blob: 7b88ec7015f5431b48a20684e81a94fd2547140b
1 | /* |
2 | * VDA H264 HW acceleration. |
3 | * |
4 | * copyright (c) 2011 Sebastien Zwickert |
5 | * |
6 | * This file is part of FFmpeg. |
7 | * |
8 | * FFmpeg is free software; you can redistribute it and/or |
9 | * modify it under the terms of the GNU Lesser General Public |
10 | * License as published by the Free Software Foundation; either |
11 | * version 2.1 of the License, or (at your option) any later version. |
12 | * |
13 | * FFmpeg is distributed in the hope that it will be useful, |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
16 | * Lesser General Public License for more details. |
17 | * |
18 | * You should have received a copy of the GNU Lesser General Public |
19 | * License along with FFmpeg; if not, write to the Free Software |
20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
21 | */ |
22 | |
23 | #include <CoreFoundation/CFDictionary.h> |
24 | #include <CoreFoundation/CFNumber.h> |
25 | #include <CoreFoundation/CFData.h> |
26 | |
27 | #include "vda.h" |
28 | #include "libavutil/avutil.h" |
29 | #include "h264dec.h" |
30 | |
31 | struct vda_buffer { |
32 | CVPixelBufferRef cv_buffer; |
33 | }; |
34 | #include "internal.h" |
35 | #include "vda_vt_internal.h" |
36 | |
37 | /* Decoder callback that adds the vda frame to the queue in display order. */ |
38 | static void vda_decoder_callback(void *vda_hw_ctx, |
39 | CFDictionaryRef user_info, |
40 | OSStatus status, |
41 | uint32_t infoFlags, |
42 | CVImageBufferRef image_buffer) |
43 | { |
44 | struct vda_context *vda_ctx = vda_hw_ctx; |
45 | |
46 | if (infoFlags & kVDADecodeInfo_FrameDropped) |
47 | vda_ctx->cv_buffer = NULL; |
48 | |
49 | if (!image_buffer) |
50 | return; |
51 | |
52 | if (vda_ctx->cv_pix_fmt_type != CVPixelBufferGetPixelFormatType(image_buffer)) |
53 | return; |
54 | |
55 | vda_ctx->cv_buffer = CVPixelBufferRetain(image_buffer); |
56 | } |
57 | |
58 | static int vda_sync_decode(VTContext *ctx, struct vda_context *vda_ctx) |
59 | { |
60 | OSStatus status; |
61 | CFDataRef coded_frame; |
62 | uint32_t flush_flags = 1 << 0; ///< kVDADecoderFlush_emitFrames |
63 | |
64 | coded_frame = CFDataCreate(kCFAllocatorDefault, |
65 | ctx->bitstream, |
66 | ctx->bitstream_size); |
67 | |
68 | status = VDADecoderDecode(vda_ctx->decoder, 0, coded_frame, NULL); |
69 | |
70 | if (kVDADecoderNoErr == status) |
71 | status = VDADecoderFlush(vda_ctx->decoder, flush_flags); |
72 | |
73 | CFRelease(coded_frame); |
74 | |
75 | return status; |
76 | } |
77 | |
78 | |
79 | static int vda_old_h264_start_frame(AVCodecContext *avctx, |
80 | av_unused const uint8_t *buffer, |
81 | av_unused uint32_t size) |
82 | { |
83 | VTContext *vda = avctx->internal->hwaccel_priv_data; |
84 | struct vda_context *vda_ctx = avctx->hwaccel_context; |
85 | |
86 | if (!vda_ctx->decoder) |
87 | return -1; |
88 | |
89 | vda->bitstream_size = 0; |
90 | |
91 | return 0; |
92 | } |
93 | |
94 | static int vda_old_h264_decode_slice(AVCodecContext *avctx, |
95 | const uint8_t *buffer, |
96 | uint32_t size) |
97 | { |
98 | VTContext *vda = avctx->internal->hwaccel_priv_data; |
99 | struct vda_context *vda_ctx = avctx->hwaccel_context; |
100 | void *tmp; |
101 | |
102 | if (!vda_ctx->decoder) |
103 | return -1; |
104 | |
105 | tmp = av_fast_realloc(vda->bitstream, |
106 | &vda->allocated_size, |
107 | vda->bitstream_size + size + 4); |
108 | if (!tmp) |
109 | return AVERROR(ENOMEM); |
110 | |
111 | vda->bitstream = tmp; |
112 | |
113 | AV_WB32(vda->bitstream + vda->bitstream_size, size); |
114 | memcpy(vda->bitstream + vda->bitstream_size + 4, buffer, size); |
