This hwaccel is interesting because it also works for wmv3/9 content,
which is not supported by the nvidia parser used by cuviddec.
... | ... |
@@ -13,7 +13,7 @@ version <next>: |
13 | 13 |
- PCE support for extended channel layouts in the AAC encoder |
14 | 14 |
- native aptX encoder and decoder |
15 | 15 |
- Raw aptX muxer and demuxer |
16 |
-- NVIDIA NVDEC-accelerated H.264, HEVC and VP9 hwaccel decoding |
|
16 |
+- NVIDIA NVDEC-accelerated H.264, HEVC, VC1 and VP9 hwaccel decoding |
|
17 | 17 |
- Intel QSV-accelerated overlay filter |
18 | 18 |
|
19 | 19 |
|
... | ... |
@@ -2740,6 +2740,8 @@ vc1_d3d11va2_hwaccel_select="vc1_decoder" |
2740 | 2740 |
vc1_dxva2_hwaccel_deps="dxva2" |
2741 | 2741 |
vc1_dxva2_hwaccel_select="vc1_decoder" |
2742 | 2742 |
vc1_mmal_hwaccel_deps="mmal" |
2743 |
+vc1_nvdec_hwaccel_deps="nvdec" |
|
2744 |
+vc1_nvdec_hwaccel_select="vc1_decoder" |
|
2743 | 2745 |
vc1_qsv_hwaccel_deps="libmfx" |
2744 | 2746 |
vc1_vaapi_hwaccel_deps="vaapi" |
2745 | 2747 |
vc1_vaapi_hwaccel_select="vc1_decoder" |
... | ... |
@@ -2763,6 +2765,7 @@ vp9_vaapi_hwaccel_select="vp9_decoder" |
2763 | 2763 |
wmv3_d3d11va_hwaccel_select="vc1_d3d11va_hwaccel" |
2764 | 2764 |
wmv3_d3d11va2_hwaccel_select="vc1_d3d11va2_hwaccel" |
2765 | 2765 |
wmv3_dxva2_hwaccel_select="vc1_dxva2_hwaccel" |
2766 |
+wmv3_nvdec_hwaccel_select="vc1_nvdec_hwaccel" |
|
2766 | 2767 |
wmv3_vaapi_hwaccel_select="vc1_vaapi_hwaccel" |
2767 | 2768 |
wmv3_vdpau_hwaccel_select="vc1_vdpau_hwaccel" |
2768 | 2769 |
|
... | ... |
@@ -864,6 +864,7 @@ OBJS-$(CONFIG_MPEG4_VDPAU_HWACCEL) += vdpau_mpeg4.o |
864 | 864 |
OBJS-$(CONFIG_MPEG4_VIDEOTOOLBOX_HWACCEL) += videotoolbox.o |
865 | 865 |
OBJS-$(CONFIG_VC1_D3D11VA_HWACCEL) += dxva2_vc1.o |
866 | 866 |
OBJS-$(CONFIG_VC1_DXVA2_HWACCEL) += dxva2_vc1.o |
867 |
+OBJS-$(CONFIG_VC1_NVDEC_HWACCEL) += nvdec_vc1.o |
|
867 | 868 |
OBJS-$(CONFIG_VC1_QSV_HWACCEL) += qsvdec_other.o |
868 | 869 |
OBJS-$(CONFIG_VC1_VAAPI_HWACCEL) += vaapi_vc1.o |
869 | 870 |
OBJS-$(CONFIG_VC1_VDPAU_HWACCEL) += vdpau_vc1.o |
... | ... |
@@ -111,6 +111,7 @@ static void register_all(void) |
111 | 111 |
REGISTER_HWACCEL(VC1_D3D11VA, vc1_d3d11va); |
112 | 112 |
REGISTER_HWACCEL(VC1_D3D11VA2, vc1_d3d11va2); |
113 | 113 |
REGISTER_HWACCEL(VC1_DXVA2, vc1_dxva2); |
114 |
+ REGISTER_HWACCEL(VC1_NVDEC, vc1_nvdec); |
|
114 | 115 |
REGISTER_HWACCEL(VC1_VAAPI, vc1_vaapi); |
115 | 116 |
REGISTER_HWACCEL(VC1_VDPAU, vc1_vdpau); |
116 | 117 |
REGISTER_HWACCEL(VC1_MMAL, vc1_mmal); |
... | ... |
@@ -128,6 +129,7 @@ static void register_all(void) |
128 | 128 |
REGISTER_HWACCEL(WMV3_D3D11VA, wmv3_d3d11va); |
129 | 129 |
REGISTER_HWACCEL(WMV3_D3D11VA2, wmv3_d3d11va2); |
130 | 130 |
REGISTER_HWACCEL(WMV3_DXVA2, wmv3_dxva2); |
131 |
+ REGISTER_HWACCEL(WMV3_NVDEC, wmv3_nvdec); |
|
131 | 132 |
