Browse code
Merge remote-tracking branch 'qatar/master'
* qatar/master:
swscale: remove misplaced comment.
ffmpeg: fix streaming to ffserver.
swscale: split out RGB48 output functions from yuv2packed[12X]_c().
build: move vpath directives to main Makefile
swscale: fix JPEG-range YUV scaling artifacts.
build: move ALLFFLIBS to a more logical place
ARM: factor some repetitive code into macros
Fix SVQ3 after adding 4:4:4 H.264 support
H.264: fix CODEC_FLAG_GRAY
4:4:4 H.264 decoding support
ac3enc: fix allocation of floating point samples.
Conflicts:
ffmpeg.c
libavcodec/dsputil_template.c
libavcodec/h264.c
libavcodec/mpegvideo.c
libavcodec/snow.c
libswscale/swscale.c
libswscale/swscale_internal.h
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Michael Niedermayer authored on 2011/06/15 09:15:25Showing 27 changed files
- Makefile
- common.mak
- libavcodec/ac3enc.c
- libavcodec/ac3enc.h
- libavcodec/ac3enc_template.c
- libavcodec/arm/h264dsp_init_arm.c
- libavcodec/arm/mpegvideo_armv5te_s.S
- libavcodec/arm/simple_idct_armv5te.S
- libavcodec/dsputil.h
- libavcodec/dsputil_template.c
- libavcodec/h264.c
- libavcodec/h264.h
- libavcodec/h264_cabac.c
- libavcodec/h264_cavlc.c
- libavcodec/h264_loopfilter.c
- libavcodec/h264_ps.c
- libavcodec/h264dsp.h
- libavcodec/h264idct_template.c
- libavcodec/mpegvideo.c
- libavcodec/ppc/h264_altivec.c
- libavcodec/snow.c
- libavcodec/svq3.c
- libavcodec/x86/dsputil_mmx.c
- libavcodec/x86/h264_i386.h
- libavcodec/x86/h264_idct.asm
- libavcodec/x86/h264_idct_10bit.asm
- libswscale/swscale.c
| ... | ... |
@@ -2,6 +2,11 @@ include config.mak |
| 2 | 2 |
|
| 3 | 3 |
SRC_DIR = $(SRC_PATH_BARE) |
| 4 | 4 |
|
| 5 |
+vpath %.c $(SRC_DIR) |
|
| 6 |
+vpath %.h $(SRC_DIR) |
|
| 7 |
+vpath %.S $(SRC_DIR) |
|
| 8 |
+vpath %.asm $(SRC_DIR) |
|
| 9 |
+vpath %.v $(SRC_DIR) |
|
| 5 | 10 |
vpath %.texi $(SRC_PATH_BARE) |
| 6 | 11 |
|
| 7 | 12 |
PROGS-$(CONFIG_FFMPEG) += ffmpeg |
| ... | ... |
@@ -24,6 +29,8 @@ ALLPROGS = $(BASENAMES:%=%$(EXESUF)) |
| 24 | 24 |
ALLPROGS_G = $(BASENAMES:%=%_g$(EXESUF)) |
| 25 | 25 |
ALLMANPAGES = $(BASENAMES:%=%.1) |
| 26 | 26 |
|
| 27 |
+ALLFFLIBS = avcodec avdevice avfilter avformat avutil postproc swscale |
|
| 28 |
+ |
|
| 27 | 29 |
FFLIBS-$(CONFIG_AVDEVICE) += avdevice |
| 28 | 30 |
FFLIBS-$(CONFIG_AVFILTER) += avfilter |
| 29 | 31 |
FFLIBS-$(CONFIG_AVFORMAT) += avformat |
| ... | ... |
@@ -6,11 +6,6 @@ |
| 6 | 6 |
all: all-yes |
| 7 | 7 |
|
| 8 | 8 |
ifndef SUBDIR |
| 9 |
-vpath %.c $(SRC_DIR) |
|
| 10 |
-vpath %.h $(SRC_DIR) |
|
| 11 |
-vpath %.S $(SRC_DIR) |
|
| 12 |
-vpath %.asm $(SRC_DIR) |
|
| 13 |
-vpath %.v $(SRC_DIR) |
|
| 14 | 9 |
|
| 15 | 10 |
ifndef V |
| 16 | 11 |
Q = @ |
| ... | ... |
@@ -25,8 +20,6 @@ $(foreach VAR,$(SILENT),$(eval override $(VAR) = @$($(VAR)))) |
| 25 | 25 |
$(eval INSTALL = @$(call ECHO,INSTALL,$$(^:$(SRC_DIR)/%=%)); $(INSTALL)) |
| 26 | 26 |
endif |
| 27 | 27 |
|
| 28 |
-ALLFFLIBS = avcodec avdevice avfilter avformat avutil postproc swscale |
|
| 29 |
- |
|
| 30 | 28 |
IFLAGS := -I. -I$(SRC_PATH) |
| 31 | 29 |
CPPFLAGS := $(IFLAGS) $(CPPFLAGS) |
| 32 | 30 |
CFLAGS += $(ECFLAGS) |
| ... | ... |
@@ -2215,15 +2215,9 @@ static av_cold int allocate_buffers(AVCodecContext *avctx) |
| 2215 | 2215 |
AC3EncodeContext *s = avctx->priv_data; |
| 2216 | 2216 |
int channels = s->channels + 1; /* includes coupling channel */ |
| 2217 | 2217 |
|
| 2218 |
- FF_ALLOC_OR_GOTO(avctx, s->windowed_samples, AC3_WINDOW_SIZE * |
|
| 2219 |
- sizeof(*s->windowed_samples), alloc_fail); |
|
| 2220 |
- FF_ALLOC_OR_GOTO(avctx, s->planar_samples, s->channels * sizeof(*s->planar_samples), |
|
| 2221 |
- alloc_fail); |
|
| 2222 |
- for (ch = 0; ch < s->channels; ch++) {
|
|
| 2223 |
- FF_ALLOCZ_OR_GOTO(avctx, s->planar_samples[ch], |
|
| 2224 |
- (AC3_FRAME_SIZE+AC3_BLOCK_SIZE) * sizeof(**s->planar_samples), |
|
| 2225 |
- alloc_fail); |
|
| 2226 |
- } |
|
| 2218 |
+ if (s->allocate_sample_buffers(s)) |
|
| 2219 |
+ goto alloc_fail; |
|
| 2220 |
+ |
|
| 2227 | 2221 |
FF_ALLOC_OR_GOTO(avctx, s->bap_buffer, AC3_MAX_BLOCKS * channels * |
| 2228 | 2222 |
AC3_MAX_COEFS * sizeof(*s->bap_buffer), alloc_fail); |
| 2229 | 2223 |
FF_ALLOC_OR_GOTO(avctx, s->bap1_buffer, AC3_MAX_BLOCKS * channels * |
| ... | ... |
@@ -2323,6 +2317,8 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx) |
| 2323 | 2323 |
AC3EncodeContext *s = avctx->priv_data; |
| 2324 | 2324 |
int ret, frame_size_58; |
| 2325 | 2325 |
|
| 2326 |
+ s->avctx = avctx; |
|
| 2327 |
+ |
|
| 2326 | 2328 |
s->eac3 = avctx->codec_id == CODEC_ID_EAC3; |
| 2327 | 2329 |
|
| 2328 | 2330 |
avctx->frame_size = AC3_FRAME_SIZE; |
| ... | ... |
@@ -2355,6 +2351,7 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx) |
| 2355 | 2355 |
s->apply_window = ff_ac3_fixed_apply_window; |
| 2356 | 2356 |
s->normalize_samples = ff_ac3_fixed_normalize_samples; |
| 2357 | 2357 |
s->scale_coefficients = ff_ac3_fixed_scale_coefficients; |
| 2358 |
+ s->allocate_sample_buffers = ff_ac3_fixed_allocate_sample_buffers; |
|
| 2358 | 2359 |
s->deinterleave_input_samples = ff_ac3_fixed_deinterleave_input_samples; |
| 2359 | 2360 |
s->apply_mdct = ff_ac3_fixed_apply_mdct; |
| 2360 | 2361 |
s->apply_channel_coupling = ff_ac3_fixed_apply_channel_coupling; |
| ... | ... |
@@ -2364,6 +2361,7 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx) |
| 2364 | 2364 |
s->mdct_init = ff_ac3_float_mdct_init; |
| 2365 | 2365 |
s->apply_window = ff_ac3_float_apply_window; |
| 2366 | 2366 |
s->scale_coefficients = ff_ac3_float_scale_coefficients; |
| 2367 |
+ s->allocate_sample_buffers = ff_ac3_float_allocate_sample_buffers; |
|
| 2367 | 2368 |
s->deinterleave_input_samples = ff_ac3_float_deinterleave_input_samples; |
| 2368 | 2369 |
s->apply_mdct = ff_ac3_float_apply_mdct; |
| 2369 | 2370 |
s->apply_channel_coupling = ff_ac3_float_apply_channel_coupling; |
| ... | ... |
@@ -135,6 +135,7 @@ typedef struct AC3Block {
|
| 135 | 135 |
typedef struct AC3EncodeContext {
|
| 136 | 136 |
AVClass *av_class; ///< AVClass used for AVOption |
| 137 | 137 |
AC3EncOptions options; ///< encoding options |
| 138 |
+ AVCodecContext *avctx; ///< parent AVCodecContext |
|
| 138 | 139 |
PutBitContext pb; ///< bitstream writer context |
| 139 | 140 |
DSPContext dsp; |
| 140 | 141 |
AC3DSPContext ac3dsp; ///< AC-3 optimized functions |
| ... | ... |
@@ -230,6 +231,7 @@ typedef struct AC3EncodeContext {
|
| 230 | 230 |
void (*scale_coefficients)(struct AC3EncodeContext *s); |
| 231 | 231 |
|
| 232 | 232 |
/* fixed vs. float templated function pointers */ |
| 233 |
+ int (*allocate_sample_buffers)(struct AC3EncodeContext *s); |
|
| 233 | 234 |
void (*deinterleave_input_samples)(struct AC3EncodeContext *s, |
| 234 | 235 |
const SampleType *samples); |
| 235 | 236 |
void (*apply_mdct)(struct AC3EncodeContext *s); |
| ... | ... |
@@ -276,6 +278,9 @@ void ff_ac3_float_scale_coefficients(AC3EncodeContext *s); |
| 276 | 276 |
|
| 277 | 277 |
/* prototypes for functions in ac3enc_template.c */ |
| 278 | 278 |
|
| 279 |
+int ff_ac3_fixed_allocate_sample_buffers(AC3EncodeContext *s); |
|
| 280 |
+int ff_ac3_float_allocate_sample_buffers(AC3EncodeContext *s); |
|
| 281 |
+ |
|
| 279 | 282 |
void ff_ac3_fixed_deinterleave_input_samples(AC3EncodeContext *s, |
| 280 | 283 |
const SampleType *samples); |
| 281 | 284 |
void ff_ac3_float_deinterleave_input_samples(AC3EncodeContext *s, |
| ... | ... |
@@ -31,6 +31,26 @@ |
| 31 | 31 |
#include "ac3enc.h" |
| 32 | 32 |
|
| 33 | 33 |
|
| 34 |
+int AC3_NAME(allocate_sample_buffers)(AC3EncodeContext *s) |
|
| 35 |
+{
|
|
| 36 |
+ int ch; |
|
| 37 |
+ |
|
| 38 |
+ FF_ALLOC_OR_GOTO(s->avctx, s->windowed_samples, AC3_WINDOW_SIZE * |
|
| 39 |
+ sizeof(*s->windowed_samples), alloc_fail); |
|
| 40 |
+ FF_ALLOC_OR_GOTO(s->avctx, s->planar_samples, s->channels * sizeof(*s->planar_samples), |
|
| 41 |
+ alloc_fail); |
|
| 42 |
+ for (ch = 0; ch < s->channels; ch++) {
|
|
| 43 |
+ FF_ALLOCZ_OR_GOTO(s->avctx, s->planar_samples[ch], |
|
| 44 |
+ (AC3_FRAME_SIZE+AC3_BLOCK_SIZE) * sizeof(**s->planar_samples), |
|
| 45 |
+ alloc_fail); |
|
| 46 |
+ } |
|
| 47 |
+ |
|
| 48 |
+ return 0; |
|
| 49 |
+alloc_fail: |
|
| 50 |
+ return AVERROR(ENOMEM); |
|
| 51 |
+} |
|
| 52 |
+ |
|
| 53 |
+ |
|
| 34 | 54 |
/** |
| 35 | 55 |
* Deinterleave input samples. |
| 36 | 56 |
* Channels are reordered from Libav's default order to AC-3 order. |
| ... | ... |
@@ -122,7 +122,8 @@ static void ff_h264dsp_init_neon(H264DSPContext *c, const int bit_depth) |
| 122 | 122 |
c->h264_idct_dc_add = ff_h264_idct_dc_add_neon; |
| 123 | 123 |
c->h264_idct_add16 = ff_h264_idct_add16_neon; |
| 124 | 124 |
c->h264_idct_add16intra = ff_h264_idct_add16intra_neon; |
| 125 |
- c->h264_idct_add8 = ff_h264_idct_add8_neon; |
|
| 125 |
+ //FIXME: reenable when asm is updated. |
|
| 126 |
+ //c->h264_idct_add8 = ff_h264_idct_add8_neon; |
|
| 126 | 127 |
c->h264_idct8_add = ff_h264_idct8_add_neon; |
| 127 | 128 |
c->h264_idct8_dc_add = ff_h264_idct8_dc_add_neon; |
| 128 | 129 |
c->h264_idct8_add4 = ff_h264_idct8_add4_neon; |
| ... | ... |
@@ -35,6 +35,21 @@ |
| 35 | 35 |
* |
| 36 | 36 |
* Inner loop should take 6 cycles per element on arm926ej-s (Nokia 770) |
| 37 | 37 |
*/ |
| 38 |
+ |
|
| 39 |
+.macro dequant_t dst, src, mul, add, tmp |
|
| 40 |
+ rsbs \tmp, ip, \src, asr #16 |
|
| 41 |
+ addgt \tmp, \add, #0 |
|
| 42 |
+ rsblt \tmp, \add, #0 |
|
| 43 |
+ smlatbne \dst, \src, \mul, \tmp |
|
| 44 |
+.endm |
|
| 45 |
+ |
|
| 46 |
+.macro dequant_b dst, src, mul, add, tmp |
|
| 47 |
+ rsbs \tmp, ip, \src, lsl #16 |
|
| 48 |
+ addgt \tmp, \add, #0 |
|
| 49 |
+ rsblt \tmp, \add, #0 |
|
| 50 |
+ smlabbne \dst, \src, \mul, \tmp |
|
| 51 |
+.endm |
|
| 52 |
+ |
|
| 38 | 53 |
function ff_dct_unquantize_h263_armv5te, export=1 |
| 39 | 54 |
push {r4-r9,lr}
|
| 40 | 55 |
mov ip, #0 |
| ... | ... |
@@ -44,50 +59,20 @@ function ff_dct_unquantize_h263_armv5te, export=1 |
| 44 | 44 |
1: |
| 45 | 45 |
ldrd r6, [r0, #8] |
| 46 | 46 |
|
| 47 |
- rsbs r9, ip, r4, asr #16 |
|
| 48 |
- addgt r9, r2, #0 |
|
| 49 |
- rsblt r9, r2, #0 |
|
| 50 |
- smlatbne r9, r4, r1, r9 |
|
| 51 |
- |
|
| 52 |
- rsbs lr, ip, r5, asr #16 |
|
| 53 |
- addgt lr, r2, #0 |
|
| 54 |
- rsblt lr, r2, #0 |
|
| 55 |
- smlatbne lr, r5, r1, lr |
|
| 56 |
- |
|
| 57 |
- rsbs r8, ip, r4, asl #16 |
|
| 58 |
- addgt r8, r2, #0 |
|
| 59 |
- rsblt r8, r2, #0 |
|
| 60 |
- smlabbne r4, r4, r1, r8 |
|
| 61 |
- |
|
| 62 |
- rsbs r8, ip, r5, asl #16 |
|
| 63 |
- addgt r8, r2, #0 |
|
| 64 |
- rsblt r8, r2, #0 |
|
| 65 |
- smlabbne r5, r5, r1, r8 |
|
| 47 |
+ dequant_t r9, r4, r1, r2, r9 |
|
| 48 |
+ dequant_t lr, r5, r1, r2, lr |
|
| 49 |
+ dequant_b r4, r4, r1, r2, r8 |
|
| 50 |
+ dequant_b r5, r5, r1, r2, r8 |
|
| 66 | 51 |
|
| 67 | 52 |
strh r4, [r0], #2 |
| 68 | 53 |
strh r9, [r0], #2 |
| 69 | 54 |
strh r5, [r0], #2 |
| 70 | 55 |
strh lr, [r0], #2 |
| 71 | 56 |
|
| 72 |
- rsbs r9, ip, r6, asr #16 |
|
| 73 |
- addgt r9, r2, #0 |
|
| 74 |
- rsblt r9, r2, #0 |
|
| 75 |
- smlatbne r9, r6, r1, r9 |
|
| 76 |
- |
|
| 77 |
- rsbs lr, ip, r7, asr #16 |
|
| 78 |
- addgt lr, r2, #0 |
|
| 79 |
- rsblt lr, r2, #0 |
|
| 80 |
- smlatbne lr, r7, r1, lr |
|
| 81 |
- |
|
| 82 |
- rsbs r8, ip, r6, asl #16 |
|
| 83 |
- addgt r8, r2, #0 |
|
| 84 |
- rsblt r8, r2, #0 |
|
| 85 |
- smlabbne r6, r6, r1, r8 |
|
| 86 |
- |
|
| 87 |
- rsbs r8, ip, r7, asl #16 |
|
| 88 |
- addgt r8, r2, #0 |
|
| 89 |
- rsblt r8, r2, #0 |
|
| 90 |
- smlabbne r7, r7, r1, r8 |
|
| 57 |
+ dequant_t r9, r6, r1, r2, r9 |
|
| 58 |
+ dequant_t lr, r7, r1, r2, lr |
|
| 59 |
+ dequant_b r6, r6, r1, r2, r8 |
|
| 60 |
+ dequant_b r7, r7, r1, r2, r8 |
|
| 91 | 61 |
|
| 92 | 62 |
strh r6, [r0], #2 |
| 93 | 63 |
strh r9, [r0], #2 |
| ... | ... |
@@ -333,6 +333,20 @@ function idct_col_armv5te |
| 333 | 333 |
ldr pc, [sp], #4 |
| 334 | 334 |
endfunc |
| 335 | 335 |
|
| 336 |
+.macro clip dst, src:vararg |
|
| 337 |
+ movs \dst, \src |
|
| 338 |
+ movmi \dst, #0 |
|
| 339 |
+ cmp \dst, #255 |
|
| 340 |
+ movgt \dst, #255 |
|
| 341 |
+.endm |
|
| 342 |
+ |
|
| 343 |
+.macro aclip dst, src:vararg |
|
| 344 |
+ adds \dst, \src |
|
| 345 |
+ movmi \dst, #0 |
|
| 346 |
+ cmp \dst, #255 |
|
| 347 |
+ movgt \dst, #255 |
|
| 348 |
+.endm |
|
| 349 |
+ |
|
| 336 | 350 |
function idct_col_put_armv5te |
| 337 | 351 |
str lr, [sp, #-4]! |
| 338 | 352 |
|
| ... | ... |
@@ -341,27 +355,15 @@ function idct_col_put_armv5te |
| 341 | 341 |
ldmfd sp!, {a3, a4}
|
| 342 | 342 |
ldr lr, [sp, #32] |
| 343 | 343 |
add a2, a3, v1 |
| 344 |
- movs a2, a2, asr #20 |
|
| 345 |
- movmi a2, #0 |
|
| 346 |
- cmp a2, #255 |
|
| 347 |
- movgt a2, #255 |
|
| 344 |
+ clip a2, a2, asr #20 |
|
| 348 | 345 |
add ip, a4, v2 |
| 349 |
- movs ip, ip, asr #20 |
|
| 350 |
- movmi ip, #0 |
|
| 351 |
- cmp ip, #255 |
|
| 352 |
- movgt ip, #255 |
|
| 346 |
+ clip ip, ip, asr #20 |
|
| 353 | 347 |
orr a2, a2, ip, lsl #8 |
| 354 | 348 |
sub a3, a3, v1 |
| 355 |
- movs a3, a3, asr #20 |
|
| 356 |
- movmi a3, #0 |
|
| 357 |
- cmp a3, #255 |
|
| 358 |
- movgt a3, #255 |
|
| 349 |
+ clip a3, a3, asr #20 |
|
| 359 | 350 |
sub a4, a4, v2 |
| 360 |
- movs a4, a4, asr #20 |
|
| 361 |
- movmi a4, #0 |
|
| 362 |
- cmp a4, #255 |
|
| 351 |
+ clip a4, a4, asr #20 |
|
| 363 | 352 |
ldr v1, [sp, #28] |
| 364 |
- movgt a4, #255 |
|
| 365 | 353 |
strh a2, [v1] |
| 366 | 354 |
add a2, v1, #2 |
| 367 | 355 |
str a2, [sp, #28] |
| ... | ... |
@@ -371,79 +373,43 @@ function idct_col_put_armv5te |
| 371 | 371 |
strh a2, [v2, v1]! |
| 372 | 372 |
|
| 373 | 373 |
sub a2, a3, v3 |
| 374 |
- movs a2, a2, asr #20 |
|
| 375 |
- movmi a2, #0 |
|
| 376 |
- cmp a2, #255 |
|
| 377 |
- movgt a2, #255 |
|
| 374 |
+ clip a2, a2, asr #20 |
|
| 378 | 375 |
sub ip, a4, v4 |
| 379 |
- movs ip, ip, asr #20 |
|
| 380 |
- movmi ip, #0 |
|
| 381 |
- cmp ip, #255 |
|
| 382 |
- movgt ip, #255 |
|
| 376 |
+ clip ip, ip, asr #20 |
|
| 383 | 377 |
orr a2, a2, ip, lsl #8 |
| 384 | 378 |
strh a2, [v1, lr]! |
| 385 | 379 |
add a3, a3, v3 |
| 386 |
- movs a2, a3, asr #20 |
|
| 387 |
- movmi a2, #0 |
|
| 388 |
- cmp a2, #255 |
|
| 389 |
- movgt a2, #255 |
|
| 380 |
+ clip a2, a3, asr #20 |
|
| 390 | 381 |
add a4, a4, v4 |
| 391 |
- movs a4, a4, asr #20 |
|
| 392 |
- movmi a4, #0 |
|
| 393 |
- cmp a4, #255 |
|
| 394 |
- movgt a4, #255 |
|
| 382 |
+ clip a4, a4, asr #20 |
|
| 395 | 383 |
orr a2, a2, a4, lsl #8 |
| 396 | 384 |
ldmfd sp!, {a3, a4}
|
| 397 | 385 |
strh a2, [v2, -lr]! |
| 398 | 386 |
|
| 399 | 387 |
add a2, a3, v5 |
| 400 |
- movs a2, a2, asr #20 |
|
| 401 |
- movmi a2, #0 |
|
| 402 |
- cmp a2, #255 |
|
| 403 |
- movgt a2, #255 |
|
| 388 |
+ clip a2, a2, asr #20 |
|
| 404 | 389 |
add ip, a4, v6 |
| 405 |
- movs ip, ip, asr #20 |
|
| 406 |
- movmi ip, #0 |
|
| 407 |
- cmp ip, #255 |
|
| 408 |
- movgt ip, #255 |
|
| 390 |
+ clip ip, ip, asr #20 |
|
| 409 | 391 |
orr a2, a2, ip, lsl #8 |
| 410 | 392 |
strh a2, [v1, lr]! |
| 411 | 393 |
sub a3, a3, v5 |
| 412 |
- movs a2, a3, asr #20 |
|
| 413 |
- movmi a2, #0 |
|
| 414 |
- cmp a2, #255 |
|
| 415 |
- movgt a2, #255 |
|
| 394 |
+ clip a2, a3, asr #20 |
|
| 416 | 395 |
sub a4, a4, v6 |
| 417 |
- movs a4, a4, asr #20 |
|
| 418 |
- movmi a4, #0 |
|
| 419 |
- cmp a4, #255 |
|
| 420 |
- movgt a4, #255 |
|
| 396 |
+ clip a4, a4, asr #20 |
|
| 421 | 397 |
orr a2, a2, a4, lsl #8 |
| 422 | 398 |
ldmfd sp!, {a3, a4}
|
| 423 | 399 |
strh a2, [v2, -lr]! |
| 424 | 400 |
|
| 425 | 401 |
add a2, a3, v7 |
| 426 |
- movs a2, a2, asr #20 |
|
| 427 |
- movmi a2, #0 |
|
| 428 |
- cmp a2, #255 |
|
| 429 |
- movgt a2, #255 |
|
| 402 |
+ clip a2, a2, asr #20 |
|
| 430 | 403 |
add ip, a4, fp |
| 431 |
- movs ip, ip, asr #20 |
|
| 432 |
- movmi ip, #0 |
|
| 433 |
- cmp ip, #255 |
|
| 434 |
- movgt ip, #255 |
|
| 404 |
+ clip ip, ip, asr #20 |
|
| 435 | 405 |
orr a2, a2, ip, lsl #8 |
| 436 | 406 |
strh a2, [v1, lr] |
| 437 | 407 |
sub a3, a3, v7 |
| 438 |
- movs a2, a3, asr #20 |
|
| 439 |
- movmi a2, #0 |
|
| 440 |
- cmp a2, #255 |
|
| 441 |
- movgt a2, #255 |
|
| 408 |
+ clip a2, a3, asr #20 |
|
| 442 | 409 |
sub a4, a4, fp |
| 443 |
- movs a4, a4, asr #20 |
|
| 444 |
- movmi a4, #0 |
|
| 445 |
- cmp a4, #255 |
|
| 446 |
- movgt a4, #255 |
|
| 410 |
+ clip a4, a4, asr #20 |
|
| 447 | 411 |
orr a2, a2, a4, lsl #8 |
| 448 | 412 |
strh a2, [v2, -lr] |
| 449 | 413 |
|
| ... | ... |
@@ -460,36 +426,22 @@ function idct_col_add_armv5te |
| 460 | 460 |
ldmfd sp!, {a3, a4}
|
| 461 | 461 |
ldrh ip, [lr] |
| 462 | 462 |
add a2, a3, v1 |
| 463 |
- mov a2, a2, asr #20 |
|
| 464 | 463 |
sub a3, a3, v1 |
| 465 | 464 |
and v1, ip, #255 |
| 466 |
- adds a2, a2, v1 |
|
| 467 |
- movmi a2, #0 |
|
| 468 |
- cmp a2, #255 |
|
| 469 |
- movgt a2, #255 |
|
| 465 |
+ aclip a2, v1, a2, asr #20 |
|
| 470 | 466 |
add v1, a4, v2 |
| 471 | 467 |
mov v1, v1, asr #20 |
| 472 |
- adds v1, v1, ip, lsr #8 |
|
| 473 |
- movmi v1, #0 |
|
| 474 |
- cmp v1, #255 |
|
| 475 |
- movgt v1, #255 |
|
| 468 |
+ aclip v1, v1, ip, lsr #8 |
|
| 476 | 469 |
orr a2, a2, v1, lsl #8 |
| 477 | 470 |
ldr v1, [sp, #32] |
| 478 | 471 |
sub a4, a4, v2 |
| 479 | 472 |
rsb v2, v1, v1, lsl #3 |
| 480 | 473 |
ldrh ip, [v2, lr]! |
| 481 | 474 |
strh a2, [lr] |
| 482 |
- mov a3, a3, asr #20 |
|
| 483 | 475 |
and a2, ip, #255 |
| 484 |
- adds a3, a3, a2 |
|
| 485 |
- movmi a3, #0 |
|
| 486 |
- cmp a3, #255 |
|
| 487 |
- movgt a3, #255 |
|
| 476 |
+ aclip a3, a2, a3, asr #20 |
|
| 488 | 477 |
mov a4, a4, asr #20 |
| 489 |
- adds a4, a4, ip, lsr #8 |
|
| 490 |
- movmi a4, #0 |
|
| 491 |
- cmp a4, #255 |
|
| 492 |
- movgt a4, #255 |
|
| 478 |
+ aclip a4, a4, ip, lsr #8 |
|
| 493 | 479 |
add a2, lr, #2 |
| 494 | 480 |
str a2, [sp, #28] |
| 495 | 481 |
orr a2, a3, a4, lsl #8 |
| ... | ... |
@@ -498,102 +450,60 @@ function idct_col_add_armv5te |
| 498 | 498 |
ldmfd sp!, {a3, a4}
|
| 499 | 499 |
ldrh ip, [lr, v1]! |
| 500 | 500 |
sub a2, a3, v3 |
| 501 |
- mov a2, a2, asr #20 |
|
| 502 | 501 |
add a3, a3, v3 |
| 503 | 502 |
and v3, ip, #255 |
| 504 |
- adds a2, a2, v3 |
|
| 505 |
- movmi a2, #0 |
|
| 506 |
- cmp a2, #255 |
|
| 507 |
- movgt a2, #255 |
|
| 503 |
+ aclip a2, v3, a2, asr #20 |
|
| 508 | 504 |
sub v3, a4, v4 |
| 509 | 505 |
mov v3, v3, asr #20 |
| 510 |
- adds v3, v3, ip, lsr #8 |
|
| 511 |
- movmi v3, #0 |
|
| 512 |
- cmp v3, #255 |
|
| 513 |
- movgt v3, #255 |
|
| 506 |
+ aclip v3, v3, ip, lsr #8 |
|
| 514 | 507 |
orr a2, a2, v3, lsl #8 |
| 515 | 508 |
add a4, a4, v4 |
| 516 | 509 |
ldrh ip, [v2, -v1]! |
| 517 | 510 |
strh a2, [lr] |
| 518 |
- mov a3, a3, asr #20 |
|
| 519 | 511 |
and a2, ip, #255 |
| 520 |
- adds a3, a3, a2 |
|
| 521 |
- movmi a3, #0 |
|
| 522 |
- cmp a3, #255 |
|
| 523 |
- movgt a3, #255 |
|
| 512 |
+ aclip a3, a2, a3, asr #20 |
|
| 524 | 513 |
mov a4, a4, asr #20 |
| 525 |
- adds a4, a4, ip, lsr #8 |
|
| 526 |
- movmi a4, #0 |
|
| 527 |
- cmp a4, #255 |
|
| 528 |
- movgt a4, #255 |
|
| 514 |
+ aclip a4, a4, ip, lsr #8 |
|
| 529 | 515 |
orr a2, a3, a4, lsl #8 |
| 530 | 516 |
strh a2, [v2] |
| 531 | 517 |
|
| 532 | 518 |
ldmfd sp!, {a3, a4}
|
| 533 | 519 |
ldrh ip, [lr, v1]! |
| 534 | 520 |
add a2, a3, v5 |
| 535 |
- mov a2, a2, asr #20 |
|
| 536 | 521 |
sub a3, a3, v5 |
| 537 | 522 |
and v3, ip, #255 |
| 538 |
- adds a2, a2, v3 |
|
| 539 |
- movmi a2, #0 |
|
| 540 |
- cmp a2, #255 |
|
| 541 |
- movgt a2, #255 |
|
| 523 |
+ aclip a2, v3, a2, asr #20 |
|
| 542 | 524 |
add v3, a4, v6 |
| 543 | 525 |
mov v3, v3, asr #20 |
| 544 |
- adds v3, v3, ip, lsr #8 |
|
| 545 |
- movmi v3, #0 |
|
| 546 |
- cmp v3, #255 |
|
| 547 |
- movgt v3, #255 |
|
| 526 |
+ aclip v3, v3, ip, lsr #8 |
|
| 548 | 527 |
orr a2, a2, v3, lsl #8 |
| 549 | 528 |
sub a4, a4, v6 |
| 550 | 529 |
ldrh ip, [v2, -v1]! |
| 551 | 530 |
strh a2, [lr] |
| 552 |
- mov a3, a3, asr #20 |
|
| 553 | 531 |
and a2, ip, #255 |
| 554 |
- adds a3, a3, a2 |
|
| 555 |
- movmi a3, #0 |
|
| 556 |
- cmp a3, #255 |
|
| 557 |
- movgt a3, #255 |
|
| 532 |
+ aclip a3, a2, a3, asr #20 |
|
| 558 | 533 |
mov a4, a4, asr #20 |
| 559 |
- adds a4, a4, ip, lsr #8 |
|
| 560 |
- movmi a4, #0 |
|
| 561 |
- cmp a4, #255 |
|
| 562 |
- movgt a4, #255 |
|
| 534 |
+ aclip a4, a4, ip, lsr #8 |
|
| 563 | 535 |
orr a2, a3, a4, lsl #8 |
| 564 | 536 |
strh a2, [v2] |
| 565 | 537 |
|
| 566 | 538 |
ldmfd sp!, {a3, a4}
|
| 567 | 539 |
ldrh ip, [lr, v1]! |
| 568 | 540 |
add a2, a3, v7 |
| 569 |
- mov a2, a2, asr #20 |
|
| 570 | 541 |
sub a3, a3, v7 |
| 571 | 542 |
and v3, ip, #255 |
| 572 |
- adds a2, a2, v3 |
|
| 573 |
- movmi a2, #0 |
|
| 574 |
- cmp a2, #255 |
|
| 575 |
- movgt a2, #255 |
|
| 543 |
+ aclip a2, v3, a2, asr #20 |
|
| 576 | 544 |
add v3, a4, fp |
| 577 | 545 |
mov v3, v3, asr #20 |
| 578 |
- adds v3, v3, ip, lsr #8 |
|
| 579 |
- movmi v3, #0 |
|
| 580 |
- cmp v3, #255 |
|
| 581 |
- movgt v3, #255 |
|
| 546 |
+ aclip v3, v3, ip, lsr #8 |
|
| 582 | 547 |
orr a2, a2, v3, lsl #8 |
| 583 | 548 |
sub a4, a4, fp |
| 584 | 549 |
ldrh ip, [v2, -v1]! |
| 585 | 550 |