115 | |
116 | vda->bitstream_size += size + 4; |
117 | |
118 | return 0; |
119 | } |
120 | |
121 | static void vda_h264_release_buffer(void *opaque, uint8_t *data) |
122 | { |
123 | struct vda_buffer *context = opaque; |
124 | CVPixelBufferRelease(context->cv_buffer); |
125 | av_free(context); |
126 | } |
127 | |
128 | static int vda_old_h264_end_frame(AVCodecContext *avctx) |
129 | { |
130 | H264Context *h = avctx->priv_data; |
131 | VTContext *vda = avctx->internal->hwaccel_priv_data; |
132 | struct vda_context *vda_ctx = avctx->hwaccel_context; |
133 | AVFrame *frame = h->cur_pic_ptr->f; |
134 | struct vda_buffer *context; |
135 | AVBufferRef *buffer; |
136 | int status; |
137 | |
138 | if (!vda_ctx->decoder || !vda->bitstream) |
139 | return -1; |
140 | |
141 | status = vda_sync_decode(vda, vda_ctx); |
142 | frame->data[3] = (void*)vda_ctx->cv_buffer; |
143 | |
144 | if (status) |
145 | av_log(avctx, AV_LOG_ERROR, "Failed to decode frame (%d)\n", status); |
146 | |
147 | if (!vda_ctx->use_ref_buffer || status) |
148 | return status; |
149 | |
150 | context = av_mallocz(sizeof(*context)); |
151 | buffer = av_buffer_create(NULL, 0, vda_h264_release_buffer, context, 0); |
152 | if (!context || !buffer) { |
153 | CVPixelBufferRelease(vda_ctx->cv_buffer); |
154 | av_free(context); |
155 | return -1; |
156 | } |
157 | |
158 | context->cv_buffer = vda_ctx->cv_buffer; |
159 | frame->buf[3] = buffer; |
160 | |
161 | return status; |
162 | } |
163 | |
164 | int ff_vda_create_decoder(struct vda_context *vda_ctx, |
165 | uint8_t *extradata, |
166 | int extradata_size) |
167 | { |
168 | OSStatus status; |
169 | CFNumberRef height; |
170 | CFNumberRef width; |
171 | CFNumberRef format; |
172 | CFDataRef avc_data; |
173 | CFMutableDictionaryRef config_info; |
174 | CFMutableDictionaryRef buffer_attributes; |
175 | CFMutableDictionaryRef io_surface_properties; |
176 | CFNumberRef cv_pix_fmt; |
177 | |
178 | vda_ctx->priv_bitstream = NULL; |
179 | vda_ctx->priv_allocated_size = 0; |
180 | |
181 | /* Each VCL NAL in the bitstream sent to the decoder |
182 | * is preceded by a 4 bytes length header. |
183 | * Change the avcC atom header if needed, to signal headers of 4 bytes. */ |
184 | if (extradata_size >= 4 && (extradata[4] & 0x03) != 0x03) { |
185 | uint8_t *rw_extradata; |
186 | |
187 | if (!(rw_extradata = av_malloc(extradata_size))) |
188 | return AVERROR(ENOMEM); |
189 | |
190 | memcpy(rw_extradata, extradata, extradata_size); |
191 | |
192 | rw_extradata[4] |= 0x03; |
193 | |
194 | avc_data = CFDataCreate(kCFAllocatorDefault, rw_extradata, extradata_size); |
195 | |
196 | av_freep(&rw_extradata); |
197 | } else { |
198 | avc_data = CFDataCreate(kCFAllocatorDefault, extradata, extradata_size); |
199 | } |
200 | |
201 | config_info = CFDictionaryCreateMutable(kCFAllocatorDefault, |
202 | 4, |
203 | &kCFTypeDictionaryKeyCallBacks, |
204 | &kCFTypeDictionaryValueCallBacks); |
205 | |
206 | height = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &vda_ctx->height); |
207 | width = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &vda_ctx->width); |
208 | format = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &vda_ctx->format); |
209 | |
210 | CFDictionarySetValue(config_info, kVDADecoderConfiguration_Height, height); |
211 | CFDictionarySetValue(config_info, kVDADecoderConfiguration_Width, width); |
212 | CFDictionarySetValue(config_info, kVDADecoderConfiguration_SourceFormat, format); |
213 | CFDictionarySetValue(config_info, kVDADecoderConfiguration_avcCData, avc_data); |
214 | |
