| 1 | /* |
| 2 | * Copyright (c) 2012, Xidorn Quan |
| 3 | * |
| 4 | * This file is part of FFmpeg. |
| 5 | * |
| 6 | * FFmpeg is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU Lesser General Public |
| 8 | * License as published by the Free Software Foundation; either |
| 9 | * version 2.1 of the License, or (at your option) any later version. |
| 10 | * |
| 11 | * FFmpeg is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 14 | * Lesser General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU Lesser General Public |
| 17 | * License along with FFmpeg; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 19 | */ |
| 20 | |
| 21 | /** |
| 22 | * @file |
| 23 | * H.264 decoder via VDA |
| 24 | * @author Xidorn Quan <quanxunzhen@gmail.com> |
| 25 | */ |
| 26 | |
| 27 | #include <string.h> |
| 28 | #include <CoreFoundation/CoreFoundation.h> |
| 29 | |
| 30 | #include "vda.h" |
| 31 | #include "h264.h" |
| 32 | #include "avcodec.h" |
| 33 | |
| 34 | #ifndef kCFCoreFoundationVersionNumber10_7 |
| 35 | #define kCFCoreFoundationVersionNumber10_7 635.00 |
| 36 | #endif |
| 37 | |
| 38 | extern AVCodec ff_h264_decoder, ff_h264_vda_decoder; |
| 39 | |
| 40 | static const enum AVPixelFormat vda_pixfmts_prior_10_7[] = { |
| 41 | AV_PIX_FMT_UYVY422, |
| 42 | AV_PIX_FMT_YUV420P, |
| 43 | AV_PIX_FMT_NONE |
| 44 | }; |
| 45 | |
| 46 | static const enum AVPixelFormat vda_pixfmts[] = { |
| 47 | AV_PIX_FMT_UYVY422, |
| 48 | AV_PIX_FMT_YUYV422, |
| 49 | AV_PIX_FMT_NV12, |
| 50 | AV_PIX_FMT_YUV420P, |
| 51 | AV_PIX_FMT_NONE |
| 52 | }; |
| 53 | |
| 54 | typedef struct { |
| 55 | H264Context h264ctx; |
| 56 | int h264_initialized; |
| 57 | struct vda_context vda_ctx; |
| 58 | enum AVPixelFormat pix_fmt; |
| 59 | |
| 60 | /* for backing-up fields set by user. |
| 61 | * we have to gain full control of such fields here */ |
| 62 | void *hwaccel_context; |
| 63 | enum AVPixelFormat (*get_format)(struct AVCodecContext *s, const enum AVPixelFormat * fmt); |
| 64 | int (*get_buffer2)(struct AVCodecContext *s, AVFrame *frame, int flags); |
| 65 | #if FF_API_GET_BUFFER |
| 66 | int (*get_buffer)(struct AVCodecContext *c, AVFrame *pic); |
| 67 | #endif |
| 68 | } VDADecoderContext; |
| 69 | |
| 70 | static enum AVPixelFormat get_format(struct AVCodecContext *avctx, |
| 71 | const enum AVPixelFormat *fmt) |
| 72 | { |
| 73 | return AV_PIX_FMT_VDA_VLD; |
| 74 | } |
| 75 | |
| 76 | typedef struct { |
| 77 | CVPixelBufferRef cv_buffer; |
| 78 | } VDABufferContext; |
| 79 | |
| 80 | static void release_buffer(void *opaque, uint8_t *data) |
| 81 | { |
| 82 | VDABufferContext *context = opaque; |
| 83 | CVPixelBufferUnlockBaseAddress(context->cv_buffer, 0); |
| 84 | CVPixelBufferRelease(context->cv_buffer); |
| 85 | av_free(context); |
| 86 | } |
| 87 | |
| 88 | static int get_buffer2(AVCodecContext *avctx, AVFrame *pic, int flag) |
| 89 | { |
| 90 | VDABufferContext *context = av_mallocz(sizeof(VDABufferContext)); |
| 91 | AVBufferRef *buffer = av_buffer_create(NULL, 0, release_buffer, context, 0); |
| 92 | if (!context || !buffer) { |
| 93 | av_free(context); |
| 94 | return AVERROR(ENOMEM); |