strh a2, [lr] |
| 586 |
- mov a3, a3, asr #20 |
|
| 587 | 551 |
and a2, ip, #255 |
| 588 |
- adds a3, a3, a2 |
|
| 589 |
- movmi a3, #0 |
|
| 590 |
- cmp a3, #255 |
|
| 591 |
- movgt a3, #255 |
|
| 552 |
+ aclip a3, a2, a3, asr #20 |
|
| 592 | 553 |
mov a4, a4, asr #20 |
| 593 |
- adds a4, a4, ip, lsr #8 |
|
| 594 |
- movmi a4, #0 |
|
| 595 |
- cmp a4, #255 |
|
| 596 |
- movgt a4, #255 |
|
| 554 |
+ aclip a4, a4, ip, lsr #8 |
|
| 597 | 555 |
orr a2, a3, a4, lsl #8 |
| 598 | 556 |
strh a2, [v2] |
| 599 | 557 |
|
| ... | ... |
@@ -505,7 +505,7 @@ typedef struct DSPContext {
|
| 505 | 505 |
#define BASIS_SHIFT 16 |
| 506 | 506 |
#define RECON_SHIFT 6 |
| 507 | 507 |
|
| 508 |
- void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int sides); |
|
| 508 |
+ void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides); |
|
| 509 | 509 |
#define EDGE_WIDTH 16 |
| 510 | 510 |
#define EDGE_TOP 1 |
| 511 | 511 |
#define EDGE_BOTTOM 2 |
| ... | ... |
@@ -79,7 +79,7 @@ static inline void FUNC(copy_block16)(uint8_t *dst, const uint8_t *src, int dstS |
| 79 | 79 |
|
| 80 | 80 |
/* draw the edges of width 'w' of an image of size width, height */ |
| 81 | 81 |
//FIXME check that this is ok for mpeg4 interlaced |
| 82 |
-static void FUNCC(draw_edges)(uint8_t *p_buf, int p_wrap, int width, int height, int w, int sides) |
|
| 82 |
+static void FUNCC(draw_edges)(uint8_t *p_buf, int p_wrap, int width, int height, int w, int h, int sides) |
|
| 83 | 83 |
{
|
| 84 | 84 |
pixel *buf = (pixel*)p_buf; |
| 85 | 85 |
int wrap = p_wrap / sizeof(pixel); |
| ... | ... |
@@ -106,10 +106,10 @@ static void FUNCC(draw_edges)(uint8_t *p_buf, int p_wrap, int width, int height, |
| 106 | 106 |
buf -= w; |
| 107 | 107 |
last_line = buf + (height - 1) * wrap; |
| 108 | 108 |
if (sides & EDGE_TOP) |
| 109 |
- for(i = 0; i < w; i++) |
|
| 109 |
+ for(i = 0; i < h; i++) |
|
| 110 | 110 |
memcpy(buf - (i + 1) * wrap, buf, (width + w + w) * sizeof(pixel)); // top |
| 111 | 111 |
if (sides & EDGE_BOTTOM) |
| 112 |
- for (i = 0; i < w; i++) |
|
| 112 |
+ for (i = 0; i < h; i++) |
|
| 113 | 113 |
memcpy(last_line + (i + 1) * wrap, last_line, (width + w + w) * sizeof(pixel)); // bottom |
| 114 | 114 |
} |
| 115 | 115 |
|
| ... | ... |
@@ -451,12 +451,13 @@ static inline void mc_dir_part(H264Context *h, Picture *pic, int n, int square, |
| 451 | 451 |
uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, |
| 452 | 452 |
int src_x_offset, int src_y_offset, |
| 453 | 453 |
qpel_mc_func *qpix_op, h264_chroma_mc_func chroma_op, |
| 454 |
- int pixel_shift){
|
|
| 454 |
+ int pixel_shift, int chroma444){
|
|
| 455 | 455 |
MpegEncContext * const s = &h->s; |
| 456 | 456 |
const int mx= h->mv_cache[list][ scan8[n] ][0] + src_x_offset*8; |
| 457 | 457 |
int my= h->mv_cache[list][ scan8[n] ][1] + src_y_offset*8; |
| 458 | 458 |
const int luma_xy= (mx&3) + ((my&3)<<2); |
| 459 |
- uint8_t * src_y = pic->data[0] + ((mx>>2) << pixel_shift) + (my>>2)*h->mb_linesize; |
|
| 459 |
+ int offset = ((mx>>2) << pixel_shift) + (my>>2)*h->mb_linesize; |
|
| 460 |
+ uint8_t * src_y = pic->data[0] + offset; |
|
| 460 | 461 |
uint8_t * src_cb, * src_cr; |
| 461 | 462 |
int extra_width= h->emu_edge_width; |
| 462 | 463 |
int extra_height= h->emu_edge_height; |
| ... | ... |
@@ -485,6 +486,31 @@ static inline void mc_dir_part(H264Context *h, Picture *pic, int n, int square, |
| 485 | 485 |
|
| 486 | 486 |
if(CONFIG_GRAY && s->flags&CODEC_FLAG_GRAY) return; |
| 487 | 487 |
|
| 488 |
+ if(chroma444){
|
|
| 489 |
+ src_cb = pic->data[1] + offset; |
|
| 490 |
+ if(emu){
|
|
| 491 |
+ s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cb - (2 << pixel_shift) - 2*h->mb_linesize, h->mb_linesize, |
|
| 492 |
+ 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height); |
|
| 493 |
+ src_cb= s->edge_emu_buffer + (2 << pixel_shift) + 2*h->mb_linesize; |
|
| 494 |
+ } |
|
| 495 |
+ qpix_op[luma_xy](dest_cb, src_cb, h->mb_linesize); //FIXME try variable height perhaps? |
|
| 496 |
+ if(!square){
|
|
| 497 |
+ qpix_op[luma_xy](dest_cb + delta, src_cb + delta, h->mb_linesize); |
|
| 498 |
+ } |
|
| 499 |
+ |
|
| 500 |
+ src_cr = pic->data[2] + offset; |
|
| 501 |
+ if(emu){
|
|
| 502 |
+ s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cr - (2 << pixel_shift) - 2*h->mb_linesize, h->mb_linesize, |
|
| 503 |
+ 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height); |
|
| 504 |
+ src_cr= s->edge_emu_buffer + (2 << pixel_shift) + 2*h->mb_linesize; |
|
| 505 |
+ } |
|
| 506 |
+ qpix_op[luma_xy](dest_cr, src_cr, h->mb_linesize); //FIXME try variable height perhaps? |
|
| 507 |
+ if(!square){
|
|
| 508 |
+ qpix_op[luma_xy](dest_cr + delta, src_cr + delta, h->mb_linesize); |
|
| 509 |
+ } |
|
| 510 |
+ return; |
|
| 511 |
+ } |
|
| 512 |
+ |
|
| 488 | 513 |
if(MB_FIELD){
|
| 489 | 514 |
// chroma offset when predicting from a field of opposite parity |
| 490 | 515 |
my += 2 * ((s->mb_y & 1) - (pic->reference - 1)); |
| ... | ... |
@@ -511,14 +537,19 @@ static inline void mc_part_std(H264Context *h, int n, int square, int chroma_hei |
| 511 | 511 |
int x_offset, int y_offset, |
| 512 | 512 |
qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put, |
| 513 | 513 |
qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg, |
| 514 |
- int list0, int list1, int pixel_shift){
|
|
| 514 |
+ int list0, int list1, int pixel_shift, int chroma444){
|
|
| 515 | 515 |
MpegEncContext * const s = &h->s; |
| 516 | 516 |
qpel_mc_func *qpix_op= qpix_put; |
| 517 | 517 |
h264_chroma_mc_func chroma_op= chroma_put; |
| 518 | 518 |
|
| 519 |
- dest_y += (2*x_offset << pixel_shift) + 2*y_offset*h-> mb_linesize; |
|
| 520 |
- dest_cb += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; |
|
| 521 |
- dest_cr += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; |
|
| 519 |
+ dest_y += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; |
|
| 520 |
+ if(chroma444){
|
|
| 521 |
+ dest_cb += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; |
|
| 522 |
+ dest_cr += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; |
|
| 523 |
+ }else{
|
|
| 524 |
+ dest_cb += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; |
|
| 525 |
+ dest_cr += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; |
|
| 526 |
+ } |
|
| 522 | 527 |
x_offset += 8*s->mb_x; |
| 523 | 528 |
y_offset += 8*(s->mb_y >> MB_FIELD); |
| 524 | 529 |
|
| ... | ... |
@@ -526,7 +557,7 @@ static inline void mc_part_std(H264Context *h, int n, int square, int chroma_hei |
| 526 | 526 |
Picture *ref= &h->ref_list[0][ h->ref_cache[0][ scan8[n] ] ]; |
| 527 | 527 |
mc_dir_part(h, ref, n, square, chroma_height, delta, 0, |
| 528 | 528 |
dest_y, dest_cb, dest_cr, x_offset, y_offset, |
| 529 |
- qpix_op, chroma_op, pixel_shift); |
|
| 529 |
+ qpix_op, chroma_op, pixel_shift, chroma444); |
|
| 530 | 530 |
|
| 531 | 531 |
qpix_op= qpix_avg; |
| 532 | 532 |
chroma_op= chroma_avg; |
| ... | ... |
@@ -536,7 +567,7 @@ static inline void mc_part_std(H264Context *h, int n, int square, int chroma_hei |
| 536 | 536 |
Picture *ref= &h->ref_list[1][ h->ref_cache[1][ scan8[n] ] ]; |
| 537 | 537 |
mc_dir_part(h, ref, n, square, chroma_height, delta, 1, |
| 538 | 538 |
dest_y, dest_cb, dest_cr, x_offset, y_offset, |
| 539 |
- qpix_op, chroma_op, pixel_shift); |
|
| 539 |
+ qpix_op, chroma_op, pixel_shift, chroma444); |
|
| 540 | 540 |
} |
| 541 | 541 |
} |
| 542 | 542 |
|
| ... | ... |
@@ -546,12 +577,19 @@ static inline void mc_part_weighted(H264Context *h, int n, int square, int chrom |
| 546 | 546 |
qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put, |
| 547 | 547 |
h264_weight_func luma_weight_op, h264_weight_func chroma_weight_op, |
| 548 | 548 |
h264_biweight_func luma_weight_avg, h264_biweight_func chroma_weight_avg, |
| 549 |
- int list0, int list1, int pixel_shift){
|
|
| 549 |
+ int list0, int list1, int pixel_shift, int chroma444){
|
|
| 550 | 550 |
MpegEncContext * const s = &h->s; |
| 551 | 551 |
|
| 552 |
- dest_y += (2*x_offset << pixel_shift) + 2*y_offset*h-> mb_linesize; |
|
| 553 |
- dest_cb += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; |
|
| 554 |
- dest_cr += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; |
|
| 552 |
+ dest_y += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; |
|
| 553 |
+ if(chroma444){
|
|
| 554 |
+ chroma_weight_avg = luma_weight_avg; |
|
| 555 |
+ chroma_weight_op = luma_weight_op; |
|
| 556 |
+ dest_cb += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; |
|
| 557 |
+ dest_cr += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; |
|
| 558 |
+ }else{
|
|
| 559 |
+ dest_cb += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; |
|
| 560 |
+ dest_cr += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; |
|
| 561 |
+ } |
|
| 555 | 562 |
x_offset += 8*s->mb_x; |
| 556 | 563 |
y_offset += 8*(s->mb_y >> MB_FIELD); |
| 557 | 564 |
|
| ... | ... |
@@ -559,17 +597,17 @@ static inline void mc_part_weighted(H264Context *h, int n, int square, int chrom |
| 559 | 559 |
/* don't optimize for luma-only case, since B-frames usually |
| 560 | 560 |
* use implicit weights => chroma too. */ |
| 561 | 561 |
uint8_t *tmp_cb = s->obmc_scratchpad; |
| 562 |
- uint8_t *tmp_cr = s->obmc_scratchpad + (8 << pixel_shift); |
|
| 563 |
- uint8_t *tmp_y = s->obmc_scratchpad + 8*h->mb_uvlinesize; |
|
| 562 |
+ uint8_t *tmp_cr = s->obmc_scratchpad + (16 << pixel_shift); |
|
| 563 |
+ uint8_t *tmp_y = s->obmc_scratchpad + 16*h->mb_uvlinesize; |
|
| 564 | 564 |
int refn0 = h->ref_cache[0][ scan8[n] ]; |
| 565 | 565 |
int refn1 = h->ref_cache[1][ scan8[n] ]; |
| 566 | 566 |
|
| 567 | 567 |
mc_dir_part(h, &h->ref_list[0][refn0], n, square, chroma_height, delta, 0, |
| 568 | 568 |
dest_y, dest_cb, dest_cr, |
| 569 |
- x_offset, y_offset, qpix_put, chroma_put, pixel_shift); |
|
| 569 |
+ x_offset, y_offset, qpix_put, chroma_put, pixel_shift, chroma444); |
|
| 570 | 570 |
mc_dir_part(h, &h->ref_list[1][refn1], n, square, chroma_height, delta, 1, |
| 571 | 571 |
tmp_y, tmp_cb, tmp_cr, |
| 572 |
- x_offset, y_offset, qpix_put, chroma_put, pixel_shift); |
|
| 572 |
+ x_offset, y_offset, qpix_put, chroma_put, pixel_shift, chroma444); |
|
| 573 | 573 |
|
| 574 | 574 |
if(h->use_weight == 2){
|
| 575 | 575 |
int weight0 = h->implicit_weight[refn0][refn1][s->mb_y&1]; |
| ... | ... |
@@ -594,7 +632,7 @@ static inline void mc_part_weighted(H264Context *h, int n, int square, int chrom |
| 594 | 594 |
Picture *ref= &h->ref_list[list][refn]; |
| 595 | 595 |
mc_dir_part(h, ref, n, square, chroma_height, delta, list, |
| 596 | 596 |
dest_y, dest_cb, dest_cr, x_offset, y_offset, |
| 597 |
- qpix_put, chroma_put, pixel_shift); |
|
| 597 |
+ qpix_put, chroma_put, pixel_shift, chroma444); |
|
| 598 | 598 |
|
| 599 | 599 |
luma_weight_op(dest_y, h->mb_linesize, h->luma_log2_weight_denom, |
| 600 | 600 |
h->luma_weight[refn][list][0], h->luma_weight[refn][list][1]); |
| ... | ... |
@@ -613,21 +651,21 @@ static inline void mc_part(H264Context *h, int n, int square, int chroma_height, |
| 613 | 613 |
qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put, |
| 614 | 614 |
qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg, |
| 615 | 615 |
h264_weight_func *weight_op, h264_biweight_func *weight_avg, |
| 616 |
- int list0, int list1, int pixel_shift){
|
|
| 616 |
+ int list0, int list1, int pixel_shift, int chroma444){
|
|
| 617 | 617 |
if((h->use_weight==2 && list0 && list1 |
| 618 | 618 |
&& (h->implicit_weight[ h->ref_cache[0][scan8[n]] ][ h->ref_cache[1][scan8[n]] ][h->s.mb_y&1] != 32)) |
| 619 | 619 |
|| h->use_weight==1) |
| 620 | 620 |
mc_part_weighted(h, n, square, chroma_height, delta, dest_y, dest_cb, dest_cr, |
| 621 | 621 |
x_offset, y_offset, qpix_put, chroma_put, |
| 622 | 622 |
weight_op[0], weight_op[3], weight_avg[0], |
| 623 |
- weight_avg[3], list0, list1, pixel_shift); |
|
| 623 |
+ weight_avg[3], list0, list1, pixel_shift, chroma444); |
|
| 624 | 624 |
else |
| 625 | 625 |
mc_part_std(h, n, square, chroma_height, delta, dest_y, dest_cb, dest_cr, |
| 626 | 626 |
x_offset, y_offset, qpix_put, chroma_put, qpix_avg, |
| 627 |
- chroma_avg, list0, list1, pixel_shift); |
|
| 627 |
+ chroma_avg, list0, list1, pixel_shift, chroma444); |
|
| 628 | 628 |
} |
| 629 | 629 |
|
| 630 |
-static inline void prefetch_motion(H264Context *h, int list, int pixel_shift){
|
|
| 630 |
+static inline void prefetch_motion(H264Context *h, int list, int pixel_shift, int chroma444){
|
|
| 631 | 631 |
/* fetch pixels for estimated mv 4 macroblocks ahead |
| 632 | 632 |
* optimized for 64byte cache lines */ |
| 633 | 633 |
MpegEncContext * const s = &h->s; |
| ... | ... |
@@ -638,8 +676,13 @@ static inline void prefetch_motion(H264Context *h, int list, int pixel_shift){
|
| 638 | 638 |
uint8_t **src= h->ref_list[list][refn].data; |
| 639 | 639 |
int off= ((mx+64)<<h->pixel_shift) + (my + (s->mb_x&3)*4)*h->mb_linesize; |
| 640 | 640 |
s->dsp.prefetch(src[0]+off, s->linesize, 4); |
| 641 |
- off= (((mx>>1)+64)<<h->pixel_shift) + ((my>>1) + (s->mb_x&7))*s->uvlinesize; |
|
| 642 |
- s->dsp.prefetch(src[1]+off, src[2]-src[1], 2); |
|
| 641 |
+ if(chroma444){
|
|
| 642 |
+ s->dsp.prefetch(src[1]+off, s->linesize, 4); |
|
| 643 |
+ s->dsp.prefetch(src[2]+off, s->linesize, 4); |
|
| 644 |
+ }else{
|
|
| 645 |
+ off= (((mx>>1)+64)<<pixel_shift) + ((my>>1) + (s->mb_x&7))*s->uvlinesize; |
|
| 646 |
+ s->dsp.prefetch(src[1]+off, src[2]-src[1], 2); |
|
| 647 |
+ } |
|
| 643 | 648 |
} |
| 644 | 649 |
} |
| 645 | 650 |
|
| ... | ... |
@@ -647,7 +690,7 @@ static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t |
| 647 | 647 |
qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put), |
| 648 | 648 |
qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg), |
| 649 | 649 |
h264_weight_func *weight_op, h264_biweight_func *weight_avg, |
| 650 |
- int pixel_shift){
|
|
| 650 |
+ int pixel_shift, int chroma444){
|
|
| 651 | 651 |
MpegEncContext * const s = &h->s; |
| 652 | 652 |
const int mb_xy= h->mb_xy; |
| 653 | 653 |
const int mb_type= s->current_picture.mb_type[mb_xy]; |
| ... | ... |
@@ -656,36 +699,36 @@ static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t |
| 656 | 656 |
|
| 657 | 657 |
if(HAVE_PTHREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME)) |
| 658 | 658 |
await_references(h); |
| 659 |
- prefetch_motion(h, 0, pixel_shift); |
|
| 659 |
+ prefetch_motion(h, 0, pixel_shift, chroma444); |
|
| 660 | 660 |
|
| 661 | 661 |
if(IS_16X16(mb_type)){
|
| 662 | 662 |
mc_part(h, 0, 1, 8, 0, dest_y, dest_cb, dest_cr, 0, 0, |
| 663 | 663 |
qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0], |
| 664 | 664 |
weight_op, weight_avg, |
| 665 | 665 |
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), |
| 666 |
- pixel_shift); |
|
| 666 |
+ pixel_shift, chroma444); |
|
| 667 | 667 |
}else if(IS_16X8(mb_type)){
|
| 668 | 668 |
mc_part(h, 0, 0, 4, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 0, |
| 669 | 669 |
qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0], |
| 670 | 670 |
&weight_op[1], &weight_avg[1], |
| 671 | 671 |
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), |
| 672 |
- pixel_shift); |
|
| 672 |
+ pixel_shift, chroma444); |
|
| 673 | 673 |
mc_part(h, 8, 0, 4, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 4, |
| 674 | 674 |
qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0], |
| 675 | 675 |
&weight_op[1], &weight_avg[1], |
| 676 | 676 |
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), |
| 677 |
- pixel_shift); |
|
| 677 |
+ pixel_shift, chroma444); |
|
| 678 | 678 |
}else if(IS_8X16(mb_type)){
|
| 679 | 679 |
mc_part(h, 0, 0, 8, 8*h->mb_linesize, dest_y, dest_cb, dest_cr, 0, 0, |
| 680 | 680 |
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], |
| 681 | 681 |
&weight_op[2], &weight_avg[2], |
| 682 | 682 |
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), |
| 683 |
- pixel_shift); |
|
| 683 |
+ pixel_shift, chroma444); |
|
| 684 | 684 |
mc_part(h, 4, 0, 8, 8*h->mb_linesize, dest_y, dest_cb, dest_cr, 4, 0, |
| 685 | 685 |
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], |
| 686 | 686 |
&weight_op[2], &weight_avg[2], |
| 687 | 687 |
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), |
| 688 |
- pixel_shift); |
|
| 688 |
+ pixel_shift, chroma444); |
|
| 689 | 689 |
}else{
|
| 690 | 690 |
int i; |
| 691 | 691 |
|
| ... | ... |
@@ -702,29 +745,29 @@ static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t |
| 702 | 702 |
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], |
| 703 | 703 |
&weight_op[3], &weight_avg[3], |
| 704 | 704 |
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), |
| 705 |
- pixel_shift); |
|
| 705 |
+ pixel_shift, chroma444); |
|
| 706 | 706 |
}else if(IS_SUB_8X4(sub_mb_type)){
|
| 707 | 707 |
mc_part(h, n , 0, 2, 4 << pixel_shift, dest_y, dest_cb, dest_cr, x_offset, y_offset, |
| 708 | 708 |
qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1], |
| 709 | 709 |
&weight_op[4], &weight_avg[4], |
| 710 | 710 |
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), |
| 711 |
- pixel_shift); |
|
| 711 |
+ pixel_shift, chroma444); |
|
| 712 | 712 |
mc_part(h, n+2, 0, 2, 4 << pixel_shift, dest_y, dest_cb, dest_cr, x_offset, y_offset+2, |
| 713 | 713 |
qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1], |
| 714 | 714 |
&weight_op[4], &weight_avg[4], |
| 715 | 715 |
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), |
| 716 |
- pixel_shift); |
|
| 716 |
+ pixel_shift, chroma444); |
|
| 717 | 717 |
}else if(IS_SUB_4X8(sub_mb_type)){
|
| 718 | 718 |
mc_part(h, n , 0, 4, 4*h->mb_linesize, dest_y, dest_cb, dest_cr, x_offset, y_offset, |
| 719 | 719 |
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], |
| 720 | 720 |
&weight_op[5], &weight_avg[5], |
| 721 | 721 |
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), |
| 722 |
- pixel_shift); |
|
| 722 |
+ pixel_shift, chroma444); |
|
| 723 | 723 |
mc_part(h, n+1, 0, 4, 4*h->mb_linesize, dest_y, dest_cb, dest_cr, x_offset+2, y_offset, |
| 724 | 724 |
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], |
| 725 | 725 |
&weight_op[5], &weight_avg[5], |
| 726 | 726 |
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), |
| 727 |
- pixel_shift); |
|
| 727 |
+ pixel_shift, chroma444); |
|
| 728 | 728 |
}else{
|
| 729 | 729 |
int j; |
| 730 | 730 |
assert(IS_SUB_4X4(sub_mb_type)); |
| ... | ... |
@@ -735,13 +778,13 @@ static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t |
| 735 | 735 |
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], |
| 736 | 736 |
&weight_op[6], &weight_avg[6], |
| 737 | 737 |
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), |
| 738 |
- pixel_shift); |
|
| 738 |
+ pixel_shift, chroma444); |
|
| 739 | 739 |
} |
| 740 | 740 |
} |
| 741 | 741 |
} |
| 742 | 742 |
} |
| 743 | 743 |
|
| 744 |
- prefetch_motion(h, 1, pixel_shift); |
|
| 744 |
+ prefetch_motion(h, 1, pixel_shift, chroma444); |
|
| 745 | 745 |
} |
| 746 | 746 |
|
| 747 | 747 |
#define hl_motion_fn(sh, bits) \ |
| ... | ... |
@@ -753,10 +796,11 @@ static av_always_inline void hl_motion_ ## bits(H264Context *h, \ |
| 753 | 753 |
qpel_mc_func (*qpix_avg)[16], \ |
| 754 | 754 |
h264_chroma_mc_func (*chroma_avg), \ |
| 755 | 755 |
h264_weight_func *weight_op, \ |
| 756 |
- h264_biweight_func *weight_avg) \ |
|
| 756 |
+ h264_biweight_func *weight_avg, \ |
|
| 757 |
+ int chroma444) \ |
|
| 757 | 758 |
{ \
|
| 758 | 759 |
hl_motion(h, dest_y, dest_cb, dest_cr, qpix_put, chroma_put, \ |
| 759 |
- qpix_avg, chroma_avg, weight_op, weight_avg, sh); \ |
|
| 760 |
+ qpix_avg, chroma_avg, weight_op, weight_avg, sh, chroma444); \ |
|
| 760 | 761 |
} |
| 761 | 762 |
hl_motion_fn(0, 8); |
| 762 | 763 |
hl_motion_fn(1, 16); |
| ... | ... |
@@ -796,16 +840,19 @@ static void free_tables(H264Context *h, int free_rbsp){
|
| 796 | 796 |
} |
| 797 | 797 |
|
| 798 | 798 |
static void init_dequant8_coeff_table(H264Context *h){
|
| 799 |
- int i,q,x; |
|
| 799 |
+ int i,j,q,x; |
|
| 800 | 800 |
const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8); |
| 801 |
- h->dequant8_coeff[0] = h->dequant8_buffer[0]; |
|
| 802 |
- h->dequant8_coeff[1] = h->dequant8_buffer[1]; |
|
| 803 | 801 |
|
| 804 |
- for(i=0; i<2; i++ ){
|
|
| 805 |
- if(i && !memcmp(h->pps.scaling_matrix8[0], h->pps.scaling_matrix8[1], 64*sizeof(uint8_t))){
|
|
| 806 |
- h->dequant8_coeff[1] = h->dequant8_buffer[0]; |
|
| 807 |
- break; |
|
| 802 |
+ for(i=0; i<6; i++ ){
|
|
| 803 |
+ h->dequant8_coeff[i] = h->dequant8_buffer[i]; |
|
| 804 |
+ for(j=0; j<i; j++){
|
|
| 805 |
+ if(!memcmp(h->pps.scaling_matrix8[j], h->pps.scaling_matrix8[i], 64*sizeof(uint8_t))){
|
|
| 806 |
+ h->dequant8_coeff[i] = h->dequant8_buffer[j]; |
|
| 807 |
+ break; |
|
| 808 |
+ } |
|
| 808 | 809 |
} |
| 810 |
+ if(j<i) |
|
| 811 |
+ continue; |
|
| 809 | 812 |
|
| 810 | 813 |
for(q=0; q<max_qp+1; q++){
|