215 | buffer_attributes = CFDictionaryCreateMutable(kCFAllocatorDefault, |
216 | 2, |
217 | &kCFTypeDictionaryKeyCallBacks, |
218 | &kCFTypeDictionaryValueCallBacks); |
219 | io_surface_properties = CFDictionaryCreateMutable(kCFAllocatorDefault, |
220 | 0, |
221 | &kCFTypeDictionaryKeyCallBacks, |
222 | &kCFTypeDictionaryValueCallBacks); |
223 | cv_pix_fmt = CFNumberCreate(kCFAllocatorDefault, |
224 | kCFNumberSInt32Type, |
225 | &vda_ctx->cv_pix_fmt_type); |
226 | CFDictionarySetValue(buffer_attributes, |
227 | kCVPixelBufferPixelFormatTypeKey, |
228 | cv_pix_fmt); |
229 | CFDictionarySetValue(buffer_attributes, |
230 | kCVPixelBufferIOSurfacePropertiesKey, |
231 | io_surface_properties); |
232 | |
233 | status = VDADecoderCreate(config_info, |
234 | buffer_attributes, |
235 | (VDADecoderOutputCallback *)vda_decoder_callback, |
236 | vda_ctx, |
237 | &vda_ctx->decoder); |
238 | |
239 | CFRelease(height); |
240 | CFRelease(width); |
241 | CFRelease(format); |
242 | CFRelease(avc_data); |
243 | CFRelease(config_info); |
244 | CFRelease(io_surface_properties); |
245 | CFRelease(cv_pix_fmt); |
246 | CFRelease(buffer_attributes); |
247 | |
248 | return status; |
249 | } |
250 | |
251 | int ff_vda_destroy_decoder(struct vda_context *vda_ctx) |
252 | { |
253 | OSStatus status = kVDADecoderNoErr; |
254 | |
255 | if (vda_ctx->decoder) |
256 | status = VDADecoderDestroy(vda_ctx->decoder); |
257 | |
258 | return status; |
259 | } |
260 | |
261 | AVHWAccel ff_h264_vda_old_hwaccel = { |
262 | .name = "h264_vda", |
263 | .type = AVMEDIA_TYPE_VIDEO, |
264 | .id = AV_CODEC_ID_H264, |
265 | .pix_fmt = AV_PIX_FMT_VDA_VLD, |
266 | .start_frame = vda_old_h264_start_frame, |
267 | .decode_slice = vda_old_h264_decode_slice, |
268 | .end_frame = vda_old_h264_end_frame, |
269 | .uninit = ff_videotoolbox_uninit, |
270 | .priv_data_size = sizeof(VTContext), |
271 | }; |
272 | |
273 | void ff_vda_output_callback(void *opaque, |
274 | CFDictionaryRef user_info, |
275 | OSStatus status, |
276 | uint32_t infoFlags, |
277 | CVImageBufferRef image_buffer) |
278 | { |
279 | AVCodecContext *ctx = opaque; |
280 | VTContext *vda = ctx->internal->hwaccel_priv_data; |
281 | |
282 | |
283 | if (vda->frame) { |
284 | CVPixelBufferRelease(vda->frame); |
285 | vda->frame = NULL; |
286 | } |
287 | |
288 | if (!image_buffer) |
289 | return; |
290 | |
291 | vda->frame = CVPixelBufferRetain(image_buffer); |
292 | } |
293 | |
294 | static int vda_h264_end_frame(AVCodecContext *avctx) |
295 | { |
296 | H264Context *h = avctx->priv_data; |
297 | VTContext *vda = avctx->internal->hwaccel_priv_data; |
298 | AVVDAContext *vda_ctx = avctx->hwaccel_context; |
299 | AVFrame *frame = h->cur_pic_ptr->f; |
300 | uint32_t flush_flags = 1 << 0; ///< kVDADecoderFlush_emitFrames |
301 | CFDataRef coded_frame; |
302 | OSStatus status; |
303 | |
304 | if (!vda->bitstream_size) |
305 | return AVERROR_INVALIDDATA; |
306 | |
307 | |
308 | coded_frame = CFDataCreate(kCFAllocatorDefault, |
309 | vda->bitstream, |
310 | vda->bitstream_size); |
311 | |
312 | status = VDADecoderDecode(vda_ctx->decoder, 0, coded_frame, NULL); |
313 | |
314 | if (status == kVDADecoderNoErr) |
315 | status = VDADecoderFlush(vda_ctx->decoder, flush_flags); |
316 | |
317 | CFRelease(coded_frame); |
318 | |
319 | if (!vda->frame) |
320 | return AVERROR_UNKNOWN; |
321 | |
322 | if (status != kVDADecoderNoErr) { |
323 | av_log(avctx, AV_LOG_ERROR, "Failed to decode frame (%d)\n", status); |
324 | return AVERROR_UNKNOWN; |
325 | } |
326 | |
327 | return ff_videotoolbox_buffer_create(vda, frame); |
328 | } |
329 | |
330 | int ff_vda_default_init(AVCodecContext *avctx) |
