1 /*
2 * Copyright (c) 2007-2013 Intel Corporation. All Rights Reserved.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the
6 * "Software"), to deal in the Software without restriction, including
7 * without limitation the rights to use, copy, modify, merge, publish,
8 * distribute, sub license, and/or sell copies of the Software, and to
9 * permit persons to whom the Software is furnished to do so, subject to
10 * the following conditions:
11 *
12 * The above copyright notice and this permission notice (including the
13 * next paragraph) shall be included in all copies or substantial portions
14 * of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
18 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
19 * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
20 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
21 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
22 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
23 */
24 #include "sysdeps.h"
25 #include <stdio.h>
26 #include <string.h>
27 #include <stdlib.h>
28 #include <getopt.h>
29 #include <unistd.h>
30 #include <sys/types.h>
31 #include <sys/stat.h>
32 #include <sys/time.h>
33 #include <sys/mman.h>
34 #include <fcntl.h>
35 #include <assert.h>
36 #include <pthread.h>
37 #include <errno.h>
38 #include <math.h>
39 #include <va/va.h>
40 #include <va/va_enc_h264.h>
41 #include "va_display.h"
42
43 #define CHECK_VASTATUS(va_status,func) \
44 if (va_status != VA_STATUS_SUCCESS) { \
45 fprintf(stderr,"%s:%s (%d) failed,exit\n", __func__, func, __LINE__); \
46 exit(1); \
47 }
48
49 #include "../loadsurface.h"
50
51 #define NAL_REF_IDC_NONE 0
52 #define NAL_REF_IDC_LOW 1
53 #define NAL_REF_IDC_MEDIUM 2
54 #define NAL_REF_IDC_HIGH 3
55
56 #define NAL_NON_IDR 1
57 #define NAL_IDR 5
58 #define NAL_SPS 7
59 #define NAL_PPS 8
60 #define NAL_SEI 6
61
62 #define SLICE_TYPE_P 0
63 #define SLICE_TYPE_B 1
64 #define SLICE_TYPE_I 2
65
66 #define ENTROPY_MODE_CAVLC 0
67 #define ENTROPY_MODE_CABAC 1
68
69 #define PROFILE_IDC_BASELINE 66
70 #define PROFILE_IDC_MAIN 77
71 #define PROFILE_IDC_HIGH 100
72
73 #define BITSTREAM_ALLOCATE_STEPPING 4096
74
75 #define SURFACE_NUM 16 /* 16 surfaces for source YUV */
76 #define SURFACE_NUM 16 /* 16 surfaces for reference */
77 static VADisplay va_dpy;
78 static VAProfile h264_profile = ~0;
79 static VAConfigAttrib attrib[VAConfigAttribTypeMax];
80 static VAConfigAttrib config_attrib[VAConfigAttribTypeMax];
81 static int config_attrib_num = 0;
82 static VASurfaceID src_surface[SURFACE_NUM];
83 static VABufferID coded_buf[SURFACE_NUM];
84 static VASurfaceID ref_surface[SURFACE_NUM];
85 static VAConfigID config_id;
86 static VAContextID context_id;
87 static VAEncSequenceParameterBufferH264 seq_param;
88 static VAEncPictureParameterBufferH264 pic_param;
89 static VAEncSliceParameterBufferH264 slice_param;
90 static VAPictureH264 CurrentCurrPic;
91 static VAPictureH264 ReferenceFrames[16], RefPicList0_P[32], RefPicList0_B[32], RefPicList1_B[32];
92
93 static unsigned int MaxFrameNum = (2<<16);
94 static unsigned int MaxPicOrderCntLsb = (2<<8);
95 static unsigned int Log2MaxFrameNum = 16;
96 static unsigned int Log2MaxPicOrderCntLsb = 8;
97
98 static unsigned int num_ref_frames = 2;
99 static unsigned int numShortTerm = 0;
100 static int constraint_set_flag = 0;
101 static int h264_packedheader = 0; /* support pack header? */
102 static int h264_maxref = (1<<16|1);
103 static int h264_entropy_mode = 1; /* cabac */
104
105 static char *coded_fn = NULL, *srcyuv_fn = NULL, *recyuv_fn = NULL;
106 static FILE *coded_fp = NULL, *srcyuv_fp = NULL, *recyuv_fp = NULL;
107 static unsigned long long srcyuv_frames = 0;
108 static int srcyuv_fourcc = VA_FOURCC_NV12;
109 static int calc_psnr = 0;
110
111 static int frame_width = 176;
112 static int frame_height = 144;
113 static int frame_width_mbaligned;
114 static int frame_height_mbaligned;
115 static int frame_rate = 30;
116 static unsigned int frame_count = 60;
117 static unsigned int frame_coded = 0;
118 static unsigned int frame_bitrate = 0;
119 static unsigned int frame_slices = 1;
120 static double frame_size = 0;
121 static int initial_qp = 26;
122 static int minimal_qp = 0;
123 static int intra_period = 30;
124 static int intra_idr_period = 60;
125 static int ip_period = 1;
126 static int rc_mode = VA_RC_VBR;
127 static unsigned long long current_frame_encoding = 0;
128 static unsigned long long current_frame_display = 0;
129 static unsigned long long current_IDR_display = 0;
130 static unsigned int current_frame_num = 0;
131 static int current_frame_type;
132 #define current_slot (current_frame_display % SURFACE_NUM)
133
134 static int misc_priv_type = 0;
135 static int misc_priv_value = 0;
136
137 #define MIN(a, b) ((a)>(b)?(b):(a))
138 #define MAX(a, b) ((a)>(b)?(a):(b))
139
140 /* thread to save coded data/upload source YUV */
141 struct storage_task_t {
142 void *next;
143 unsigned long long display_order;
144 unsigned long long encode_order;
145 };
146 static struct storage_task_t *storage_task_header = NULL, *storage_task_tail = NULL;
147 #define SRC_SURFACE_IN_ENCODING 0
148 #define SRC_SURFACE_IN_STORAGE 1
149 static int srcsurface_status[SURFACE_NUM];
150 static int encode_syncmode = 0;
151 static pthread_mutex_t encode_mutex = PTHREAD_MUTEX_INITIALIZER;
152 static pthread_cond_t encode_cond = PTHREAD_COND_INITIALIZER;
153 static pthread_t encode_thread;
154
155 /* for performance profiling */
156 static unsigned int UploadPictureTicks=0;
157 static unsigned int BeginPictureTicks=0;
158 static unsigned int RenderPictureTicks=0;
159 static unsigned int EndPictureTicks=0;
160 static unsigned int SyncPictureTicks=0;
161 static unsigned int SavePictureTicks=0;
162 static unsigned int TotalTicks=0;
163
164 struct __bitstream {
165 unsigned int *buffer;
166 int bit_offset;
167 int max_size_in_dword;
168 };
169 typedef struct __bitstream bitstream;
170
171
172 static unsigned int
va_swap32(unsigned int val)173 va_swap32(unsigned int val)
174 {
175 unsigned char *pval = (unsigned char *)&val;
176
177 return ((pval[0] << 24) |
178 (pval[1] << 16) |
179 (pval[2] << 8) |
180 (pval[3] << 0));
181 }
182
183 static void
bitstream_start(bitstream * bs)184 bitstream_start(bitstream *bs)
185 {
186 bs->max_size_in_dword = BITSTREAM_ALLOCATE_STEPPING;
187 bs->buffer = calloc(bs->max_size_in_dword * sizeof(int), 1);
188 bs->bit_offset = 0;
189 }
190
191 static void
bitstream_end(bitstream * bs)192 bitstream_end(bitstream *bs)
193 {
194 int pos = (bs->bit_offset >> 5);
195 int bit_offset = (bs->bit_offset & 0x1f);
196 int bit_left = 32 - bit_offset;
197
198 if (bit_offset) {
199 bs->buffer[pos] = va_swap32((bs->buffer[pos] << bit_left));
200 }
201 }
202
203 static void
bitstream_put_ui(bitstream * bs,unsigned int val,int size_in_bits)204 bitstream_put_ui(bitstream *bs, unsigned int val, int size_in_bits)
205 {
206 int pos = (bs->bit_offset >> 5);
207 int bit_offset = (bs->bit_offset & 0x1f);
208 int bit_left = 32 - bit_offset;
209
210 if (!size_in_bits)
211 return;
212
213 bs->bit_offset += size_in_bits;
214
215 if (bit_left > size_in_bits) {
216 bs->buffer[pos] = (bs->buffer[pos] << size_in_bits | val);
217 } else {
218 size_in_bits -= bit_left;
219 bs->buffer[pos] = (bs->buffer[pos] << bit_left) | (val >> size_in_bits);
220 bs->buffer[pos] = va_swap32(bs->buffer[pos]);
221
222 if (pos + 1 == bs->max_size_in_dword) {
223 bs->max_size_in_dword += BITSTREAM_ALLOCATE_STEPPING;
224 bs->buffer = realloc(bs->buffer, bs->max_size_in_dword * sizeof(unsigned int));
225 }
226
227 bs->buffer[pos + 1] = val;
228 }
229 }
230
231 static void
bitstream_put_ue(bitstream * bs,unsigned int val)232 bitstream_put_ue(bitstream *bs, unsigned int val)
233 {
234 int size_in_bits = 0;
235 int tmp_val = ++val;
236
237 while (tmp_val) {
238 tmp_val >>= 1;
239 size_in_bits++;
240 }
241
242 bitstream_put_ui(bs, 0, size_in_bits - 1); // leading zero
243 bitstream_put_ui(bs, val, size_in_bits);
244 }
245
246 static void
bitstream_put_se(bitstream * bs,int val)247 bitstream_put_se(bitstream *bs, int val)
248 {
249 unsigned int new_val;
250
251 if (val <= 0)
252 new_val = -2 * val;
253 else
254 new_val = 2 * val - 1;
255
256 bitstream_put_ue(bs, new_val);
257 }
258
259 static void
bitstream_byte_aligning(bitstream * bs,int bit)260 bitstream_byte_aligning(bitstream *bs, int bit)
261 {
262 int bit_offset = (bs->bit_offset & 0x7);
263 int bit_left = 8 - bit_offset;
264 int new_val;
265
266 if (!bit_offset)
267 return;
268
269 assert(bit == 0 || bit == 1);
270
271 if (bit)
272 new_val = (1 << bit_left) - 1;
273 else
274 new_val = 0;
275
276 bitstream_put_ui(bs, new_val, bit_left);
277 }
278
279 static void
rbsp_trailing_bits(bitstream * bs)280 rbsp_trailing_bits(bitstream *bs)
281 {
282 bitstream_put_ui(bs, 1, 1);
283 bitstream_byte_aligning(bs, 0);
284 }
285
nal_start_code_prefix(bitstream * bs)286 static void nal_start_code_prefix(bitstream *bs)
287 {
288 bitstream_put_ui(bs, 0x00000001, 32);
289 }
290
nal_header(bitstream * bs,int nal_ref_idc,int nal_unit_type)291 static void nal_header(bitstream *bs, int nal_ref_idc, int nal_unit_type)
292 {
293 bitstream_put_ui(bs, 0, 1); /* forbidden_zero_bit: 0 */
294 bitstream_put_ui(bs, nal_ref_idc, 2);
295 bitstream_put_ui(bs, nal_unit_type, 5);
296 }
297
sps_rbsp(bitstream * bs)298 static void sps_rbsp(bitstream *bs)
299 {
300 int profile_idc = PROFILE_IDC_BASELINE;
301
302 if (h264_profile == VAProfileH264High)
303 profile_idc = PROFILE_IDC_HIGH;
304 else if (h264_profile == VAProfileH264Main)
305 profile_idc = PROFILE_IDC_MAIN;
306
307 bitstream_put_ui(bs, profile_idc, 8); /* profile_idc */
308 bitstream_put_ui(bs, !!(constraint_set_flag & 1), 1); /* constraint_set0_flag */
309 bitstream_put_ui(bs, !!(constraint_set_flag & 2), 1); /* constraint_set1_flag */
310 bitstream_put_ui(bs, !!(constraint_set_flag & 4), 1); /* constraint_set2_flag */
311 bitstream_put_ui(bs, !!(constraint_set_flag & 8), 1); /* constraint_set3_flag */
312 bitstream_put_ui(bs, 0, 4); /* reserved_zero_4bits */