| 811 | 814 |
int shift = div6[q]; |
| ... | ... |
@@ -853,7 +900,7 @@ static void init_dequant_tables(H264Context *h){
|
| 853 | 853 |
for(x=0; x<16; x++) |
| 854 | 854 |
h->dequant4_coeff[i][0][x] = 1<<6; |
| 855 | 855 |
if(h->pps.transform_8x8_mode) |
| 856 |
- for(i=0; i<2; i++) |
|
| 856 |
+ for(i=0; i<6; i++) |
|
| 857 | 857 |
for(x=0; x<64; x++) |
| 858 | 858 |
h->dequant8_coeff[i][0][x] = 1<<6; |
| 859 | 859 |
} |
| ... | ... |
@@ -868,7 +915,7 @@ int ff_h264_alloc_tables(H264Context *h){
|
| 868 | 868 |
|
| 869 | 869 |
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->intra4x4_pred_mode, row_mb_num * 8 * sizeof(uint8_t), fail) |
| 870 | 870 |
|
| 871 |
- FF_ALLOCZ_OR_GOTO(h->s.avctx, h->non_zero_count , big_mb_num * 32 * sizeof(uint8_t), fail) |
|
| 871 |
+ FF_ALLOCZ_OR_GOTO(h->s.avctx, h->non_zero_count , big_mb_num * 48 * sizeof(uint8_t), fail) |
|
| 872 | 872 |
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->slice_table_base , (big_mb_num+s->mb_stride) * sizeof(*h->slice_table_base), fail) |
| 873 | 873 |
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->cbp_table, big_mb_num * sizeof(uint16_t), fail) |
| 874 | 874 |
|
| ... | ... |
@@ -930,8 +977,8 @@ static void clone_tables(H264Context *dst, H264Context *src, int i){
|
| 930 | 930 |
* Allocate buffers which are not shared amongst multiple threads. |
| 931 | 931 |
*/ |
| 932 | 932 |
static int context_init(H264Context *h){
|
| 933 |
- FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[0], h->s.mb_width * (16+8+8) * sizeof(uint8_t)*2, fail) |
|
| 934 |
- FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[1], h->s.mb_width * (16+8+8) * sizeof(uint8_t)*2, fail) |
|
| 933 |
+ FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[0], h->s.mb_width * 16*3 * sizeof(uint8_t)*2, fail) |
|
| 934 |
+ FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[1], h->s.mb_width * 16*3 * sizeof(uint8_t)*2, fail) |
|
| 935 | 935 |
|
| 936 | 936 |
h->ref_cache[0][scan8[5 ]+1] = h->ref_cache[0][scan8[7 ]+1] = h->ref_cache[0][scan8[13]+1] = |
| 937 | 937 |
h->ref_cache[1][scan8[5 ]+1] = h->ref_cache[1][scan8[7 ]+1] = h->ref_cache[1][scan8[13]+1] = PART_NOT_AVAILABLE; |
| ... | ... |
@@ -1130,9 +1177,10 @@ static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContex |
| 1130 | 1130 |
|
| 1131 | 1131 |
// frame_start may not be called for the next thread (if it's decoding a bottom field) |
| 1132 | 1132 |
// so this has to be allocated here |
| 1133 |
- h->s.obmc_scratchpad = av_malloc(16*2*s->linesize + 8*2*s->uvlinesize); |
|
| 1133 |
+ h->s.obmc_scratchpad = av_malloc(16*6*s->linesize); |
|
| 1134 | 1134 |
|
| 1135 | 1135 |
s->dsp.clear_blocks(h->mb); |
| 1136 |
+ s->dsp.clear_blocks(h->mb+(24*16<<h->pixel_shift)); |
|
| 1136 | 1137 |
} |
| 1137 | 1138 |
|
| 1138 | 1139 |
//extradata/NAL handling |
| ... | ... |
@@ -1151,7 +1199,7 @@ static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContex |
| 1151 | 1151 |
for(i=0; i<6; i++) |
| 1152 | 1152 |
h->dequant4_coeff[i] = h->dequant4_buffer[0] + (h1->dequant4_coeff[i] - h1->dequant4_buffer[0]); |
| 1153 | 1153 |
|
| 1154 |
- for(i=0; i<2; i++) |
|
| 1154 |
+ for(i=0; i<6; i++) |
|
| 1155 | 1155 |
h->dequant8_coeff[i] = h->dequant8_buffer[0] + (h1->dequant8_coeff[i] - h1->dequant8_buffer[0]); |
| 1156 | 1156 |
|
| 1157 | 1157 |
h->dequant_coeff_pps = h1->dequant_coeff_pps; |
| ... | ... |
@@ -1206,20 +1254,20 @@ int ff_h264_frame_start(H264Context *h){
|
| 1206 | 1206 |
|
| 1207 | 1207 |
for(i=0; i<16; i++){
|
| 1208 | 1208 |
h->block_offset[i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 4*s->linesize*((scan8[i] - scan8[0])>>3); |
| 1209 |
- h->block_offset[24+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 8*s->linesize*((scan8[i] - scan8[0])>>3); |
|
| 1209 |
+ h->block_offset[48+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 8*s->linesize*((scan8[i] - scan8[0])>>3); |
|
| 1210 | 1210 |
} |
| 1211 |
- for(i=0; i<4; i++){
|
|
| 1211 |
+ for(i=0; i<16; i++){
|
|
| 1212 | 1212 |
h->block_offset[16+i]= |
| 1213 |
- h->block_offset[20+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 4*s->uvlinesize*((scan8[i] - scan8[0])>>3); |
|
| 1214 |
- h->block_offset[24+16+i]= |
|
| 1215 |
- h->block_offset[24+20+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 8*s->uvlinesize*((scan8[i] - scan8[0])>>3); |
|
| 1213 |
+ h->block_offset[32+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 4*s->uvlinesize*((scan8[i] - scan8[0])>>3); |
|
| 1214 |
+ h->block_offset[48+16+i]= |
|
| 1215 |
+ h->block_offset[48+32+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 8*s->uvlinesize*((scan8[i] - scan8[0])>>3); |
|
| 1216 | 1216 |
} |
| 1217 | 1217 |
|
| 1218 | 1218 |
/* can't be in alloc_tables because linesize isn't known there. |
| 1219 | 1219 |
* FIXME: redo bipred weight to not require extra buffer? */ |
| 1220 | 1220 |
for(i = 0; i < thread_count; i++) |
| 1221 | 1221 |
if(h->thread_context[i] && !h->thread_context[i]->s.obmc_scratchpad) |
| 1222 |
- h->thread_context[i]->s.obmc_scratchpad = av_malloc(16*2*s->linesize + 8*2*s->uvlinesize); |
|
| 1222 |
+ h->thread_context[i]->s.obmc_scratchpad = av_malloc(16*6*s->linesize); |
|
| 1223 | 1223 |
|
| 1224 | 1224 |
/* some macroblocks can be accessed before they're available in case of lost slices, mbaff or threading*/ |
| 1225 | 1225 |
memset(h->slice_table, -1, (s->mb_height*s->mb_stride-1) * sizeof(*h->slice_table)); |
| ... | ... |
@@ -1404,7 +1452,7 @@ static void decode_postinit(H264Context *h, int setup_finished){
|
| 1404 | 1404 |
ff_thread_finish_setup(s->avctx); |
| 1405 | 1405 |
} |
| 1406 | 1406 |
|
| 1407 |
-static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int simple){
|
|
| 1407 |
+static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int chroma444, int simple){
|
|
| 1408 | 1408 |
MpegEncContext * const s = &h->s; |
| 1409 | 1409 |
uint8_t *top_border; |
| 1410 | 1410 |
int top_idx = 1; |
| ... | ... |
@@ -1422,12 +1470,24 @@ static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src |
| 1422 | 1422 |
if (pixel_shift) |
| 1423 | 1423 |
AV_COPY128(top_border+16, src_y+15*linesize+16); |
| 1424 | 1424 |
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
|
| 1425 |
- if (pixel_shift) {
|
|
| 1426 |
- AV_COPY128(top_border+32, src_cb+7*uvlinesize); |
|
| 1427 |
- AV_COPY128(top_border+48, src_cr+7*uvlinesize); |
|
| 1425 |
+ if(chroma444){
|
|
| 1426 |
+ if (pixel_shift){
|
|
| 1427 |
+ AV_COPY128(top_border+32, src_cb + 15*uvlinesize); |
|
| 1428 |
+ AV_COPY128(top_border+48, src_cb + 15*uvlinesize+16); |
|
| 1429 |
+ AV_COPY128(top_border+64, src_cr + 15*uvlinesize); |
|
| 1430 |
+ AV_COPY128(top_border+80, src_cr + 15*uvlinesize+16); |
|
| 1431 |
+ } else {
|
|
| 1432 |
+ AV_COPY128(top_border+16, src_cb + 15*uvlinesize); |
|
| 1433 |
+ AV_COPY128(top_border+32, src_cr + 15*uvlinesize); |
|
| 1434 |
+ } |
|
| 1428 | 1435 |
} else {
|
| 1429 |
- AV_COPY64(top_border+16, src_cb+7*uvlinesize); |
|
| 1430 |
- AV_COPY64(top_border+24, src_cr+7*uvlinesize); |
|
| 1436 |
+ if (pixel_shift) {
|
|
| 1437 |
+ AV_COPY128(top_border+32, src_cb+7*uvlinesize); |
|
| 1438 |
+ AV_COPY128(top_border+48, src_cr+7*uvlinesize); |
|
| 1439 |
+ } else {
|
|
| 1440 |
+ AV_COPY64(top_border+16, src_cb+7*uvlinesize); |
|
| 1441 |
+ AV_COPY64(top_border+24, src_cr+7*uvlinesize); |
|
| 1442 |
+ } |
|
| 1431 | 1443 |
} |
| 1432 | 1444 |
} |
| 1433 | 1445 |
} |
| ... | ... |
@@ -1445,12 +1505,24 @@ static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src |
| 1445 | 1445 |
AV_COPY128(top_border+16, src_y+16*linesize+16); |
| 1446 | 1446 |
|
| 1447 | 1447 |
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
|
| 1448 |
- if (pixel_shift) {
|
|
| 1449 |
- AV_COPY128(top_border+32, src_cb+8*uvlinesize); |
|
| 1450 |
- AV_COPY128(top_border+48, src_cr+8*uvlinesize); |
|
| 1448 |
+ if(chroma444){
|
|
| 1449 |
+ if (pixel_shift){
|
|
| 1450 |
+ AV_COPY128(top_border+32, src_cb + 16*linesize); |
|
| 1451 |
+ AV_COPY128(top_border+48, src_cb + 16*linesize+16); |
|
| 1452 |
+ AV_COPY128(top_border+64, src_cr + 16*linesize); |
|
| 1453 |
+ AV_COPY128(top_border+80, src_cr + 16*linesize+16); |
|
| 1454 |
+ } else {
|
|
| 1455 |
+ AV_COPY128(top_border+16, src_cb + 16*linesize); |
|
| 1456 |
+ AV_COPY128(top_border+32, src_cr + 16*linesize); |
|
| 1457 |
+ } |
|
| 1451 | 1458 |
} else {
|
| 1452 |
- AV_COPY64(top_border+16, src_cb+8*uvlinesize); |
|
| 1453 |
- AV_COPY64(top_border+24, src_cr+8*uvlinesize); |
|
| 1459 |
+ if (pixel_shift) {
|
|
| 1460 |
+ AV_COPY128(top_border+32, src_cb+8*uvlinesize); |
|
| 1461 |
+ AV_COPY128(top_border+48, src_cr+8*uvlinesize); |
|
| 1462 |
+ } else {
|
|
| 1463 |
+ AV_COPY64(top_border+16, src_cb+8*uvlinesize); |
|
| 1464 |
+ AV_COPY64(top_border+24, src_cr+8*uvlinesize); |
|
| 1465 |
+ } |
|
| 1454 | 1466 |
} |
| 1455 | 1467 |
} |
| 1456 | 1468 |
} |
| ... | ... |
@@ -1458,7 +1530,8 @@ static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src |
| 1458 | 1458 |
static inline void xchg_mb_border(H264Context *h, uint8_t *src_y, |
| 1459 | 1459 |
uint8_t *src_cb, uint8_t *src_cr, |
| 1460 | 1460 |
int linesize, int uvlinesize, |
| 1461 |
- int xchg, int simple, int pixel_shift){
|
|
| 1461 |
+ int xchg, int chroma444, |
|
| 1462 |
+ int simple, int pixel_shift){
|
|
| 1462 | 1463 |
MpegEncContext * const s = &h->s; |
| 1463 | 1464 |
int deblock_topleft; |
| 1464 | 1465 |
int deblock_top; |
| ... | ... |
@@ -1513,13 +1586,28 @@ else AV_COPY64(b,a); |
| 1513 | 1513 |
} |
| 1514 | 1514 |
} |
| 1515 | 1515 |
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
|
| 1516 |
- if(deblock_top){
|
|
| 1516 |
+ if(chroma444){
|
|
| 1517 | 1517 |
if(deblock_topleft){
|
| 1518 |
- XCHG(top_border_m1 + (16 << pixel_shift), src_cb - (7 << pixel_shift), 1); |
|
| 1519 |
- XCHG(top_border_m1 + (24 << pixel_shift), src_cr - (7 << pixel_shift), 1); |
|
| 1518 |
+ XCHG(top_border_m1 + (24 << pixel_shift), src_cb - (7 << pixel_shift), 1); |
|
| 1519 |
+ XCHG(top_border_m1 + (40 << pixel_shift), src_cr - (7 << pixel_shift), 1); |
|
| 1520 |
+ } |
|
| 1521 |
+ XCHG(top_border + (16 << pixel_shift), src_cb + (1 << pixel_shift), xchg); |
|
| 1522 |
+ XCHG(top_border + (24 << pixel_shift), src_cb + (9 << pixel_shift), 1); |
|
| 1523 |
+ XCHG(top_border + (32 << pixel_shift), src_cr + (1 << pixel_shift), xchg); |
|
| 1524 |
+ XCHG(top_border + (40 << pixel_shift), src_cr + (9 << pixel_shift), 1); |
|
| 1525 |
+ if(s->mb_x+1 < s->mb_width){
|
|
| 1526 |
+ XCHG(h->top_borders[top_idx][s->mb_x+1] + (16 << pixel_shift), src_cb + (17 << pixel_shift), 1); |
|
| 1527 |
+ XCHG(h->top_borders[top_idx][s->mb_x+1] + (32 << pixel_shift), src_cr + (17 << pixel_shift), 1); |
|
| 1528 |
+ } |
|
| 1529 |
+ } else {
|
|
| 1530 |
+ if(deblock_top){
|
|
| 1531 |
+ if(deblock_topleft){
|
|
| 1532 |
+ XCHG(top_border_m1 + (16 << pixel_shift), src_cb - (7 << pixel_shift), 1); |
|
| 1533 |
+ XCHG(top_border_m1 + (24 << pixel_shift), src_cr - (7 << pixel_shift), 1); |
|
| 1534 |
+ } |
|
| 1535 |
+ XCHG(top_border + (16 << pixel_shift), src_cb+1+pixel_shift, 1); |
|
| 1536 |
+ XCHG(top_border + (24 << pixel_shift), src_cr+1+pixel_shift, 1); |
|
| 1520 | 1537 |
} |
| 1521 |
- XCHG(top_border + (16 << pixel_shift), src_cb+1+pixel_shift, 1); |
|
| 1522 |
- XCHG(top_border + (24 << pixel_shift), src_cr+1+pixel_shift, 1); |
|
| 1523 | 1538 |
} |
| 1524 | 1539 |
} |
| 1525 | 1540 |
} |
| ... | ... |
@@ -1538,6 +1626,159 @@ static av_always_inline void dctcoef_set(DCTELEM *mb, int high_bit_depth, int in |
| 1538 | 1538 |
AV_WN16A(mb + index, value); |
| 1539 | 1539 |
} |
| 1540 | 1540 |
|
| 1541 |
+static av_always_inline void hl_decode_mb_predict_luma(H264Context *h, int mb_type, int is_h264, int simple, int transform_bypass, |
|
| 1542 |
+ int pixel_shift, int *block_offset, int linesize, uint8_t *dest_y, int p) |
|
| 1543 |
+{
|
|
| 1544 |
+ MpegEncContext * const s = &h->s; |
|
| 1545 |
+ void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride); |
|
| 1546 |
+ void (*idct_dc_add)(uint8_t *dst, DCTELEM *block, int stride); |
|
| 1547 |
+ int i; |
|
| 1548 |
+ int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1]; |
|
| 1549 |
+ block_offset += 16*p; |
|
| 1550 |
+ if(IS_INTRA4x4(mb_type)){
|
|
| 1551 |
+ if(simple || !s->encoding){
|
|
| 1552 |
+ if(IS_8x8DCT(mb_type)){
|
|
| 1553 |
+ if(transform_bypass){
|
|
| 1554 |
+ idct_dc_add = |
|
| 1555 |
+ idct_add = s->dsp.add_pixels8; |
|
| 1556 |
+ }else{
|
|
| 1557 |
+ idct_dc_add = h->h264dsp.h264_idct8_dc_add; |
|
| 1558 |
+ idct_add = h->h264dsp.h264_idct8_add; |
|
| 1559 |
+ } |
|
| 1560 |
+ for(i=0; i<16; i+=4){
|
|
| 1561 |
+ uint8_t * const ptr= dest_y + block_offset[i]; |
|
| 1562 |
+ const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ]; |
|
| 1563 |
+ if(transform_bypass && h->sps.profile_idc==244 && dir<=1){
|
|
| 1564 |
+ h->hpc.pred8x8l_add[dir](ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); |
|
| 1565 |
+ }else{
|
|
| 1566 |
+ const int nnz = h->non_zero_count_cache[ scan8[i+p*16] ]; |
|
| 1567 |
+ h->hpc.pred8x8l[ dir ](ptr, (h->topleft_samples_available<<i)&0x8000, |
|
| 1568 |
+ (h->topright_samples_available<<i)&0x4000, linesize); |
|
| 1569 |
+ if(nnz){
|
|
| 1570 |
+ if(nnz == 1 && dctcoef_get(h->mb, pixel_shift, i*16+p*256)) |
|
| 1571 |
+ idct_dc_add(ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); |
|
| 1572 |
+ else |
|
| 1573 |
+ idct_add (ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); |
|
| 1574 |
+ } |
|
| 1575 |
+ } |
|
| 1576 |
+ } |
|
| 1577 |
+ }else{
|
|
| 1578 |
+ if(transform_bypass){
|
|
| 1579 |
+ idct_dc_add = |
|
| 1580 |
+ idct_add = s->dsp.add_pixels4; |
|
| 1581 |
+ }else{
|
|
| 1582 |
+ idct_dc_add = h->h264dsp.h264_idct_dc_add; |
|
| 1583 |
+ idct_add = h->h264dsp.h264_idct_add; |
|
| 1584 |
+ } |
|
| 1585 |
+ for(i=0; i<16; i++){
|
|
| 1586 |
+ uint8_t * const ptr= dest_y + block_offset[i]; |
|
| 1587 |
+ const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ]; |
|
| 1588 |
+ |
|
| 1589 |
+ if(transform_bypass && h->sps.profile_idc==244 && dir<=1){
|
|
| 1590 |
+ h->hpc.pred4x4_add[dir](ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); |
|
| 1591 |
+ }else{
|
|
| 1592 |
+ uint8_t *topright; |
|
| 1593 |
+ int nnz, tr; |
|
| 1594 |
+ uint64_t tr_high; |
|
| 1595 |
+ if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){
|
|
| 1596 |
+ const int topright_avail= (h->topright_samples_available<<i)&0x8000; |
|
| 1597 |
+ assert(mb_y || linesize <= block_offset[i]); |
|
| 1598 |
+ if(!topright_avail){
|
|
| 1599 |
+ if (pixel_shift) {
|
|
| 1600 |
+ tr_high= ((uint16_t*)ptr)[3 - linesize/2]*0x0001000100010001ULL; |
|
| 1601 |
+ topright= (uint8_t*) &tr_high; |
|
| 1602 |
+ } else {
|
|
| 1603 |
+ tr= ptr[3 - linesize]*0x01010101; |
|
| 1604 |
+ topright= (uint8_t*) &tr; |
|
| 1605 |
+ } |
|
| 1606 |
+ }else |
|
| 1607 |
+ topright= ptr + (4 << pixel_shift) - linesize; |
|
| 1608 |
+ }else |
|
| 1609 |
+ topright= NULL; |
|
| 1610 |
+ |
|
| 1611 |
+ h->hpc.pred4x4[ dir ](ptr, topright, linesize); |
|
| 1612 |
+ nnz = h->non_zero_count_cache[ scan8[i+p*16] ]; |
|
| 1613 |
+ if(nnz){
|
|
| 1614 |
+ if(is_h264){
|
|
| 1615 |
+ if(nnz == 1 && dctcoef_get(h->mb, pixel_shift, i*16+p*256)) |
|
| 1616 |
+ idct_dc_add(ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); |
|
| 1617 |
+ else |
|
| 1618 |
+ idct_add (ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); |
|
| 1619 |
+ }else |
|
| 1620 |
+ ff_svq3_add_idct_c(ptr, h->mb + i*16+p*256, linesize, qscale, 0); |
|
| 1621 |
+ } |
|
| 1622 |
+ } |
|
| 1623 |
+ } |
|
| 1624 |
+ } |
|
| 1625 |
+ } |
|
| 1626 |
+ }else{
|
|
| 1627 |
+ h->hpc.pred16x16[ h->intra16x16_pred_mode ](dest_y , linesize); |
|
| 1628 |
+ if(is_h264){
|
|
| 1629 |
+ if(h->non_zero_count_cache[ scan8[LUMA_DC_BLOCK_INDEX+p] ]){
|
|
| 1630 |
+ if(!transform_bypass) |
|
| 1631 |
+ h->h264dsp.h264_luma_dc_dequant_idct(h->mb+(p*256 << pixel_shift), h->mb_luma_dc[p], h->dequant4_coeff[p][qscale][0]); |
|
| 1632 |
+ else{
|
|
| 1633 |
+ static const uint8_t dc_mapping[16] = { 0*16, 1*16, 4*16, 5*16, 2*16, 3*16, 6*16, 7*16,
|
|
| 1634 |
+ 8*16, 9*16,12*16,13*16,10*16,11*16,14*16,15*16}; |
|
| 1635 |
+ for(i = 0; i < 16; i++) |
|
| 1636 |
+ dctcoef_set(h->mb+p*256, pixel_shift, dc_mapping[i], dctcoef_get(h->mb_luma_dc[p], pixel_shift, i)); |
|
| 1637 |
+ } |
|
| 1638 |
+ } |
|
| 1639 |
+ }else |
|
| 1640 |
+ ff_svq3_luma_dc_dequant_idct_c(h->mb+p*256, h->mb_luma_dc[p], qscale); |
|
| 1641 |
+ } |
|
| 1642 |
+} |
|
| 1643 |
+ |
|
| 1644 |
+static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, int mb_type, int is_h264, int simple, int transform_bypass, |
|
| 1645 |
+ int pixel_shift, int *block_offset, int linesize, uint8_t *dest_y, int p) |
|
| 1646 |
+{
|
|
| 1647 |
+ MpegEncContext * const s = &h->s; |
|
| 1648 |
+ void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride); |
|
| 1649 |
+ int i; |
|
| 1650 |
+ block_offset += 16*p; |
|
| 1651 |
+ if(!IS_INTRA4x4(mb_type)){
|
|
| 1652 |
+ if(is_h264){
|
|
| 1653 |
+ if(IS_INTRA16x16(mb_type)){
|
|
| 1654 |
+ if(transform_bypass){
|
|
| 1655 |
+ if(h->sps.profile_idc==244 && (h->intra16x16_pred_mode==VERT_PRED8x8 || h->intra16x16_pred_mode==HOR_PRED8x8)){
|
|
| 1656 |
+ h->hpc.pred16x16_add[h->intra16x16_pred_mode](dest_y, block_offset, h->mb + (p*256 << pixel_shift), linesize); |
|
| 1657 |
+ }else{
|
|
| 1658 |
+ for(i=0; i<16; i++){
|
|
| 1659 |
+ if(h->non_zero_count_cache[ scan8[i+p*16] ] || dctcoef_get(h->mb, pixel_shift, i*16)) |
|
| 1660 |
+ s->dsp.add_pixels4(dest_y + block_offset[i], h->mb + (i*16+p*256 << pixel_shift), linesize); |
|
| 1661 |
+ } |
|
| 1662 |
+ } |
|
| 1663 |
+ }else{
|
|
| 1664 |
+ h->h264dsp.h264_idct_add16intra(dest_y, block_offset, h->mb + (p*256 << pixel_shift), linesize, h->non_zero_count_cache+p*5*8); |
|
| 1665 |
+ } |
|
| 1666 |
+ }else if(h->cbp&15){
|
|
| 1667 |
+ if(transform_bypass){
|
|
| 1668 |
+ const int di = IS_8x8DCT(mb_type) ? 4 : 1; |
|
| 1669 |
+ idct_add= IS_8x8DCT(mb_type) ? s->dsp.add_pixels8 : s->dsp.add_pixels4; |
|
| 1670 |
+ for(i=0; i<16; i+=di){
|
|
| 1671 |
+ if(h->non_zero_count_cache[ scan8[i+p*16] ]){
|
|
| 1672 |
+ idct_add(dest_y + block_offset[i], h->mb + (i*16+p*256 << pixel_shift), linesize); |
|
| 1673 |
+ } |
|
| 1674 |
+ } |
|
| 1675 |
+ }else{
|
|
| 1676 |
+ if(IS_8x8DCT(mb_type)){
|
|
| 1677 |
+ h->h264dsp.h264_idct8_add4(dest_y, block_offset, h->mb + (p*256 << pixel_shift), linesize, h->non_zero_count_cache+p*5*8); |
|
| 1678 |
+ }else{
|
|
| 1679 |
+ h->h264dsp.h264_idct_add16(dest_y, block_offset, h->mb + (p*256 << pixel_shift), linesize, h->non_zero_count_cache+p*5*8); |
|
| 1680 |
+ } |
|
| 1681 |
+ } |
|
| 1682 |
+ } |
|
| 1683 |
+ }else{
|
|
| 1684 |
+ for(i=0; i<16; i++){
|
|
| 1685 |
+ if(h->non_zero_count_cache[ scan8[i+p*16] ] || h->mb[i*16+p*256]){ //FIXME benchmark weird rule, & below
|
|
| 1686 |
+ uint8_t * const ptr= dest_y + block_offset[i]; |
|
| 1687 |
+ ff_svq3_add_idct_c(ptr, h->mb + i*16 + p*256, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0); |
|
| 1688 |
+ } |
|
| 1689 |
+ } |
|
| 1690 |
+ } |
|
| 1691 |
+ } |
|
| 1692 |
+} |
|
| 1693 |
+ |
|
| 1541 | 1694 |
static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, int pixel_shift){
|
| 1542 | 1695 |
MpegEncContext * const s = &h->s; |
| 1543 | 1696 |
const int mb_x= s->mb_x; |
| ... | ... |
@@ -1546,13 +1787,12 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i |
| 1546 | 1546 |
const int mb_type= s->current_picture.mb_type[mb_xy]; |
| 1547 | 1547 |
uint8_t *dest_y, *dest_cb, *dest_cr; |
| 1548 | 1548 |
int linesize, uvlinesize /*dct_offset*/; |
| 1549 |
- int i; |
|
| 1549 |
+ int i, j; |
|
| 1550 | 1550 |
int *block_offset = &h->block_offset[0]; |
| 1551 | 1551 |
const int transform_bypass = !simple && (s->qscale == 0 && h->sps.transform_bypass); |
| 1552 | 1552 |
/* is_h264 should always be true if SVQ3 is disabled. */ |
| 1553 | 1553 |
const int is_h264 = !CONFIG_SVQ3_DECODER || simple || s->codec_id == CODEC_ID_H264; |
| 1554 | 1554 |
void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride); |
| 1555 |
- void (*idct_dc_add)(uint8_t *dst, DCTELEM *block, int stride); |
|
| 1556 | 1555 |
|
| 1557 | 1556 |
dest_y = s->current_picture.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16; |
| 1558 | 1557 |
dest_cb = s->current_picture.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8; |
| ... | ... |
@@ -1566,7 +1806,7 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i |
| 1566 | 1566 |
if (!simple && MB_FIELD) {
|
| 1567 | 1567 |
linesize = h->mb_linesize = s->linesize * 2; |
| 1568 | 1568 |
uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2; |
| 1569 |
- block_offset = &h->block_offset[24]; |
|
| 1569 |
+ block_offset = &h->block_offset[48]; |
|
| 1570 | 1570 |
if(mb_y&1){ //FIXME move out of this function?