REGISTER_HWACCEL(WMV3_VAAPI, wmv3_vaapi); |
132 | 133 |
REGISTER_HWACCEL(WMV3_VDPAU, wmv3_vdpau); |
133 | 134 |
|
... | ... |
@@ -54,7 +54,9 @@ static int map_avcodec_id(enum AVCodecID id) |
54 | 54 |
switch (id) { |
55 | 55 |
case AV_CODEC_ID_H264: return cudaVideoCodec_H264; |
56 | 56 |
case AV_CODEC_ID_HEVC: return cudaVideoCodec_HEVC; |
57 |
+ case AV_CODEC_ID_VC1: return cudaVideoCodec_VC1; |
|
57 | 58 |
case AV_CODEC_ID_VP9: return cudaVideoCodec_VP9; |
59 |
+ case AV_CODEC_ID_WMV3: return cudaVideoCodec_VC1; |
|
58 | 60 |
} |
59 | 61 |
return -1; |
60 | 62 |
} |
61 | 63 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,184 @@ |
0 |
+/* |
|
1 |
+ * VC1 HW decode acceleration through NVDEC |
|
2 |
+ * |
|
3 |
+ * Copyright (c) 2017 Philip Langdale |
|
4 |
+ * |
|
5 |
+ * This file is part of FFmpeg. |
|
6 |
+ * |
|
7 |
+ * FFmpeg is free software; you can redistribute it and/or |
|
8 |
+ * modify it under the terms of the GNU Lesser General Public |
|
9 |
+ * License as published by the Free Software Foundation; either |
|
10 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
11 |
+ * |
|
12 |
+ * FFmpeg is distributed in the hope that it will be useful, |
|
13 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
14 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
15 |
+ * Lesser General Public License for more details. |
|
16 |
+ * |
|
17 |
+ * You should have received a copy of the GNU Lesser General Public |
|
18 |
+ * License along with FFmpeg; if not, write to the Free Software |
|
19 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
20 |
+ */ |
|
21 |
+ |
|
22 |
+#include "avcodec.h" |
|
23 |
+#include "nvdec.h" |
|
24 |
+#include "decode.h" |
|
25 |
+#include "vc1.h" |
|
26 |
+ |
|
27 |
+static unsigned char get_ref_idx(AVFrame *frame) |
|
28 |
+{ |
|
29 |
+ FrameDecodeData *fdd; |
|
30 |
+ NVDECFrame *cf; |
|
31 |
+ |
|
32 |
+ if (!frame || !frame->private_ref) |
|
33 |
+ return 255; |
|
34 |
+ |
|
35 |
+ fdd = (FrameDecodeData*)frame->private_ref->data; |
|
36 |
+ cf = (NVDECFrame*)fdd->hwaccel_priv; |
|
37 |
+ |
|
38 |
+ return cf->idx; |
|
39 |
+} |
|
40 |
+ |
|
41 |
+static int nvdec_vc1_start_frame(AVCodecContext *avctx, const uint8_t *buffer, uint32_t size) |
|
42 |
+{ |
|
43 |
+ VC1Context *v = avctx->priv_data; |
|
44 |
+ MpegEncContext *s = &v->s; |
|
45 |
+ |
|
46 |
+ NVDECContext *ctx = avctx->internal->hwaccel_priv_data; |
|
47 |
+ CUVIDPICPARAMS *pp = &ctx->pic_params; |
|
48 |
+ FrameDecodeData *fdd; |
|
49 |
+ NVDECFrame *cf; |
|
50 |
+ AVFrame *cur_frame = s->current_picture.f; |
|
51 |
+ |
|
52 |
+ int ret; |
|
53 |
+ |
|
54 |
+ ret = ff_nvdec_start_frame(avctx, cur_frame); |
|
55 |
+ if (ret < 0) |
|
56 |
+ return ret; |
|
57 |
+ |
|
58 |
+ fdd = (FrameDecodeData*)cur_frame->private_ref->data; |