| 95 | } |
| 96 | |
| 97 | pic->buf[0] = buffer; |
| 98 | pic->data[0] = (void *)1; |
| 99 | return 0; |
| 100 | } |
| 101 | |
| 102 | static inline void set_context(AVCodecContext *avctx) |
| 103 | { |
| 104 | VDADecoderContext *ctx = avctx->priv_data; |
| 105 | ctx->hwaccel_context = avctx->hwaccel_context; |
| 106 | avctx->hwaccel_context = &ctx->vda_ctx; |
| 107 | ctx->get_format = avctx->get_format; |
| 108 | avctx->get_format = get_format; |
| 109 | ctx->get_buffer2 = avctx->get_buffer2; |
| 110 | avctx->get_buffer2 = get_buffer2; |
| 111 | #if FF_API_GET_BUFFER |
| 112 | ctx->get_buffer = avctx->get_buffer; |
| 113 | avctx->get_buffer = NULL; |
| 114 | #endif |
| 115 | } |
| 116 | |
| 117 | static inline void restore_context(AVCodecContext *avctx) |
| 118 | { |
| 119 | VDADecoderContext *ctx = avctx->priv_data; |
| 120 | avctx->hwaccel_context = ctx->hwaccel_context; |
| 121 | avctx->get_format = ctx->get_format; |
| 122 | avctx->get_buffer2 = ctx->get_buffer2; |
| 123 | #if FF_API_GET_BUFFER |
| 124 | avctx->get_buffer = ctx->get_buffer; |
| 125 | #endif |
| 126 | } |
| 127 | |
| 128 | static int vdadec_decode(AVCodecContext *avctx, |
| 129 | void *data, int *got_frame, AVPacket *avpkt) |
| 130 | { |
| 131 | VDADecoderContext *ctx = avctx->priv_data; |
| 132 | AVFrame *pic = data; |
| 133 | int ret; |
| 134 | |
| 135 | set_context(avctx); |
| 136 | ret = ff_h264_decoder.decode(avctx, data, got_frame, avpkt); |
| 137 | restore_context(avctx); |
| 138 | if (*got_frame) { |
| 139 | AVBufferRef *buffer = pic->buf[0]; |
| 140 | VDABufferContext *context = av_buffer_get_opaque(buffer); |
| 141 | CVPixelBufferRef cv_buffer = (CVPixelBufferRef)pic->data[3]; |
| 142 | |
| 143 | CVPixelBufferRetain(cv_buffer); |
| 144 | CVPixelBufferLockBaseAddress(cv_buffer, 0); |
| 145 | context->cv_buffer = cv_buffer; |
| 146 | pic->format = ctx->pix_fmt; |
| 147 | if (CVPixelBufferIsPlanar(cv_buffer)) { |
| 148 | int i, count = CVPixelBufferGetPlaneCount(cv_buffer); |
| 149 | av_assert0(count < 4); |
| 150 | for (i = 0; i < count; i++) { |
| 151 | pic->data[i] = CVPixelBufferGetBaseAddressOfPlane(cv_buffer, i); |
| 152 | pic->linesize[i] = CVPixelBufferGetBytesPerRowOfPlane(cv_buffer, i); |
| 153 | } |
| 154 | } else { |
| 155 | pic->data[0] = CVPixelBufferGetBaseAddress(cv_buffer); |
| 156 | pic->linesize[0] = CVPixelBufferGetBytesPerRow(cv_buffer); |
| 157 | } |
| 158 | } |
| 159 | avctx->pix_fmt = ctx->pix_fmt; |
| 160 | |
| 161 | return ret; |
| 162 | } |
| 163 | |
| 164 | static av_cold int vdadec_close(AVCodecContext *avctx) |
| 165 | { |
| 166 | VDADecoderContext *ctx = avctx->priv_data; |
| 167 | /* release buffers and decoder */ |
| 168 | ff_vda_destroy_decoder(&ctx->vda_ctx); |
| 169 | /* close H.264 decoder */ |
| 170 | if (ctx->h264_initialized) { |
| 171 | set_context(avctx); |
| 172 | ff_h264_decoder.close(avctx); |
| 173 | restore_context(avctx); |
| 174 | } |
| 175 | return 0; |
| 176 | } |
| 177 | |
| 178 | static av_cold int vdadec_init(AVCodecContext *avctx) |
| 179 | { |
| 180 | VDADecoderContext *ctx = avctx->priv_data; |
| 181 | struct vda_context *vda_ctx = &ctx->vda_ctx; |