|
| 1571 | 1571 |
dest_y -= s->linesize*15; |
| 1572 | 1572 |
dest_cb-= s->uvlinesize*7; |
| ... | ... |
@@ -1607,227 +1847,95 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i |
| 1607 | 1607 |
for (j = 0; j < 16; j++) |
| 1608 | 1608 |
tmp_y[j] = get_bits(&gb, bit_depth); |
| 1609 | 1609 |
} |
| 1610 |
- for (i = 0; i < 8; i++) {
|
|
| 1611 |
- uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize); |
|
| 1612 |
- for (j = 0; j < 8; j++) |
|
| 1613 |
- tmp_cb[j] = get_bits(&gb, bit_depth); |
|
| 1614 |
- } |
|
| 1615 |
- for (i = 0; i < 8; i++) {
|
|
| 1616 |
- uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize); |
|
| 1617 |
- for (j = 0; j < 8; j++) |
|
| 1618 |
- tmp_cr[j] = get_bits(&gb, bit_depth); |
|
| 1610 |
+ if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
|
|
| 1611 |
+ for (i = 0; i < 8; i++) {
|
|
| 1612 |
+ uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize); |
|
| 1613 |
+ for (j = 0; j < 8; j++) |
|
| 1614 |
+ tmp_cb[j] = get_bits(&gb, bit_depth); |
|
| 1615 |
+ } |
|
| 1616 |
+ for (i = 0; i < 8; i++) {
|
|
| 1617 |
+ uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize); |
|
| 1618 |
+ for (j = 0; j < 8; j++) |
|
| 1619 |
+ tmp_cr[j] = get_bits(&gb, bit_depth); |
|
| 1620 |
+ } |
|
| 1619 | 1621 |
} |
| 1620 | 1622 |
} else {
|
| 1621 |
- for (i=0; i<16; i++) {
|
|
| 1622 |
- memcpy(dest_y + i* linesize, h->mb + i*8, 16); |
|
| 1623 |
- } |
|
| 1624 |
- for (i=0; i<8; i++) {
|
|
| 1625 |
- memcpy(dest_cb+ i*uvlinesize, h->mb + 128 + i*4, 8); |
|
| 1626 |
- memcpy(dest_cr+ i*uvlinesize, h->mb + 160 + i*4, 8); |
|
| 1627 |
- } |
|
| 1623 |
+ for (i=0; i<16; i++) {
|
|
| 1624 |
+ memcpy(dest_y + i* linesize, h->mb + i*8, 16); |
|
| 1625 |
+ } |
|
| 1626 |
+ if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
|
|
| 1627 |
+ for (i=0; i<8; i++) {
|
|
| 1628 |
+ memcpy(dest_cb+ i*uvlinesize, h->mb + 128 + i*4, 8); |
|
| 1629 |
+ memcpy(dest_cr+ i*uvlinesize, h->mb + 160 + i*4, 8); |
|
| 1630 |
+ } |
|
| 1631 |
+ } |
|
| 1628 | 1632 |
} |
| 1629 | 1633 |
} else {
|
| 1630 | 1634 |
if(IS_INTRA(mb_type)){
|
| 1631 | 1635 |
if(h->deblocking_filter) |
| 1632 |
- xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1, simple, pixel_shift); |
|
| 1636 |
+ xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1, 0, simple, pixel_shift); |
|
| 1633 | 1637 |
|
| 1634 | 1638 |
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
|
| 1635 | 1639 |
h->hpc.pred8x8[ h->chroma_pred_mode ](dest_cb, uvlinesize); |
| 1636 | 1640 |
h->hpc.pred8x8[ h->chroma_pred_mode ](dest_cr, uvlinesize); |
| 1637 | 1641 |
} |
| 1638 | 1642 |
|
| 1639 |
- if(IS_INTRA4x4(mb_type)){
|
|
| 1640 |
- if(simple || !s->encoding){
|
|
| 1641 |
- if(IS_8x8DCT(mb_type)){
|
|
| 1642 |
- if(transform_bypass){
|
|
| 1643 |
- idct_dc_add = |
|
| 1644 |
- idct_add = s->dsp.add_pixels8; |
|
| 1645 |
- }else{
|
|
| 1646 |
- idct_dc_add = h->h264dsp.h264_idct8_dc_add; |
|
| 1647 |
- idct_add = h->h264dsp.h264_idct8_add; |
|
| 1648 |
- } |
|
| 1649 |
- for(i=0; i<16; i+=4){
|
|
| 1650 |
- uint8_t * const ptr= dest_y + block_offset[i]; |
|
| 1651 |
- const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ]; |
|
| 1652 |
- if(transform_bypass && h->sps.profile_idc==244 && dir<=1){
|
|
| 1653 |
- h->hpc.pred8x8l_add[dir](ptr, h->mb + (i*16 << pixel_shift), linesize); |
|
| 1654 |
- }else{
|
|
| 1655 |
- const int nnz = h->non_zero_count_cache[ scan8[i] ]; |
|
| 1656 |
- h->hpc.pred8x8l[ dir ](ptr, (h->topleft_samples_available<<i)&0x8000, |
|
| 1657 |
- (h->topright_samples_available<<i)&0x4000, linesize); |
|
| 1658 |
- if(nnz){
|
|
| 1659 |
- if(nnz == 1 && dctcoef_get(h->mb, pixel_shift, i*16)) |
|
| 1660 |
- idct_dc_add(ptr, h->mb + (i*16 << pixel_shift), linesize); |
|
| 1661 |
- else |
|
| 1662 |
- idct_add (ptr, h->mb + (i*16 << pixel_shift), linesize); |
|
| 1663 |
- } |
|
| 1664 |
- } |
|
| 1665 |
- } |
|
| 1666 |
- }else{
|
|
| 1667 |
- if(transform_bypass){
|
|
| 1668 |
- idct_dc_add = |
|
| 1669 |
- idct_add = s->dsp.add_pixels4; |
|
| 1670 |
- }else{
|
|
| 1671 |
- idct_dc_add = h->h264dsp.h264_idct_dc_add; |
|
| 1672 |
- idct_add = h->h264dsp.h264_idct_add; |
|
| 1673 |
- } |
|
| 1674 |
- for(i=0; i<16; i++){
|
|
| 1675 |
- uint8_t * const ptr= dest_y + block_offset[i]; |
|
| 1676 |
- const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ]; |
|
| 1677 |
- |
|
| 1678 |
- if(transform_bypass && h->sps.profile_idc==244 && dir<=1){
|
|
| 1679 |
- h->hpc.pred4x4_add[dir](ptr, h->mb + (i*16 << pixel_shift), linesize); |
|
| 1680 |
- }else{
|
|
| 1681 |
- uint8_t *topright; |
|
| 1682 |
- int nnz, tr; |
|
| 1683 |
- uint64_t tr_high; |
|
| 1684 |
- if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){
|
|
| 1685 |
- const int topright_avail= (h->topright_samples_available<<i)&0x8000; |
|
| 1686 |
- assert(mb_y || linesize <= block_offset[i]); |
|
| 1687 |
- if(!topright_avail){
|
|
| 1688 |
- if (pixel_shift) {
|
|
| 1689 |
- tr_high= ((uint16_t*)ptr)[3 - linesize/2]*0x0001000100010001ULL; |
|
| 1690 |
- topright= (uint8_t*) &tr_high; |
|
| 1691 |
- } else {
|
|
| 1692 |
- tr= ptr[3 - linesize]*0x01010101; |
|
| 1693 |
- topright= (uint8_t*) &tr; |
|
| 1694 |
- } |
|
| 1695 |
- }else |
|
| 1696 |
- topright= ptr + (4 << pixel_shift) - linesize; |
|
| 1697 |
- }else |
|
| 1698 |
- topright= NULL; |
|
| 1699 |
- |
|
| 1700 |
- h->hpc.pred4x4[ dir ](ptr, topright, linesize); |
|
| 1701 |
- nnz = h->non_zero_count_cache[ scan8[i] ]; |
|
| 1702 |
- if(nnz){
|
|
| 1703 |
- if(is_h264){
|
|
| 1704 |
- if(nnz == 1 && dctcoef_get(h->mb, pixel_shift, i*16)) |
|
| 1705 |
- idct_dc_add(ptr, h->mb + (i*16 << pixel_shift), linesize); |
|
| 1706 |
- else |
|
| 1707 |
- idct_add (ptr, h->mb + (i*16<<pixel_shift), linesize); |
|
| 1708 |
- } |
|
| 1709 |
-#if CONFIG_SVQ3_DECODER |
|
| 1710 |
- else |
|
| 1711 |
- ff_svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, 0); |
|
| 1712 |
-#endif |
|
| 1713 |
- } |
|
| 1714 |
- } |
|
| 1715 |
- } |
|
| 1716 |
- } |
|
| 1717 |
- } |
|
| 1718 |
- }else{
|
|
| 1719 |
- h->hpc.pred16x16[ h->intra16x16_pred_mode ](dest_y , linesize); |
|
| 1720 |
- if(is_h264){
|
|
| 1721 |
- if(h->non_zero_count_cache[ scan8[LUMA_DC_BLOCK_INDEX] ]){
|
|
| 1722 |
- if(!transform_bypass) |
|
| 1723 |
- h->h264dsp.h264_luma_dc_dequant_idct(h->mb, h->mb_luma_dc, h->dequant4_coeff[0][s->qscale][0]); |
|
| 1724 |
- else{
|
|
| 1725 |
- static const uint8_t dc_mapping[16] = { 0*16, 1*16, 4*16, 5*16, 2*16, 3*16, 6*16, 7*16,
|
|
| 1726 |
- 8*16, 9*16,12*16,13*16,10*16,11*16,14*16,15*16}; |
|
| 1727 |
- for(i = 0; i < 16; i++) |
|
| 1728 |
- dctcoef_set(h->mb, pixel_shift, dc_mapping[i], dctcoef_get(h->mb_luma_dc, pixel_shift, i)); |
|
| 1729 |
- } |
|
| 1730 |
- } |
|
| 1731 |
- } |
|
| 1732 |
-#if CONFIG_SVQ3_DECODER |
|
| 1733 |
- else |
|
| 1734 |
- ff_svq3_luma_dc_dequant_idct_c(h->mb, h->mb_luma_dc, s->qscale); |
|
| 1735 |
-#endif |
|
| 1736 |
- } |
|
| 1643 |
+ hl_decode_mb_predict_luma(h, mb_type, is_h264, simple, transform_bypass, pixel_shift, block_offset, linesize, dest_y, 0); |
|
| 1644 |
+ |
|
| 1737 | 1645 |
if(h->deblocking_filter) |
| 1738 |
- xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0, simple, pixel_shift); |
|
| 1646 |
+ xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0, 0, simple, pixel_shift); |
|
| 1739 | 1647 |
}else if(is_h264){
|
| 1740 | 1648 |
if (pixel_shift) {
|
| 1741 | 1649 |
hl_motion_16(h, dest_y, dest_cb, dest_cr, |
| 1742 | 1650 |
s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, |
| 1743 | 1651 |
s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, |
| 1744 | 1652 |
h->h264dsp.weight_h264_pixels_tab, |
| 1745 |
- h->h264dsp.biweight_h264_pixels_tab); |
|
| 1653 |
+ h->h264dsp.biweight_h264_pixels_tab, 0); |
|
| 1746 | 1654 |
} else |
| 1747 | 1655 |
hl_motion_8(h, dest_y, dest_cb, dest_cr, |
| 1748 | 1656 |
s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, |
| 1749 | 1657 |
s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, |
| 1750 | 1658 |
h->h264dsp.weight_h264_pixels_tab, |
| 1751 |
- h->h264dsp.biweight_h264_pixels_tab); |
|
| 1659 |
+ h->h264dsp.biweight_h264_pixels_tab, 0); |
|
| 1752 | 1660 |
} |
| 1753 | 1661 |
|
| 1754 |
- |
|
| 1755 |
- if(!IS_INTRA4x4(mb_type)){
|
|
| 1756 |
- if(is_h264){
|
|
| 1757 |
- if(IS_INTRA16x16(mb_type)){
|
|
| 1758 |
- if(transform_bypass){
|
|
| 1759 |
- if(h->sps.profile_idc==244 && (h->intra16x16_pred_mode==VERT_PRED8x8 || h->intra16x16_pred_mode==HOR_PRED8x8)){
|
|
| 1760 |
- h->hpc.pred16x16_add[h->intra16x16_pred_mode](dest_y, block_offset, h->mb, linesize); |
|
| 1761 |
- }else{
|
|
| 1762 |
- for(i=0; i<16; i++){
|
|
| 1763 |
- if(h->non_zero_count_cache[ scan8[i] ] || dctcoef_get(h->mb, pixel_shift, i*16)) |
|
| 1764 |
- s->dsp.add_pixels4(dest_y + block_offset[i], h->mb + (i*16 << pixel_shift), linesize); |
|
| 1765 |
- } |
|
| 1766 |
- } |
|
| 1767 |
- }else{
|
|
| 1768 |
- h->h264dsp.h264_idct_add16intra(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache); |
|
| 1769 |
- } |
|
| 1770 |
- }else if(h->cbp&15){
|
|
| 1771 |
- if(transform_bypass){
|
|
| 1772 |
- const int di = IS_8x8DCT(mb_type) ? 4 : 1; |
|
| 1773 |
- idct_add= IS_8x8DCT(mb_type) ? s->dsp.add_pixels8 : s->dsp.add_pixels4; |
|
| 1774 |
- for(i=0; i<16; i+=di){
|
|
| 1775 |
- if(h->non_zero_count_cache[ scan8[i] ]){
|
|
| 1776 |
- idct_add(dest_y + block_offset[i], h->mb + (i*16 << pixel_shift), linesize); |
|
| 1777 |
- } |
|
| 1778 |
- } |
|
| 1779 |
- }else{
|
|
| 1780 |
- if(IS_8x8DCT(mb_type)){
|
|
| 1781 |
- h->h264dsp.h264_idct8_add4(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache); |
|
| 1782 |
- }else{
|
|
| 1783 |
- h->h264dsp.h264_idct_add16(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache); |
|
| 1784 |
- } |
|
| 1785 |
- } |
|
| 1786 |
- } |
|
| 1787 |
- } |
|
| 1788 |
-#if CONFIG_SVQ3_DECODER |
|
| 1789 |
- else{
|
|
| 1790 |
- for(i=0; i<16; i++){
|
|
| 1791 |
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
|
|
| 1792 |
- uint8_t * const ptr= dest_y + block_offset[i]; |
|
| 1793 |
- ff_svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0); |
|
| 1794 |
- } |
|
| 1795 |
- } |
|
| 1796 |
- } |
|
| 1797 |
-#endif |
|
| 1798 |
- } |
|
| 1662 |
+ hl_decode_mb_idct_luma(h, mb_type, is_h264, simple, transform_bypass, pixel_shift, block_offset, linesize, dest_y, 0); |
|
| 1799 | 1663 |
|
| 1800 | 1664 |
if((simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)) && (h->cbp&0x30)){
|
| 1801 | 1665 |
uint8_t *dest[2] = {dest_cb, dest_cr};
|
| 1802 | 1666 |
if(transform_bypass){
|
| 1803 | 1667 |
if(IS_INTRA(mb_type) && h->sps.profile_idc==244 && (h->chroma_pred_mode==VERT_PRED8x8 || h->chroma_pred_mode==HOR_PRED8x8)){
|
| 1804 |
- h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0], block_offset + 16, h->mb + (16*16 << pixel_shift), uvlinesize); |
|
| 1805 |
- h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1], block_offset + 20, h->mb + (20*16 << pixel_shift), uvlinesize); |
|
| 1668 |
+ h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0], block_offset + 16, h->mb + (16*16*1 << pixel_shift), uvlinesize); |
|
| 1669 |
+ h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1], block_offset + 32, h->mb + (16*16*2 << pixel_shift), uvlinesize); |
|
| 1806 | 1670 |
}else{
|
| 1807 | 1671 |
idct_add = s->dsp.add_pixels4; |
| 1808 |
- for(i=16; i<16+8; i++){
|
|
| 1809 |
- if(h->non_zero_count_cache[ scan8[i] ] || dctcoef_get(h->mb, pixel_shift, i*16)) |
|
| 1810 |
- idct_add (dest[(i&4)>>2] + block_offset[i], h->mb + (i*16 << pixel_shift), uvlinesize); |
|
| 1672 |
+ for(j=1; j<3; j++){
|
|
| 1673 |
+ for(i=j*16; i<j*16+4; i++){
|
|
| 1674 |
+ if(h->non_zero_count_cache[ scan8[i] ] || dctcoef_get(h->mb, pixel_shift, i*16)) |
|
| 1675 |
+ idct_add (dest[j-1] + block_offset[i], h->mb + (i*16 << pixel_shift), uvlinesize); |
|
| 1676 |
+ } |
|
| 1811 | 1677 |
} |
| 1812 | 1678 |
} |
| 1813 | 1679 |
}else{
|
| 1814 | 1680 |
if(is_h264){
|
| 1815 | 1681 |
if(h->non_zero_count_cache[ scan8[CHROMA_DC_BLOCK_INDEX+0] ]) |
| 1816 |
- h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16 << pixel_shift) , h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]); |
|
| 1682 |
+ h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*1 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]); |
|
| 1817 | 1683 |
if(h->non_zero_count_cache[ scan8[CHROMA_DC_BLOCK_INDEX+1] ]) |
| 1818 |
- h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + ((16*16+4*16) << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]); |
|
| 1684 |
+ h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*2 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]); |
|
| 1819 | 1685 |
h->h264dsp.h264_idct_add8(dest, block_offset, |
| 1820 | 1686 |
h->mb, uvlinesize, |
| 1821 | 1687 |
h->non_zero_count_cache); |
| 1822 | 1688 |
} |
| 1823 | 1689 |
#if CONFIG_SVQ3_DECODER |
| 1824 | 1690 |
else{
|
| 1825 |
- h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16*16 , h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]); |
|
| 1826 |
- h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16*16+4*16, h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]); |
|
| 1827 |
- for(i=16; i<16+8; i++){
|
|
| 1828 |
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
|
|
| 1829 |
- uint8_t * const ptr= dest[(i&4)>>2] + block_offset[i]; |
|
| 1830 |
- ff_svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, ff_h264_chroma_qp[0][s->qscale + 12] - 12, 2); |
|
| 1691 |
+ h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16*16*1, h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]); |
|
| 1692 |
+ h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16*16*2, h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]); |
|
| 1693 |
+ for(j=1; j<3; j++){
|
|
| 1694 |
+ for(i=j*16; i<j*16+4; i++){
|
|
| 1695 |
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
|
|
| 1696 |
+ uint8_t * const ptr= dest[j-1] + block_offset[i]; |
|
| 1697 |
+ ff_svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, ff_h264_chroma_qp[0][s->qscale + 12] - 12, 2); |
|
| 1698 |
+ } |
|
| 1831 | 1699 |
} |
| 1832 | 1700 |
} |
| 1833 | 1701 |
} |
| ... | ... |
@@ -1836,7 +1944,113 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i |
| 1836 | 1836 |
} |
| 1837 | 1837 |
} |
| 1838 | 1838 |
if(h->cbp || IS_INTRA(mb_type)) |
| 1839 |
+ {
|
|
| 1840 |
+ s->dsp.clear_blocks(h->mb); |
|
| 1841 |
+ s->dsp.clear_blocks(h->mb+(24*16<<pixel_shift)); |
|
| 1842 |
+ } |
|
| 1843 |
+} |
|
| 1844 |
+ |
|
| 1845 |
+static av_always_inline void hl_decode_mb_444_internal(H264Context *h, int simple, int pixel_shift){
|
|
| 1846 |
+ MpegEncContext * const s = &h->s; |
|
| 1847 |
+ const int mb_x= s->mb_x; |
|
| 1848 |
+ const int mb_y= s->mb_y; |
|
| 1849 |
+ const int mb_xy= h->mb_xy; |
|
| 1850 |
+ const int mb_type= s->current_picture.mb_type[mb_xy]; |
|
| 1851 |
+ uint8_t *dest[3]; |
|
| 1852 |
+ int linesize; |
|
| 1853 |
+ int i, j, p; |
|
| 1854 |
+ int *block_offset = &h->block_offset[0]; |
|
| 1855 |
+ const int transform_bypass = !simple && (s->qscale == 0 && h->sps.transform_bypass); |
|
| 1856 |
+ const int plane_count = (simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)) ? 3 : 1; |
|
| 1857 |
+ |
|
| 1858 |
+ for (p = 0; p < plane_count; p++) |
|
| 1859 |
+ {
|
|
| 1860 |
+ dest[p] = s->current_picture.data[p] + ((mb_x << pixel_shift) + mb_y * s->linesize) * 16; |
|
| 1861 |
+ s->dsp.prefetch(dest[p] + (s->mb_x&3)*4*s->linesize + (64 << pixel_shift), s->linesize, 4); |
|
| 1862 |
+ } |
|
| 1863 |
+ |
|
| 1864 |
+ h->list_counts[mb_xy]= h->list_count; |
|
| 1865 |
+ |
|
| 1866 |
+ if (!simple && MB_FIELD) {
|
|
| 1867 |
+ linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize * 2; |
|
| 1868 |
+ block_offset = &h->block_offset[48]; |
|
| 1869 |
+ if(mb_y&1) //FIXME move out of this function? |
|
| 1870 |
+ for (p = 0; p < 3; p++) |
|
| 1871 |
+ dest[p] -= s->linesize*15; |
|
| 1872 |
+ if(FRAME_MBAFF) {
|
|
| 1873 |
+ int list; |
|
| 1874 |
+ for(list=0; list<h->list_count; list++){
|
|
| 1875 |
+ if(!USES_LIST(mb_type, list)) |
|
| 1876 |
+ continue; |
|
| 1877 |
+ if(IS_16X16(mb_type)){
|
|
| 1878 |
+ int8_t *ref = &h->ref_cache[list][scan8[0]]; |
|
| 1879 |
+ fill_rectangle(ref, 4, 4, 8, (16+*ref)^(s->mb_y&1), 1); |
|
| 1880 |
+ }else{
|
|
| 1881 |
+ for(i=0; i<16; i+=4){
|
|
| 1882 |
+ int ref = h->ref_cache[list][scan8[i]]; |
|
| 1883 |
+ if(ref >= 0) |
|
| 1884 |
+ fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, 8, (16+ref)^(s->mb_y&1), 1); |
|
| 1885 |
+ } |
|
| 1886 |
+ } |
|
| 1887 |
+ } |
|
| 1888 |
+ } |
|
| 1889 |
+ } else {
|
|
| 1890 |
+ linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize; |
|
| 1891 |
+ } |
|
| 1892 |
+ |
|
| 1893 |
+ if (!simple && IS_INTRA_PCM(mb_type)) {
|
|
| 1894 |
+ if (pixel_shift) {
|
|
| 1895 |
+ const int bit_depth = h->sps.bit_depth_luma; |
|
| 1896 |
+ GetBitContext gb; |
|
| 1897 |
+ init_get_bits(&gb, (uint8_t*)h->mb, 768*bit_depth); |
|
| 1898 |
+ |
|
| 1899 |
+ for (p = 0; p < plane_count; p++) {
|
|
| 1900 |
+ for (i = 0; i < 16; i++) {
|
|
| 1901 |
+ uint16_t *tmp = (uint16_t*)(dest[p] + i*linesize); |
|
| 1902 |
+ for (j = 0; j < 16; j++) |
|
| 1903 |
+ tmp[j] = get_bits(&gb, bit_depth); |
|
| 1904 |
+ } |
|
| 1905 |
+ } |
|
| 1906 |
+ } else {
|
|
| 1907 |
+ for (p = 0; p < plane_count; p++) {
|
|
| 1908 |
+ for (i = 0; i < 16; i++) {
|
|
| 1909 |
+ memcpy(dest[p] + i*linesize, h->mb + p*128 + i*8, 16); |
|
| 1910 |
+ } |
|
| 1911 |
+ } |
|
| 1912 |
+ } |
|
| 1913 |
+ } else {
|
|
| 1914 |
+ if(IS_INTRA(mb_type)){
|
|
| 1915 |
+ if(h->deblocking_filter) |
|
| 1916 |
+ xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, linesize, 1, 1, simple, pixel_shift); |
|
| 1917 |
+ |
|
| 1918 |
+ for (p = 0; p < plane_count; p++) |
|
| 1919 |
+ hl_decode_mb_predict_luma(h, mb_type, 1, simple, transform_bypass, pixel_shift, block_offset, linesize, dest[p], p); |
|
| 1920 |
+ |
|
| 1921 |
+ if(h->deblocking_filter) |
|
| 1922 |
+ xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, linesize, 0, 1, simple, pixel_shift); |
|
| 1923 |
+ }else{
|
|
| 1924 |
+ if (pixel_shift) {
|
|
| 1925 |
+ hl_motion_16(h, dest[0], dest[1], dest[2], |
|
| 1926 |
+ s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, |
|
| 1927 |
+ s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, |
|
| 1928 |
+ h->h264dsp.weight_h264_pixels_tab, |
|
| 1929 |
+ h->h264dsp.biweight_h264_pixels_tab, 1); |
|
| 1930 |
+ } else |
|
| 1931 |
+ hl_motion_8(h, dest[0], dest[1], dest[2], |
|
| 1932 |