331 | { |
332 | AVVDAContext *vda_ctx = avctx->hwaccel_context; |
333 | OSStatus status = kVDADecoderNoErr; |
334 | CFNumberRef height; |
335 | CFNumberRef width; |
336 | CFNumberRef format; |
337 | CFDataRef avc_data; |
338 | CFMutableDictionaryRef config_info; |
339 | CFMutableDictionaryRef buffer_attributes; |
340 | CFMutableDictionaryRef io_surface_properties; |
341 | CFNumberRef cv_pix_fmt; |
342 | int32_t fmt = 'avc1', pix_fmt = vda_ctx->cv_pix_fmt_type; |
343 | |
344 | // kCVPixelFormatType_420YpCbCr8Planar; |
345 | |
346 | avc_data = ff_videotoolbox_avcc_extradata_create(avctx); |
347 | |
348 | config_info = CFDictionaryCreateMutable(kCFAllocatorDefault, |
349 | 4, |
350 | &kCFTypeDictionaryKeyCallBacks, |
351 | &kCFTypeDictionaryValueCallBacks); |
352 | |
353 | height = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &avctx->height); |
354 | width = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &avctx->width); |
355 | format = CFNumberCreate(kCFAllocatorDefault, kCFNumberSInt32Type, &fmt); |
356 | CFDictionarySetValue(config_info, kVDADecoderConfiguration_Height, height); |
357 | CFDictionarySetValue(config_info, kVDADecoderConfiguration_Width, width); |
358 | CFDictionarySetValue(config_info, kVDADecoderConfiguration_avcCData, avc_data); |
359 | CFDictionarySetValue(config_info, kVDADecoderConfiguration_SourceFormat, format); |
360 | |
361 | buffer_attributes = CFDictionaryCreateMutable(kCFAllocatorDefault, |
362 | 2, |
363 | &kCFTypeDictionaryKeyCallBacks, |
364 | &kCFTypeDictionaryValueCallBacks); |
365 | io_surface_properties = CFDictionaryCreateMutable(kCFAllocatorDefault, |
366 | 0, |
367 | &kCFTypeDictionaryKeyCallBacks, |
368 | &kCFTypeDictionaryValueCallBacks); |
369 | cv_pix_fmt = CFNumberCreate(kCFAllocatorDefault, |
370 | kCFNumberSInt32Type, |
371 | &pix_fmt); |
372 | |
373 | CFDictionarySetValue(buffer_attributes, |
374 | kCVPixelBufferPixelFormatTypeKey, |
375 | cv_pix_fmt); |
376 | CFDictionarySetValue(buffer_attributes, |
377 | kCVPixelBufferIOSurfacePropertiesKey, |
378 | io_surface_properties); |
379 | |
380 | status = VDADecoderCreate(config_info, |
381 | buffer_attributes, |
382 | (VDADecoderOutputCallback *)ff_vda_output_callback, |
383 | avctx, |
384 | &vda_ctx->decoder); |
385 | |
386 | CFRelease(format); |
387 | CFRelease(height); |
388 | CFRelease(width); |
389 | CFRelease(avc_data); |
390 | CFRelease(config_info); |
391 | CFRelease(cv_pix_fmt); |
392 | CFRelease(io_surface_properties); |
393 | CFRelease(buffer_attributes); |
394 | |
395 | if (status != kVDADecoderNoErr) { |
396 | av_log(avctx, AV_LOG_ERROR, "Cannot initialize VDA %d\n", status); |
397 | } |
398 | |
399 | switch (status) { |
400 | case kVDADecoderHardwareNotSupportedErr: |
401 | case kVDADecoderFormatNotSupportedErr: |
402 | return AVERROR(ENOSYS); |
403 | case kVDADecoderConfigurationError: |
404 | return AVERROR(EINVAL); |
405 | case kVDADecoderDecoderFailedErr: |
406 | return AVERROR_INVALIDDATA; |
407 | case kVDADecoderNoErr: |
408 | return 0; |
409 | default: |
410 | return AVERROR_UNKNOWN; |
411 | } |
412 | } |
413 | |
414 | AVHWAccel ff_h264_vda_hwaccel = { |
415 | .name = "h264_vda", |
416 | .type = AVMEDIA_TYPE_VIDEO, |
417 | .id = AV_CODEC_ID_H264, |
418 | .pix_fmt = AV_PIX_FMT_VDA, |
419 | .alloc_frame = ff_videotoolbox_alloc_frame, |
420 | .start_frame = ff_videotoolbox_h264_start_frame, |
421 | .decode_slice = ff_videotoolbox_h264_decode_slice, |
422 | .end_frame = vda_h264_end_frame, |
423 | .uninit = ff_videotoolbox_uninit, |
424 | .priv_data_size = sizeof(VTContext), |
425 | }; |
426 |