313 bitstream_put_ui(bs, seq_param.level_idc, 8); /* level_idc */
314 bitstream_put_ue(bs, seq_param.seq_parameter_set_id); /* seq_parameter_set_id */
315
316 if ( profile_idc == PROFILE_IDC_HIGH) {
317 bitstream_put_ue(bs, 1); /* chroma_format_idc = 1, 4:2:0 */
318 bitstream_put_ue(bs, 0); /* bit_depth_luma_minus8 */
319 bitstream_put_ue(bs, 0); /* bit_depth_chroma_minus8 */
320 bitstream_put_ui(bs, 0, 1); /* qpprime_y_zero_transform_bypass_flag */
321 bitstream_put_ui(bs, 0, 1); /* seq_scaling_matrix_present_flag */
322 }
323
324 bitstream_put_ue(bs, seq_param.seq_fields.bits.log2_max_frame_num_minus4); /* log2_max_frame_num_minus4 */
325 bitstream_put_ue(bs, seq_param.seq_fields.bits.pic_order_cnt_type); /* pic_order_cnt_type */
326
327 if (seq_param.seq_fields.bits.pic_order_cnt_type == 0)
328 bitstream_put_ue(bs, seq_param.seq_fields.bits.log2_max_pic_order_cnt_lsb_minus4); /* log2_max_pic_order_cnt_lsb_minus4 */
329 else {
330 assert(0);
331 }
332
333 bitstream_put_ue(bs, seq_param.max_num_ref_frames); /* num_ref_frames */
334 bitstream_put_ui(bs, 0, 1); /* gaps_in_frame_num_value_allowed_flag */
335
336 bitstream_put_ue(bs, seq_param.picture_width_in_mbs - 1); /* pic_width_in_mbs_minus1 */
337 bitstream_put_ue(bs, seq_param.picture_height_in_mbs - 1); /* pic_height_in_map_units_minus1 */
338 bitstream_put_ui(bs, seq_param.seq_fields.bits.frame_mbs_only_flag, 1); /* frame_mbs_only_flag */
339
340 if (!seq_param.seq_fields.bits.frame_mbs_only_flag) {
341 assert(0);
342 }
343
344 bitstream_put_ui(bs, seq_param.seq_fields.bits.direct_8x8_inference_flag, 1); /* direct_8x8_inference_flag */
345 bitstream_put_ui(bs, seq_param.frame_cropping_flag, 1); /* frame_cropping_flag */
346
347 if (seq_param.frame_cropping_flag) {
348 bitstream_put_ue(bs, seq_param.frame_crop_left_offset); /* frame_crop_left_offset */
349 bitstream_put_ue(bs, seq_param.frame_crop_right_offset); /* frame_crop_right_offset */
350 bitstream_put_ue(bs, seq_param.frame_crop_top_offset); /* frame_crop_top_offset */
351 bitstream_put_ue(bs, seq_param.frame_crop_bottom_offset); /* frame_crop_bottom_offset */
352 }
353
354 //if ( frame_bit_rate < 0 ) { //TODO EW: the vui header isn't correct
355 if ( 1 ) {
356 bitstream_put_ui(bs, 0, 1); /* vui_parameters_present_flag */
357 } else {
358 bitstream_put_ui(bs, 1, 1); /* vui_parameters_present_flag */
359 bitstream_put_ui(bs, 0, 1); /* aspect_ratio_info_present_flag */
360 bitstream_put_ui(bs, 0, 1); /* overscan_info_present_flag */
361 bitstream_put_ui(bs, 0, 1); /* video_signal_type_present_flag */
362 bitstream_put_ui(bs, 0, 1); /* chroma_loc_info_present_flag */
363 bitstream_put_ui(bs, 1, 1); /* timing_info_present_flag */
364 {
365 bitstream_put_ui(bs, 15, 32);
366 bitstream_put_ui(bs, 900, 32);
367 bitstream_put_ui(bs, 1, 1);
368 }
369 bitstream_put_ui(bs, 1, 1); /* nal_hrd_parameters_present_flag */
370 {
371 // hrd_parameters
372 bitstream_put_ue(bs, 0); /* cpb_cnt_minus1 */
373 bitstream_put_ui(bs, 4, 4); /* bit_rate_scale */
374 bitstream_put_ui(bs, 6, 4); /* cpb_size_scale */
375
376 bitstream_put_ue(bs, frame_bitrate - 1); /* bit_rate_value_minus1[0] */
377 bitstream_put_ue(bs, frame_bitrate*8 - 1); /* cpb_size_value_minus1[0] */
378 bitstream_put_ui(bs, 1, 1); /* cbr_flag[0] */
379
380 bitstream_put_ui(bs, 23, 5); /* initial_cpb_removal_delay_length_minus1 */
381 bitstream_put_ui(bs, 23, 5); /* cpb_removal_delay_length_minus1 */
382 bitstream_put_ui(bs, 23, 5); /* dpb_output_delay_length_minus1 */
383 bitstream_put_ui(bs, 23, 5); /* time_offset_length */
384 }
385 bitstream_put_ui(bs, 0, 1); /* vcl_hrd_parameters_present_flag */
386 bitstream_put_ui(bs, 0, 1); /* low_delay_hrd_flag */
387
388 bitstream_put_ui(bs, 0, 1); /* pic_struct_present_flag */
389 bitstream_put_ui(bs, 0, 1); /* bitstream_restriction_flag */
390 }
391
392 rbsp_trailing_bits(bs); /* rbsp_trailing_bits */
393 }
394
395
pps_rbsp(bitstream * bs)396 static void pps_rbsp(bitstream *bs)
397 {
398 bitstream_put_ue(bs, pic_param.pic_parameter_set_id); /* pic_parameter_set_id */
399 bitstream_put_ue(bs, pic_param.seq_parameter_set_id); /* seq_parameter_set_id */
400
401 bitstream_put_ui(bs, pic_param.pic_fields.bits.entropy_coding_mode_flag, 1); /* entropy_coding_mode_flag */
402
403 bitstream_put_ui(bs, 0, 1); /* pic_order_present_flag: 0 */
404
405 bitstream_put_ue(bs, 0); /* num_slice_groups_minus1 */
406
407 bitstream_put_ue(bs, pic_param.num_ref_idx_l0_active_minus1); /* num_ref_idx_l0_active_minus1 */
408 bitstream_put_ue(bs, pic_param.num_ref_idx_l1_active_minus1); /* num_ref_idx_l1_active_minus1 1 */
409
410 bitstream_put_ui(bs, pic_param.pic_fields.bits.weighted_pred_flag, 1); /* weighted_pred_flag: 0 */
411 bitstream_put_ui(bs, pic_param.pic_fields.bits.weighted_bipred_idc, 2); /* weighted_bipred_idc: 0 */
412
413 bitstream_put_se(bs, pic_param.pic_init_qp - 26); /* pic_init_qp_minus26 */
414 bitstream_put_se(bs, 0); /* pic_init_qs_minus26 */
415 bitstream_put_se(bs, 0); /* chroma_qp_index_offset */
416
417 bitstream_put_ui(bs, pic_param.pic_fields.bits.deblocking_filter_control_present_flag, 1); /* deblocking_filter_control_present_flag */
418 bitstream_put_ui(bs, 0, 1); /* constrained_intra_pred_flag */
419 bitstream_put_ui(bs, 0, 1); /* redundant_pic_cnt_present_flag */
420
421 /* more_rbsp_data */
422 bitstream_put_ui(bs, pic_param.pic_fields.bits.transform_8x8_mode_flag, 1); /*transform_8x8_mode_flag */
423 bitstream_put_ui(bs, 0, 1); /* pic_scaling_matrix_present_flag */
424 bitstream_put_se(bs, pic_param.second_chroma_qp_index_offset ); /*second_chroma_qp_index_offset */
425
426 rbsp_trailing_bits(bs);
427 }
428
429
430 static int
build_packed_pic_buffer(unsigned char ** header_buffer)431 build_packed_pic_buffer(unsigned char **header_buffer)
432 {
433 bitstream bs;
434
435 bitstream_start(&bs);
436 nal_start_code_prefix(&bs);
437 nal_header(&bs, NAL_REF_IDC_HIGH, NAL_PPS);
438 pps_rbsp(&bs);
439 bitstream_end(&bs);
440
441 *header_buffer = (unsigned char *)bs.buffer;
442 return bs.bit_offset;
443 }
444
445 static int
build_packed_seq_buffer(unsigned char ** header_buffer)446 build_packed_seq_buffer(unsigned char **header_buffer)
447 {
448 bitstream bs;
449
450 bitstream_start(&bs);
451 nal_start_code_prefix(&bs);
452 nal_header(&bs, NAL_REF_IDC_HIGH, NAL_SPS);
453 sps_rbsp(&bs);
454 bitstream_end(&bs);
455
456 *header_buffer = (unsigned char *)bs.buffer;
457 return bs.bit_offset;
458 }
459
460 static int
build_packed_sei_buffer_timing(unsigned int init_cpb_removal_length,unsigned int init_cpb_removal_delay,unsigned int init_cpb_removal_delay_offset,unsigned int cpb_removal_length,unsigned int cpb_removal_delay,unsigned int dpb_output_length,unsigned int dpb_output_delay,unsigned char ** sei_buffer)461 build_packed_sei_buffer_timing(unsigned int init_cpb_removal_length,
462 unsigned int init_cpb_removal_delay,
463 unsigned int init_cpb_removal_delay_offset,
464 unsigned int cpb_removal_length,
465 unsigned int cpb_removal_delay,
466 unsigned int dpb_output_length,
467 unsigned int dpb_output_delay,
468 unsigned char **sei_buffer)
469 {
470 unsigned char *byte_buf;
471 int bp_byte_size, i, pic_byte_size;
472
473 bitstream nal_bs;
474 bitstream sei_bp_bs, sei_pic_bs;
475
476 bitstream_start(&sei_bp_bs);
477 bitstream_put_ue(&sei_bp_bs, 0); /*seq_parameter_set_id*/
478 bitstream_put_ui(&sei_bp_bs, init_cpb_removal_delay, cpb_removal_length);
479 bitstream_put_ui(&sei_bp_bs, init_cpb_removal_delay_offset, cpb_removal_length);
480 if ( sei_bp_bs.bit_offset & 0x7) {
481 bitstream_put_ui(&sei_bp_bs, 1, 1);
482 }
483 bitstream_end(&sei_bp_bs);
484 bp_byte_size = (sei_bp_bs.bit_offset + 7) / 8;
485
486 bitstream_start(&sei_pic_bs);
487 bitstream_put_ui(&sei_pic_bs, cpb_removal_delay, cpb_removal_length);
488 bitstream_put_ui(&sei_pic_bs, dpb_output_delay, dpb_output_length);
489 if ( sei_pic_bs.bit_offset & 0x7) {
490 bitstream_put_ui(&sei_pic_bs, 1, 1);
491 }
492 bitstream_end(&sei_pic_bs);
493 pic_byte_size = (sei_pic_bs.bit_offset + 7) / 8;
494
495 bitstream_start(&nal_bs);
496 nal_start_code_prefix(&nal_bs);
497 nal_header(&nal_bs, NAL_REF_IDC_NONE, NAL_SEI);
498
499 /* Write the SEI buffer period data */
500 bitstream_put_ui(&nal_bs, 0, 8);
501 bitstream_put_ui(&nal_bs, bp_byte_size, 8);
502
503 byte_buf = (unsigned char *)sei_bp_bs.buffer;
504 for(i = 0; i < bp_byte_size; i++) {
505 bitstream_put_ui(&nal_bs, byte_buf[i], 8);
506 }
507 free(byte_buf);
508 /* write the SEI timing data */
509 bitstream_put_ui(&nal_bs, 0x01, 8);
510 bitstream_put_ui(&nal_bs, pic_byte_size, 8);
511
512 byte_buf = (unsigned char *)sei_pic_bs.buffer;
513 for(i = 0; i < pic_byte_size; i++) {
514 bitstream_put_ui(&nal_bs, byte_buf[i], 8);
515 }
516 free(byte_buf);
517
518 rbsp_trailing_bits(&nal_bs);
519 bitstream_end(&nal_bs);
520
521 *sei_buffer = (unsigned char *)nal_bs.buffer;
522
523 return nal_bs.bit_offset;
524 }
525
526
527
528 /*
529 * Helper function for profiling purposes
530 */
GetTickCount()531 static unsigned int GetTickCount()
532 {
533 struct timeval tv;
534 if (gettimeofday(&tv, NULL))
535 return 0;
536 return tv.tv_usec/1000+tv.tv_sec*1000;
537 }
538
539 /*
540 Assume frame sequence is: Frame#0,#1,#2,...,#M,...,#X,... (encoding order)
541 1) period between Frame #X and Frame #N = #X - #N
542 2) 0 means infinite for intra_period/intra_idr_period, and 0 is invalid for ip_period
543 3) intra_idr_period % intra_period (intra_period > 0) and intra_period % ip_period must be 0
544 4) intra_period and intra_idr_period take precedence over ip_period
545 5) if ip_period > 1, intra_period and intra_idr_period are not the strict periods
546 of I/IDR frames, see bellow examples
547 -------------------------------------------------------------------
548 intra_period intra_idr_period ip_period frame sequence (intra_period/intra_idr_period/ip_period)
549 0 ignored 1 IDRPPPPPPP ... (No IDR/I any more)
550 0 ignored >=2 IDR(PBB)(PBB)... (No IDR/I any more)
551 1 0 ignored IDRIIIIIII... (No IDR any more)
552 1 1 ignored IDR IDR IDR IDR...