+ s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, |
|
| 1933 |
+ s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, |
|
| 1934 |
+ h->h264dsp.weight_h264_pixels_tab, |
|
| 1935 |
+ h->h264dsp.biweight_h264_pixels_tab, 1); |
|
| 1936 |
+ } |
|
| 1937 |
+ |
|
| 1938 |
+ for (p = 0; p < plane_count; p++) |
|
| 1939 |
+ hl_decode_mb_idct_luma(h, mb_type, 1, simple, transform_bypass, pixel_shift, block_offset, linesize, dest[p], p); |
|
| 1940 |
+ } |
|
| 1941 |
+ if(h->cbp || IS_INTRA(mb_type)) |
|
| 1942 |
+ {
|
|
| 1839 | 1943 |
s->dsp.clear_blocks(h->mb); |
| 1944 |
+ s->dsp.clear_blocks(h->mb+(24*16<<pixel_shift)); |
|
| 1945 |
+ } |
|
| 1840 | 1946 |
} |
| 1841 | 1947 |
|
| 1842 | 1948 |
/** |
| ... | ... |
@@ -1856,13 +2070,26 @@ static void av_noinline hl_decode_mb_complex(H264Context *h){
|
| 1856 | 1856 |
hl_decode_mb_internal(h, 0, h->pixel_shift); |
| 1857 | 1857 |
} |
| 1858 | 1858 |
|
| 1859 |
+static void av_noinline hl_decode_mb_444_complex(H264Context *h){
|
|
| 1860 |
+ hl_decode_mb_444_internal(h, 0, h->pixel_shift); |
|
| 1861 |
+} |
|
| 1862 |
+ |
|
| 1863 |
+static void av_noinline hl_decode_mb_444_simple(H264Context *h){
|
|
| 1864 |
+ hl_decode_mb_444_internal(h, 1, 0); |
|
| 1865 |
+} |
|
| 1866 |
+ |
|
| 1859 | 1867 |
void ff_h264_hl_decode_mb(H264Context *h){
|
| 1860 | 1868 |
MpegEncContext * const s = &h->s; |
| 1861 | 1869 |
const int mb_xy= h->mb_xy; |
| 1862 | 1870 |
const int mb_type= s->current_picture.mb_type[mb_xy]; |
| 1863 | 1871 |
int is_complex = CONFIG_SMALL || h->is_complex || IS_INTRA_PCM(mb_type) || s->qscale == 0; |
| 1864 | 1872 |
|
| 1865 |
- if (is_complex) {
|
|
| 1873 |
+ if (CHROMA444) {
|
|
| 1874 |
+ if(is_complex || h->pixel_shift) |
|
| 1875 |
+ hl_decode_mb_444_complex(h); |
|
| 1876 |
+ else |
|
| 1877 |
+ hl_decode_mb_444_simple(h); |
|
| 1878 |
+ } else if (is_complex) {
|
|
| 1866 | 1879 |
hl_decode_mb_complex(h); |
| 1867 | 1880 |
} else if (h->pixel_shift) {
|
| 1868 | 1881 |
hl_decode_mb_simple_16(h); |
| ... | ... |
@@ -1878,7 +2105,7 @@ static int pred_weight_table(H264Context *h){
|
| 1878 | 1878 |
h->use_weight= 0; |
| 1879 | 1879 |
h->use_weight_chroma= 0; |
| 1880 | 1880 |
h->luma_log2_weight_denom= get_ue_golomb(&s->gb); |
| 1881 |
- if(CHROMA) |
|
| 1881 |
+ if(h->sps.chroma_format_idc) |
|
| 1882 | 1882 |
h->chroma_log2_weight_denom= get_ue_golomb(&s->gb); |
| 1883 | 1883 |
luma_def = 1<<h->luma_log2_weight_denom; |
| 1884 | 1884 |
chroma_def = 1<<h->chroma_log2_weight_denom; |
| ... | ... |
@@ -1903,7 +2130,7 @@ static int pred_weight_table(H264Context *h){
|
| 1903 | 1903 |
h->luma_weight[i][list][1]= 0; |
| 1904 | 1904 |
} |
| 1905 | 1905 |
|
| 1906 |
- if(CHROMA){
|
|
| 1906 |
+ if(h->sps.chroma_format_idc){
|
|
| 1907 | 1907 |
chroma_weight_flag= get_bits1(&s->gb); |
| 1908 | 1908 |
if(chroma_weight_flag){
|
| 1909 | 1909 |
int j; |
| ... | ... |
@@ -2333,11 +2560,11 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
|
| 2333 | 2333 |
|
| 2334 | 2334 |
h->b_stride= s->mb_width*4; |
| 2335 | 2335 |
|
| 2336 |
- s->width = 16*s->mb_width - 2*FFMIN(h->sps.crop_right, 7); |
|
| 2336 |
+ s->width = 16*s->mb_width - (2>>CHROMA444)*FFMIN(h->sps.crop_right, (8<<CHROMA444)-1); |
|
| 2337 | 2337 |
if(h->sps.frame_mbs_only_flag) |
| 2338 |
- s->height= 16*s->mb_height - 2*FFMIN(h->sps.crop_bottom, 7); |
|
| 2338 |
+ s->height= 16*s->mb_height - (2>>CHROMA444)*FFMIN(h->sps.crop_bottom, (8<<CHROMA444)-1); |
|
| 2339 | 2339 |
else |
| 2340 |
- s->height= 16*s->mb_height - 4*FFMIN(h->sps.crop_bottom, 7); |
|
| 2340 |
+ s->height= 16*s->mb_height - (4>>CHROMA444)*FFMIN(h->sps.crop_bottom, (8<<CHROMA444)-1); |
|
| 2341 | 2341 |
|
| 2342 | 2342 |
if (s->context_initialized |
| 2343 | 2343 |
&& ( s->width != s->avctx->width || s->height != s->avctx->height |
| ... | ... |
@@ -2382,18 +2609,22 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
|
| 2382 | 2382 |
|
| 2383 | 2383 |
switch (h->sps.bit_depth_luma) {
|
| 2384 | 2384 |
case 9 : |
| 2385 |
- s->avctx->pix_fmt = PIX_FMT_YUV420P9; |
|
| 2385 |
+ s->avctx->pix_fmt = CHROMA444 ? PIX_FMT_YUV444P9 : PIX_FMT_YUV420P9; |
|
| 2386 | 2386 |
break; |
| 2387 | 2387 |
case 10 : |
| 2388 |
- s->avctx->pix_fmt = PIX_FMT_YUV420P10; |
|
| 2388 |
+ s->avctx->pix_fmt = CHROMA444 ? PIX_FMT_YUV444P10 : PIX_FMT_YUV420P10; |
|
| 2389 | 2389 |
break; |
| 2390 | 2390 |
default: |
| 2391 |
- s->avctx->pix_fmt = s->avctx->get_format(s->avctx, |
|
| 2392 |
- s->avctx->codec->pix_fmts ? |
|
| 2393 |
- s->avctx->codec->pix_fmts : |
|
| 2394 |
- s->avctx->color_range == AVCOL_RANGE_JPEG ? |
|
| 2395 |
- hwaccel_pixfmt_list_h264_jpeg_420 : |
|
| 2396 |
- ff_hwaccel_pixfmt_list_420); |
|
| 2391 |
+ if (CHROMA444){
|
|
| 2392 |
+ s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ444P : PIX_FMT_YUV444P; |
|
| 2393 |
+ }else{
|
|
| 2394 |
+ s->avctx->pix_fmt = s->avctx->get_format(s->avctx, |
|
| 2395 |
+ s->avctx->codec->pix_fmts ? |
|
| 2396 |
+ s->avctx->codec->pix_fmts : |
|
| 2397 |
+ s->avctx->color_range == AVCOL_RANGE_JPEG ? |
|
| 2398 |
+ hwaccel_pixfmt_list_h264_jpeg_420 : |
|
| 2399 |
+ ff_hwaccel_pixfmt_list_420); |
|
| 2400 |
+ } |
|
| 2397 | 2401 |
} |
| 2398 | 2402 |
|
| 2399 | 2403 |
s->avctx->hwaccel = ff_find_hwaccel(s->avctx->codec->id, s->avctx->pix_fmt); |
| ... | ... |
@@ -2896,10 +3127,10 @@ static int fill_filter_caches(H264Context *h, int mb_type){
|
| 2896 | 2896 |
if(IS_INTRA(mb_type)) |
| 2897 | 2897 |
return 0; |
| 2898 | 2898 |
|
| 2899 |
- AV_COPY32(&h->non_zero_count_cache[4+8*1], &h->non_zero_count[mb_xy][ 4]); |
|
| 2900 |
- AV_COPY32(&h->non_zero_count_cache[4+8*2], &h->non_zero_count[mb_xy][12]); |
|
| 2901 |
- AV_COPY32(&h->non_zero_count_cache[4+8*3], &h->non_zero_count[mb_xy][20]); |
|
| 2902 |
- AV_COPY32(&h->non_zero_count_cache[4+8*4], &h->non_zero_count[mb_xy][28]); |
|
| 2899 |
+ AV_COPY32(&h->non_zero_count_cache[4+8* 1], &h->non_zero_count[mb_xy][ 0]); |
|
| 2900 |
+ AV_COPY32(&h->non_zero_count_cache[4+8* 2], &h->non_zero_count[mb_xy][ 4]); |
|
| 2901 |
+ AV_COPY32(&h->non_zero_count_cache[4+8* 3], &h->non_zero_count[mb_xy][ 8]); |
|
| 2902 |
+ AV_COPY32(&h->non_zero_count_cache[4+8* 4], &h->non_zero_count[mb_xy][12]); |
|
| 2903 | 2903 |
|
| 2904 | 2904 |
h->cbp= h->cbp_table[mb_xy]; |
| 2905 | 2905 |
|
| ... | ... |
@@ -2951,45 +3182,45 @@ static int fill_filter_caches(H264Context *h, int mb_type){
|
| 2951 | 2951 |
*/ |
| 2952 | 2952 |
//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec) |
| 2953 | 2953 |
if(top_type){
|
| 2954 |
- AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][4+3*8]); |
|
| 2954 |
+ AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][3*4]); |
|
| 2955 | 2955 |
} |
| 2956 | 2956 |
|
| 2957 | 2957 |
if(left_type[0]){
|
| 2958 |
- h->non_zero_count_cache[3+8*1]= h->non_zero_count[left_xy[0]][7+0*8]; |
|
| 2959 |
- h->non_zero_count_cache[3+8*2]= h->non_zero_count[left_xy[0]][7+1*8]; |
|
| 2960 |
- h->non_zero_count_cache[3+8*3]= h->non_zero_count[left_xy[0]][7+2*8]; |
|
| 2961 |
- h->non_zero_count_cache[3+8*4]= h->non_zero_count[left_xy[0]][7+3*8]; |
|
| 2958 |
+ h->non_zero_count_cache[3+8*1]= h->non_zero_count[left_xy[0]][3+0*4]; |
|
| 2959 |
+ h->non_zero_count_cache[3+8*2]= h->non_zero_count[left_xy[0]][3+1*4]; |
|
| 2960 |
+ h->non_zero_count_cache[3+8*3]= h->non_zero_count[left_xy[0]][3+2*4]; |
|
| 2961 |
+ h->non_zero_count_cache[3+8*4]= h->non_zero_count[left_xy[0]][3+3*4]; |
|
| 2962 | 2962 |
} |
| 2963 | 2963 |
|
| 2964 | 2964 |
// CAVLC 8x8dct requires NNZ values for residual decoding that differ from what the loop filter needs |
| 2965 | 2965 |
if(!CABAC && h->pps.transform_8x8_mode){
|
| 2966 | 2966 |
if(IS_8x8DCT(top_type)){
|
| 2967 | 2967 |
h->non_zero_count_cache[4+8*0]= |
| 2968 |
- h->non_zero_count_cache[5+8*0]= h->cbp_table[top_xy] & 4; |
|
| 2968 |
+ h->non_zero_count_cache[5+8*0]= (h->cbp_table[top_xy] & 0x4000) >> 12; |
|
| 2969 | 2969 |
h->non_zero_count_cache[6+8*0]= |
| 2970 |
- h->non_zero_count_cache[7+8*0]= h->cbp_table[top_xy] & 8; |
|
| 2970 |
+ h->non_zero_count_cache[7+8*0]= (h->cbp_table[top_xy] & 0x8000) >> 12; |
|
| 2971 | 2971 |
} |
| 2972 | 2972 |
if(IS_8x8DCT(left_type[0])){
|
| 2973 | 2973 |
h->non_zero_count_cache[3+8*1]= |
| 2974 |
- h->non_zero_count_cache[3+8*2]= h->cbp_table[left_xy[0]]&2; //FIXME check MBAFF |
|
| 2974 |
+ h->non_zero_count_cache[3+8*2]= (h->cbp_table[left_xy[0]]&0x2000) >> 12; //FIXME check MBAFF |
|
| 2975 | 2975 |
} |
| 2976 | 2976 |
if(IS_8x8DCT(left_type[1])){
|
| 2977 | 2977 |
h->non_zero_count_cache[3+8*3]= |
| 2978 |
- h->non_zero_count_cache[3+8*4]= h->cbp_table[left_xy[1]]&8; //FIXME check MBAFF |
|
| 2978 |
+ h->non_zero_count_cache[3+8*4]= (h->cbp_table[left_xy[1]]&0x8000) >> 12; //FIXME check MBAFF |
|
| 2979 | 2979 |
} |
| 2980 | 2980 |
|
| 2981 | 2981 |
if(IS_8x8DCT(mb_type)){
|
| 2982 | 2982 |
h->non_zero_count_cache[scan8[0 ]]= h->non_zero_count_cache[scan8[1 ]]= |
| 2983 |
- h->non_zero_count_cache[scan8[2 ]]= h->non_zero_count_cache[scan8[3 ]]= h->cbp & 1; |
|
| 2983 |
+ h->non_zero_count_cache[scan8[2 ]]= h->non_zero_count_cache[scan8[3 ]]= (h->cbp & 0x1000) >> 12; |
|
| 2984 | 2984 |
|
| 2985 | 2985 |
h->non_zero_count_cache[scan8[0+ 4]]= h->non_zero_count_cache[scan8[1+ 4]]= |
| 2986 |
- h->non_zero_count_cache[scan8[2+ 4]]= h->non_zero_count_cache[scan8[3+ 4]]= h->cbp & 2; |
|
| 2986 |
+ h->non_zero_count_cache[scan8[2+ 4]]= h->non_zero_count_cache[scan8[3+ 4]]= (h->cbp & 0x2000) >> 12; |
|
| 2987 | 2987 |
|
| 2988 | 2988 |
h->non_zero_count_cache[scan8[0+ 8]]= h->non_zero_count_cache[scan8[1+ 8]]= |
| 2989 |
- h->non_zero_count_cache[scan8[2+ 8]]= h->non_zero_count_cache[scan8[3+ 8]]= h->cbp & 4; |
|
| 2989 |
+ h->non_zero_count_cache[scan8[2+ 8]]= h->non_zero_count_cache[scan8[3+ 8]]= (h->cbp & 0x4000) >> 12; |
|
| 2990 | 2990 |
|
| 2991 | 2991 |
h->non_zero_count_cache[scan8[0+12]]= h->non_zero_count_cache[scan8[1+12]]= |
| 2992 |
- h->non_zero_count_cache[scan8[2+12]]= h->non_zero_count_cache[scan8[3+12]]= h->cbp & 8; |
|
| 2992 |
+ h->non_zero_count_cache[scan8[2+12]]= h->non_zero_count_cache[scan8[3+12]]= (h->cbp & 0x8000) >> 12; |
|
| 2993 | 2993 |
} |
| 2994 | 2994 |
} |
| 2995 | 2995 |
|
| ... | ... |
@@ -3063,8 +3294,8 @@ static void loop_filter(H264Context *h, int start_x, int end_x){
|
| 3063 | 3063 |
s->mb_x= mb_x; |
| 3064 | 3064 |
s->mb_y= mb_y; |
| 3065 | 3065 |
dest_y = s->current_picture.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16; |
| 3066 |
- dest_cb = s->current_picture.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8; |
|
| 3067 |
- dest_cr = s->current_picture.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8; |
|
| 3066 |
+ dest_cb = s->current_picture.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444); |
|
| 3067 |
+ dest_cr = s->current_picture.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444); |
|
| 3068 | 3068 |
//FIXME simplify above |
| 3069 | 3069 |
|
| 3070 | 3070 |
if (MB_FIELD) {
|
| ... | ... |
@@ -3079,7 +3310,7 @@ static void loop_filter(H264Context *h, int start_x, int end_x){
|
| 3079 | 3079 |
linesize = h->mb_linesize = s->linesize; |
| 3080 | 3080 |
uvlinesize = h->mb_uvlinesize = s->uvlinesize; |
| 3081 | 3081 |
} |
| 3082 |
- backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0); |
|
| 3082 |
+ backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, CHROMA444, 0); |
|
| 3083 | 3083 |
if(fill_filter_caches(h, mb_type)) |
| 3084 | 3084 |
continue; |
| 3085 | 3085 |
h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.qscale_table[mb_xy]); |
| ... | ... |
@@ -39,9 +39,6 @@ |
| 39 | 39 |
#define interlaced_dct interlaced_dct_is_a_bad_name |
| 40 | 40 |
#define mb_intra mb_intra_is_not_initialized_see_mb_type |
| 41 | 41 |
|
| 42 |
-#define LUMA_DC_BLOCK_INDEX 24 |
|
| 43 |
-#define CHROMA_DC_BLOCK_INDEX 25 |
|
| 44 |
- |
|
| 45 | 42 |
#define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8 |
| 46 | 43 |
#define COEFF_TOKEN_VLC_BITS 8 |
| 47 | 44 |
#define TOTAL_ZEROS_VLC_BITS 9 |
| ... | ... |
@@ -60,8 +57,6 @@ |
| 60 | 60 |
* of progressive decoding by about 2%. */ |
| 61 | 61 |
#define ALLOW_INTERLACE |
| 62 | 62 |
|
| 63 |
-#define ALLOW_NOCHROMA |
|
| 64 |
- |
|
| 65 | 63 |
#define FMO 0 |
| 66 | 64 |
|
| 67 | 65 |
/** |
| ... | ... |
@@ -85,16 +80,12 @@ |
| 85 | 85 |
#endif |
| 86 | 86 |
#define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE) |
| 87 | 87 |
|
| 88 |
-#ifdef ALLOW_NOCHROMA |
|
| 89 |
-#define CHROMA h->sps.chroma_format_idc |
|
| 90 |
-#else |
|
| 91 |
-#define CHROMA 1 |
|
| 92 |
-#endif |
|
| 93 |
- |
|
| 94 | 88 |
#ifndef CABAC |
| 95 | 89 |
#define CABAC h->pps.cabac |
| 96 | 90 |
#endif |
| 97 | 91 |
|
| 92 |
+#define CHROMA444 (h->sps.chroma_format_idc == 3) |
|
| 93 |
+ |
|
| 98 | 94 |
#define EXTENDED_SAR 255 |
| 99 | 95 |
|
| 100 | 96 |
#define MB_TYPE_REF0 MB_TYPE_ACPRED //dirty but it fits in 16 bit |
| ... | ... |
@@ -198,7 +189,7 @@ typedef struct SPS{
|
| 198 | 198 |
int num_reorder_frames; |
| 199 | 199 |
int scaling_matrix_present; |
| 200 | 200 |
uint8_t scaling_matrix4[6][16]; |
| 201 |
- uint8_t scaling_matrix8[2][64]; |
|
| 201 |
+ uint8_t scaling_matrix8[6][64]; |
|
| 202 | 202 |
int nal_hrd_parameters_present_flag; |
| 203 | 203 |
int vcl_hrd_parameters_present_flag; |
| 204 | 204 |
int pic_struct_present_flag; |
| ... | ... |
@@ -233,7 +224,7 @@ typedef struct PPS{
|
| 233 | 233 |
int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag |
| 234 | 234 |
int transform_8x8_mode; ///< transform_8x8_mode_flag |
| 235 | 235 |
uint8_t scaling_matrix4[6][16]; |
| 236 |
- uint8_t scaling_matrix8[2][64]; |
|
| 236 |
+ uint8_t scaling_matrix8[6][64]; |
|
| 237 | 237 |
uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table |
| 238 | 238 |
int chroma_qp_diff; |
| 239 | 239 |
}PPS; |
| ... | ... |
@@ -298,21 +289,15 @@ typedef struct H264Context{
|
| 298 | 298 |
unsigned int top_samples_available; |
| 299 | 299 |
unsigned int topright_samples_available; |
| 300 | 300 |
unsigned int left_samples_available; |
| 301 |
- uint8_t (*top_borders[2])[(16+2*8)*2]; |
|
| 301 |
+ uint8_t (*top_borders[2])[(16*3)*2]; |
|
| 302 | 302 |
|
| 303 | 303 |
/** |
| 304 | 304 |
* non zero coeff count cache. |
| 305 | 305 |
* is 64 if not available. |
| 306 | 306 |
*/ |
| 307 |
- DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[6*8]; |
|
| 307 |
+ DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15*8]; |
|
| 308 | 308 |
|
| 309 |
- /* |
|
| 310 |
- .UU.YYYY |
|
| 311 |
- .UU.YYYY |
|
| 312 |
- .vv.YYYY |
|
| 313 |
- .VV.YYYY |
|
| 314 |
- */ |
|
| 315 |
- uint8_t (*non_zero_count)[32]; |
|
| 309 |
+ uint8_t (*non_zero_count)[48]; |
|
| 316 | 310 |
|
| 317 | 311 |
/** |
| 318 | 312 |
* Motion vector cache. |
| ... | ... |
@@ -336,7 +321,7 @@ typedef struct H264Context{
|
| 336 | 336 |
* block_offset[ 0..23] for frame macroblocks |
| 337 | 337 |
* block_offset[24..47] for field macroblocks |
| 338 | 338 |
*/ |
| 339 |
- int block_offset[2*(16+8)]; |
|
| 339 |
+ int block_offset[2*(16*3)]; |
|
| 340 | 340 |
|
| 341 | 341 |
uint32_t *mb2b_xy; //FIXME are these 4 a good idea? |
| 342 | 342 |
uint32_t *mb2br_xy; |
| ... | ... |
@@ -356,9 +341,9 @@ typedef struct H264Context{
|
| 356 | 356 |
PPS pps; //FIXME move to Picture perhaps? (->no) do we need that? |
| 357 | 357 |
|
| 358 | 358 |
uint32_t dequant4_buffer[6][QP_MAX_NUM+1][16]; //FIXME should these be moved down? |
| 359 |
- uint32_t dequant8_buffer[2][QP_MAX_NUM+1][64]; |
|
| 359 |
+ uint32_t dequant8_buffer[6][QP_MAX_NUM+1][64]; |
|
| 360 | 360 |
uint32_t (*dequant4_coeff[6])[16]; |
| 361 |
- uint32_t (*dequant8_coeff[2])[64]; |
|
| 361 |
+ uint32_t (*dequant8_coeff[6])[64]; |
|
| 362 | 362 |
|
| 363 | 363 |
int slice_num; |
| 364 | 364 |
uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1 |
| ... | ... |
@@ -408,15 +393,15 @@ typedef struct H264Context{
|
| 408 | 408 |
GetBitContext *intra_gb_ptr; |
| 409 | 409 |
GetBitContext *inter_gb_ptr; |
| 410 | 410 |
|
| 411 |
- DECLARE_ALIGNED(16, DCTELEM, mb)[16*24*2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space. |
|
| 412 |
- DECLARE_ALIGNED(16, DCTELEM, mb_luma_dc)[16*2]; |
|
| 411 |
+ DECLARE_ALIGNED(16, DCTELEM, mb)[16*48*2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space. |
|
| 412 |
+ DECLARE_ALIGNED(16, DCTELEM, mb_luma_dc)[3][16*2]; |
|
| 413 | 413 |
DCTELEM mb_padding[256*2]; ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either check that i is not too large or ensure that there is some unused stuff after mb |
| 414 | 414 |
|
| 415 | 415 |
/** |
| 416 | 416 |
* Cabac |
| 417 | 417 |
*/ |
| 418 | 418 |
CABACContext cabac; |
| 419 |
- uint8_t cabac_state[460]; |
|
| 419 |
+ uint8_t cabac_state[1024]; |
|
| 420 | 420 |
|
| 421 | 421 |
/* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */ |
| 422 | 422 |
uint16_t *cbp_table; |
| ... | ... |
@@ -721,27 +706,43 @@ o-o o-o |
| 721 | 721 |
*/ |
| 722 | 722 |
|
| 723 | 723 |
/* Scan8 organization: |
| 724 |
- * 0 1 2 3 4 5 6 7 |
|
| 725 |
- * 0 u u y y y y y |
|
| 726 |
- * 1 u U U y Y Y Y Y |
|
| 727 |
- * 2 u U U y Y Y Y Y |
|
| 728 |
- * 3 v v y Y Y Y Y |
|
| 729 |
- * 4 v V V y Y Y Y Y |
|
| 730 |
- * 5 v V V DYDUDV |
|
| 724 |
+ * 0 1 2 3 4 5 6 7 |
|
| 725 |
+ * 0 DY y y y y y |
|
| 726 |
+ * 1 y Y Y Y Y |
|
| 727 |
+ * 2 y Y Y Y Y |
|
| 728 |
+ * 3 y Y Y Y Y |
|
| 729 |
+ * 4 y Y Y Y Y |
|
| 730 |
+ * 5 DU u u u u u |
|
| 731 |
+ * 6 u U U U U |
|
| 732 |
+ * 7 u U U U U |
|
| 733 |
+ * 8 u U U U U |
|
| 734 |
+ * 9 u U U U U |
|
| 735 |
+ * 10 DV v v v v v |
|
| 736 |
+ * 11 v V V V V |
|
| 737 |
+ * 12 v V V V V |
|
| 738 |
+ * 13 v V V V V |
|
| 739 |
+ * 14 v V V V V |
|
| 731 | 740 |
* DY/DU/DV are for luma/chroma DC. |
| 732 | 741 |
*/ |
| 733 | 742 |
|
| 743 |
+#define LUMA_DC_BLOCK_INDEX 48 |
|
| 744 |
+#define CHROMA_DC_BLOCK_INDEX 49 |
|
| 745 |
+ |
|
| 734 | 746 |
//This table must be here because scan8[constant] must be known at compiletime |
| 735 |
-static const uint8_t scan8[16 + 2*4 + 3]={
|
|
| 736 |
- 4+1*8, 5+1*8, 4+2*8, 5+2*8, |
|
| 737 |
- 6+1*8, 7+1*8, 6+2*8, 7+2*8, |
|
| 738 |
- 4+3*8, 5+3*8, 4+4*8, 5+4*8, |
|
| 739 |
- 6+3*8, 7+3*8, 6+4*8, 7+4*8, |
|
| 740 |
- 1+1*8, 2+1*8, |
|
| 741 |
- 1+2*8, 2+2*8, |
|
| 742 |
- 1+4*8, 2+4*8, |
|
| 743 |
- 1+5*8, 2+5*8, |
|
| 744 |
- 4+5*8, 5+5*8, 6+5*8 |
|
| 747 |
+static const uint8_t scan8[16*3 + 3]={
|
|
| 748 |
+ 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8, |
|
| 749 |
+ 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8, |
|
| 750 |
+ 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8, |
|