|
59 |
+ cf = (NVDECFrame*)fdd->hwaccel_priv; |
|
60 |
+ |
|
61 |
+ *pp = (CUVIDPICPARAMS) { |
|
62 |
+ .PicWidthInMbs = (cur_frame->width + 15) / 16, |
|
63 |
+ .FrameHeightInMbs = (cur_frame->height + 15) / 16, |
|
64 |
+ .CurrPicIdx = cf->idx, |
|
65 |
+ .field_pic_flag = v->field_mode, |
|
66 |
+ .bottom_field_flag = v->cur_field_type, |
|
67 |
+ .second_field = v->second_field, |
|
68 |
+ |
|
69 |
+ .intra_pic_flag = s->pict_type == AV_PICTURE_TYPE_I || |
|
70 |
+ s->pict_type == AV_PICTURE_TYPE_BI, |
|
71 |
+ .ref_pic_flag = s->pict_type == AV_PICTURE_TYPE_I || |
|
72 |
+ s->pict_type == AV_PICTURE_TYPE_P, |
|
73 |
+ |
|
74 |
+ .CodecSpecific.vc1 = { |
|
75 |
+ .ForwardRefIdx = get_ref_idx(s->last_picture.f), |
|
76 |
+ .BackwardRefIdx = get_ref_idx(s->next_picture.f), |
|
77 |
+ .FrameWidth = cur_frame->width, |
|
78 |
+ .FrameHeight = cur_frame->height, |
|
79 |
+ |
|
80 |
+ .intra_pic_flag = s->pict_type == AV_PICTURE_TYPE_I || |
|
81 |
+ s->pict_type == AV_PICTURE_TYPE_BI, |
|
82 |
+ .ref_pic_flag = s->pict_type == AV_PICTURE_TYPE_I || |
|
83 |
+ s->pict_type == AV_PICTURE_TYPE_P, |
|
84 |
+ .progressive_fcm = v->fcm == 0, |
|
85 |
+ |
|
86 |
+ .profile = v->profile, |
|
87 |
+ .postprocflag = v->postprocflag, |
|
88 |
+ .pulldown = v->broadcast, |
|
89 |
+ .interlace = v->interlace, |
|
90 |
+ .tfcntrflag = v->tfcntrflag, |
|
91 |
+ .finterpflag = v->finterpflag, |
|
92 |
+ .psf = v->psf, |
|
93 |
+ .multires = v->multires, |
|
94 |
+ .syncmarker = v->resync_marker, |
|
95 |
+ .rangered = v->rangered, |
|
96 |
+ .maxbframes = s->max_b_frames, |
|
97 |
+ |
|
98 |
+ .panscan_flag = v->panscanflag, |
|
99 |
+ .refdist_flag = v->refdist_flag, |
|
100 |
+ .extended_mv = v->extended_mv, |
|
101 |
+ .dquant = v->dquant, |
|
102 |
+ .vstransform = v->vstransform, |
|
103 |
+ .loopfilter = v->s.loop_filter, |
|
104 |
+ .fastuvmc = v->fastuvmc, |
|
105 |
+ .overlap = v->overlap, |
|
106 |
+ .quantizer = v->quantizer_mode, |
|
107 |
+ .extended_dmv = v->extended_dmv, |
|
108 |
+ .range_mapy_flag = v->range_mapy_flag, |
|
109 |
+ .range_mapy = v->range_mapy, |
|
110 |
+ .range_mapuv_flag = v->range_mapuv_flag, |
|
111 |
+ .range_mapuv = v->range_mapuv, |
|
112 |
+ .rangeredfrm = v->rangeredfrm, |
|
113 |
+ } |
|
114 |
+ }; |
|
115 |
+ |
|
116 |
+ return 0; |
|
117 |
+} |
|
118 |
+ |
|
119 |
+static int nvdec_vc1_end_frame(AVCodecContext *avctx) |
|
120 |
+{ |
|
121 |
+ NVDECContext *ctx = avctx->internal->hwaccel_priv_data; |
|
122 |
+ int ret = ff_nvdec_end_frame(avctx); |
|
123 |
+ ctx->bitstream = NULL; |
|
124 |
+ return ret; |
|
125 |
+} |
|
126 |
+ |
|
127 |
+static int nvdec_vc1_decode_slice(AVCodecContext *avctx, const uint8_t *buffer, uint32_t size) |
|
128 |
+{ |
|
129 |
+ NVDECContext *ctx = avctx->internal->hwaccel_priv_data; |