553 1 >=2 ignored IDRII IDRII IDR... (1/3/ignore)
554 >=2 0 1 IDRPPP IPPP I... (3/0/1)
555 >=2 0 >=2 IDR(PBB)(PBB)(IBB) (6/0/3)
556 (PBB)(IBB)(PBB)(IBB)...
557 >=2 >=2 1 IDRPPPPP IPPPPP IPPPPP (6/18/1)
558 IDRPPPPP IPPPPP IPPPPP...
559 >=2 >=2 >=2 {IDR(PBB)(PBB)(IBB)(PBB)(IBB)(PBB)} (6/18/3)
560 {IDR(PBB)(PBB)(IBB)(PBB)(IBB)(PBB)}...
561 {IDR(PBB)(PBB)(IBB)(PBB)} (6/12/3)
562 {IDR(PBB)(PBB)(IBB)(PBB)}...
563 {IDR(PBB)(PBB)} (6/6/3)
564 {IDR(PBB)(PBB)}.
565 */
566
567 /*
568 * Return displaying order with specified periods and encoding order
569 * displaying_order: displaying order
570 * frame_type: frame type
571 */
572 #define FRAME_P 0
573 #define FRAME_B 1
574 #define FRAME_I 2
575 #define FRAME_IDR 7
encoding2display_order(unsigned long long encoding_order,int intra_period,int intra_idr_period,int ip_period,unsigned long long * displaying_order,int * frame_type)576 void encoding2display_order(
577 unsigned long long encoding_order,int intra_period,
578 int intra_idr_period,int ip_period,
579 unsigned long long *displaying_order,
580 int *frame_type)
581 {
582 int encoding_order_gop = 0;
583
584 if (intra_period == 1) { /* all are I/IDR frames */
585 *displaying_order = encoding_order;
586 if (intra_idr_period == 0)
587 *frame_type = (encoding_order == 0)?FRAME_IDR:FRAME_I;
588 else
589 *frame_type = (encoding_order % intra_idr_period == 0)?FRAME_IDR:FRAME_I;
590 return;
591 }
592
593 if (intra_period == 0)
594 intra_idr_period = 0;
595
596 /* new sequence like
597 * IDR PPPPP IPPPPP
598 * IDR (PBB)(PBB)(IBB)(PBB)
599 */
600 encoding_order_gop = (intra_idr_period == 0)? encoding_order:
601 (encoding_order % (intra_idr_period + ((ip_period == 1)?0:1)));
602
603 if (encoding_order_gop == 0) { /* the first frame */
604 *frame_type = FRAME_IDR;
605 *displaying_order = encoding_order;
606 } else if (((encoding_order_gop - 1) % ip_period) != 0) { /* B frames */
607 *frame_type = FRAME_B;
608 *displaying_order = encoding_order - 1;
609 } else if ((intra_period != 0) && /* have I frames */
610 (encoding_order_gop >= 2) &&
611 ((ip_period == 1 && encoding_order_gop % intra_period == 0) || /* for IDR PPPPP IPPPP */
612 /* for IDR (PBB)(PBB)(IBB) */
613 (ip_period >= 2 && ((encoding_order_gop - 1) / ip_period % (intra_period / ip_period)) == 0))) {
614 *frame_type = FRAME_I;
615 *displaying_order = encoding_order + ip_period - 1;
616 } else {
617 *frame_type = FRAME_P;
618 *displaying_order = encoding_order + ip_period - 1;
619 }
620 }
621
622
fourcc_to_string(int fourcc)623 static char *fourcc_to_string(int fourcc)
624 {
625 switch (fourcc) {
626 case VA_FOURCC_NV12:
627 return "NV12";
628 case VA_FOURCC_IYUV:
629 return "IYUV";
630 case VA_FOURCC_YV12:
631 return "YV12";
632 case VA_FOURCC_UYVY:
633 return "UYVY";
634 default:
635 return "Unknown";
636 }
637 }
638
string_to_fourcc(char * str)639 static int string_to_fourcc(char *str)
640 {
641 int fourcc;
642
643 if (!strncmp(str, "NV12", 4))
644 fourcc = VA_FOURCC_NV12;
645 else if (!strncmp(str, "IYUV", 4))
646 fourcc = VA_FOURCC_IYUV;
647 else if (!strncmp(str, "YV12", 4))
648 fourcc = VA_FOURCC_YV12;
649 else if (!strncmp(str, "UYVY", 4))
650 fourcc = VA_FOURCC_UYVY;
651 else {
652 printf("Unknow FOURCC\n");
653 fourcc = -1;
654 }
655 return fourcc;
656 }
657
658
rc_to_string(int rcmode)659 static char *rc_to_string(int rcmode)
660 {
661 switch (rc_mode) {
662 case VA_RC_NONE:
663 return "NONE";
664 case VA_RC_CBR:
665 return "CBR";
666 case VA_RC_VBR:
667 return "VBR";
668 case VA_RC_VCM:
669 return "VCM";
670 case VA_RC_CQP:
671 return "CQP";
672 case VA_RC_VBR_CONSTRAINED:
673 return "VBR_CONSTRAINED";
674 default:
675 return "Unknown";
676 }
677 }
678
string_to_rc(char * str)679 static int string_to_rc(char *str)
680 {
681 int rc_mode;
682
683 if (!strncmp(str, "NONE", 4))
684 rc_mode = VA_RC_NONE;
685 else if (!strncmp(str, "CBR", 3))
686 rc_mode = VA_RC_CBR;
687 else if (!strncmp(str, "VBR", 3))
688 rc_mode = VA_RC_VBR;
689 else if (!strncmp(str, "VCM", 3))
690 rc_mode = VA_RC_VCM;
691 else if (!strncmp(str, "CQP", 3))
692 rc_mode = VA_RC_CQP;
693 else if (!strncmp(str, "VBR_CONSTRAINED", 15))
694 rc_mode = VA_RC_VBR_CONSTRAINED;
695 else {
696 printf("Unknown RC mode\n");
697 rc_mode = -1;
698 }
699 return rc_mode;
700 }
701
702
print_help(void)703 static int print_help(void)
704 {
705 printf("./h264encode <options>\n");
706 printf(" -w <width> -h <height>\n");
707 printf(" -framecount <frame number>\n");
708 printf(" -n <frame number>\n");
709 printf(" if set to 0 and srcyuv is set, the frame count is from srcuv file\n");
710 printf(" -o <coded file>\n");
711 printf(" -f <frame rate>\n");
712 printf(" --intra_period <number>\n");
713 printf(" --idr_period <number>\n");
714 printf(" --ip_period <number>\n");
715 printf(" --bitrate <bitrate>\n");
716 printf(" --initialqp <number>\n");
717 printf(" --minqp <number>\n");
718 printf(" --rcmode <NONE|CBR|VBR|VCM|CQP|VBR_CONTRAINED>\n");
719 printf(" --syncmode: sequentially upload source, encoding, save result, no multi-thread\n");
720 printf(" --srcyuv <filename> load YUV from a file\n");
721 printf(" --fourcc <NV12|IYUV|YV12> source YUV fourcc\n");
722 printf(" --recyuv <filename> save reconstructed YUV into a file\n");
723 printf(" --enablePSNR calculate PSNR of recyuv vs. srcyuv\n");
724 printf(" --entropy <0|1>, 1 means cabac, 0 cavlc\n");
725 printf(" --profile <BP|MP|HP>\n");
726 return 0;
727 }
728
process_cmdline(int argc,char * argv[])729 static int process_cmdline(int argc, char *argv[])
730 {
731 char c;
732 const struct option long_opts[] = {
733 {"help", no_argument, NULL, 0 },
734 {"bitrate", required_argument, NULL, 1 },
735 {"minqp", required_argument, NULL, 2 },
736 {"initialqp", required_argument, NULL, 3 },
737 {"intra_period", required_argument, NULL, 4 },
738 {"idr_period", required_argument, NULL, 5 },
739 {"ip_period", required_argument, NULL, 6 },
740 {"rcmode", required_argument, NULL, 7 },
741 {"srcyuv", required_argument, NULL, 9 },
742 {"recyuv", required_argument, NULL, 10 },
743 {"fourcc", required_argument, NULL, 11 },
744 {"syncmode", no_argument, NULL, 12 },
745 {"enablePSNR", no_argument, NULL, 13 },
746 {"prit", required_argument, NULL, 14 },
747 {"priv", required_argument, NULL, 15 },
748 {"framecount", required_argument, NULL, 16 },
749 {"entropy", required_argument, NULL, 17 },
750 {"profile", required_argument, NULL, 18 },
751 {NULL, no_argument, NULL, 0 }};
752 int long_index;
753
754 while ((c =getopt_long_only(argc,argv,"w:h:n:f:o:?",long_opts,&long_index)) != EOF) {
755 switch (c) {
756 case 'w':
757 frame_width = atoi(optarg);
758 break;
759 case 'h':
760 frame_height = atoi(optarg);
761 break;
762 case 'n':
763 case 16:
764 frame_count = atoi(optarg);
765 break;
766 case 'f':
767 frame_rate = atoi(optarg);
768 break;
769 case 'o':
770 coded_fn = strdup(optarg);
771 break;
772 case 0:
773 print_help();
774 exit(0);
775 case 1:
776 frame_bitrate = atoi(optarg);
777 break;
778 case 2:
779 minimal_qp = atoi(optarg);
780 break;
781 case 3:
782 initial_qp = atoi(optarg);
783 break;
784 case 4:
785 intra_period = atoi(optarg);
786 break;
787 case 5:
788 intra_idr_period = atoi(optarg);
789 break;
790 case 6:
791 ip_period = atoi(optarg);
792 break;
793 case 7:
794 rc_mode = string_to_rc(optarg);
795 if (rc_mode < 0) {
796 print_help();
797 exit(1);
798 }
799 break;
800 case 9:
801 srcyuv_fn = strdup(optarg);
802 break;
803 case 10:
804 recyuv_fn = strdup(optarg);
805 break;
806 case 11:
807 srcyuv_fourcc = string_to_fourcc(optarg);
808 if (srcyuv_fourcc <= 0) {
809 print_help();
810 exit(1);
811 }
812 break;
813 case 12:
814 encode_syncmode = 1;
815 break;
816 case 13:
817 calc_psnr = 1;
818 break;
819 case 14:
820 misc_priv_type = strtol(optarg, NULL, 0);
821 break;
822 case 15:
823 misc_priv_value = strtol(optarg, NULL, 0);
824 break;
825 case 17:
826 h264_entropy_mode = atoi(optarg) ? 1: 0;
827 break;
828 case 18:
829 if (strncmp(optarg, "BP", 2) == 0)
830 h264_profile = VAProfileH264Baseline;
831 else if (strncmp(optarg, "MP", 2) == 0)
832 h264_profile = VAProfileH264Main;
833 else if (strncmp(optarg, "HP", 2) == 0)
834 h264_profile = VAProfileH264High;
835 else
836 h264_profile = 0;
837 break;
838 case ':':
839 case '?':
840 print_help();
841 exit(0);
842 }
843 }
844
845 if (ip_period < 1) {
846 printf(" ip_period must be greater than 0\n");
847 exit(0);
848 }
849 if (intra_period != 1 && intra_period % ip_period != 0) {
850 printf(" intra_period must be a multiplier of ip_period\n");
851 exit(0);
852 }
853 if (intra_period != 0 && intra_idr_period % intra_period != 0) {
854 printf(" intra_idr_period must be a multiplier of intra_period\n");
855 exit(0);
856 }
857
858 if (frame_bitrate == 0)
859 frame_bitrate = frame_width * frame_height * 12 * frame_rate / 50;
860
861 /* open source file */
862 if (srcyuv_fn) {
863 srcyuv_fp = fopen(srcyuv_fn,"r");
864
865 if (srcyuv_fp == NULL)
866 printf("Open source YUV file %s failed, use auto-generated YUV data\n", srcyuv_fn);
867 else {
868 struct stat tmp;
869
870 fstat(fileno(srcyuv_fp), &tmp);
871 srcyuv_frames = tmp.st_size / (frame_width * frame_height * 1.5);
872 printf("Source YUV file %s with %llu frames\n", srcyuv_fn, srcyuv_frames);
873
874 if (frame_count == 0)
875 frame_count = srcyuv_frames;
876 }
877 }
878
879 /* open source file */
880 if (recyuv_fn) {
881 recyuv_fp = fopen(recyuv_fn,"w+");
882
883 if (recyuv_fp == NULL)
884 printf("Open reconstructed YUV file %s failed\n", recyuv_fn);