| 751 |
+ 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8, |
|
| 752 |
+ 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8, |
|
| 753 |
+ 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8, |
|
| 754 |
+ 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8, |
|
| 755 |
+ 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8, |
|
| 756 |
+ 4+11*8, 5+11*8, 4+12*8, 5+12*8, |
|
| 757 |
+ 6+11*8, 7+11*8, 6+12*8, 7+12*8, |
|
| 758 |
+ 4+13*8, 5+13*8, 4+14*8, 5+14*8, |
|
| 759 |
+ 6+13*8, 7+13*8, 6+14*8, 7+14*8, |
|
| 760 |
+ 0+ 0*8, 0+ 5*8, 0+10*8 |
|
| 745 | 761 |
}; |
| 746 | 762 |
|
| 747 | 763 |
static av_always_inline uint32_t pack16to32(int a, int b){
|
| ... | ... |
@@ -773,11 +774,11 @@ static void fill_decode_neighbors(H264Context *h, int mb_type){
|
| 773 | 773 |
MpegEncContext * const s = &h->s; |
| 774 | 774 |
const int mb_xy= h->mb_xy; |
| 775 | 775 |
int topleft_xy, top_xy, topright_xy, left_xy[2]; |
| 776 |
- static const uint8_t left_block_options[4][16]={
|
|
| 777 |
- {0,1,2,3,7,10,8,11,7+0*8, 7+1*8, 7+2*8, 7+3*8, 2+0*8, 2+3*8, 2+1*8, 2+2*8},
|
|
| 778 |
- {2,2,3,3,8,11,8,11,7+2*8, 7+2*8, 7+3*8, 7+3*8, 2+1*8, 2+2*8, 2+1*8, 2+2*8},
|
|
| 779 |
- {0,0,1,1,7,10,7,10,7+0*8, 7+0*8, 7+1*8, 7+1*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8},
|
|
| 780 |
- {0,2,0,2,7,10,7,10,7+0*8, 7+2*8, 7+0*8, 7+2*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8}
|
|
| 776 |
+ static const uint8_t left_block_options[4][32]={
|
|
| 777 |
+ {0,1,2,3,7,10,8,11,3+0*4, 3+1*4, 3+2*4, 3+3*4, 1+4*4, 1+8*4, 1+5*4, 1+9*4},
|
|
| 778 |
+ {2,2,3,3,8,11,8,11,3+2*4, 3+2*4, 3+3*4, 3+3*4, 1+5*4, 1+9*4, 1+5*4, 1+9*4},
|
|
| 779 |
+ {0,0,1,1,7,10,7,10,3+0*4, 3+0*4, 3+1*4, 3+1*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4},
|
|
| 780 |
+ {0,2,0,2,7,10,7,10,3+0*4, 3+2*4, 3+0*4, 3+2*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4}
|
|
| 781 | 781 |
}; |
| 782 | 782 |
|
| 783 | 783 |
h->topleft_partition= -1; |
| ... | ... |
@@ -947,32 +948,41 @@ static void fill_decode_caches(H264Context *h, int mb_type){
|
| 947 | 947 |
*/ |
| 948 | 948 |
//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec) |
| 949 | 949 |
if(top_type){
|
| 950 |
- AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][4+3*8]); |
|
| 951 |
- h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][1+1*8]; |
|
| 952 |
- h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][2+1*8]; |
|
| 953 |
- |
|
| 954 |
- h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][1+2*8]; |
|
| 955 |
- h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][2+2*8]; |
|
| 956 |
- }else {
|
|
| 957 |
- h->non_zero_count_cache[1+8*0]= |
|
| 958 |
- h->non_zero_count_cache[2+8*0]= |
|
| 959 |
- |
|
| 960 |
- h->non_zero_count_cache[1+8*3]= |
|
| 961 |
- h->non_zero_count_cache[2+8*3]= |
|
| 962 |
- AV_WN32A(&h->non_zero_count_cache[4+8*0], CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040); |
|
| 950 |
+ AV_COPY32(&h->non_zero_count_cache[4+8* 0], &h->non_zero_count[top_xy][4*3]); |
|
| 951 |
+ if(CHROMA444){
|
|
| 952 |
+ AV_COPY32(&h->non_zero_count_cache[4+8* 5], &h->non_zero_count[top_xy][4* 7]); |
|
| 953 |
+ AV_COPY32(&h->non_zero_count_cache[4+8*10], &h->non_zero_count[top_xy][4*11]); |
|
| 954 |
+ }else{
|
|
| 955 |
+ AV_COPY32(&h->non_zero_count_cache[4+8* 5], &h->non_zero_count[top_xy][4* 5]); |
|
| 956 |
+ AV_COPY32(&h->non_zero_count_cache[4+8*10], &h->non_zero_count[top_xy][4* 9]); |
|
| 957 |
+ } |
|
| 958 |
+ }else{
|
|
| 959 |
+ uint32_t top_empty = CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040; |
|
| 960 |
+ AV_WN32A(&h->non_zero_count_cache[4+8* 0], top_empty); |
|
| 961 |
+ AV_WN32A(&h->non_zero_count_cache[4+8* 5], top_empty); |
|
| 962 |
+ AV_WN32A(&h->non_zero_count_cache[4+8*10], top_empty); |
|
| 963 | 963 |
} |
| 964 | 964 |
|
| 965 | 965 |
for (i=0; i<2; i++) {
|
| 966 | 966 |
if(left_type[i]){
|
| 967 |
- h->non_zero_count_cache[3+8*1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]]; |
|
| 968 |
- h->non_zero_count_cache[3+8*2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]]; |
|
| 969 |
- h->non_zero_count_cache[0+8*1 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]]; |
|
| 970 |
- h->non_zero_count_cache[0+8*4 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]]; |
|
| 967 |
+ h->non_zero_count_cache[3+8* 1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]]; |
|
| 968 |
+ h->non_zero_count_cache[3+8* 2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]]; |
|
| 969 |
+ if(CHROMA444){
|
|
| 970 |
+ h->non_zero_count_cache[3+8* 6 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]+4*4]; |
|
| 971 |
+ h->non_zero_count_cache[3+8* 7 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]+4*4]; |
|
| 972 |
+ h->non_zero_count_cache[3+8*11 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]+8*4]; |
|
| 973 |
+ h->non_zero_count_cache[3+8*12 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]+8*4]; |
|
| 974 |
+ }else{
|
|
| 975 |
+ h->non_zero_count_cache[3+8* 6 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]]; |
|
| 976 |
+ h->non_zero_count_cache[3+8*11 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]]; |
|
| 977 |
+ } |
|
| 971 | 978 |
}else{
|
| 972 |
- h->non_zero_count_cache[3+8*1 + 2*8*i]= |
|
| 973 |
- h->non_zero_count_cache[3+8*2 + 2*8*i]= |
|
| 974 |
- h->non_zero_count_cache[0+8*1 + 8*i]= |
|
| 975 |
- h->non_zero_count_cache[0+8*4 + 8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64; |
|
| 979 |
+ h->non_zero_count_cache[3+8* 1 + 2*8*i]= |
|
| 980 |
+ h->non_zero_count_cache[3+8* 2 + 2*8*i]= |
|
| 981 |
+ h->non_zero_count_cache[3+8* 6 + 2*8*i]= |
|
| 982 |
+ h->non_zero_count_cache[3+8* 7 + 2*8*i]= |
|
| 983 |
+ h->non_zero_count_cache[3+8*11 + 2*8*i]= |
|
| 984 |
+ h->non_zero_count_cache[3+8*12 + 2*8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64; |
|
| 976 | 985 |
} |
| 977 | 986 |
} |
| 978 | 987 |
|
| ... | ... |
@@ -981,15 +991,15 @@ static void fill_decode_caches(H264Context *h, int mb_type){
|
| 981 | 981 |
if(top_type) {
|
| 982 | 982 |
h->top_cbp = h->cbp_table[top_xy]; |
| 983 | 983 |
} else {
|
| 984 |
- h->top_cbp = IS_INTRA(mb_type) ? 0x1CF : 0x00F; |
|
| 984 |
+ h->top_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F; |
|
| 985 | 985 |
} |
| 986 | 986 |
// left_cbp |
| 987 | 987 |
if (left_type[0]) {
|
| 988 |
- h->left_cbp = (h->cbp_table[left_xy[0]] & 0x1f0) |
|
| 988 |
+ h->left_cbp = (h->cbp_table[left_xy[0]] & 0x7F0) |
|
| 989 | 989 |
| ((h->cbp_table[left_xy[0]]>>(left_block[0]&(~1)))&2) |
| 990 | 990 |
| (((h->cbp_table[left_xy[1]]>>(left_block[2]&(~1)))&2) << 2); |
| 991 | 991 |
} else {
|
| 992 |
- h->left_cbp = IS_INTRA(mb_type) ? 0x1CF : 0x00F; |
|
| 992 |
+ h->left_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F; |
|
| 993 | 993 |
} |
| 994 | 994 |
} |
| 995 | 995 |
} |
| ... | ... |
@@ -1190,11 +1200,21 @@ static inline int pred_intra_mode(H264Context *h, int n){
|
| 1190 | 1190 |
static inline void write_back_non_zero_count(H264Context *h){
|
| 1191 | 1191 |
const int mb_xy= h->mb_xy; |
| 1192 | 1192 |
|
| 1193 |
- AV_COPY64(&h->non_zero_count[mb_xy][ 0], &h->non_zero_count_cache[0+8*1]); |
|
| 1194 |
- AV_COPY64(&h->non_zero_count[mb_xy][ 8], &h->non_zero_count_cache[0+8*2]); |
|
| 1195 |
- AV_COPY32(&h->non_zero_count[mb_xy][16], &h->non_zero_count_cache[0+8*5]); |
|
| 1196 |
- AV_COPY32(&h->non_zero_count[mb_xy][20], &h->non_zero_count_cache[4+8*3]); |
|
| 1197 |
- AV_COPY64(&h->non_zero_count[mb_xy][24], &h->non_zero_count_cache[0+8*4]); |
|
| 1193 |
+ AV_COPY32(&h->non_zero_count[mb_xy][ 0], &h->non_zero_count_cache[4+8* 1]); |
|
| 1194 |
+ AV_COPY32(&h->non_zero_count[mb_xy][ 4], &h->non_zero_count_cache[4+8* 2]); |
|
| 1195 |
+ AV_COPY32(&h->non_zero_count[mb_xy][ 8], &h->non_zero_count_cache[4+8* 3]); |
|
| 1196 |
+ AV_COPY32(&h->non_zero_count[mb_xy][12], &h->non_zero_count_cache[4+8* 4]); |
|
| 1197 |
+ AV_COPY32(&h->non_zero_count[mb_xy][16], &h->non_zero_count_cache[4+8* 6]); |
|
| 1198 |
+ AV_COPY32(&h->non_zero_count[mb_xy][20], &h->non_zero_count_cache[4+8* 7]); |
|
| 1199 |
+ AV_COPY32(&h->non_zero_count[mb_xy][32], &h->non_zero_count_cache[4+8*11]); |
|
| 1200 |
+ AV_COPY32(&h->non_zero_count[mb_xy][36], &h->non_zero_count_cache[4+8*12]); |
|
| 1201 |
+ |
|
| 1202 |
+ if(CHROMA444){
|
|
| 1203 |
+ AV_COPY32(&h->non_zero_count[mb_xy][24], &h->non_zero_count_cache[4+8* 8]); |
|
| 1204 |
+ AV_COPY32(&h->non_zero_count[mb_xy][28], &h->non_zero_count_cache[4+8* 9]); |
|
| 1205 |
+ AV_COPY32(&h->non_zero_count[mb_xy][40], &h->non_zero_count_cache[4+8*13]); |
|
| 1206 |
+ AV_COPY32(&h->non_zero_count[mb_xy][44], &h->non_zero_count_cache[4+8*14]); |
|
| 1207 |
+ } |
|
| 1198 | 1208 |
} |
| 1199 | 1209 |
|
| 1200 | 1210 |
static inline void write_back_motion(H264Context *h, int mb_type){
|
| ... | ... |
@@ -1267,8 +1287,7 @@ static void av_unused decode_mb_skip(H264Context *h){
|
| 1267 | 1267 |
const int mb_xy= h->mb_xy; |
| 1268 | 1268 |
int mb_type=0; |
| 1269 | 1269 |
|
| 1270 |
- memset(h->non_zero_count[mb_xy], 0, 32); |
|
| 1271 |
- memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui |
|
| 1270 |
+ memset(h->non_zero_count[mb_xy], 0, 48); |
|
| 1272 | 1271 |
|
| 1273 | 1272 |
if(MB_FIELD) |
| 1274 | 1273 |
mb_type|= MB_TYPE_INTERLACED; |
| ... | ... |
@@ -45,7 +45,7 @@ |
| 45 | 45 |
|
| 46 | 46 |
/* Cabac pre state table */ |
| 47 | 47 |
|
| 48 |
-static const int8_t cabac_context_init_I[460][2] = |
|
| 48 |
+static const int8_t cabac_context_init_I[1024][2] = |
|
| 49 | 49 |
{
|
| 50 | 50 |
/* 0 - 10 */ |
| 51 | 51 |
{ 20, -15 }, { 2, 54 }, { 3, 74 }, { 20, -15 },
|
| ... | ... |
@@ -211,10 +211,153 @@ static const int8_t cabac_context_init_I[460][2] = |
| 211 | 211 |
{ -10, 73 }, { -10, 70 }, { -10, 69 }, { -5, 66 },
|
| 212 | 212 |
{ -9, 64 }, { -5, 58 }, { 2, 59 }, { 21, -10 },
|
| 213 | 213 |
{ 24, -11 }, { 28, -8 }, { 28, -1 }, { 29, 3 },
|
| 214 |
- { 29, 9 }, { 35, 20 }, { 29, 36 }, { 14, 67 }
|
|
| 214 |
+ { 29, 9 }, { 35, 20 }, { 29, 36 }, { 14, 67 },
|
|
| 215 |
+ |
|
| 216 |
+ /* 460 -> 1024 */ |
|
| 217 |
+ { -17, 123 }, { -12, 115 }, { -16, 122 }, { -11, 115 },
|
|
| 218 |
+ { -12, 63 }, { -2, 68 }, { -15, 84 }, { -13, 104 },
|
|
| 219 |
+ { -3, 70 }, { -8, 93 }, { -10, 90 }, { -30, 127 },
|
|
| 220 |
+ { -17, 123 }, { -12, 115 }, { -16, 122 }, { -11, 115 },
|
|
| 221 |
+ { -12, 63 }, { -2, 68 }, { -15, 84 }, { -13, 104 },
|
|
| 222 |
+ { -3, 70 }, { -8, 93 }, { -10, 90 }, { -30, 127 },
|
|
| 223 |
+ { -7, 93 }, { -11, 87 }, { -3, 77 }, { -5, 71 },
|
|
| 224 |
+ { -4, 63 }, { -4, 68 }, { -12, 84 }, { -7, 62 },
|
|
| 225 |
+ { -7, 65 }, { 8, 61 }, { 5, 56 }, { -2, 66 },
|
|
| 226 |
+ { 1, 64 }, { 0, 61 }, { -2, 78 }, { 1, 50 },
|
|
| 227 |
+ { 7, 52 }, { 10, 35 }, { 0, 44 }, { 11, 38 },
|
|
| 228 |
+ { 1, 45 }, { 0, 46 }, { 5, 44 }, { 31, 17 },
|
|
| 229 |
+ { 1, 51 }, { 7, 50 }, { 28, 19 }, { 16, 33 },
|
|
| 230 |
+ { 14, 62 }, { -13, 108 }, { -15, 100 }, { -13, 101 },
|
|
| 231 |
+ { -13, 91 }, { -12, 94 }, { -10, 88 }, { -16, 84 },
|
|
| 232 |
+ { -10, 86 }, { -7, 83 }, { -13, 87 }, { -19, 94 },
|
|
| 233 |
+ { 1, 70 }, { 0, 72 }, { -5, 74 }, { 18, 59 },
|
|
| 234 |
+ { -7, 93 }, { -11, 87 }, { -3, 77 }, { -5, 71 },
|
|
| 235 |
+ { -4, 63 }, { -4, 68 }, { -12, 84 }, { -7, 62 },
|
|
| 236 |
+ { -7, 65 }, { 8, 61 }, { 5, 56 }, { -2, 66 },
|
|
| 237 |
+ { 1, 64 }, { 0, 61 }, { -2, 78 }, { 1, 50 },
|
|
| 238 |
+ { 7, 52 }, { 10, 35 }, { 0, 44 }, { 11, 38 },
|
|
| 239 |
+ { 1, 45 }, { 0, 46 }, { 5, 44 }, { 31, 17 },
|
|
| 240 |
+ { 1, 51 }, { 7, 50 }, { 28, 19 }, { 16, 33 },
|
|
| 241 |
+ { 14, 62 }, { -13, 108 }, { -15, 100 }, { -13, 101 },
|
|
| 242 |
+ { -13, 91 }, { -12, 94 }, { -10, 88 }, { -16, 84 },
|
|
| 243 |
+ { -10, 86 }, { -7, 83 }, { -13, 87 }, { -19, 94 },
|
|
| 244 |
+ { 1, 70 }, { 0, 72 }, { -5, 74 }, { 18, 59 },
|
|
| 245 |
+ { 24, 0 }, { 15, 9 }, { 8, 25 }, { 13, 18 },
|
|
| 246 |
+ { 15, 9 }, { 13, 19 }, { 10, 37 }, { 12, 18 },
|
|
| 247 |
+ { 6, 29 }, { 20, 33 }, { 15, 30 }, { 4, 45 },
|
|
| 248 |
+ { 1, 58 }, { 0, 62 }, { 7, 61 }, { 12, 38 },
|
|
| 249 |
+ { 11, 45 }, { 15, 39 }, { 11, 42 }, { 13, 44 },
|
|
| 250 |
+ { 16, 45 }, { 12, 41 }, { 10, 49 }, { 30, 34 },
|
|
| 251 |
+ { 18, 42 }, { 10, 55 }, { 17, 51 }, { 17, 46 },
|
|
| 252 |
+ { 0, 89 }, { 26, -19 }, { 22, -17 }, { 26, -17 },
|
|
| 253 |
+ { 30, -25 }, { 28, -20 }, { 33, -23 }, { 37, -27 },
|
|
| 254 |
+ { 33, -23 }, { 40, -28 }, { 38, -17 }, { 33, -11 },
|
|
| 255 |
+ { 40, -15 }, { 41, -6 }, { 38, 1 }, { 41, 17 },
|
|
| 256 |
+ { 24, 0 }, { 15, 9 }, { 8, 25 }, { 13, 18 },
|
|
| 257 |
+ { 15, 9 }, { 13, 19 }, { 10, 37 }, { 12, 18 },
|
|
| 258 |
+ { 6, 29 }, { 20, 33 }, { 15, 30 }, { 4, 45 },
|
|
| 259 |
+ { 1, 58 }, { 0, 62 }, { 7, 61 }, { 12, 38 },
|
|
| 260 |
+ { 11, 45 }, { 15, 39 }, { 11, 42 }, { 13, 44 },
|
|
| 261 |
+ { 16, 45 }, { 12, 41 }, { 10, 49 }, { 30, 34 },
|
|
| 262 |
+ { 18, 42 }, { 10, 55 }, { 17, 51 }, { 17, 46 },
|
|
| 263 |
+ { 0, 89 }, { 26, -19 }, { 22, -17 }, { 26, -17 },
|
|
| 264 |
+ { 30, -25 }, { 28, -20 }, { 33, -23 }, { 37, -27 },
|
|
| 265 |
+ { 33, -23 }, { 40, -28 }, { 38, -17 }, { 33, -11 },
|
|
| 266 |
+ { 40, -15 }, { 41, -6 }, { 38, 1 }, { 41, 17 },
|
|
| 267 |
+ { -17, 120 }, { -20, 112 }, { -18, 114 }, { -11, 85 },
|
|
| 268 |
+ { -15, 92 }, { -14, 89 }, { -26, 71 }, { -15, 81 },
|
|
| 269 |
+ { -14, 80 }, { 0, 68 }, { -14, 70 }, { -24, 56 },
|
|
| 270 |
+ { -23, 68 }, { -24, 50 }, { -11, 74 }, { -14, 106 },
|
|
| 271 |
+ { -13, 97 }, { -15, 90 }, { -12, 90 }, { -18, 88 },
|
|
| 272 |
+ { -10, 73 }, { -9, 79 }, { -14, 86 }, { -10, 73 },
|
|
| 273 |
+ { -10, 70 }, { -10, 69 }, { -5, 66 }, { -9, 64 },
|
|
| 274 |
+ { -5, 58 }, { 2, 59 }, { 23, -13 }, { 26, -13 },
|
|
| 275 |
+ { 40, -15 }, { 49, -14 }, { 44, 3 }, { 45, 6 },
|
|
| 276 |
+ { 44, 34 }, { 33, 54 }, { 19, 82 }, { 21, -10 },
|
|
| 277 |
+ { 24, -11 }, { 28, -8 }, { 28, -1 }, { 29, 3 },
|
|
| 278 |
+ { 29, 9 }, { 35, 20 }, { 29, 36 }, { 14, 67 },
|
|
| 279 |
+ { -3, 75 }, { -1, 23 }, { 1, 34 }, { 1, 43 },
|
|
| 280 |
+ { 0, 54 }, { -2, 55 }, { 0, 61 }, { 1, 64 },
|
|
| 281 |
+ { 0, 68 }, { -9, 92 }, { -17, 120 }, { -20, 112 },
|
|
| 282 |
+ { -18, 114 }, { -11, 85 }, { -15, 92 }, { -14, 89 },
|
|
| 283 |
+ { -26, 71 }, { -15, 81 }, { -14, 80 }, { 0, 68 },
|
|
| 284 |
+ { -14, 70 }, { -24, 56 }, { -23, 68 }, { -24, 50 },
|
|
| 285 |
+ { -11, 74 }, { -14, 106 }, { -13, 97 }, { -15, 90 },
|
|
| 286 |
+ { -12, 90 }, { -18, 88 }, { -10, 73 }, { -9, 79 },
|
|
| 287 |
+ { -14, 86 }, { -10, 73 }, { -10, 70 }, { -10, 69 },
|
|
| 288 |
+ { -5, 66 }, { -9, 64 }, { -5, 58 }, { 2, 59 },
|
|
| 289 |
+ { 23, -13 }, { 26, -13 }, { 40, -15 }, { 49, -14 },
|
|
| 290 |
+ { 44, 3 }, { 45, 6 }, { 44, 34 }, { 33, 54 },
|
|
| 291 |
+ { 19, 82 }, { 21, -10 }, { 24, -11 }, { 28, -8 },
|
|
| 292 |
+ { 28, -1 }, { 29, 3 }, { 29, 9 }, { 35, 20 },
|
|
| 293 |
+ { 29, 36 }, { 14, 67 }, { -3, 75 }, { -1, 23 },
|
|
| 294 |
+ { 1, 34 }, { 1, 43 }, { 0, 54 }, { -2, 55 },
|
|
| 295 |
+ { 0, 61 }, { 1, 64 }, { 0, 68 }, { -9, 92 },
|
|
| 296 |
+ { -6, 93 }, { -6, 84 }, { -8, 79 }, { 0, 66 },
|
|
| 297 |
+ { -1, 71 }, { 0, 62 }, { -2, 60 }, { -2, 59 },
|
|
| 298 |
+ { -5, 75 }, { -3, 62 }, { -4, 58 }, { -9, 66 },
|
|
| 299 |
+ { -1, 79 }, { 0, 71 }, { 3, 68 }, { 10, 44 },
|
|
| 300 |
+ { -7, 62 }, { 15, 36 }, { 14, 40 }, { 16, 27 },
|
|
| 301 |
+ { 12, 29 }, { 1, 44 }, { 20, 36 }, { 18, 32 },
|
|
| 302 |
+ { 5, 42 }, { 1, 48 }, { 10, 62 }, { 17, 46 },
|
|
| 303 |
+ { 9, 64 }, { -12, 104 }, { -11, 97 }, { -16, 96 },
|
|
| 304 |
+ { -7, 88 }, { -8, 85 }, { -7, 85 }, { -9, 85 },
|
|
| 305 |
+ { -13, 88 }, { 4, 66 }, { -3, 77 }, { -3, 76 },
|
|
| 306 |
+ { -6, 76 }, { 10, 58 }, { -1, 76 }, { -1, 83 },
|
|
| 307 |
+ { -6, 93 }, { -6, 84 }, { -8, 79 }, { 0, 66 },
|
|
| 308 |
+ { -1, 71 }, { 0, 62 }, { -2, 60 }, { -2, 59 },
|
|
| 309 |
+ { -5, 75 }, { -3, 62 }, { -4, 58 }, { -9, 66 },
|
|
| 310 |
+ { -1, 79 }, { 0, 71 }, { 3, 68 }, { 10, 44 },
|
|
| 311 |
+ { -7, 62 }, { 15, 36 }, { 14, 40 }, { 16, 27 },
|
|
| 312 |
+ { 12, 29 }, { 1, 44 }, { 20, 36 }, { 18, 32 },
|
|
| 313 |
+ { 5, 42 }, { 1, 48 }, { 10, 62 }, { 17, 46 },
|
|
| 314 |
+ { 9, 64 }, { -12, 104 }, { -11, 97 }, { -16, 96 },
|
|
| 315 |
+ { -7, 88 }, { -8, 85 }, { -7, 85 }, { -9, 85 },
|
|
| 316 |
+ { -13, 88 }, { 4, 66 }, { -3, 77 }, { -3, 76 },
|
|
| 317 |
+ { -6, 76 }, { 10, 58 }, { -1, 76 }, { -1, 83 },
|
|
| 318 |
+ { 15, 6 }, { 6, 19 }, { 7, 16 }, { 12, 14 },
|
|
| 319 |
+ { 18, 13 }, { 13, 11 }, { 13, 15 }, { 15, 16 },
|
|
| 320 |
+ { 12, 23 }, { 13, 23 }, { 15, 20 }, { 14, 26 },
|
|
| 321 |
+ { 14, 44 }, { 17, 40 }, { 17, 47 }, { 24, 17 },
|
|
| 322 |
+ { 21, 21 }, { 25, 22 }, { 31, 27 }, { 22, 29 },
|
|
| 323 |
+ { 19, 35 }, { 14, 50 }, { 10, 57 }, { 7, 63 },
|
|
| 324 |
+ { -2, 77 }, { -4, 82 }, { -3, 94 }, { 9, 69 },
|
|
| 325 |
+ { -12, 109 }, { 36, -35 }, { 36, -34 }, { 32, -26 },
|
|
| 326 |
+ { 37, -30 }, { 44, -32 }, { 34, -18 }, { 34, -15 },
|
|
| 327 |
+ { 40, -15 }, { 33, -7 }, { 35, -5 }, { 33, 0 },
|
|
| 328 |
+ { 38, 2 }, { 33, 13 }, { 23, 35 }, { 13, 58 },
|
|
| 329 |
+ { 15, 6 }, { 6, 19 }, { 7, 16 }, { 12, 14 },
|
|
| 330 |
+ { 18, 13 }, { 13, 11 }, { 13, 15 }, { 15, 16 },
|
|
| 331 |
+ { 12, 23 }, { 13, 23 }, { 15, 20 }, { 14, 26 },
|
|
| 332 |
+ { 14, 44 }, { 17, 40 }, { 17, 47 }, { 24, 17 },
|
|
| 333 |
+ { 21, 21 }, { 25, 22 }, { 31, 27 }, { 22, 29 },
|
|
| 334 |
+ { 19, 35 }, { 14, 50 }, { 10, 57 }, { 7, 63 },
|
|
| 335 |
+ { -2, 77 }, { -4, 82 }, { -3, 94 }, { 9, 69 },
|
|
| 336 |
+ { -12, 109 }, { 36, -35 }, { 36, -34 }, { 32, -26 },
|
|
| 337 |
+ { 37, -30 }, { 44, -32 }, { 34, -18 }, { 34, -15 },
|
|
| 338 |
+ { 40, -15 }, { 33, -7 }, { 35, -5 }, { 33, 0 },
|
|
| 339 |
+ { 38, 2 }, { 33, 13 }, { 23, 35 }, { 13, 58 },
|
|
| 340 |
+ { -3, 71 }, { -6, 42 }, { -5, 50 }, { -3, 54 },
|
|
| 341 |
+ { -2, 62 }, { 0, 58 }, { 1, 63 }, { -2, 72 },
|
|
| 342 |
+ { -1, 74 }, { -9, 91 }, { -5, 67 }, { -5, 27 },
|
|
| 343 |
+ { -3, 39 }, { -2, 44 }, { 0, 46 }, { -16, 64 },
|
|
| 344 |
+ { -8, 68 }, { -10, 78 }, { -6, 77 }, { -10, 86 },
|
|
| 345 |
+ { -12, 92 }, { -15, 55 }, { -10, 60 }, { -6, 62 },
|
|
| 346 |
+ { -4, 65 }, { -12, 73 }, { -8, 76 }, { -7, 80 },
|
|
| 347 |
+ { -9, 88 }, { -17, 110 }, { -3, 71 }, { -6, 42 },
|
|
| 348 |
+ { -5, 50 }, { -3, 54 }, { -2, 62 }, { 0, 58 },
|
|
| 349 |
+ { 1, 63 }, { -2, 72 }, { -1, 74 }, { -9, 91 },
|
|
| 350 |
+ { -5, 67 }, { -5, 27 }, { -3, 39 }, { -2, 44 },
|
|
| 351 |
+ { 0, 46 }, { -16, 64 }, { -8, 68 }, { -10, 78 },