| 182 | OSStatus status; |
| 183 | int ret, i; |
| 184 | |
| 185 | ctx->h264_initialized = 0; |
| 186 | |
| 187 | /* init pix_fmts of codec */ |
| 188 | if (!ff_h264_vda_decoder.pix_fmts) { |
| 189 | if (kCFCoreFoundationVersionNumber < kCFCoreFoundationVersionNumber10_7) |
| 190 | ff_h264_vda_decoder.pix_fmts = vda_pixfmts_prior_10_7; |
| 191 | else |
| 192 | ff_h264_vda_decoder.pix_fmts = vda_pixfmts; |
| 193 | } |
| 194 | |
| 195 | /* init vda */ |
| 196 | memset(vda_ctx, 0, sizeof(struct vda_context)); |
| 197 | vda_ctx->width = avctx->width; |
| 198 | vda_ctx->height = avctx->height; |
| 199 | vda_ctx->format = 'avc1'; |
| 200 | vda_ctx->use_sync_decoding = 1; |
| 201 | vda_ctx->use_ref_buffer = 1; |
| 202 | ctx->pix_fmt = avctx->get_format(avctx, avctx->codec->pix_fmts); |
| 203 | switch (ctx->pix_fmt) { |
| 204 | case AV_PIX_FMT_UYVY422: |
| 205 | vda_ctx->cv_pix_fmt_type = '2vuy'; |
| 206 | break; |
| 207 | case AV_PIX_FMT_YUYV422: |
| 208 | vda_ctx->cv_pix_fmt_type = 'yuvs'; |
| 209 | break; |
| 210 | case AV_PIX_FMT_NV12: |
| 211 | vda_ctx->cv_pix_fmt_type = '420v'; |
| 212 | break; |
| 213 | case AV_PIX_FMT_YUV420P: |
| 214 | vda_ctx->cv_pix_fmt_type = 'y420'; |
| 215 | break; |
| 216 | default: |
| 217 | av_log(avctx, AV_LOG_ERROR, "Unsupported pixel format: %d\n", avctx->pix_fmt); |
| 218 | goto failed; |
| 219 | } |
| 220 | status = ff_vda_create_decoder(vda_ctx, |
| 221 | avctx->extradata, avctx->extradata_size); |
| 222 | if (status != kVDADecoderNoErr) { |
| 223 | av_log(avctx, AV_LOG_ERROR, |
| 224 | "Failed to init VDA decoder: %d.\n", status); |
| 225 | goto failed; |
| 226 | } |
| 227 | |
| 228 | /* init H.264 decoder */ |
| 229 | set_context(avctx); |
| 230 | ret = ff_h264_decoder.init(avctx); |
| 231 | restore_context(avctx); |
| 232 | if (ret < 0) { |
| 233 | av_log(avctx, AV_LOG_ERROR, "Failed to open H.264 decoder.\n"); |
| 234 | goto failed; |
| 235 | } |
| 236 | ctx->h264_initialized = 1; |
| 237 | |
| 238 | for (i = 0; i < MAX_SPS_COUNT; i++) { |
| 239 | SPS *sps = ctx->h264ctx.sps_buffers[i]; |
| 240 | if (sps && (sps->bit_depth_luma != 8 || |
| 241 | sps->chroma_format_idc == 2 || |
| 242 | sps->chroma_format_idc == 3)) { |
| 243 | av_log(avctx, AV_LOG_ERROR, "Format is not supported.\n"); |
| 244 | goto failed; |
| 245 | } |
| 246 | } |
| 247 | |
| 248 | return 0; |
| 249 | |
| 250 | failed: |
| 251 | vdadec_close(avctx); |
| 252 | return -1; |
| 253 | } |
| 254 | |
| 255 | static void vdadec_flush(AVCodecContext *avctx) |
| 256 | { |
| 257 | set_context(avctx); |
| 258 | ff_h264_decoder.flush(avctx); |
| 259 | restore_context(avctx); |
| 260 | } |
| 261 | |
| 262 | AVCodec ff_h264_vda_decoder = { |
| 263 | .name = "h264_vda", |
| 264 | .type = AVMEDIA_TYPE_VIDEO, |
| 265 | .id = AV_CODEC_ID_H264, |
| 266 | .priv_data_size = sizeof(VDADecoderContext), |
| 267 | .init = vdadec_init, |
| 268 | .close = vdadec_close, |
| 269 | .decode = vdadec_decode, |
| 270 | .capabilities = CODEC_CAP_DELAY, |
| 271 | .flush = vdadec_flush, |
| 272 | .long_name = NULL_IF_CONFIG_SMALL("H.264 (VDA acceleration)"), |
| 273 | }; |