885 }
886
887 if (coded_fn == NULL) {
888 struct stat buf;
889 if (stat("/tmp", &buf) == 0)
890 coded_fn = strdup("/tmp/test.264");
891 else if (stat("/sdcard", &buf) == 0)
892 coded_fn = strdup("/sdcard/test.264");
893 else
894 coded_fn = strdup("./test.264");
895 }
896
897 /* store coded data into a file */
898 coded_fp = fopen(coded_fn,"w+");
899 if (coded_fp == NULL) {
900 printf("Open file %s failed, exit\n", coded_fn);
901 exit(1);
902 }
903
904 frame_width_mbaligned = (frame_width + 15) & (~15);
905 frame_height_mbaligned = (frame_height + 15) & (~15);
906 if (frame_width != frame_width_mbaligned ||
907 frame_height != frame_height_mbaligned) {
908 printf("Source frame is %dx%d and will code clip to %dx%d with crop\n",
909 frame_width, frame_height,
910 frame_width_mbaligned, frame_height_mbaligned
911 );
912 }
913
914 return 0;
915 }
916
init_va(void)917 static int init_va(void)
918 {
919 VAProfile profile_list[]={VAProfileH264High,VAProfileH264Main,VAProfileH264Baseline,VAProfileH264ConstrainedBaseline};
920 VAEntrypoint *entrypoints;
921 int num_entrypoints, slice_entrypoint;
922 int support_encode = 0;
923 int major_ver, minor_ver;
924 VAStatus va_status;
925 unsigned int i;
926
927 va_dpy = va_open_display();
928 va_status = vaInitialize(va_dpy, &major_ver, &minor_ver);
929 CHECK_VASTATUS(va_status, "vaInitialize");
930
931 num_entrypoints = vaMaxNumEntrypoints(va_dpy);
932 entrypoints = malloc(num_entrypoints * sizeof(*entrypoints));
933 if (!entrypoints) {
934 fprintf(stderr, "error: failed to initialize VA entrypoints array\n");
935 exit(1);
936 }
937
938 /* use the highest profile */
939 for (i = 0; i < sizeof(profile_list)/sizeof(profile_list[0]); i++) {
940 if ((h264_profile != ~0) && h264_profile != profile_list[i])
941 continue;
942
943 h264_profile = profile_list[i];
944 vaQueryConfigEntrypoints(va_dpy, h264_profile, entrypoints, &num_entrypoints);
945 for (slice_entrypoint = 0; slice_entrypoint < num_entrypoints; slice_entrypoint++) {
946 if (entrypoints[slice_entrypoint] == VAEntrypointEncSlice) {
947 support_encode = 1;
948 break;
949 }
950 }
951 if (support_encode == 1)
952 break;
953 }
954
955 if (support_encode == 0) {
956 printf("Can't find VAEntrypointEncSlice for H264 profiles\n");
957 exit(1);
958 } else {
959 switch (h264_profile) {
960 case VAProfileH264Baseline:
961 printf("Use profile VAProfileH264Baseline\n");
962 ip_period = 1;
963 constraint_set_flag |= (1 << 0); /* Annex A.2.1 */
964 h264_entropy_mode = 0;
965 break;
966 case VAProfileH264ConstrainedBaseline:
967 printf("Use profile VAProfileH264ConstrainedBaseline\n");
968 constraint_set_flag |= (1 << 0 | 1 << 1); /* Annex A.2.2 */
969 ip_period = 1;
970 break;
971
972 case VAProfileH264Main:
973 printf("Use profile VAProfileH264Main\n");
974 constraint_set_flag |= (1 << 1); /* Annex A.2.2 */
975 break;
976
977 case VAProfileH264High:
978 constraint_set_flag |= (1 << 3); /* Annex A.2.4 */
979 printf("Use profile VAProfileH264High\n");
980 break;
981 default:
982 printf("unknow profile. Set to Baseline");
983 h264_profile = VAProfileH264Baseline;
984 ip_period = 1;
985 constraint_set_flag |= (1 << 0); /* Annex A.2.1 */
986 break;
987 }
988 }
989
990 /* find out the format for the render target, and rate control mode */
991 for (i = 0; i < VAConfigAttribTypeMax; i++)
992 attrib[i].type = i;
993
994 va_status = vaGetConfigAttributes(va_dpy, h264_profile, VAEntrypointEncSlice,
995 &attrib[0], VAConfigAttribTypeMax);
996 CHECK_VASTATUS(va_status, "vaGetConfigAttributes");
997 /* check the interested configattrib */
998 if ((attrib[VAConfigAttribRTFormat].value & VA_RT_FORMAT_YUV420) == 0) {
999 printf("Not find desired YUV420 RT format\n");
1000 exit(1);
1001 } else {
1002 config_attrib[config_attrib_num].type = VAConfigAttribRTFormat;
1003 config_attrib[config_attrib_num].value = VA_RT_FORMAT_YUV420;
1004 config_attrib_num++;
1005 }
1006
1007 if (attrib[VAConfigAttribRateControl].value != VA_ATTRIB_NOT_SUPPORTED) {
1008 int tmp = attrib[VAConfigAttribRateControl].value;
1009
1010 printf("Support rate control mode (0x%x):", tmp);
1011
1012 if (tmp & VA_RC_NONE)
1013 printf("NONE ");
1014 if (tmp & VA_RC_CBR)
1015 printf("CBR ");
1016 if (tmp & VA_RC_VBR)
1017 printf("VBR ");
1018 if (tmp & VA_RC_VCM)
1019 printf("VCM ");
1020 if (tmp & VA_RC_CQP)
1021 printf("CQP ");
1022 if (tmp & VA_RC_VBR_CONSTRAINED)
1023 printf("VBR_CONSTRAINED ");
1024
1025 printf("\n");
1026
1027 /* need to check if support rc_mode */
1028 config_attrib[config_attrib_num].type = VAConfigAttribRateControl;
1029 config_attrib[config_attrib_num].value = rc_mode;
1030 config_attrib_num++;
1031 }
1032
1033
1034 if (attrib[VAConfigAttribEncPackedHeaders].value != VA_ATTRIB_NOT_SUPPORTED) {
1035 int tmp = attrib[VAConfigAttribEncPackedHeaders].value;
1036
1037 printf("Support VAConfigAttribEncPackedHeaders\n");
1038
1039 h264_packedheader = 1;
1040 config_attrib[config_attrib_num].type = VAConfigAttribEncPackedHeaders;
1041 config_attrib[config_attrib_num].value = VA_ENC_PACKED_HEADER_NONE;
1042
1043 if (tmp & VA_ENC_PACKED_HEADER_SEQUENCE) {
1044 printf("Support packed sequence headers\n");
1045 config_attrib[config_attrib_num].value |= VA_ENC_PACKED_HEADER_SEQUENCE;
1046 }
1047
1048 if (tmp & VA_ENC_PACKED_HEADER_PICTURE) {
1049 printf("Support packed picture headers\n");
1050 config_attrib[config_attrib_num].value |= VA_ENC_PACKED_HEADER_PICTURE;
1051 }
1052
1053 if (tmp & VA_ENC_PACKED_HEADER_SLICE) {
1054 printf("Support packed slice headers\n");
1055 config_attrib[config_attrib_num].value |= VA_ENC_PACKED_HEADER_SLICE;
1056 }
1057
1058 if (tmp & VA_ENC_PACKED_HEADER_MISC) {
1059 printf("Support packed misc headers\n");
1060 config_attrib[config_attrib_num].value |= VA_ENC_PACKED_HEADER_MISC;
1061 }
1062
1063 config_attrib_num++;
1064 }
1065
1066 if (attrib[VAConfigAttribEncInterlaced].value != VA_ATTRIB_NOT_SUPPORTED) {
1067 int tmp = attrib[VAConfigAttribEncInterlaced].value;
1068
1069 printf("Support VAConfigAttribEncInterlaced\n");
1070
1071 if (tmp & VA_ENC_INTERLACED_FRAME)
1072 printf("support VA_ENC_INTERLACED_FRAME\n");
1073 if (tmp & VA_ENC_INTERLACED_FIELD)
1074 printf("Support VA_ENC_INTERLACED_FIELD\n");
1075 if (tmp & VA_ENC_INTERLACED_MBAFF)
1076 printf("Support VA_ENC_INTERLACED_MBAFF\n");
1077 if (tmp & VA_ENC_INTERLACED_PAFF)
1078 printf("Support VA_ENC_INTERLACED_PAFF\n");
1079
1080 config_attrib[config_attrib_num].type = VAConfigAttribEncInterlaced;
1081 config_attrib[config_attrib_num].value = VA_ENC_PACKED_HEADER_NONE;
1082 config_attrib_num++;
1083 }
1084
1085 if (attrib[VAConfigAttribEncMaxRefFrames].value != VA_ATTRIB_NOT_SUPPORTED) {
1086 h264_maxref = attrib[VAConfigAttribEncMaxRefFrames].value;
1087
1088 printf("Support %d RefPicList0 and %d RefPicList1\n",
1089 h264_maxref & 0xffff, (h264_maxref >> 16) & 0xffff );
1090 }
1091
1092 if (attrib[VAConfigAttribEncMaxSlices].value != VA_ATTRIB_NOT_SUPPORTED)
1093 printf("Support %d slices\n", attrib[VAConfigAttribEncMaxSlices].value);
1094
1095 if (attrib[VAConfigAttribEncSliceStructure].value != VA_ATTRIB_NOT_SUPPORTED) {
1096 int tmp = attrib[VAConfigAttribEncSliceStructure].value;
1097
1098 printf("Support VAConfigAttribEncSliceStructure\n");
1099
1100 if (tmp & VA_ENC_SLICE_STRUCTURE_ARBITRARY_ROWS)
1101 printf("Support VA_ENC_SLICE_STRUCTURE_ARBITRARY_ROWS\n");
1102 if (tmp & VA_ENC_SLICE_STRUCTURE_POWER_OF_TWO_ROWS)
1103 printf("Support VA_ENC_SLICE_STRUCTURE_POWER_OF_TWO_ROWS\n");
1104 if (tmp & VA_ENC_SLICE_STRUCTURE_ARBITRARY_MACROBLOCKS)
1105 printf("Support VA_ENC_SLICE_STRUCTURE_ARBITRARY_MACROBLOCKS\n");
1106 }
1107 if (attrib[VAConfigAttribEncMacroblockInfo].value != VA_ATTRIB_NOT_SUPPORTED) {
1108 printf("Support VAConfigAttribEncMacroblockInfo\n");
1109 }
1110
1111 free(entrypoints);
1112 return 0;
1113 }
1114
setup_encode()1115 static int setup_encode()
1116 {
1117 VAStatus va_status;
1118 VASurfaceID *tmp_surfaceid;
1119 int codedbuf_size, i;
1120
1121 va_status = vaCreateConfig(va_dpy, h264_profile, VAEntrypointEncSlice,
1122 &config_attrib[0], config_attrib_num, &config_id);
1123 CHECK_VASTATUS(va_status, "vaCreateConfig");
1124
1125 /* create source surfaces */
1126 va_status = vaCreateSurfaces(va_dpy,
1127 VA_RT_FORMAT_YUV420, frame_width_mbaligned, frame_height_mbaligned,
1128 &src_surface[0], SURFACE_NUM,
1129 NULL, 0);
1130 CHECK_VASTATUS(va_status, "vaCreateSurfaces");
1131
1132 /* create reference surfaces */
1133 va_status = vaCreateSurfaces(
1134 va_dpy,
1135 VA_RT_FORMAT_YUV420, frame_width_mbaligned, frame_height_mbaligned,
1136 &ref_surface[0], SURFACE_NUM,
1137 NULL, 0
1138 );
1139 CHECK_VASTATUS(va_status, "vaCreateSurfaces");
1140
1141 tmp_surfaceid = calloc(2 * SURFACE_NUM, sizeof(VASurfaceID));
1142 memcpy(tmp_surfaceid, src_surface, SURFACE_NUM * sizeof(VASurfaceID));
1143 memcpy(tmp_surfaceid + SURFACE_NUM, ref_surface, SURFACE_NUM * sizeof(VASurfaceID));
1144
1145 /* Create a context for this encode pipe */
1146 va_status = vaCreateContext(va_dpy, config_id,
1147 frame_width_mbaligned, frame_height_mbaligned,
1148 VA_PROGRESSIVE,
1149 tmp_surfaceid, 2 * SURFACE_NUM,
1150 &context_id);
1151 CHECK_VASTATUS(va_status, "vaCreateContext");
1152 free(tmp_surfaceid);
1153
1154 codedbuf_size = (frame_width_mbaligned * frame_height_mbaligned * 400) / (16*16);
1155
1156 for (i = 0; i < SURFACE_NUM; i++) {
1157 /* create coded buffer once for all
1158 * other VA buffers which won't be used again after vaRenderPicture.