|
130 |
+ void *tmp; |
|
131 |
+ |
|
132 |
+ tmp = av_fast_realloc(ctx->slice_offsets, &ctx->slice_offsets_allocated, |
|
133 |
+ (ctx->nb_slices + 1) * sizeof(*ctx->slice_offsets)); |
|
134 |
+ if (!tmp) |
|
135 |
+ return AVERROR(ENOMEM); |
|
136 |
+ ctx->slice_offsets = tmp; |
|
137 |
+ |
|
138 |
+ if (!ctx->bitstream) |
|
139 |
+ ctx->bitstream = (uint8_t*)buffer; |
|
140 |
+ |
|
141 |
+ ctx->slice_offsets[ctx->nb_slices] = buffer - ctx->bitstream; |
|
142 |
+ ctx->bitstream_len += size; |
|
143 |
+ ctx->nb_slices++; |
|
144 |
+ |
|
145 |
+ return 0; |
|
146 |
+} |
|
147 |
+ |
|
148 |
+static int nvdec_vc1_frame_params(AVCodecContext *avctx, |
|
149 |
+ AVBufferRef *hw_frames_ctx) |
|
150 |
+{ |
|
151 |
+ // Each frame can at most have one P and one B reference |
|
152 |
+ return ff_nvdec_frame_params(avctx, hw_frames_ctx, 2); |
|
153 |
+} |
|
154 |
+ |
|
155 |
+AVHWAccel ff_vc1_nvdec_hwaccel = { |
|
156 |
+ .name = "vc1_nvdec", |
|
157 |
+ .type = AVMEDIA_TYPE_VIDEO, |
|
158 |
+ .id = AV_CODEC_ID_VC1, |
|
159 |
+ .pix_fmt = AV_PIX_FMT_CUDA, |
|
160 |
+ .start_frame = nvdec_vc1_start_frame, |
|
161 |
+ .end_frame = nvdec_vc1_end_frame, |
|
162 |
+ .decode_slice = nvdec_vc1_decode_slice, |
|
163 |
+ .frame_params = nvdec_vc1_frame_params, |
|
164 |
+ .init = ff_nvdec_decode_init, |
|
165 |
+ .uninit = ff_nvdec_decode_uninit, |
|
166 |
+ .priv_data_size = sizeof(NVDECContext), |
|
167 |
+}; |
|
168 |
+ |
|
169 |
+#if CONFIG_WMV3_NVDEC_HWACCEL |
|
170 |
+AVHWAccel ff_wmv3_nvdec_hwaccel = { |
|
171 |
+ .name = "wmv3_nvdec", |
|
172 |
+ .type = AVMEDIA_TYPE_VIDEO, |
|
173 |
+ .id = AV_CODEC_ID_WMV3, |
|
174 |
+ .pix_fmt = AV_PIX_FMT_CUDA, |
|
175 |
+ .start_frame = nvdec_vc1_start_frame, |
|
176 |
+ .end_frame = nvdec_vc1_end_frame, |
|
177 |
+ .decode_slice = nvdec_vc1_decode_slice, |
|
178 |
+ .frame_params = nvdec_vc1_frame_params, |
|
179 |
+ .init = ff_nvdec_decode_init, |
|
180 |
+ .uninit = ff_nvdec_decode_uninit, |
|
181 |
+ .priv_data_size = sizeof(NVDECContext), |
|
182 |
+}; |
|
183 |
+#endif |
... | ... |
@@ -1119,6 +1119,9 @@ static const enum AVPixelFormat vc1_hwaccel_pixfmt_list_420[] = { |
1119 | 1119 |
AV_PIX_FMT_D3D11VA_VLD, |
1120 | 1120 |
AV_PIX_FMT_D3D11, |
1121 | 1121 |
#endif |
1122 |
+#if CONFIG_VC1_NVDEC_HWACCEL |
|
1123 |
+ AV_PIX_FMT_CUDA, |
|
1124 |
+#endif |
|
1122 | 1125 |
#if CONFIG_VC1_VAAPI_HWACCEL |
1123 | 1126 |
AV_PIX_FMT_VAAPI, |
1124 | 1127 |
#endif |
... | ... |
@@ -29,7 +29,7 @@ |
29 | 29 |
|
30 | 30 |
#define LIBAVCODEC_VERSION_MAJOR 58 |
31 | 31 |
#define LIBAVCODEC_VERSION_MINOR 3 |
32 |
-#define LIBAVCODEC_VERSION_MICRO 101 |
|
32 |
+#define LIBAVCODEC_VERSION_MICRO 102 |
|
33 | 33 |
|
34 | 34 |
#define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ |
35 | 35 |
LIBAVCODEC_VERSION_MINOR, \ |