mov_write_eac3_tag(AVFormatCo
}
info = track->eac3_priv;
- size = 2 + ((32 * (info->num_ind_sub + 1) + 7) >> 3);
+ size = 2 + ((32 * (info->num_ind_sub + 1) + 7 + 1 + 8) >> 3);
buf = av_malloc(size);
if (!buf) {
return AVERROR(ENOMEM);
@@ -639,6 +645,11 @@ static int mov_write_eac3_tag(AVFormatCo
put_bits(&pbc, 9, info->substream[i].chan_loc);
}
}
+ if (info->flag_ec3_extension_type_a == 1) {
+ put_bits(&pbc, 7, 0); /* reserved */
+ put_bits(&pbc, 1, info->flag_ec3_extension_type_a);
+ put_bits(&pbc, 8, info->complexity_index_type_a);
+ }
flush_put_bits(&pbc);
size = put_bytes_output(&pbc);
debian/patches/0086-workaround-amd-hevc-vaapi-encoder-out-of-band-vtag-hvc1.patch
Vendored
+13
View File
@@ -0,0 +1,13 @@
Index: FFmpeg/libavformat/movenc.c
===================================================================
--- FFmpeg.orig/libavformat/movenc.c
+++ FFmpeg/libavformat/movenc.c
@@ -6929,7 +6929,7 @@ int ff_mov_write_packet(AVFormatContext
} else if (par->codec_id == AV_CODEC_ID_HEVC && trk->extradata_size[trk->last_stsd_index] > 6 &&
(AV_RB24(trk->extradata[trk->last_stsd_index]) == 1 || AV_RB32(trk->extradata[trk->last_stsd_index]) == 1)) {
/* extradata is Annex B, assume the bitstream is too and convert it */
- int filter_ps = (trk->tag == MKTAG('h','v','c','1'));
+ int filter_ps = 0; // Always disable it as a Wa for AMD hevc_vaapi encoder
if (trk->hint_track >= 0 && trk->hint_track < mov->nb_tracks) {
ret = ff_hevc_annexb2mp4_buf(pkt->data, &reformatted_data,
&size, filter_ps, NULL);
debian/patches/0094-backport-a-fix-from-upstream-for-magicyuv-decoder.patch → debian/patches/0087-backport-a-fix-from-upstream-for-magicyuv-decoder.patch
Vendored
View File

Some files were not shown because too many files have changed in this diff Show More