1159 * so APP can always vaCreateBuffer for every frame
1160 * but coded buffer need to be mapped and accessed after vaRenderPicture/vaEndPicture
1161 * so VA won't maintain the coded buffer
1162 */
1163 va_status = vaCreateBuffer(va_dpy,context_id,VAEncCodedBufferType,
1164 codedbuf_size, 1, NULL, &coded_buf[i]);
1165 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1166 }
1167
1168 return 0;
1169 }
1170
1171
1172
1173 #define partition(ref, field, key, ascending) \
1174 while (i <= j) { \
1175 if (ascending) { \
1176 while (ref[i].field < key) \
1177 i++; \
1178 while (ref[j].field > key) \
1179 j--; \
1180 } else { \
1181 while (ref[i].field > key) \
1182 i++; \
1183 while (ref[j].field < key) \
1184 j--; \
1185 } \
1186 if (i <= j) { \
1187 tmp = ref[i]; \
1188 ref[i] = ref[j]; \
1189 ref[j] = tmp; \
1190 i++; \
1191 j--; \
1192 } \
1193 } \
1194
sort_one(VAPictureH264 ref[],int left,int right,int ascending,int frame_idx)1195 static void sort_one(VAPictureH264 ref[], int left, int right,
1196 int ascending, int frame_idx)
1197 {
1198 int i = left, j = right;
1199 unsigned int key;
1200 VAPictureH264 tmp;
1201
1202 if (frame_idx) {
1203 key = ref[(left + right) / 2].frame_idx;
1204 partition(ref, frame_idx, key, ascending);
1205 } else {
1206 key = ref[(left + right) / 2].TopFieldOrderCnt;
1207 partition(ref, TopFieldOrderCnt, (signed int)key, ascending);
1208 }
1209
1210 /* recursion */
1211 if (left < j)
1212 sort_one(ref, left, j, ascending, frame_idx);
1213
1214 if (i < right)
1215 sort_one(ref, i, right, ascending, frame_idx);
1216 }
1217
sort_two(VAPictureH264 ref[],int left,int right,unsigned int key,unsigned int frame_idx,int partition_ascending,int list0_ascending,int list1_ascending)1218 static void sort_two(VAPictureH264 ref[], int left, int right, unsigned int key, unsigned int frame_idx,
1219 int partition_ascending, int list0_ascending, int list1_ascending)
1220 {
1221 int i = left, j = right;
1222 VAPictureH264 tmp;
1223
1224 if (frame_idx) {
1225 partition(ref, frame_idx, key, partition_ascending);
1226 } else {
1227 partition(ref, TopFieldOrderCnt, (signed int)key, partition_ascending);
1228 }
1229
1230
1231 sort_one(ref, left, i-1, list0_ascending, frame_idx);
1232 sort_one(ref, j+1, right, list1_ascending, frame_idx);
1233 }
1234
update_ReferenceFrames(void)1235 static int update_ReferenceFrames(void)
1236 {
1237 int i;
1238
1239 if (current_frame_type == FRAME_B)
1240 return 0;
1241
1242 CurrentCurrPic.flags = VA_PICTURE_H264_SHORT_TERM_REFERENCE;
1243 numShortTerm++;
1244 if (numShortTerm > num_ref_frames)
1245 numShortTerm = num_ref_frames;
1246 for (i=numShortTerm-1; i>0; i--)
1247 ReferenceFrames[i] = ReferenceFrames[i-1];
1248 ReferenceFrames[0] = CurrentCurrPic;
1249
1250 if (current_frame_type != FRAME_B)
1251 current_frame_num++;
1252 if (current_frame_num > MaxFrameNum)
1253 current_frame_num = 0;
1254
1255 return 0;
1256 }
1257
1258
update_RefPicList(void)1259 static int update_RefPicList(void)
1260 {
1261 unsigned int current_poc = CurrentCurrPic.TopFieldOrderCnt;
1262
1263 if (current_frame_type == FRAME_P) {
1264 memcpy(RefPicList0_P, ReferenceFrames, numShortTerm * sizeof(VAPictureH264));
1265 sort_one(RefPicList0_P, 0, numShortTerm-1, 0, 1);
1266 }
1267
1268 if (current_frame_type == FRAME_B) {
1269 memcpy(RefPicList0_B, ReferenceFrames, numShortTerm * sizeof(VAPictureH264));
1270 sort_two(RefPicList0_B, 0, numShortTerm-1, current_poc, 0,
1271 1, 0, 1);
1272
1273 memcpy(RefPicList1_B, ReferenceFrames, numShortTerm * sizeof(VAPictureH264));
1274 sort_two(RefPicList1_B, 0, numShortTerm-1, current_poc, 0,
1275 0, 1, 0);
1276 }
1277
1278 return 0;
1279 }
1280
1281
render_sequence(void)1282 static int render_sequence(void)
1283 {
1284 VABufferID seq_param_buf, rc_param_buf, misc_param_tmpbuf, render_id[2];
1285 VAStatus va_status;
1286 VAEncMiscParameterBuffer *misc_param, *misc_param_tmp;
1287 VAEncMiscParameterRateControl *misc_rate_ctrl;
1288
1289 seq_param.level_idc = 41 /*SH_LEVEL_3*/;
1290 seq_param.picture_width_in_mbs = frame_width_mbaligned / 16;
1291 seq_param.picture_height_in_mbs = frame_height_mbaligned / 16;
1292 seq_param.bits_per_second = frame_bitrate;
1293
1294 seq_param.intra_period = intra_period;
1295 seq_param.intra_idr_period = intra_idr_period;
1296 seq_param.ip_period = ip_period;
1297
1298 seq_param.max_num_ref_frames = num_ref_frames;
1299 seq_param.seq_fields.bits.frame_mbs_only_flag = 1;
1300 seq_param.time_scale = 900;
1301 seq_param.num_units_in_tick = 15; /* Tc = num_units_in_tick / time_sacle */
1302 seq_param.seq_fields.bits.log2_max_pic_order_cnt_lsb_minus4 = Log2MaxPicOrderCntLsb - 4;
1303 seq_param.seq_fields.bits.log2_max_frame_num_minus4 = Log2MaxFrameNum - 4;;
1304 seq_param.seq_fields.bits.frame_mbs_only_flag = 1;
1305 seq_param.seq_fields.bits.chroma_format_idc = 1;
1306 seq_param.seq_fields.bits.direct_8x8_inference_flag = 1;
1307
1308 if (frame_width != frame_width_mbaligned ||
1309 frame_height != frame_height_mbaligned) {
1310 seq_param.frame_cropping_flag = 1;
1311 seq_param.frame_crop_left_offset = 0;
1312 seq_param.frame_crop_right_offset = (frame_width_mbaligned - frame_width)/2;
1313 seq_param.frame_crop_top_offset = 0;
1314 seq_param.frame_crop_bottom_offset = (frame_height_mbaligned - frame_height)/2;
1315 }
1316
1317 va_status = vaCreateBuffer(va_dpy, context_id,
1318 VAEncSequenceParameterBufferType,
1319 sizeof(seq_param),1,&seq_param,&seq_param_buf);
1320 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1321
1322 va_status = vaCreateBuffer(va_dpy, context_id,
1323 VAEncMiscParameterBufferType,
1324 sizeof(VAEncMiscParameterBuffer) + sizeof(VAEncMiscParameterRateControl),
1325 1,NULL,&rc_param_buf);
1326 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1327
1328 vaMapBuffer(va_dpy, rc_param_buf,(void **)&misc_param);
1329 misc_param->type = VAEncMiscParameterTypeRateControl;
1330 misc_rate_ctrl = (VAEncMiscParameterRateControl *)misc_param->data;
1331 memset(misc_rate_ctrl, 0, sizeof(*misc_rate_ctrl));
1332 misc_rate_ctrl->bits_per_second = frame_bitrate;
1333 misc_rate_ctrl->target_percentage = 66;
1334 misc_rate_ctrl->window_size = 1000;
1335 misc_rate_ctrl->initial_qp = initial_qp;
1336 misc_rate_ctrl->min_qp = minimal_qp;
1337 misc_rate_ctrl->basic_unit_size = 0;
1338 vaUnmapBuffer(va_dpy, rc_param_buf);
1339
1340 render_id[0] = seq_param_buf;
1341 render_id[1] = rc_param_buf;
1342
1343 va_status = vaRenderPicture(va_dpy,context_id, &render_id[0], 2);
1344 CHECK_VASTATUS(va_status,"vaRenderPicture");;
1345
1346 if (misc_priv_type != 0) {
1347 va_status = vaCreateBuffer(va_dpy, context_id,
1348 VAEncMiscParameterBufferType,
1349 sizeof(VAEncMiscParameterBuffer),
1350 1, NULL, &misc_param_tmpbuf);
1351 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1352 vaMapBuffer(va_dpy, misc_param_tmpbuf,(void **)&misc_param_tmp);
1353 misc_param_tmp->type = misc_priv_type;
1354 misc_param_tmp->data[0] = misc_priv_value;
1355 vaUnmapBuffer(va_dpy, misc_param_tmpbuf);
1356
1357 va_status = vaRenderPicture(va_dpy,context_id, &misc_param_tmpbuf, 1);
1358 }
1359
1360 return 0;
1361 }
1362
calc_poc(int pic_order_cnt_lsb)1363 static int calc_poc(int pic_order_cnt_lsb)
1364 {
1365 static int PicOrderCntMsb_ref = 0, pic_order_cnt_lsb_ref = 0;
1366 int prevPicOrderCntMsb, prevPicOrderCntLsb;
1367 int PicOrderCntMsb, TopFieldOrderCnt;
1368
1369 if (current_frame_type == FRAME_IDR)
1370 prevPicOrderCntMsb = prevPicOrderCntLsb = 0;
1371 else {
1372 prevPicOrderCntMsb = PicOrderCntMsb_ref;
1373 prevPicOrderCntLsb = pic_order_cnt_lsb_ref;
1374 }
1375
1376 if ((pic_order_cnt_lsb < prevPicOrderCntLsb) &&
1377 ((prevPicOrderCntLsb - pic_order_cnt_lsb) >= (int)(MaxPicOrderCntLsb / 2)))
1378 PicOrderCntMsb = prevPicOrderCntMsb + MaxPicOrderCntLsb;
1379 else if ((pic_order_cnt_lsb > prevPicOrderCntLsb) &&
1380 ((pic_order_cnt_lsb - prevPicOrderCntLsb) > (int)(MaxPicOrderCntLsb / 2)))
1381 PicOrderCntMsb = prevPicOrderCntMsb - MaxPicOrderCntLsb;
1382 else
1383 PicOrderCntMsb = prevPicOrderCntMsb;
1384
1385 TopFieldOrderCnt = PicOrderCntMsb + pic_order_cnt_lsb;
1386
1387 if (current_frame_type != FRAME_B) {
1388 PicOrderCntMsb_ref = PicOrderCntMsb;
1389 pic_order_cnt_lsb_ref = pic_order_cnt_lsb;
1390 }
1391
1392 return TopFieldOrderCnt;
1393 }
1394
render_picture(void)1395 static int render_picture(void)
1396 {
1397 VABufferID pic_param_buf;
1398 VAStatus va_status;
1399 int i = 0;
1400
1401 pic_param.CurrPic.picture_id = ref_surface[current_slot];
1402 pic_param.CurrPic.frame_idx = current_frame_num;
1403 pic_param.CurrPic.flags = 0;
1404 pic_param.CurrPic.TopFieldOrderCnt = calc_poc((current_frame_display - current_IDR_display) % MaxPicOrderCntLsb);
1405 pic_param.CurrPic.BottomFieldOrderCnt = pic_param.CurrPic.TopFieldOrderCnt;
1406 CurrentCurrPic = pic_param.CurrPic;
1407
1408 if (getenv("TO_DEL")) { /* set RefPicList into ReferenceFrames */
1409 update_RefPicList(); /* calc RefPicList */