|
|
| 352 |
+ { -6, 77 }, { -10, 86 }, { -12, 92 }, { -15, 55 },
|
|
| 353 |
+ { -10, 60 }, { -6, 62 }, { -4, 65 }, { -12, 73 },
|
|
| 354 |
+ { -8, 76 }, { -7, 80 }, { -9, 88 }, { -17, 110 },
|
|
| 355 |
+ { -3, 70 }, { -8, 93 }, { -10, 90 }, { -30, 127 },
|
|
| 356 |
+ { -3, 70 }, { -8, 93 }, { -10, 90 }, { -30, 127 },
|
|
| 357 |
+ { -3, 70 }, { -8, 93 }, { -10, 90 }, { -30, 127 }
|
|
| 215 | 358 |
}; |
| 216 | 359 |
|
| 217 |
-static const int8_t cabac_context_init_PB[3][460][2] = |
|
| 360 |
+static const int8_t cabac_context_init_PB[3][1024][2] = |
|
| 218 | 361 |
{
|
| 219 | 362 |
/* i_cabac_init_idc == 0 */ |
| 220 | 363 |
{
|
| ... | ... |
@@ -370,6 +513,149 @@ static const int8_t cabac_context_init_PB[3][460][2] = |
| 370 | 370 |
{ -14, 66 }, { 0, 59 }, { 2, 59 }, { 21, -13 },
|
| 371 | 371 |
{ 33, -14 }, { 39, -7 }, { 46, -2 }, { 51, 2 },
|
| 372 | 372 |
{ 60, 6 }, { 61, 17 }, { 55, 34 }, { 42, 62 },
|
| 373 |
+ |
|
| 374 |
+ /* 460 - 1024 */ |
|
| 375 |
+ { -7, 92 }, { -5, 89 }, { -7, 96 }, { -13, 108 },
|
|
| 376 |
+ { -3, 46 }, { -1, 65 }, { -1, 57 }, { -9, 93 },
|
|
| 377 |
+ { -3, 74 }, { -9, 92 }, { -8, 87 }, { -23, 126 },
|
|
| 378 |
+ { -7, 92 }, { -5, 89 }, { -7, 96 }, { -13, 108 },
|
|
| 379 |
+ { -3, 46 }, { -1, 65 }, { -1, 57 }, { -9, 93 },
|
|
| 380 |
+ { -3, 74 }, { -9, 92 }, { -8, 87 }, { -23, 126 },
|
|
| 381 |
+ { -2, 85 }, { -6, 78 }, { -1, 75 }, { -7, 77 },
|
|
| 382 |
+ { 2, 54 }, { 5, 50 }, { -3, 68 }, { 1, 50 },
|
|
| 383 |
+ { 6, 42 }, { -4, 81 }, { 1, 63 }, { -4, 70 },
|
|
| 384 |
+ { 0, 67 }, { 2, 57 }, { -2, 76 }, { 11, 35 },
|
|
| 385 |
+ { 4, 64 }, { 1, 61 }, { 11, 35 }, { 18, 25 },
|
|
| 386 |
+ { 12, 24 }, { 13, 29 }, { 13, 36 }, { -10, 93 },
|
|
| 387 |
+ { -7, 73 }, { -2, 73 }, { 13, 46 }, { 9, 49 },
|
|
| 388 |
+ { -7, 100 }, { 9, 53 }, { 2, 53 }, { 5, 53 },
|
|
| 389 |
+ { -2, 61 }, { 0, 56 }, { 0, 56 }, { -13, 63 },
|
|
| 390 |
+ { -5, 60 }, { -1, 62 }, { 4, 57 }, { -6, 69 },
|
|
| 391 |
+ { 4, 57 }, { 14, 39 }, { 4, 51 }, { 13, 68 },
|
|
| 392 |
+ { -2, 85 }, { -6, 78 }, { -1, 75 }, { -7, 77 },
|
|
| 393 |
+ { 2, 54 }, { 5, 50 }, { -3, 68 }, { 1, 50 },
|
|
| 394 |
+ { 6, 42 }, { -4, 81 }, { 1, 63 }, { -4, 70 },
|
|
| 395 |
+ { 0, 67 }, { 2, 57 }, { -2, 76 }, { 11, 35 },
|
|
| 396 |
+ { 4, 64 }, { 1, 61 }, { 11, 35 }, { 18, 25 },
|
|
| 397 |
+ { 12, 24 }, { 13, 29 }, { 13, 36 }, { -10, 93 },
|
|
| 398 |
+ { -7, 73 }, { -2, 73 }, { 13, 46 }, { 9, 49 },
|
|
| 399 |
+ { -7, 100 }, { 9, 53 }, { 2, 53 }, { 5, 53 },
|
|
| 400 |
+ { -2, 61 }, { 0, 56 }, { 0, 56 }, { -13, 63 },
|
|
| 401 |
+ { -5, 60 }, { -1, 62 }, { 4, 57 }, { -6, 69 },
|
|
| 402 |
+ { 4, 57 }, { 14, 39 }, { 4, 51 }, { 13, 68 },
|
|
| 403 |
+ { 11, 28 }, { 2, 40 }, { 3, 44 }, { 0, 49 },
|
|
| 404 |
+ { 0, 46 }, { 2, 44 }, { 2, 51 }, { 0, 47 },
|
|
| 405 |
+ { 4, 39 }, { 2, 62 }, { 6, 46 }, { 0, 54 },
|
|
| 406 |
+ { 3, 54 }, { 2, 58 }, { 4, 63 }, { 6, 51 },
|
|
| 407 |
+ { 6, 57 }, { 7, 53 }, { 6, 52 }, { 6, 55 },
|
|
| 408 |
+ { 11, 45 }, { 14, 36 }, { 8, 53 }, { -1, 82 },
|
|
| 409 |
+ { 7, 55 }, { -3, 78 }, { 15, 46 }, { 22, 31 },
|
|
| 410 |
+ { -1, 84 }, { 25, 7 }, { 30, -7 }, { 28, 3 },
|
|
| 411 |
+ { 28, 4 }, { 32, 0 }, { 34, -1 }, { 30, 6 },
|
|
| 412 |
+ { 30, 6 }, { 32, 9 }, { 31, 19 }, { 26, 27 },
|
|
| 413 |
+ { 26, 30 }, { 37, 20 }, { 28, 34 }, { 17, 70 },
|
|
| 414 |
+ { 11, 28 }, { 2, 40 }, { 3, 44 }, { 0, 49 },
|
|
| 415 |
+ { 0, 46 }, { 2, 44 }, { 2, 51 }, { 0, 47 },
|
|
| 416 |
+ { 4, 39 }, { 2, 62 }, { 6, 46 }, { 0, 54 },
|
|
| 417 |
+ { 3, 54 }, { 2, 58 }, { 4, 63 }, { 6, 51 },
|
|
| 418 |
+ { 6, 57 }, { 7, 53 }, { 6, 52 }, { 6, 55 },
|
|
| 419 |
+ { 11, 45 }, { 14, 36 }, { 8, 53 }, { -1, 82 },
|
|
| 420 |
+ { 7, 55 }, { -3, 78 }, { 15, 46 }, { 22, 31 },
|
|
| 421 |
+ { -1, 84 }, { 25, 7 }, { 30, -7 }, { 28, 3 },
|
|
| 422 |
+ { 28, 4 }, { 32, 0 }, { 34, -1 }, { 30, 6 },
|
|
| 423 |
+ { 30, 6 }, { 32, 9 }, { 31, 19 }, { 26, 27 },
|
|
| 424 |
+ { 26, 30 }, { 37, 20 }, { 28, 34 }, { 17, 70 },
|
|
| 425 |
+ { -4, 79 }, { -7, 71 }, { -5, 69 }, { -9, 70 },
|
|
| 426 |
+ { -8, 66 }, { -10, 68 }, { -19, 73 }, { -12, 69 },
|
|
| 427 |
+ { -16, 70 }, { -15, 67 }, { -20, 62 }, { -19, 70 },
|
|
| 428 |
+ { -16, 66 }, { -22, 65 }, { -20, 63 }, { -5, 85 },
|
|
| 429 |
+ { -6, 81 }, { -10, 77 }, { -7, 81 }, { -17, 80 },
|
|
| 430 |
+ { -18, 73 }, { -4, 74 }, { -10, 83 }, { -9, 71 },
|
|
| 431 |
+ { -9, 67 }, { -1, 61 }, { -8, 66 }, { -14, 66 },
|
|
| 432 |
+ { 0, 59 }, { 2, 59 }, { 9, -2 }, { 26, -9 },
|
|
| 433 |
+ { 33, -9 }, { 39, -7 }, { 41, -2 }, { 45, 3 },
|
|
| 434 |
+ { 49, 9 }, { 45, 27 }, { 36, 59 }, { 21, -13 },
|
|
| 435 |
+ { 33, -14 }, { 39, -7 }, { 46, -2 }, { 51, 2 },
|
|
| 436 |
+ { 60, 6 }, { 61, 17 }, { 55, 34 }, { 42, 62 },
|
|
| 437 |
+ { -6, 66 }, { -7, 35 }, { -7, 42 }, { -8, 45 },
|
|
| 438 |
+ { -5, 48 }, { -12, 56 }, { -6, 60 }, { -5, 62 },
|
|
| 439 |
+ { -8, 66 }, { -8, 76 }, { -4, 79 }, { -7, 71 },
|
|
| 440 |
+ { -5, 69 }, { -9, 70 }, { -8, 66 }, { -10, 68 },
|
|
| 441 |
+ { -19, 73 }, { -12, 69 }, { -16, 70 }, { -15, 67 },
|
|
| 442 |
+ { -20, 62 }, { -19, 70 }, { -16, 66 }, { -22, 65 },
|
|
| 443 |
+ { -20, 63 }, { -5, 85 }, { -6, 81 }, { -10, 77 },
|
|
| 444 |
+ { -7, 81 }, { -17, 80 }, { -18, 73 }, { -4, 74 },
|
|
| 445 |
+ { -10, 83 }, { -9, 71 }, { -9, 67 }, { -1, 61 },
|
|
| 446 |
+ { -8, 66 }, { -14, 66 }, { 0, 59 }, { 2, 59 },
|
|
| 447 |
+ { 9, -2 }, { 26, -9 }, { 33, -9 }, { 39, -7 },
|
|
| 448 |
+ { 41, -2 }, { 45, 3 }, { 49, 9 }, { 45, 27 },
|
|
| 449 |
+ { 36, 59 }, { 21, -13 }, { 33, -14 }, { 39, -7 },
|
|
| 450 |
+ { 46, -2 }, { 51, 2 }, { 60, 6 }, { 61, 17 },
|
|
| 451 |
+ { 55, 34 }, { 42, 62 }, { -6, 66 }, { -7, 35 },
|
|
| 452 |
+ { -7, 42 }, { -8, 45 }, { -5, 48 }, { -12, 56 },
|
|
| 453 |
+ { -6, 60 }, { -5, 62 }, { -8, 66 }, { -8, 76 },
|
|
| 454 |
+ { -13, 106 }, { -16, 106 }, { -10, 87 }, { -21, 114 },
|
|
| 455 |
+ { -18, 110 }, { -14, 98 }, { -22, 110 }, { -21, 106 },
|
|
| 456 |
+ { -18, 103 }, { -21, 107 }, { -23, 108 }, { -26, 112 },
|
|
| 457 |
+ { -10, 96 }, { -12, 95 }, { -5, 91 }, { -9, 93 },
|
|
| 458 |
+ { -22, 94 }, { -5, 86 }, { 9, 67 }, { -4, 80 },
|
|
| 459 |
+ { -10, 85 }, { -1, 70 }, { 7, 60 }, { 9, 58 },
|
|
| 460 |
+ { 5, 61 }, { 12, 50 }, { 15, 50 }, { 18, 49 },
|
|
| 461 |
+ { 17, 54 }, { 10, 41 }, { 7, 46 }, { -1, 51 },
|
|
| 462 |
+ { 7, 49 }, { 8, 52 }, { 9, 41 }, { 6, 47 },
|
|
| 463 |
+ { 2, 55 }, { 13, 41 }, { 10, 44 }, { 6, 50 },
|
|
| 464 |
+ { 5, 53 }, { 13, 49 }, { 4, 63 }, { 6, 64 },
|
|
| 465 |
+ { -13, 106 }, { -16, 106 }, { -10, 87 }, { -21, 114 },
|
|
| 466 |
+ { -18, 110 }, { -14, 98 }, { -22, 110 }, { -21, 106 },
|
|
| 467 |
+ { -18, 103 }, { -21, 107 }, { -23, 108 }, { -26, 112 },
|
|
| 468 |
+ { -10, 96 }, { -12, 95 }, { -5, 91 }, { -9, 93 },
|
|
| 469 |
+ { -22, 94 }, { -5, 86 }, { 9, 67 }, { -4, 80 },
|
|
| 470 |
+ { -10, 85 }, { -1, 70 }, { 7, 60 }, { 9, 58 },
|
|
| 471 |
+ { 5, 61 }, { 12, 50 }, { 15, 50 }, { 18, 49 },
|
|
| 472 |
+ { 17, 54 }, { 10, 41 }, { 7, 46 }, { -1, 51 },
|
|
| 473 |
+ { 7, 49 }, { 8, 52 }, { 9, 41 }, { 6, 47 },
|
|
| 474 |
+ { 2, 55 }, { 13, 41 }, { 10, 44 }, { 6, 50 },
|
|
| 475 |
+ { 5, 53 }, { 13, 49 }, { 4, 63 }, { 6, 64 },
|
|
| 476 |
+ { 14, 11 }, { 11, 14 }, { 9, 11 }, { 18, 11 },
|
|
| 477 |
+ { 21, 9 }, { 23, -2 }, { 32, -15 }, { 32, -15 },
|
|
| 478 |
+ { 34, -21 }, { 39, -23 }, { 42, -33 }, { 41, -31 },
|
|
| 479 |
+ { 46, -28 }, { 38, -12 }, { 21, 29 }, { 45, -24 },
|
|
| 480 |
+ { 53, -45 }, { 48, -26 }, { 65, -43 }, { 43, -19 },
|
|
| 481 |
+ { 39, -10 }, { 30, 9 }, { 18, 26 }, { 20, 27 },
|
|
| 482 |
+ { 0, 57 }, { -14, 82 }, { -5, 75 }, { -19, 97 },
|
|
| 483 |
+ { -35, 125 }, { 27, 0 }, { 28, 0 }, { 31, -4 },
|
|
| 484 |
+ { 27, 6 }, { 34, 8 }, { 30, 10 }, { 24, 22 },
|
|
| 485 |
+ { 33, 19 }, { 22, 32 }, { 26, 31 }, { 21, 41 },
|
|
| 486 |
+ { 26, 44 }, { 23, 47 }, { 16, 65 }, { 14, 71 },
|
|
| 487 |
+ { 14, 11 }, { 11, 14 }, { 9, 11 }, { 18, 11 },
|
|
| 488 |
+ { 21, 9 }, { 23, -2 }, { 32, -15 }, { 32, -15 },
|
|
| 489 |
+ { 34, -21 }, { 39, -23 }, { 42, -33 }, { 41, -31 },
|
|
| 490 |
+ { 46, -28 }, { 38, -12 }, { 21, 29 }, { 45, -24 },
|
|
| 491 |
+ { 53, -45 }, { 48, -26 }, { 65, -43 }, { 43, -19 },
|
|
| 492 |
+ { 39, -10 }, { 30, 9 }, { 18, 26 }, { 20, 27 },
|
|
| 493 |
+ { 0, 57 }, { -14, 82 }, { -5, 75 }, { -19, 97 },
|
|
| 494 |
+ { -35, 125 }, { 27, 0 }, { 28, 0 }, { 31, -4 },
|
|
| 495 |
+ { 27, 6 }, { 34, 8 }, { 30, 10 }, { 24, 22 },
|
|
| 496 |
+ { 33, 19 }, { 22, 32 }, { 26, 31 }, { 21, 41 },
|
|
| 497 |
+ { 26, 44 }, { 23, 47 }, { 16, 65 }, { 14, 71 },
|
|
| 498 |
+ { -6, 76 }, { -2, 44 }, { 0, 45 }, { 0, 52 },
|
|
| 499 |
+ { -3, 64 }, { -2, 59 }, { -4, 70 }, { -4, 75 },
|
|
| 500 |
+ { -8, 82 }, { -17, 102 }, { -9, 77 }, { 3, 24 },
|
|
| 501 |
+ { 0, 42 }, { 0, 48 }, { 0, 55 }, { -6, 59 },
|
|
| 502 |
+ { -7, 71 }, { -12, 83 }, { -11, 87 }, { -30, 119 },
|
|
| 503 |
+ { 1, 58 }, { -3, 29 }, { -1, 36 }, { 1, 38 },
|
|
| 504 |
+ { 2, 43 }, { -6, 55 }, { 0, 58 }, { 0, 64 },
|
|
| 505 |
+ { -3, 74 }, { -10, 90 }, { -6, 76 }, { -2, 44 },
|
|
| 506 |
+ { 0, 45 }, { 0, 52 }, { -3, 64 }, { -2, 59 },
|
|
| 507 |
+ { -4, 70 }, { -4, 75 }, { -8, 82 }, { -17, 102 },
|
|
| 508 |
+ { -9, 77 }, { 3, 24 }, { 0, 42 }, { 0, 48 },
|
|
| 509 |
+ { 0, 55 }, { -6, 59 }, { -7, 71 }, { -12, 83 },
|
|
| 510 |
+ { -11, 87 }, { -30, 119 }, { 1, 58 }, { -3, 29 },
|
|
| 511 |
+ { -1, 36 }, { 1, 38 }, { 2, 43 }, { -6, 55 },
|
|
| 512 |
+ { 0, 58 }, { 0, 64 }, { -3, 74 }, { -10, 90 },
|
|
| 513 |
+ { -3, 74 }, { -9, 92 }, { -8, 87 }, { -23, 126 },
|
|
| 514 |
+ { -3, 74 }, { -9, 92 }, { -8, 87 }, { -23, 126 },
|
|
| 515 |
+ { -3, 74 }, { -9, 92 }, { -8, 87 }, { -23, 126 }
|
|
| 373 | 516 |
}, |
| 374 | 517 |
|
| 375 | 518 |
/* i_cabac_init_idc == 1 */ |
| ... | ... |
@@ -526,6 +812,149 @@ static const int8_t cabac_context_init_PB[3][460][2] = |
| 526 | 526 |
{ -9, 60 }, { 1, 54 }, { 2, 58 }, { 17, -10 },
|
| 527 | 527 |
{ 32, -13 }, { 42, -9 }, { 49, -5 }, { 53, 0 },
|
| 528 | 528 |
{ 64, 3 }, { 68, 10 }, { 66, 27 }, { 47, 57 },
|
| 529 |
+ |
|
| 530 |
+ /* 460 - 1024 */ |
|
| 531 |
+ { 0, 80 }, { -5, 89 }, { -7, 94 }, { -4, 92 },
|
|
| 532 |
+ { 0, 39 }, { 0, 65 }, { -15, 84 }, { -35, 127 },
|
|
| 533 |
+ { -2, 73 }, { -12, 104 }, { -9, 91 }, { -31, 127 },
|
|
| 534 |
+ { 0, 80 }, { -5, 89 }, { -7, 94 }, { -4, 92 },
|
|
| 535 |
+ { 0, 39 }, { 0, 65 }, { -15, 84 }, { -35, 127 },
|
|
| 536 |
+ { -2, 73 }, { -12, 104 }, { -9, 91 }, { -31, 127 },
|
|
| 537 |
+ { -13, 103 }, { -13, 91 }, { -9, 89 }, { -14, 92 },
|
|
| 538 |
+ { -8, 76 }, { -12, 87 }, { -23, 110 }, { -24, 105 },
|
|
| 539 |
+ { -10, 78 }, { -20, 112 }, { -17, 99 }, { -78, 127 },
|
|
| 540 |
+ { -70, 127 }, { -50, 127 }, { -46, 127 }, { -4, 66 },
|
|
| 541 |
+ { -5, 78 }, { -4, 71 }, { -8, 72 }, { 2, 59 },
|
|
| 542 |
+ { -1, 55 }, { -7, 70 }, { -6, 75 }, { -8, 89 },
|
|
| 543 |
+ { -34, 119 }, { -3, 75 }, { 32, 20 }, { 30, 22 },
|
|
| 544 |
+ { -44, 127 }, { 0, 54 }, { -5, 61 }, { 0, 58 },
|
|
| 545 |
+ { -1, 60 }, { -3, 61 }, { -8, 67 }, { -25, 84 },
|
|
| 546 |
+ { -14, 74 }, { -5, 65 }, { 5, 52 }, { 2, 57 },
|
|
| 547 |
+ { 0, 61 }, { -9, 69 }, { -11, 70 }, { 18, 55 },
|
|
| 548 |
+ { -13, 103 }, { -13, 91 }, { -9, 89 }, { -14, 92 },
|
|
| 549 |
+ { -8, 76 }, { -12, 87 }, { -23, 110 }, { -24, 105 },
|
|
| 550 |
+ { -10, 78 }, { -20, 112 }, { -17, 99 }, { -78, 127 },
|
|
| 551 |
+ { -70, 127 }, { -50, 127 }, { -46, 127 }, { -4, 66 },
|
|
| 552 |
+ { -5, 78 }, { -4, 71 }, { -8, 72 }, { 2, 59 },
|
|
| 553 |
+ { -1, 55 }, { -7, 70 }, { -6, 75 }, { -8, 89 },
|
|
| 554 |
+ { -34, 119 }, { -3, 75 }, { 32, 20 }, { 30, 22 },
|
|
| 555 |
+ { -44, 127 }, { 0, 54 }, { -5, 61 }, { 0, 58 },
|
|
| 556 |
+ { -1, 60 }, { -3, 61 }, { -8, 67 }, { -25, 84 },
|
|
| 557 |
+ { -14, 74 }, { -5, 65 }, { 5, 52 }, { 2, 57 },
|
|
| 558 |
+ { 0, 61 }, { -9, 69 }, { -11, 70 }, { 18, 55 },
|
|
| 559 |
+ { 4, 45 }, { 10, 28 }, { 10, 31 }, { 33, -11 },
|
|
| 560 |
+ { 52, -43 }, { 18, 15 }, { 28, 0 }, { 35, -22 },
|
|
| 561 |
+ { 38, -25 }, { 34, 0 }, { 39, -18 }, { 32, -12 },
|
|
| 562 |
+ { 102, -94 }, { 0, 0 }, { 56, -15 }, { 33, -4 },
|
|
| 563 |
+ { 29, 10 }, { 37, -5 }, { 51, -29 }, { 39, -9 },
|
|
| 564 |
+ { 52, -34 }, { 69, -58 }, { 67, -63 }, { 44, -5 },
|
|
| 565 |
+ { 32, 7 }, { 55, -29 }, { 32, 1 }, { 0, 0 },
|
|
| 566 |
+ { 27, 36 }, { 33, -25 }, { 34, -30 }, { 36, -28 },
|
|
| 567 |
+ { 38, -28 }, { 38, -27 }, { 34, -18 }, { 35, -16 },
|
|
| 568 |
+ { 34, -14 }, { 32, -8 }, { 37, -6 }, { 35, 0 },
|
|
| 569 |
+ { 30, 10 }, { 28, 18 }, { 26, 25 }, { 29, 41 },
|
|
| 570 |
+ { 4, 45 }, { 10, 28 }, { 10, 31 }, { 33, -11 },
|
|
| 571 |
+ { 52, -43 }, { 18, 15 }, { 28, 0 }, { 35, -22 },
|
|
| 572 |
+ { 38, -25 }, { 34, 0 }, { 39, -18 }, { 32, -12 },
|
|
| 573 |
+ { 102, -94 }, { 0, 0 }, { 56, -15 }, { 33, -4 },
|
|
| 574 |
+ { 29, 10 }, { 37, -5 }, { 51, -29 }, { 39, -9 },
|
|
| 575 |
+ { 52, -34 }, { 69, -58 }, { 67, -63 }, { 44, -5 },
|
|
| 576 |
+ { 32, 7 }, { 55, -29 }, { 32, 1 }, { 0, 0 },
|
|
| 577 |
+ { 27, 36 }, { 33, -25 }, { 34, -30 }, { 36, -28 },
|
|
| 578 |
+ { 38, -28 }, { 38, -27 }, { 34, -18 }, { 35, -16 },
|
|
| 579 |
+ { 34, -14 }, { 32, -8 }, { 37, -6 }, { 35, 0 },
|
|
| 580 |
+ { 30, 10 }, { 28, 18 }, { 26, 25 }, { 29, 41 },
|
|
| 581 |
+ { -5, 85 }, { -6, 81 }, { -10, 77 }, { -7, 81 },
|
|
| 582 |
+ { -17, 80 }, { -18, 73 }, { -4, 74 }, { -10, 83 },
|
|
| 583 |
+ { -9, 71 }, { -9, 67 }, { -1, 61 }, { -8, 66 },
|
|
| 584 |
+ { -14, 66 }, { 0, 59 }, { 2, 59 }, { -3, 81 },
|
|
| 585 |
+ { -3, 76 }, { -7, 72 }, { -6, 78 }, { -12, 72 },
|
|
| 586 |
+ { -14, 68 }, { -3, 70 }, { -6, 76 }, { -5, 66 },
|
|
| 587 |
+ { -5, 62 }, { 0, 57 }, { -4, 61 }, { -9, 60 },
|
|
| 588 |
+ { 1, 54 }, { 2, 58 }, { 17, -10 }, { 32, -13 },
|
|
| 589 |
+ { 42, -9 }, { 49, -5 }, { 53, 0 }, { 64, 3 },
|
|
| 590 |
+ { 68, 10 }, { 66, 27 }, { 47, 57 }, { 17, -10 },
|
|
| 591 |
+ { 32, -13 }, { 42, -9 }, { 49, -5 }, { 53, 0 },
|
|
| 592 |
+ { 64, 3 }, { 68, 10 }, { 66, 27 }, { 47, 57 },
|
|
| 593 |
+ { -5, 71 }, { 0, 24 }, { -1, 36 }, { -2, 42 },
|
|
| 594 |
+ { -2, 52 }, { -9, 57 }, { -6, 63 }, { -4, 65 },
|
|
| 595 |
+ { -4, 67 }, { -7, 82 }, { -5, 85 }, { -6, 81 },
|
|
| 596 |
+ { -10, 77 }, { -7, 81 }, { -17, 80 }, { -18, 73 },
|
|
| 597 |
+ { -4, 74 }, { -10, 83 }, { -9, 71 }, { -9, 67 },
|
|
| 598 |
+ { -1, 61 }, { -8, 66 }, { -14, 66 }, { 0, 59 },
|
|
| 599 |
+ { 2, 59 }, { -3, 81 }, { -3, 76 }, { -7, 72 },
|
|
| 600 |
+ { -6, 78 }, { -12, 72 }, { -14, 68 }, { -3, 70 },
|
|
| 601 |
+ { -6, 76 }, { -5, 66 }, { -5, 62 }, { 0, 57 },
|
|
| 602 |
+ { -4, 61 }, { -9, 60 }, { 1, 54 }, { 2, 58 },
|
|
| 603 |
+ { 17, -10 }, { 32, -13 }, { 42, -9 }, { 49, -5 },
|
|
| 604 |
+ { 53, 0 }, { 64, 3 }, { 68, 10 }, { 66, 27 },
|
|
| 605 |
+ { 47, 57 }, { 17, -10 }, { 32, -13 }, { 42, -9 },
|
|
| 606 |
+ { 49, -5 }, { 53, 0 }, { 64, 3 }, { 68, 10 },
|
|
| 607 |
+ { 66, 27 }, { 47, 57 }, { -5, 71 }, { 0, 24 },
|
|
| 608 |
+ { -1, 36 }, { -2, 42 }, { -2, 52 }, { -9, 57 },
|
|
| 609 |
+ { -6, 63 }, { -4, 65 }, { -4, 67 }, { -7, 82 },
|
|
| 610 |
+ { -21, 126 }, { -23, 124 }, { -20, 110 }, { -26, 126 },
|
|
| 611 |
+ { -25, 124 }, { -17, 105 }, { -27, 121 }, { -27, 117 },
|
|
| 612 |
+ { -17, 102 }, { -26, 117 }, { -27, 116 }, { -33, 122 },
|
|
| 613 |
+ { -10, 95 }, { -14, 100 }, { -8, 95 }, { -17, 111 },
|
|
| 614 |
+ { -28, 114 }, { -6, 89 }, { -2, 80 }, { -4, 82 },
|
|
| 615 |
+ { -9, 85 }, { -8, 81 }, { -1, 72 }, { 5, 64 },
|
|
| 616 |
+ { 1, 67 }, { 9, 56 }, { 0, 69 }, { 1, 69 },
|
|
| 617 |
+ { 7, 69 }, { -7, 69 }, { -6, 67 }, { -16, 77 },
|
|
| 618 |
+ { -2, 64 }, { 2, 61 }, { -6, 67 }, { -3, 64 },
|
|
| 619 |
+ { 2, 57 }, { -3, 65 }, { -3, 66 }, { 0, 62 },
|
|
| 620 |
+ { 9, 51 }, { -1, 66 }, { -2, 71 }, { -2, 75 },
|
|
| 621 |
+ { -21, 126 }, { -23, 124 }, { -20, 110 }, { -26, 126 },
|
|
| 622 |
+ { -25, 124 }, { -17, 105 }, { -27, 121 }, { -27, 117 },
|
|
| 623 |
+ { -17, 102 }, { -26, 117 }, { -27, 116 }, { -33, 122 },
|
|
| 624 |
+ { -10, 95 }, { -14, 100 }, { -8, 95 }, { -17, 111 },
|
|
| 625 |
+ { -28, 114 }, { -6, 89 }, { -2, 80 }, { -4, 82 },
|
|
| 626 |
+ { -9, 85 }, { -8, 81 }, { -1, 72 }, { 5, 64 },
|
|
| 627 |
+ { 1, 67 }, { 9, 56 }, { 0, 69 }, { 1, 69 },
|
|
| 628 |
+ { 7, 69 }, { -7, 69 }, { -6, 67 }, { -16, 77 },
|
|
| 629 |
+ { -2, 64 }, { 2, 61 }, { -6, 67 }, { -3, 64 },
|
|
| 630 |
+ { 2, 57 }, { -3, 65 }, { -3, 66 }, { 0, 62 },
|
|
| 631 |
+ { 9, 51 }, { -1, 66 }, { -2, 71 }, { -2, 75 },
|
|
| 632 |
+ { 19, -6 }, { 18, -6 }, { 14, 0 }, { 26, -12 },
|
|
| 633 |
+ { 31, -16 }, { 33, -25 }, { 33, -22 }, { 37, -28 },
|
|
| 634 |
+ { 39, -30 }, { 42, -30 }, { 47, -42 }, { 45, -36 },
|
|
| 635 |
+ { 49, -34 }, { 41, -17 }, { 32, 9 }, { 69, -71 },
|
|
| 636 |
+ { 63, -63 }, { 66, -64 }, { 77, -74 }, { 54, -39 },
|
|
| 637 |
+ { 52, -35 }, { 41, -10 }, { 36, 0 }, { 40, -1 },
|
|
| 638 |
+ { 30, 14 }, { 28, 26 }, { 23, 37 }, { 12, 55 },
|
|
| 639 |
+ { 11, 65 }, { 37, -33 }, { 39, -36 }, { 40, -37 },
|
|
| 640 |
+ { 38, -30 }, { 46, -33 }, { 42, -30 }, { 40, -24 },
|
|
| 641 |
+ { 49, -29 }, { 38, -12 }, { 40, -10 }, { 38, -3 },
|
|
| 642 |
+ { 46, -5 }, { 31, 20 }, { 29, 30 }, { 25, 44 },
|
|
| 643 |
+ { 19, -6 }, { 18, -6 }, { 14, 0 }, { 26, -12 },
|
|
| 644 |
+ { 31, -16 }, { 33, -25 }, { 33, -22 }, { 37, -28 },
|
|
| 645 |
+ { 39, -30 }, { 42, -30 }, { 47, -42 }, { 45, -36 },
|
|
| 646 |
+ { 49, -34 }, { 41, -17 }, { 32, 9 }, { 69, -71 },
|
|
| 647 |
+ { 63, -63 }, { 66, -64 }, { 77, -74 }, { 54, -39 },
|
|
| 648 |
+ { 52, -35 }, { 41, -10 }, { 36, 0 }, { 40, -1 },
|
|
| 649 |
+ { 30, 14 }, { 28, 26 }, { 23, 37 }, { 12, 55 },
|
|
| 650 |
+ { 11, 65 }, { 37, -33 }, { 39, -36 }, { 40, -37 },
|
|
| 651 |
+ { 38, -30 }, { 46, -33 }, { 42, -30 }, { 40, -24 },
|
|
| 652 |
+ { 49, -29 }, { 38, -12 }, { 40, -10 }, { 38, -3 },
|
|
| 653 |
+ { 46, -5 }, { 31, 20 }, { 29, 30 }, { 25, 44 },
|
|
| 654 |
+ { -23, 112 }, { -15, 71 }, { -7, 61 }, { 0, 53 },
|
|
| 655 |
+ { -5, 66 }, { -11, 77 }, { -9, 80 }, { -9, 84 },
|
|
| 656 |
+ { -10, 87 }, { -34, 127 }, { -21, 101 }, { -3, 39 },
|
|
| 657 |
+ { -5, 53 }, { -7, 61 }, { -11, 75 }, { -15, 77 },
|
|
| 658 |
+ { -17, 91 }, { -25, 107 }, { -25, 111 }, { -28, 122 },
|
|
| 659 |
+ { -11, 76 }, { -10, 44 }, { -10, 52 }, { -10, 57 },
|
|
| 660 |
+ { -9, 58 }, { -16, 72 }, { -7, 69 }, { -4, 69 },
|
|
| 661 |
+ { -5, 74 }, { -9, 86 }, { -23, 112 }, { -15, 71 },
|
|
| 662 |
+ { -7, 61 }, { 0, 53 }, { -5, 66 }, { -11, 77 },
|
|
| 663 |