1410 memset(pic_param.ReferenceFrames, 0xff, 16 * sizeof(VAPictureH264)); /* invalid all */
1411 if (current_frame_type == FRAME_P) {
1412 pic_param.ReferenceFrames[0] = RefPicList0_P[0];
1413 } else if (current_frame_type == FRAME_B) {
1414 pic_param.ReferenceFrames[0] = RefPicList0_B[0];
1415 pic_param.ReferenceFrames[1] = RefPicList1_B[0];
1416 }
1417 } else {
1418 memcpy(pic_param.ReferenceFrames, ReferenceFrames, numShortTerm*sizeof(VAPictureH264));
1419 for (i = numShortTerm; i < SURFACE_NUM; i++) {
1420 pic_param.ReferenceFrames[i].picture_id = VA_INVALID_SURFACE;
1421 pic_param.ReferenceFrames[i].flags = VA_PICTURE_H264_INVALID;
1422 }
1423 }
1424
1425 pic_param.pic_fields.bits.idr_pic_flag = (current_frame_type == FRAME_IDR);
1426 pic_param.pic_fields.bits.reference_pic_flag = (current_frame_type != FRAME_B);
1427 pic_param.pic_fields.bits.entropy_coding_mode_flag = h264_entropy_mode;
1428 pic_param.pic_fields.bits.deblocking_filter_control_present_flag = 1;
1429 pic_param.frame_num = current_frame_num;
1430 pic_param.coded_buf = coded_buf[current_slot];
1431 pic_param.last_picture = (current_frame_encoding == frame_count);
1432 pic_param.pic_init_qp = initial_qp;
1433
1434 va_status = vaCreateBuffer(va_dpy, context_id,VAEncPictureParameterBufferType,
1435 sizeof(pic_param),1,&pic_param, &pic_param_buf);
1436 CHECK_VASTATUS(va_status,"vaCreateBuffer");;
1437
1438 va_status = vaRenderPicture(va_dpy,context_id, &pic_param_buf, 1);
1439 CHECK_VASTATUS(va_status,"vaRenderPicture");
1440
1441 return 0;
1442 }
1443
render_packedsequence(void)1444 static int render_packedsequence(void)
1445 {
1446 VAEncPackedHeaderParameterBuffer packedheader_param_buffer;
1447 VABufferID packedseq_para_bufid, packedseq_data_bufid, render_id[2];
1448 unsigned int length_in_bits;
1449 unsigned char *packedseq_buffer = NULL;
1450 VAStatus va_status;
1451
1452 length_in_bits = build_packed_seq_buffer(&packedseq_buffer);
1453
1454 packedheader_param_buffer.type = VAEncPackedHeaderSequence;
1455
1456 packedheader_param_buffer.bit_length = length_in_bits; /*length_in_bits*/
1457 packedheader_param_buffer.has_emulation_bytes = 0;
1458 va_status = vaCreateBuffer(va_dpy,
1459 context_id,
1460 VAEncPackedHeaderParameterBufferType,
1461 sizeof(packedheader_param_buffer), 1, &packedheader_param_buffer,
1462 &packedseq_para_bufid);
1463 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1464
1465 va_status = vaCreateBuffer(va_dpy,
1466 context_id,
1467 VAEncPackedHeaderDataBufferType,
1468 (length_in_bits + 7) / 8, 1, packedseq_buffer,
1469 &packedseq_data_bufid);
1470 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1471
1472 render_id[0] = packedseq_para_bufid;
1473 render_id[1] = packedseq_data_bufid;
1474 va_status = vaRenderPicture(va_dpy,context_id, render_id, 2);
1475 CHECK_VASTATUS(va_status,"vaRenderPicture");
1476
1477 free(packedseq_buffer);
1478
1479 return 0;
1480 }
1481
1482
render_packedpicture(void)1483 static int render_packedpicture(void)
1484 {
1485 VAEncPackedHeaderParameterBuffer packedheader_param_buffer;
1486 VABufferID packedpic_para_bufid, packedpic_data_bufid, render_id[2];
1487 unsigned int length_in_bits;
1488 unsigned char *packedpic_buffer = NULL;
1489 VAStatus va_status;
1490
1491 length_in_bits = build_packed_pic_buffer(&packedpic_buffer);
1492 packedheader_param_buffer.type = VAEncPackedHeaderPicture;
1493 packedheader_param_buffer.bit_length = length_in_bits;
1494 packedheader_param_buffer.has_emulation_bytes = 0;
1495
1496 va_status = vaCreateBuffer(va_dpy,
1497 context_id,
1498 VAEncPackedHeaderParameterBufferType,
1499 sizeof(packedheader_param_buffer), 1, &packedheader_param_buffer,
1500 &packedpic_para_bufid);
1501 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1502
1503 va_status = vaCreateBuffer(va_dpy,
1504 context_id,
1505 VAEncPackedHeaderDataBufferType,
1506 (length_in_bits + 7) / 8, 1, packedpic_buffer,
1507 &packedpic_data_bufid);
1508 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1509
1510 render_id[0] = packedpic_para_bufid;
1511 render_id[1] = packedpic_data_bufid;
1512 va_status = vaRenderPicture(va_dpy,context_id, render_id, 2);
1513 CHECK_VASTATUS(va_status,"vaRenderPicture");
1514
1515 free(packedpic_buffer);
1516
1517 return 0;
1518 }
1519
render_packedsei(void)1520 static void render_packedsei(void)
1521 {
1522 VAEncPackedHeaderParameterBuffer packed_header_param_buffer;
1523 VABufferID packed_sei_header_param_buf_id, packed_sei_buf_id, render_id[2];
1524 unsigned int length_in_bits /*offset_in_bytes*/;
1525 unsigned char *packed_sei_buffer = NULL;
1526 VAStatus va_status;
1527 int init_cpb_size, target_bit_rate, i_initial_cpb_removal_delay_length, i_initial_cpb_removal_delay;
1528 int i_cpb_removal_delay, i_dpb_output_delay_length, i_cpb_removal_delay_length;
1529
1530 /* it comes for the bps defined in SPS */
1531 target_bit_rate = frame_bitrate;
1532 init_cpb_size = (target_bit_rate * 8) >> 10;
1533 i_initial_cpb_removal_delay = init_cpb_size * 0.5 * 1024 / target_bit_rate * 90000;
1534
1535 i_cpb_removal_delay = 2;
1536 i_initial_cpb_removal_delay_length = 24;
1537 i_cpb_removal_delay_length = 24;
1538 i_dpb_output_delay_length = 24;
1539
1540
1541 length_in_bits = build_packed_sei_buffer_timing(
1542 i_initial_cpb_removal_delay_length,
1543 i_initial_cpb_removal_delay,
1544 0,
1545 i_cpb_removal_delay_length,
1546 i_cpb_removal_delay * current_frame_encoding,
1547 i_dpb_output_delay_length,
1548 0,
1549 &packed_sei_buffer);
1550
1551 //offset_in_bytes = 0;
1552 packed_header_param_buffer.type = VAEncPackedHeaderH264_SEI;
1553 packed_header_param_buffer.bit_length = length_in_bits;
1554 packed_header_param_buffer.has_emulation_bytes = 0;
1555
1556 va_status = vaCreateBuffer(va_dpy,
1557 context_id,
1558 VAEncPackedHeaderParameterBufferType,
1559 sizeof(packed_header_param_buffer), 1, &packed_header_param_buffer,
1560 &packed_sei_header_param_buf_id);
1561 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1562
1563 va_status = vaCreateBuffer(va_dpy,
1564 context_id,
1565 VAEncPackedHeaderDataBufferType,
1566 (length_in_bits + 7) / 8, 1, packed_sei_buffer,
1567 &packed_sei_buf_id);
1568 CHECK_VASTATUS(va_status,"vaCreateBuffer");
1569
1570
1571 render_id[0] = packed_sei_header_param_buf_id;
1572 render_id[1] = packed_sei_buf_id;
1573 va_status = vaRenderPicture(va_dpy,context_id, render_id, 2);
1574 CHECK_VASTATUS(va_status,"vaRenderPicture");
1575
1576
1577 free(packed_sei_buffer);
1578
1579 return;
1580 }
1581
1582
render_hrd(void)1583 static int render_hrd(void)
1584 {
1585 VABufferID misc_parameter_hrd_buf_id;
1586 VAStatus va_status;
1587 VAEncMiscParameterBuffer *misc_param;
1588 VAEncMiscParameterHRD *misc_hrd_param;
1589
1590 va_status = vaCreateBuffer(va_dpy, context_id,
1591 VAEncMiscParameterBufferType,
1592 sizeof(VAEncMiscParameterBuffer) + sizeof(VAEncMiscParameterHRD),
1593 1,
1594 NULL,
1595 &misc_parameter_hrd_buf_id);
1596 CHECK_VASTATUS(va_status, "vaCreateBuffer");
1597
1598 vaMapBuffer(va_dpy,
1599 misc_parameter_hrd_buf_id,
1600 (void **)&misc_param);
1601 misc_param->type = VAEncMiscParameterTypeHRD;
1602 misc_hrd_param = (VAEncMiscParameterHRD *)misc_param->data;
1603
1604 if (frame_bitrate > 0) {
1605 misc_hrd_param->initial_buffer_fullness = frame_bitrate * 1024 * 4;
1606 misc_hrd_param->buffer_size = frame_bitrate * 1024 * 8;
1607 } else {
1608 misc_hrd_param->initial_buffer_fullness = 0;
1609 misc_hrd_param->buffer_size = 0;
1610 }
1611 vaUnmapBuffer(va_dpy, misc_parameter_hrd_buf_id);
1612
1613 va_status = vaRenderPicture(va_dpy,context_id, &misc_parameter_hrd_buf_id, 1);
1614 CHECK_VASTATUS(va_status,"vaRenderPicture");;
1615
1616 return 0;
1617 }
1618
render_slice(void)1619 static int render_slice(void)
1620 {
1621 VABufferID slice_param_buf;
1622 VAStatus va_status;
1623 int i;
1624
1625 update_RefPicList();
1626
1627 /* one frame, one slice */
1628 slice_param.macroblock_address = 0;
1629 slice_param.num_macroblocks = frame_width_mbaligned * frame_height_mbaligned/(16*16); /* Measured by MB */
1630 slice_param.slice_type = (current_frame_type == FRAME_IDR)?2:current_frame_type;
1631 if (current_frame_type == FRAME_IDR) {
1632 if (current_frame_encoding != 0)
1633 ++slice_param.idr_pic_id;
1634 } else if (current_frame_type == FRAME_P) {
1635 int refpiclist0_max = h264_maxref & 0xffff;
1636 memcpy(slice_param.RefPicList0, RefPicList0_P, refpiclist0_max*sizeof(VAPictureH264));
1637
1638 for (i = refpiclist0_max; i < 32; i++) {
1639 slice_param.RefPicList0[i].picture_id = VA_INVALID_SURFACE;
1640 slice_param.RefPicList0[i].flags = VA_PICTURE_H264_INVALID;
1641 }
1642 } else if (current_frame_type == FRAME_B) {
1643 int refpiclist0_max = h264_maxref & 0xffff;
1644 int refpiclist1_max = (h264_maxref >> 16) & 0xffff;
1645
1646 memcpy(slice_param.RefPicList0, RefPicList0_B, refpiclist0_max*sizeof(VAPictureH264));
1647 for (i = refpiclist0_max; i < 32; i++) {
1648 slice_param.RefPicList0[i].picture_id = VA_INVALID_SURFACE;
1649 slice_param.RefPicList0[i].flags = VA_PICTURE_H264_INVALID;
1650 }
1651
1652 memcpy(slice_param.RefPicList1, RefPicList1_B, refpiclist1_max*sizeof(VAPictureH264));
1653 for (i = refpiclist1_max; i < 32; i++) {
1654 slice_param.RefPicList1[i].picture_id = VA_INVALID_SURFACE;
1655 slice_param.RefPicList1[i].flags = VA_PICTURE_H264_INVALID;
1656 }
1657 }
1658
1659 slice_param.slice_alpha_c0_offset_div2 = 0;
1660 slice_param.slice_beta_offset_div2 = 0;
1661 slice_param.direct_spatial_mv_pred_flag = 1;
1662 slice_param.pic_order_cnt_lsb = (current_frame_display - current_IDR_display) % MaxPicOrderCntLsb;
1663
1664 va_status = vaCreateBuffer(va_dpy,context_id,VAEncSliceParameterBufferType,
1665 sizeof(slice_param),1,&slice_param,&slice_param_buf);
1666 CHECK_VASTATUS(va_status,"vaCreateBuffer");;
1667
1668 va_status = vaRenderPicture(va_dpy,context_id, &slice_param_buf, 1);
1669 CHECK_VASTATUS(va_status,"vaRenderPicture");
1670
1671 return 0;
1672 }
1673
1674
upload_source_YUV_once_for_all()1675 static int upload_source_YUV_once_for_all()
1676 {
1677 int box_width=8;
1678 int row_shift=0;
1679 int i;
1680
1681 for (i = 0; i < SURFACE_NUM; i++) {
1682 printf("\rLoading data into surface %d.....", i);
1683 upload_surface(va_dpy, src_surface[i], box_width, row_shift, 0);
1684
1685 row_shift++;
1686 if (row_shift==(2*box_width)) row_shift= 0;
1687 }
1688 printf("Complete surface loading\n");
1689
1690 return 0;
1691 }
1692
load_surface(VASurfaceID surface_id,unsigned long long display_order)1693 static int load_surface(VASurfaceID surface_id, unsigned long long display_order)
1694 {
1695 unsigned char *srcyuv_ptr = NULL, *src_Y = NULL, *src_U = NULL, *src_V = NULL;
1696 unsigned long long frame_start, mmap_start;
1697 char *mmap_ptr = NULL;
1698 int frame_size, mmap_size;
1699
1700 if (srcyuv_fp == NULL)
1701 return 0;
1702
1703 /* allow encoding more than srcyuv_frames */
1704 display_order = display_order % srcyuv_frames;
1705 frame_size = frame_width * frame_height * 3 / 2; /* for YUV420 */
1706 frame_start = display_order * frame_size;
1707
1708 mmap_start = frame_start & (~0xfff);
1709 mmap_size = (frame_size + (frame_start & 0xfff) + 0xfff) & (~0xfff);
1710 mmap_ptr = mmap(0, mmap_size, PROT_READ, MAP_SHARED,
1711 fileno(srcyuv_fp), mmap_start);
1712 if (mmap_ptr == MAP_FAILED) {
1713 printf("Failed to mmap YUV file (%s)\n", strerror(errno));
1714 return 1;
1715 }
1716 srcyuv_ptr = (unsigned char *)mmap_ptr + (frame_start & 0xfff);
1717 if (srcyuv_fourcc == VA_FOURCC_NV12) {
1718 src_Y = srcyuv_ptr;
1719 src_U = src_Y + frame_width * frame_height;
1720 src_V = NULL;
1721 } else if (srcyuv_fourcc == VA_FOURCC_IYUV ||
1722 srcyuv_fourcc == VA_FOURCC_YV12) {
1723 src_Y = srcyuv_ptr;
1724 if (srcyuv_fourcc == VA_FOURCC_IYUV) {
1725 src_U = src_Y + frame_width * frame_height;
1726 src_V = src_U + (frame_width/2) * (frame_height/2);
1727 } else { /* YV12 */
1728 src_V = src_Y + frame_width * frame_height;
1729 src_U = src_V + (frame_width/2) * (frame_height/2);
1730 }
1731 } else {
1732 printf("Unsupported source YUV format\n");
1733 exit(1);
1734 }
1735
1736 upload_surface_yuv(va_dpy, surface_id,
1737 srcyuv_fourcc, frame_width, frame_height,
1738 src_Y, src_U, src_V);
1739 if (mmap_ptr)
1740 munmap(mmap_ptr, mmap_size);
1741
1742 return 0;
1743 }
1744
1745
save_recyuv(VASurfaceID surface_id,unsigned long long display_order,unsigned long long encode_order)1746 static int save_recyuv(VASurfaceID surface_id,
1747 unsigned long long display_order,
1748 unsigned long long encode_order)
1749 {
1750 unsigned char *dst_Y = NULL, *dst_U = NULL, *dst_V = NULL;
1751
1752 if (recyuv_fp == NULL)
1753 return 0;
1754
1755 if (srcyuv_fourcc == VA_FOURCC_NV12) {
1756 int uv_size = 2 * (frame_width/2) * (frame_height/2);
1757 dst_Y = malloc(2*uv_size);
1758 dst_U = malloc(uv_size);
1759 } else if (srcyuv_fourcc == VA_FOURCC_IYUV ||
1760 srcyuv_fourcc == VA_FOURCC_YV12) {
1761 int uv_size = (frame_width/2) * (frame_height/2);
1762 dst_Y = malloc(4*uv_size);
1763 dst_U = malloc(uv_size);
1764 dst_V = malloc(uv_size);
1765 } else {
1766 printf("Unsupported source YUV format\n");
1767 exit(1);
1768 }
1769
1770 download_surface_yuv(va_dpy, surface_id,
1771 srcyuv_fourcc, frame_width, frame_height,
1772 dst_Y, dst_U, dst_V);
1773 fseek(recyuv_fp, display_order * frame_width * frame_height * 1.5, SEEK_SET);
1774
1775 if (srcyuv_fourcc == VA_FOURCC_NV12) {
1776 int uv_size = 2 * (frame_width/2) * (frame_height/2);
1777 fwrite(dst_Y, uv_size * 2, 1, recyuv_fp);
1778 fwrite(dst_U, uv_size, 1, recyuv_fp);
1779 } else if (srcyuv_fourcc == VA_FOURCC_IYUV ||
1780 srcyuv_fourcc == VA_FOURCC_YV12) {
1781 int uv_size = (frame_width/2) * (frame_height/2);
1782 fwrite(dst_Y, uv_size * 4, 1, recyuv_fp);
1783
1784 if (srcyuv_fourcc == VA_FOURCC_IYUV) {
1785 fwrite(dst_U, uv_size, 1, recyuv_fp);
1786 fwrite(dst_V, uv_size, 1, recyuv_fp);
1787 } else {
1788 fwrite(dst_V, uv_size, 1, recyuv_fp);
1789 fwrite(dst_U, uv_size, 1, recyuv_fp);
1790 }
1791 } else {
1792 printf("Unsupported YUV format\n");
1793 exit(1);
1794 }
1795
1796 if (dst_Y)
1797 free(dst_Y);
1798 if (dst_U)
1799 free(dst_U);
1800 if (dst_V)
1801 free(dst_V);
1802
1803 fflush(recyuv_fp);
1804
1805 return 0;
1806 }
1807
1808
save_codeddata(unsigned long long display_order,unsigned long long encode_order)1809 static int save_codeddata(unsigned long long display_order, unsigned long long encode_order)
1810 {
1811 VACodedBufferSegment *buf_list = NULL;
1812 VAStatus va_status;
1813 unsigned int coded_size = 0;
1814
1815 va_status = vaMapBuffer(va_dpy,coded_buf[display_order % SURFACE_NUM],(void **)(&buf_list));
1816 CHECK_VASTATUS(va_status,"vaMapBuffer");
1817 while (buf_list != NULL) {
1818 coded_size += fwrite(buf_list->buf, 1, buf_list->size, coded_fp);
1819 buf_list = (VACodedBufferSegment *) buf_list->next;
1820
1821 frame_size += coded_size;
1822 }
1823 vaUnmapBuffer(va_dpy,coded_buf[display_order % SURFACE_NUM]);
1824
1825 printf("\r "); /* return back to startpoint */
1826 switch (encode_order % 4) {
1827 case 0:
1828 printf("|");
1829 break;
1830 case 1:
1831 printf("/");
1832 break;
1833 case 2:
1834 printf("-");
1835 break;
1836 case 3:
1837 printf("\\");
1838 break;
1839 }
1840 printf("%08lld", encode_order);
1841 printf("(%06d bytes coded)",coded_size);
1842
1843 fflush(coded_fp);
1844
1845 return 0;
1846 }
1847
1848
storage_task_dequeue(void)1849 static struct storage_task_t * storage_task_dequeue(void)
1850 {
1851 struct storage_task_t *header;
1852
1853 pthread_mutex_lock(&encode_mutex);
1854
1855 header = storage_task_header;
1856 if (storage_task_header != NULL) {
1857 if (storage_task_tail == storage_task_header)
1858 storage_task_tail = NULL;
1859 storage_task_header = header->next;
1860 }
1861
1862 pthread_mutex_unlock(&encode_mutex);
1863
1864 return header;
1865 }
1866
storage_task_queue(unsigned long long display_order,unsigned long long encode_order)1867 static int storage_task_queue(unsigned long long display_order, unsigned long long encode_order)
1868 {
1869 struct storage_task_t *tmp;
1870
1871 tmp = calloc(1, sizeof(struct storage_task_t));
1872 tmp->display_order = display_order;
1873 tmp->encode_order = encode_order;
1874
1875 pthread_mutex_lock(&encode_mutex);
1876
1877 if (storage_task_header == NULL) {
1878 storage_task_header = tmp;
1879 storage_task_tail = tmp;
1880 } else {
1881 storage_task_tail->next = tmp;
1882 storage_task_tail = tmp;
1883 }
1884
1885 srcsurface_status[display_order % SURFACE_NUM] = SRC_SURFACE_IN_STORAGE;
1886 pthread_cond_signal(&encode_cond);
1887
1888 pthread_mutex_unlock(&encode_mutex);
1889
1890 return 0;
1891 }
1892
storage_task(unsigned long long display_order,unsigned long long encode_order)1893 static void storage_task(unsigned long long display_order, unsigned long long encode_order)
1894 {
1895 unsigned int tmp;
1896 VAStatus va_status;
1897
1898 tmp = GetTickCount();
1899 va_status = vaSyncSurface(va_dpy, src_surface[display_order % SURFACE_NUM]);
1900 CHECK_VASTATUS(va_status,"vaSyncSurface");
1901 SyncPictureTicks += GetTickCount() - tmp;
1902 tmp = GetTickCount();
1903 save_codeddata(display_order, encode_order);
1904 SavePictureTicks += GetTickCount() - tmp;
1905
1906 save_recyuv(ref_surface[display_order % SURFACE_NUM], display_order, encode_order);
1907
1908 /* reload a new frame data */
1909 tmp = GetTickCount();
1910 if (srcyuv_fp != NULL)
1911 load_surface(src_surface[display_order % SURFACE_NUM], display_order + SURFACE_NUM);
1912 UploadPictureTicks += GetTickCount() - tmp;
1913
1914 pthread_mutex_lock(&encode_mutex);
1915 srcsurface_status[display_order % SURFACE_NUM] = SRC_SURFACE_IN_ENCODING;
1916 pthread_mutex_unlock(&encode_mutex);
1917 }
1918
1919
storage_task_thread(void * t)1920 static void * storage_task_thread(void *t)
1921 {
1922 while (1) {
1923 struct storage_task_t *current;
1924
1925 current = storage_task_dequeue();
1926 if (current == NULL) {
1927 pthread_mutex_lock(&encode_mutex);
1928 pthread_cond_wait(&encode_cond, &encode_mutex);
1929 pthread_mutex_unlock(&encode_mutex);
1930 continue;
1931 }
1932
1933 storage_task(current->display_order, current->encode_order);
1934
1935 free(current);
1936
1937 /* all frames are saved, exit the thread */
1938 if (++frame_coded >= frame_count)
1939 break;
1940 }
1941
1942 return 0;
1943 }
1944
1945
encode_frames(void)1946 static int encode_frames(void)
1947 {
1948 unsigned int i, tmp;
1949 VAStatus va_status;
1950 //VASurfaceStatus surface_status;
1951
1952 /* upload RAW YUV data into all surfaces */
1953 tmp = GetTickCount();
1954 if (srcyuv_fp != NULL) {
1955 for (i = 0; i < SURFACE_NUM; i++)
1956 load_surface(src_surface[i], i);
1957 } else
1958 upload_source_YUV_once_for_all();
1959 UploadPictureTicks += GetTickCount() - tmp;
1960
1961 /* ready for encoding */
1962 memset(srcsurface_status, SRC_SURFACE_IN_ENCODING, sizeof(srcsurface_status));
1963
1964 memset(&seq_param, 0, sizeof(seq_param));
1965 memset(&pic_param, 0, sizeof(pic_param));
1966 memset(&slice_param, 0, sizeof(slice_param));
1967
1968 if (encode_syncmode == 0)
1969 pthread_create(&encode_thread, NULL, storage_task_thread, NULL);
1970
1971 for (current_frame_encoding = 0; current_frame_encoding < frame_count; current_frame_encoding++) {
1972 encoding2display_order(current_frame_encoding, intra_period, intra_idr_period, ip_period,
1973 ¤t_frame_display, ¤t_frame_type);
1974 if (current_frame_type == FRAME_IDR) {
1975 numShortTerm = 0;
1976 current_frame_num = 0;
1977 current_IDR_display = current_frame_display;
1978 }
1979
1980 /* check if the source frame is ready */
1981 while (srcsurface_status[current_slot] != SRC_SURFACE_IN_ENCODING) {
1982 usleep(1);
1983 }
1984
1985 tmp = GetTickCount();
1986 va_status = vaBeginPicture(va_dpy, context_id, src_surface[current_slot]);
1987 CHECK_VASTATUS(va_status,"vaBeginPicture");
1988 BeginPictureTicks += GetTickCount() - tmp;
1989
1990 tmp = GetTickCount();
1991 if (current_frame_type == FRAME_IDR) {
1992 render_sequence();
1993 render_picture();
1994 if (h264_packedheader) {
1995 render_packedsequence();
1996 render_packedpicture();
1997 }
1998 //if (rc_mode == VA_RC_CBR)
1999 // render_packedsei();
2000 //render_hrd();
2001 } else {
2002 //render_sequence();
2003 render_picture();
2004 //if (rc_mode == VA_RC_CBR)
2005 // render_packedsei();
2006 //render_hrd();
2007 }
2008 render_slice();
2009 RenderPictureTicks += GetTickCount() - tmp;
2010
2011 tmp = GetTickCount();
2012 va_status = vaEndPicture(va_dpy,context_id);
2013 CHECK_VASTATUS(va_status,"vaEndPicture");;
2014 EndPictureTicks += GetTickCount() - tmp;
2015
2016 if (encode_syncmode)
2017 storage_task(current_frame_display, current_frame_encoding);
2018 else /* queue the storage task queue */
2019 storage_task_queue(current_frame_display, current_frame_encoding);
2020
2021 update_ReferenceFrames();
2022 }
2023
2024 if (encode_syncmode == 0) {
2025 int ret;
2026 pthread_join(encode_thread, (void **)&ret);
2027 }
2028
2029 return 0;
2030 }
2031
2032
release_encode()2033 static int release_encode()
2034 {
2035 int i;
2036
2037 vaDestroySurfaces(va_dpy,&src_surface[0],SURFACE_NUM);
2038 vaDestroySurfaces(va_dpy,&ref_surface[0],SURFACE_NUM);
2039
2040 for (i = 0; i < SURFACE_NUM; i++)
2041 vaDestroyBuffer(va_dpy,coded_buf[i]);
2042
2043 vaDestroyContext(va_dpy,context_id);
2044 vaDestroyConfig(va_dpy,config_id);
2045
2046 return 0;
2047 }
2048
deinit_va()2049 static int deinit_va()
2050 {
2051 vaTerminate(va_dpy);
2052
2053 va_close_display(va_dpy);
2054
2055 return 0;
2056 }
2057
2058
print_input()2059 static int print_input()
2060 {
2061 printf("\n\nINPUT:Try to encode H264...\n");
2062 printf("INPUT: RateControl : %s\n", rc_to_string(rc_mode));
2063 printf("INPUT: Resolution : %dx%d, %d frames\n",
2064 frame_width, frame_height, frame_count);
2065 printf("INPUT: FrameRate : %d\n", frame_rate);
2066 printf("INPUT: Bitrate : %d\n", frame_bitrate);
2067 printf("INPUT: Slieces : %d\n", frame_slices);
2068 printf("INPUT: IntraPeriod : %d\n", intra_period);
2069 printf("INPUT: IDRPeriod : %d\n", intra_idr_period);
2070 printf("INPUT: IpPeriod : %d\n", ip_period);
2071 printf("INPUT: Initial QP : %d\n", initial_qp);
2072 printf("INPUT: Min QP : %d\n", minimal_qp);
2073 printf("INPUT: Source YUV : %s", srcyuv_fp?"FILE":"AUTO generated");
2074 if (srcyuv_fp)
2075 printf(":%s (fourcc %s)\n", srcyuv_fn, fourcc_to_string(srcyuv_fourcc));
2076 else
2077 printf("\n");
2078 printf("INPUT: Coded Clip : %s\n", coded_fn);
2079 if (recyuv_fp == NULL)
2080 printf("INPUT: Rec Clip : %s\n", "Not save reconstructed frame");
2081 else
2082 printf("INPUT: Rec Clip : Save reconstructed frame into %s (fourcc %s)\n", recyuv_fn,
2083 fourcc_to_string(srcyuv_fourcc));
2084
2085 printf("\n\n"); /* return back to startpoint */
2086
2087 return 0;
2088 }
2089
calc_PSNR(double * psnr)2090 static int calc_PSNR(double *psnr)
2091 {
2092 char *srcyuv_ptr = NULL, *recyuv_ptr = NULL, tmp;
2093 unsigned long long min_size;
2094 unsigned long long i, sse=0;
2095 double ssemean;
2096 int fourM = 0x400000; /* 4M */
2097
2098 min_size = MIN(srcyuv_frames, frame_count) * frame_width * frame_height * 1.5;
2099 for (i=0; i<min_size; i++) {
2100 unsigned long long j = i % fourM;
2101
2102 if ((i % fourM) == 0) {
2103 if (srcyuv_ptr)
2104 munmap(srcyuv_ptr, fourM);
2105 if (recyuv_ptr)
2106 munmap(recyuv_ptr, fourM);
2107
2108 srcyuv_ptr = mmap(0, fourM, PROT_READ, MAP_SHARED, fileno(srcyuv_fp), i);
2109 recyuv_ptr = mmap(0, fourM, PROT_READ, MAP_SHARED, fileno(recyuv_fp), i);
2110 if ((srcyuv_ptr == MAP_FAILED) || (recyuv_ptr == MAP_FAILED)) {
2111 printf("Failed to mmap YUV files\n");
2112 return 1;
2113 }
2114 }
2115 tmp = srcyuv_ptr[j] - recyuv_ptr[j];
2116 sse += tmp * tmp;
2117 }
2118 ssemean = (double)sse/(double)min_size;
2119 *psnr = 20.0*log10(255) - 10.0*log10(ssemean);
2120
2121 if (srcyuv_ptr)
2122 munmap(srcyuv_ptr, fourM);
2123 if (recyuv_ptr)
2124 munmap(recyuv_ptr, fourM);
2125
2126 return 0;
2127 }
2128
print_performance(unsigned int PictureCount)2129 static int print_performance(unsigned int PictureCount)
2130 {
2131 unsigned int psnr_ret = 1, others = 0;
2132 double psnr = 0, total_size = frame_width * frame_height * 1.5 * frame_count;
2133
2134 if (calc_psnr && srcyuv_fp && recyuv_fp)
2135 psnr_ret = calc_PSNR(&psnr);
2136
2137 others = TotalTicks - UploadPictureTicks - BeginPictureTicks
2138 - RenderPictureTicks - EndPictureTicks - SyncPictureTicks - SavePictureTicks;
2139
2140 printf("\n\n");
2141
2142 printf("PERFORMANCE: Frame Rate : %.2f fps (%d frames, %d ms (%.2f ms per frame))\n",
2143 (double) 1000*PictureCount / TotalTicks, PictureCount,
2144 TotalTicks, ((double) TotalTicks) / (double) PictureCount);
2145 printf("PERFORMANCE: Compression ratio : %d:1\n", (unsigned int)(total_size / frame_size));
2146 if (psnr_ret == 0)
2147 printf("PERFORMANCE: PSNR : %.2f (%lld frames calculated)\n",
2148 psnr, MIN(frame_count, srcyuv_frames));
2149
2150 printf("PERFORMANCE: UploadPicture : %d ms (%.2f, %.2f%% percent)\n",
2151 (int) UploadPictureTicks, ((double) UploadPictureTicks) / (double) PictureCount,
2152 UploadPictureTicks/(double) TotalTicks/0.01);
2153 printf("PERFORMANCE: vaBeginPicture : %d ms (%.2f, %.2f%% percent)\n",
2154 (int) BeginPictureTicks, ((double) BeginPictureTicks) / (double) PictureCount,
2155 BeginPictureTicks/(double) TotalTicks/0.01);
2156 printf("PERFORMANCE: vaRenderHeader : %d ms (%.2f, %.2f%% percent)\n",
2157 (int) RenderPictureTicks, ((double) RenderPictureTicks) / (double) PictureCount,
2158 RenderPictureTicks/(double) TotalTicks/0.01);
2159 printf("PERFORMANCE: vaEndPicture : %d ms (%.2f, %.2f%% percent)\n",
2160 (int) EndPictureTicks, ((double) EndPictureTicks) / (double) PictureCount,
2161 EndPictureTicks/(double) TotalTicks/0.01);
2162 printf("PERFORMANCE: vaSyncSurface : %d ms (%.2f, %.2f%% percent)\n",
2163 (int) SyncPictureTicks, ((double) SyncPictureTicks) / (double) PictureCount,
2164 SyncPictureTicks/(double) TotalTicks/0.01);
2165 printf("PERFORMANCE: SavePicture : %d ms (%.2f, %.2f%% percent)\n",
2166 (int) SavePictureTicks, ((double) SavePictureTicks) / (double) PictureCount,
2167 SavePictureTicks/(double) TotalTicks/0.01);
2168 printf("PERFORMANCE: Others : %d ms (%.2f, %.2f%% percent)\n",
2169 (int) others, ((double) others) / (double) PictureCount,
2170 others/(double) TotalTicks/0.01);
2171
2172 if (encode_syncmode == 0)
2173 printf("(Multithread enabled, the timing is only for reference)\n");
2174
2175 return 0;
2176 }
2177
2178
main(int argc,char ** argv)2179 int main(int argc,char **argv)
2180 {
2181 unsigned int start;
2182
2183 process_cmdline(argc, argv);
2184
2185 print_input();
2186
2187 start = GetTickCount();
2188
2189 init_va();
2190 setup_encode();
2191
2192 encode_frames();
2193
2194 release_encode();
2195 deinit_va();
2196
2197 TotalTicks += GetTickCount() - start;
2198 print_performance(frame_count);
2199
2200 return 0;
2201 }
2202