+ { -9, 80 }, { -9, 84 }, { -10, 87 }, { -34, 127 },
|
|
| 664 |
+ { -21, 101 }, { -3, 39 }, { -5, 53 }, { -7, 61 },
|
|
| 665 |
+ { -11, 75 }, { -15, 77 }, { -17, 91 }, { -25, 107 },
|
|
| 666 |
+ { -25, 111 }, { -28, 122 }, { -11, 76 }, { -10, 44 },
|
|
| 667 |
+ { -10, 52 }, { -10, 57 }, { -9, 58 }, { -16, 72 },
|
|
| 668 |
+ { -7, 69 }, { -4, 69 }, { -5, 74 }, { -9, 86 },
|
|
| 669 |
+ { -2, 73 }, { -12, 104 }, { -9, 91 }, { -31, 127 },
|
|
| 670 |
+ { -2, 73 }, { -12, 104 }, { -9, 91 }, { -31, 127 },
|
|
| 671 |
+ { -2, 73 }, { -12, 104 }, { -9, 91 }, { -31, 127 }
|
|
| 529 | 672 |
}, |
| 530 | 673 |
|
| 531 | 674 |
/* i_cabac_init_idc == 2 */ |
| ... | ... |
@@ -682,6 +1111,149 @@ static const int8_t cabac_context_init_PB[3][460][2] = |
| 682 | 682 |
{ -14, 59 }, { -9, 52 }, { -11, 68 }, { 9, -2 },
|
| 683 | 683 |
{ 30, -10 }, { 31, -4 }, { 33, -1 }, { 33, 7 },
|
| 684 | 684 |
{ 31, 12 }, { 37, 23 }, { 31, 38 }, { 20, 64 },
|
| 685 |
+ |
|
| 686 |
+ /* 460 - 1024 */ |
|
| 687 |
+ { 11, 80 }, { 5, 76 }, { 2, 84 }, { 5, 78 },
|
|
| 688 |
+ { -6, 55 }, { 4, 61 }, { -14, 83 }, { -37, 127 },
|
|
| 689 |
+ { -5, 79 }, { -11, 104 }, { -11, 91 }, { -30, 127 },
|
|
| 690 |
+ { 11, 80 }, { 5, 76 }, { 2, 84 }, { 5, 78 },
|
|
| 691 |
+ { -6, 55 }, { 4, 61 }, { -14, 83 }, { -37, 127 },
|
|
| 692 |
+ { -5, 79 }, { -11, 104 }, { -11, 91 }, { -30, 127 },
|
|
| 693 |
+ { -4, 86 }, { -12, 88 }, { -5, 82 }, { -3, 72 },
|
|
| 694 |
+ { -4, 67 }, { -8, 72 }, { -16, 89 }, { -9, 69 },
|
|
| 695 |
+ { -1, 59 }, { 5, 66 }, { 4, 57 }, { -4, 71 },
|
|
| 696 |
+ { -2, 71 }, { 2, 58 }, { -1, 74 }, { -4, 44 },
|
|
| 697 |
+ { -1, 69 }, { 0, 62 }, { -7, 51 }, { -4, 47 },
|
|
| 698 |
+ { -6, 42 }, { -3, 41 }, { -6, 53 }, { 8, 76 },
|
|
| 699 |
+ { -9, 78 }, { -11, 83 }, { 9, 52 }, { 0, 67 },
|
|
| 700 |
+ { -5, 90 }, { 1, 67 }, { -15, 72 }, { -5, 75 },
|
|
| 701 |
+ { -8, 80 }, { -21, 83 }, { -21, 64 }, { -13, 31 },
|
|
| 702 |
+ { -25, 64 }, { -29, 94 }, { 9, 75 }, { 17, 63 },
|
|
| 703 |
+ { -8, 74 }, { -5, 35 }, { -2, 27 }, { 13, 91 },
|
|
| 704 |
+ { -4, 86 }, { -12, 88 }, { -5, 82 }, { -3, 72 },
|
|
| 705 |
+ { -4, 67 }, { -8, 72 }, { -16, 89 }, { -9, 69 },
|
|
| 706 |
+ { -1, 59 }, { 5, 66 }, { 4, 57 }, { -4, 71 },
|
|
| 707 |
+ { -2, 71 }, { 2, 58 }, { -1, 74 }, { -4, 44 },
|
|
| 708 |
+ { -1, 69 }, { 0, 62 }, { -7, 51 }, { -4, 47 },
|
|
| 709 |
+ { -6, 42 }, { -3, 41 }, { -6, 53 }, { 8, 76 },
|
|
| 710 |
+ { -9, 78 }, { -11, 83 }, { 9, 52 }, { 0, 67 },
|
|
| 711 |
+ { -5, 90 }, { 1, 67 }, { -15, 72 }, { -5, 75 },
|
|
| 712 |
+ { -8, 80 }, { -21, 83 }, { -21, 64 }, { -13, 31 },
|
|
| 713 |
+ { -25, 64 }, { -29, 94 }, { 9, 75 }, { 17, 63 },
|
|
| 714 |
+ { -8, 74 }, { -5, 35 }, { -2, 27 }, { 13, 91 },
|
|
| 715 |
+ { 4, 39 }, { 0, 42 }, { 7, 34 }, { 11, 29 },
|
|
| 716 |
+ { 8, 31 }, { 6, 37 }, { 7, 42 }, { 3, 40 },
|
|
| 717 |
+ { 8, 33 }, { 13, 43 }, { 13, 36 }, { 4, 47 },
|
|
| 718 |
+ { 3, 55 }, { 2, 58 }, { 6, 60 }, { 8, 44 },
|
|
| 719 |
+ { 11, 44 }, { 14, 42 }, { 7, 48 }, { 4, 56 },
|
|
| 720 |
+ { 4, 52 }, { 13, 37 }, { 9, 49 }, { 19, 58 },
|
|
| 721 |
+ { 10, 48 }, { 12, 45 }, { 0, 69 }, { 20, 33 },
|
|
| 722 |
+ { 8, 63 }, { 35, -18 }, { 33, -25 }, { 28, -3 },
|
|
| 723 |
+ { 24, 10 }, { 27, 0 }, { 34, -14 }, { 52, -44 },
|
|
| 724 |
+ { 39, -24 }, { 19, 17 }, { 31, 25 }, { 36, 29 },
|
|
| 725 |
+ { 24, 33 }, { 34, 15 }, { 30, 20 }, { 22, 73 },
|
|
| 726 |
+ { 4, 39 }, { 0, 42 }, { 7, 34 }, { 11, 29 },
|
|
| 727 |
+ { 8, 31 }, { 6, 37 }, { 7, 42 }, { 3, 40 },
|
|
| 728 |
+ { 8, 33 }, { 13, 43 }, { 13, 36 }, { 4, 47 },
|
|
| 729 |
+ { 3, 55 }, { 2, 58 }, { 6, 60 }, { 8, 44 },
|
|
| 730 |
+ { 11, 44 }, { 14, 42 }, { 7, 48 }, { 4, 56 },
|
|
| 731 |
+ { 4, 52 }, { 13, 37 }, { 9, 49 }, { 19, 58 },
|
|
| 732 |
+ { 10, 48 }, { 12, 45 }, { 0, 69 }, { 20, 33 },
|
|
| 733 |
+ { 8, 63 }, { 35, -18 }, { 33, -25 }, { 28, -3 },
|
|
| 734 |
+ { 24, 10 }, { 27, 0 }, { 34, -14 }, { 52, -44 },
|
|
| 735 |
+ { 39, -24 }, { 19, 17 }, { 31, 25 }, { 36, 29 },
|
|
| 736 |
+ { 24, 33 }, { 34, 15 }, { 30, 20 }, { 22, 73 },
|
|
| 737 |
+ { -3, 78 }, { -8, 74 }, { -9, 72 }, { -10, 72 },
|
|
| 738 |
+ { -18, 75 }, { -12, 71 }, { -11, 63 }, { -5, 70 },
|
|
| 739 |
+ { -17, 75 }, { -14, 72 }, { -16, 67 }, { -8, 53 },
|
|
| 740 |
+ { -14, 59 }, { -9, 52 }, { -11, 68 }, { -3, 78 },
|
|
| 741 |
+ { -8, 74 }, { -9, 72 }, { -10, 72 }, { -18, 75 },
|
|
| 742 |
+ { -12, 71 }, { -11, 63 }, { -5, 70 }, { -17, 75 },
|
|
| 743 |
+ { -14, 72 }, { -16, 67 }, { -8, 53 }, { -14, 59 },
|
|
| 744 |
+ { -9, 52 }, { -11, 68 }, { 9, -2 }, { 30, -10 },
|
|
| 745 |
+ { 31, -4 }, { 33, -1 }, { 33, 7 }, { 31, 12 },
|
|
| 746 |
+ { 37, 23 }, { 31, 38 }, { 20, 64 }, { 9, -2 },
|
|
| 747 |
+ { 30, -10 }, { 31, -4 }, { 33, -1 }, { 33, 7 },
|
|
| 748 |
+ { 31, 12 }, { 37, 23 }, { 31, 38 }, { 20, 64 },
|
|
| 749 |
+ { -9, 71 }, { -7, 37 }, { -8, 44 }, { -11, 49 },
|
|
| 750 |
+ { -10, 56 }, { -12, 59 }, { -8, 63 }, { -9, 67 },
|
|
| 751 |
+ { -6, 68 }, { -10, 79 }, { -3, 78 }, { -8, 74 },
|
|
| 752 |
+ { -9, 72 }, { -10, 72 }, { -18, 75 }, { -12, 71 },
|
|
| 753 |
+ { -11, 63 }, { -5, 70 }, { -17, 75 }, { -14, 72 },
|
|
| 754 |
+ { -16, 67 }, { -8, 53 }, { -14, 59 }, { -9, 52 },
|
|
| 755 |
+ { -11, 68 }, { -3, 78 }, { -8, 74 }, { -9, 72 },
|
|
| 756 |
+ { -10, 72 }, { -18, 75 }, { -12, 71 }, { -11, 63 },
|
|
| 757 |
+ { -5, 70 }, { -17, 75 }, { -14, 72 }, { -16, 67 },
|
|
| 758 |
+ { -8, 53 }, { -14, 59 }, { -9, 52 }, { -11, 68 },
|
|
| 759 |
+ { 9, -2 }, { 30, -10 }, { 31, -4 }, { 33, -1 },
|
|
| 760 |
+ { 33, 7 }, { 31, 12 }, { 37, 23 }, { 31, 38 },
|
|
| 761 |
+ { 20, 64 }, { 9, -2 }, { 30, -10 }, { 31, -4 },
|
|
| 762 |
+ { 33, -1 }, { 33, 7 }, { 31, 12 }, { 37, 23 },
|
|
| 763 |
+ { 31, 38 }, { 20, 64 }, { -9, 71 }, { -7, 37 },
|
|
| 764 |
+ { -8, 44 }, { -11, 49 }, { -10, 56 }, { -12, 59 },
|
|
| 765 |
+ { -8, 63 }, { -9, 67 }, { -6, 68 }, { -10, 79 },
|
|
| 766 |
+ { -22, 127 }, { -25, 127 }, { -25, 120 }, { -27, 127 },
|
|
| 767 |
+ { -19, 114 }, { -23, 117 }, { -25, 118 }, { -26, 117 },
|
|
| 768 |
+ { -24, 113 }, { -28, 118 }, { -31, 120 }, { -37, 124 },
|
|
| 769 |
+ { -10, 94 }, { -15, 102 }, { -10, 99 }, { -13, 106 },
|
|
| 770 |
+ { -50, 127 }, { -5, 92 }, { 17, 57 }, { -5, 86 },
|
|
| 771 |
+ { -13, 94 }, { -12, 91 }, { -2, 77 }, { 0, 71 },
|
|
| 772 |
+ { -1, 73 }, { 4, 64 }, { -7, 81 }, { 5, 64 },
|
|
| 773 |
+ { 15, 57 }, { 1, 67 }, { 0, 68 }, { -10, 67 },
|
|
| 774 |
+ { 1, 68 }, { 0, 77 }, { 2, 64 }, { 0, 68 },
|
|
| 775 |
+ { -5, 78 }, { 7, 55 }, { 5, 59 }, { 2, 65 },
|
|
| 776 |
+ { 14, 54 }, { 15, 44 }, { 5, 60 }, { 2, 70 },
|
|
| 777 |
+ { -22, 127 }, { -25, 127 }, { -25, 120 }, { -27, 127 },
|
|
| 778 |
+ { -19, 114 }, { -23, 117 }, { -25, 118 }, { -26, 117 },
|
|
| 779 |
+ { -24, 113 }, { -28, 118 }, { -31, 120 }, { -37, 124 },
|
|
| 780 |
+ { -10, 94 }, { -15, 102 }, { -10, 99 }, { -13, 106 },
|
|
| 781 |
+ { -50, 127 }, { -5, 92 }, { 17, 57 }, { -5, 86 },
|
|
| 782 |
+ { -13, 94 }, { -12, 91 }, { -2, 77 }, { 0, 71 },
|
|
| 783 |
+ { -1, 73 }, { 4, 64 }, { -7, 81 }, { 5, 64 },
|
|
| 784 |
+ { 15, 57 }, { 1, 67 }, { 0, 68 }, { -10, 67 },
|
|
| 785 |
+ { 1, 68 }, { 0, 77 }, { 2, 64 }, { 0, 68 },
|
|
| 786 |
+ { -5, 78 }, { 7, 55 }, { 5, 59 }, { 2, 65 },
|
|
| 787 |
+ { 14, 54 }, { 15, 44 }, { 5, 60 }, { 2, 70 },
|
|
| 788 |
+ { 17, -13 }, { 16, -9 }, { 17, -12 }, { 27, -21 },
|
|
| 789 |
+ { 37, -30 }, { 41, -40 }, { 42, -41 }, { 48, -47 },
|
|
| 790 |
+ { 39, -32 }, { 46, -40 }, { 52, -51 }, { 46, -41 },
|
|
| 791 |
+ { 52, -39 }, { 43, -19 }, { 32, 11 }, { 61, -55 },
|
|
| 792 |
+ { 56, -46 }, { 62, -50 }, { 81, -67 }, { 45, -20 },
|
|
| 793 |
+ { 35, -2 }, { 28, 15 }, { 34, 1 }, { 39, 1 },
|
|
| 794 |
+ { 30, 17 }, { 20, 38 }, { 18, 45 }, { 15, 54 },
|
|
| 795 |
+ { 0, 79 }, { 36, -16 }, { 37, -14 }, { 37, -17 },
|
|
| 796 |
+ { 32, 1 }, { 34, 15 }, { 29, 15 }, { 24, 25 },
|
|
| 797 |
+ { 34, 22 }, { 31, 16 }, { 35, 18 }, { 31, 28 },
|
|
| 798 |
+ { 33, 41 }, { 36, 28 }, { 27, 47 }, { 21, 62 },
|
|
| 799 |
+ { 17, -13 }, { 16, -9 }, { 17, -12 }, { 27, -21 },
|
|
| 800 |
+ { 37, -30 }, { 41, -40 }, { 42, -41 }, { 48, -47 },
|
|
| 801 |
+ { 39, -32 }, { 46, -40 }, { 52, -51 }, { 46, -41 },
|
|
| 802 |
+ { 52, -39 }, { 43, -19 }, { 32, 11 }, { 61, -55 },
|
|
| 803 |
+ { 56, -46 }, { 62, -50 }, { 81, -67 }, { 45, -20 },
|
|
| 804 |
+ { 35, -2 }, { 28, 15 }, { 34, 1 }, { 39, 1 },
|
|
| 805 |
+ { 30, 17 }, { 20, 38 }, { 18, 45 }, { 15, 54 },
|
|
| 806 |
+ { 0, 79 }, { 36, -16 }, { 37, -14 }, { 37, -17 },
|
|
| 807 |
+ { 32, 1 }, { 34, 15 }, { 29, 15 }, { 24, 25 },
|
|
| 808 |
+ { 34, 22 }, { 31, 16 }, { 35, 18 }, { 31, 28 },
|
|
| 809 |
+ { 33, 41 }, { 36, 28 }, { 27, 47 }, { 21, 62 },
|
|
| 810 |
+ { -24, 115 }, { -22, 82 }, { -9, 62 }, { 0, 53 },
|
|
| 811 |
+ { 0, 59 }, { -14, 85 }, { -13, 89 }, { -13, 94 },
|
|
| 812 |
+ { -11, 92 }, { -29, 127 }, { -21, 100 }, { -14, 57 },
|
|
| 813 |
+ { -12, 67 }, { -11, 71 }, { -10, 77 }, { -21, 85 },
|
|
| 814 |
+ { -16, 88 }, { -23, 104 }, { -15, 98 }, { -37, 127 },
|
|
| 815 |
+ { -10, 82 }, { -8, 48 }, { -8, 61 }, { -8, 66 },
|
|
| 816 |
+ { -7, 70 }, { -14, 75 }, { -10, 79 }, { -9, 83 },
|
|
| 817 |
+ { -12, 92 }, { -18, 108 }, { -24, 115 }, { -22, 82 },
|
|
| 818 |
+ { -9, 62 }, { 0, 53 }, { 0, 59 }, { -14, 85 },
|
|
| 819 |
+ { -13, 89 }, { -13, 94 }, { -11, 92 }, { -29, 127 },
|
|
| 820 |
+ { -21, 100 }, { -14, 57 }, { -12, 67 }, { -11, 71 },
|
|
| 821 |
+ { -10, 77 }, { -21, 85 }, { -16, 88 }, { -23, 104 },
|
|
| 822 |
+ { -15, 98 }, { -37, 127 }, { -10, 82 }, { -8, 48 },
|
|
| 823 |
+ { -8, 61 }, { -8, 66 }, { -7, 70 }, { -14, 75 },
|
|
| 824 |
+ { -10, 79 }, { -9, 83 }, { -12, 92 }, { -18, 108 },
|
|
| 825 |
+ { -5, 79 }, { -11, 104 }, { -11, 91 }, { -30, 127 },
|
|
| 826 |
+ { -5, 79 }, { -11, 104 }, { -11, 91 }, { -30, 127 },
|
|
| 827 |
+ { -5, 79 }, { -11, 104 }, { -11, 91 }, { -30, 127 }
|
|
| 685 | 828 |
} |
| 686 | 829 |
}; |
| 687 | 830 |
|
| ... | ... |
@@ -695,7 +1267,7 @@ void ff_h264_init_cabac_states(H264Context *h) {
|
| 695 | 695 |
else tab = cabac_context_init_PB[h->cabac_init_idc]; |
| 696 | 696 |
|
| 697 | 697 |
/* calculate pre-state */ |
| 698 |
- for( i= 0; i < 460; i++ ) {
|
|
| 698 |
+ for( i= 0; i < 1024; i++ ) {
|
|
| 699 | 699 |
int pre = 2*(((tab[i][0] * slice_qp) >>4 ) + tab[i][1]) - 127; |
| 700 | 700 |
|
| 701 | 701 |
pre^= pre>>31; |
| ... | ... |
@@ -957,21 +1529,22 @@ static int decode_cabac_mb_mvd( H264Context *h, int ctxbase, int amvd, int *mvda |
| 957 | 957 |
my += decode_cabac_mb_mvd( h, 47, amvd1, &mpy );\ |
| 958 | 958 |
} |
| 959 | 959 |
|
| 960 |
-static av_always_inline int get_cabac_cbf_ctx( H264Context *h, int cat, int idx, int is_dc ) {
|
|
| 960 |
+static av_always_inline int get_cabac_cbf_ctx( H264Context *h, int cat, int idx, int max_coeff, int is_dc ) {
|
|
| 961 | 961 |
int nza, nzb; |
| 962 | 962 |
int ctx = 0; |
| 963 |
+ static const uint16_t base_ctx[14] = {85,89,93,97,101,1012,460,464,468,1016,472,476,480,1020};
|
|
| 963 | 964 |
|
| 964 | 965 |
if( is_dc ) {
|
| 965 |
- if( cat == 0 ) {
|
|
| 966 |
- nza = h->left_cbp&0x100; |
|
| 967 |
- nzb = h-> top_cbp&0x100; |
|
| 968 |
- } else {
|
|
| 966 |
+ if( cat == 3 ) {
|
|
| 969 | 967 |
idx -= CHROMA_DC_BLOCK_INDEX; |
| 970 | 968 |
nza = (h->left_cbp>>(6+idx))&0x01; |
| 971 | 969 |
nzb = (h-> top_cbp>>(6+idx))&0x01; |
| 970 |
+ } else {
|
|
| 971 |
+ idx -= LUMA_DC_BLOCK_INDEX; |
|
| 972 |
+ nza = h->left_cbp&(0x100<<idx); |
|
| 973 |
+ nzb = h-> top_cbp&(0x100<<idx); |
|
| 972 | 974 |
} |
| 973 | 975 |
} else {
|
| 974 |
- assert(cat == 1 || cat == 2 || cat == 4); |
|
| 975 | 976 |
nza = h->non_zero_count_cache[scan8[idx] - 1]; |
| 976 | 977 |
nzb = h->non_zero_count_cache[scan8[idx] - 8]; |
| 977 | 978 |
} |
| ... | ... |
@@ -982,7 +1555,7 @@ static av_always_inline int get_cabac_cbf_ctx( H264Context *h, int cat, int idx, |
| 982 | 982 |
if( nzb > 0 ) |
| 983 | 983 |
ctx += 2; |
| 984 | 984 |
|
| 985 |
- return ctx + 4 * cat; |
|
| 985 |
+ return base_ctx[cat] + ctx; |
|
| 986 | 986 |
} |
| 987 | 987 |
|
| 988 | 988 |
DECLARE_ASM_CONST(1, uint8_t, last_coeff_flag_offset_8x8)[63] = {
|
| ... | ... |
@@ -993,16 +1566,16 @@ DECLARE_ASM_CONST(1, uint8_t, last_coeff_flag_offset_8x8)[63] = {
|
| 993 | 993 |
}; |
| 994 | 994 |
|
| 995 | 995 |
static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff, int is_dc ) {
|
| 996 |
- static const int significant_coeff_flag_offset[2][6] = {
|
|
| 997 |
- { 105+0, 105+15, 105+29, 105+44, 105+47, 402 },
|
|
| 998 |
- { 277+0, 277+15, 277+29, 277+44, 277+47, 436 }
|
|
| 996 |
+ static const int significant_coeff_flag_offset[2][14] = {
|
|
| 997 |
+ { 105+0, 105+15, 105+29, 105+44, 105+47, 402, 484+0, 484+15, 484+29, 660, 528+0, 528+15, 528+29, 718 },
|
|
| 998 |
+ { 277+0, 277+15, 277+29, 277+44, 277+47, 436, 776+0, 776+15, 776+29, 675, 820+0, 820+15, 820+29, 733 }
|
|
| 999 | 999 |
}; |
| 1000 |
- static const int last_coeff_flag_offset[2][6] = {
|
|
| 1001 |
- { 166+0, 166+15, 166+29, 166+44, 166+47, 417 },
|
|
| 1002 |
- { 338+0, 338+15, 338+29, 338+44, 338+47, 451 }
|
|
| 1000 |
+ static const int last_coeff_flag_offset[2][14] = {
|
|
| 1001 |
+ { 166+0, 166+15, 166+29, 166+44, 166+47, 417, 572+0, 572+15, 572+29, 690, 616+0, 616+15, 616+29, 748 },
|
|
| 1002 |
+ { 338+0, 338+15, 338+29, 338+44, 338+47, 451, 864+0, 864+15, 864+29, 699, 908+0, 908+15, 908+29, 757 }
|
|
| 1003 | 1003 |
}; |
| 1004 |
- static const int coeff_abs_level_m1_offset[6] = {
|
|
| 1005 |
- 227+0, 227+10, 227+20, 227+30, 227+39, 426 |
|
| 1004 |
+ static const int coeff_abs_level_m1_offset[14] = {
|
|
| 1005 |
+ 227+0, 227+10, 227+20, 227+30, 227+39, 426, 952+0, 952+10, 952+20, 708, 982+0, 982+10, 982+20, 766 |
|
| 1006 | 1006 |
}; |
| 1007 | 1007 |
static const uint8_t significant_coeff_flag_offset_8x8[2][63] = {
|
| 1008 | 1008 |
{ 0, 1, 2, 3, 4, 5, 5, 4, 4, 3, 3, 4, 4, 4, 5, 5,
|
| ... | ... |
@@ -1057,7 +1630,7 @@ static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCT |
| 1057 | 1057 |
abs_level_m1_ctx_base = h->cabac_state |
| 1058 | 1058 |
+ coeff_abs_level_m1_offset[cat]; |
| 1059 | 1059 |
|
| 1060 |
- if( !is_dc && cat == 5 ) {
|
|
| 1060 |
+ if( !is_dc && max_coeff == 64 ) {
|
|
| 1061 | 1061 |
#define DECODE_SIGNIFICANCE( coefs, sig_off, last_off ) \ |
| 1062 | 1062 |
for(last= 0; last < coefs; last++) { \
|
| 1063 | 1063 |
uint8_t *sig_ctx = significant_coeff_ctx_base + sig_off; \ |
| ... | ... |
@@ -1075,9 +1648,11 @@ static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCT |
| 1075 | 1075 |
} |
| 1076 | 1076 |
const uint8_t *sig_off = significant_coeff_flag_offset_8x8[MB_FIELD]; |
| 1077 | 1077 |
#if ARCH_X86 && HAVE_7REGS && HAVE_EBX_AVAILABLE && !defined(BROKEN_RELOCATIONS) |
| 1078 |
- coeff_count= decode_significance_8x8_x86(CC, significant_coeff_ctx_base, index, sig_off); |
|
| 1078 |
+ coeff_count= decode_significance_8x8_x86(CC, significant_coeff_ctx_base, index, |
|
| 1079 |
+ last_coeff_ctx_base-significant_coeff_ctx_base, sig_off); |
|
| 1079 | 1080 |
} else {
|
| 1080 |
- coeff_count= decode_significance_x86(CC, max_coeff, significant_coeff_ctx_base, index); |
|
| 1081 |
+ coeff_count= decode_significance_x86(CC, max_coeff, significant_coeff_ctx_base, index, |
|
| 1082 |
+ last_coeff_ctx_base-significant_coeff_ctx_base); |
|
| 1081 | 1083 |
#else |
| 1082 | 1084 |
DECODE_SIGNIFICANCE( 63, sig_off[last], last_coeff_flag_offset_8x8[last] ); |
| 1083 | 1085 |
} else {
|
| ... | ... |
@@ -1087,16 +1662,16 @@ static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCT |
| 1087 | 1087 |
assert(coeff_count > 0); |
| 1088 | 1088 |
|
| 1089 | 1089 |
if( is_dc ) {
|
| 1090 |
- if( cat == 0 ) |
|
| 1091 |
- h->cbp_table[h->mb_xy] |= 0x100; |
|
| 1092 |
- else |
|
| 1090 |
+ if( cat == 3 ) |
|
| 1093 | 1091 |
h->cbp_table[h->mb_xy] |= 0x40 << (n - CHROMA_DC_BLOCK_INDEX); |
| 1092 |
+ else |
|
| 1093 |
+ h->cbp_table[h->mb_xy] |= 0x100 << (n - LUMA_DC_BLOCK_INDEX); |
|
| 1094 | 1094 |
h->non_zero_count_cache[scan8[n]] = coeff_count; |
| 1095 | 1095 |
} else {
|
| 1096 |
- if( cat == 5 ) |
|
| 1096 |
+ if( max_coeff == 64 ) |
|
| 1097 | 1097 |
fill_rectangle(&h->non_zero_count_cache[scan8[n]], 2, 2, 8, coeff_count, 1); |
| 1098 | 1098 |
else {
|
| 1099 |
- assert( cat == 1 || cat == 2 || cat == 4 ); |
|
| 1099 |
+ assert( cat == 1 || cat == 2 || cat == 4 || cat == 7 || cat == 8 || cat == 11 || cat == 12 ); |
|
| 1100 | 1100 |
h->non_zero_count_cache[scan8[n]] = coeff_count; |
| 1101 | 1101 |
} |
| 1102 | 1102 |
} |
| ... | ... |
@@ -1180,7 +1755,7 @@ static void decode_cabac_residual_nondc_internal( H264Context *h, DCTELEM *block |
| 1180 | 1180 |
|
| 1181 | 1181 |
static av_always_inline void decode_cabac_residual_dc( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, int max_coeff ) {
|
| 1182 | 1182 |
/* read coded block flag */ |
| 1183 |
- if( get_cabac( &h->cabac, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n, 1 ) ] ) == 0 ) {
|
|
| 1183 |
+ if( get_cabac( &h->cabac, &h->cabac_state[get_cabac_cbf_ctx( h, cat, n, max_coeff, 1 ) ] ) == 0 ) {
|
|
| 1184 | 1184 |
h->non_zero_count_cache[scan8[n]] = 0; |
| 1185 | 1185 |
return; |
| 1186 | 1186 |
} |
| ... | ... |
@@ -1189,13 +1764,68 @@ static av_always_inline void decode_cabac_residual_dc( H264Context *h, DCTELEM * |
| 1189 | 1189 |
|
| 1190 | 1190 |
static av_always_inline void decode_cabac_residual_nondc( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff ) {
|
| 1191 | 1191 |
/* read coded block flag */ |
| 1192 |
- if( cat != 5 && get_cabac( &h->cabac, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n, 0 ) ] ) == 0 ) {
|
|
| 1193 |
- h->non_zero_count_cache[scan8[n]] = 0; |
|
| 1192 |
+ if( (cat != 5 || CHROMA444) && get_cabac( &h->cabac, &h->cabac_state[get_cabac_cbf_ctx( h, cat, n, max_coeff, 0 ) ] ) == 0 ) {
|
|
| 1193 |
+ if( max_coeff == 64 ) {
|
|
| 1194 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[n]], 2, 2, 8, 0, 1); |
|
| 1195 |
+ } else {
|
|
| 1196 |
+ h->non_zero_count_cache[scan8[n]] = 0; |
|
| 1197 |
+ } |
|
| 1194 | 1198 |
return; |
| 1195 | 1199 |
} |
| 1196 | 1200 |
decode_cabac_residual_nondc_internal( h, block, cat, n, scantable, qmul, max_coeff ); |
| 1197 | 1201 |
} |
| 1198 | 1202 |
|
| 1203 |
+static av_always_inline void decode_cabac_luma_residual( H264Context *h, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p ) |
|
| 1204 |
+{
|
|
| 1205 |
+ static const uint8_t ctx_cat[4][3] = {{0,6,10},{1,7,11},{2,8,12},{5,9,13}};
|
|
| 1206 |
+ const uint32_t *qmul; |
|
| 1207 |
+ int i8x8, i4x4; |
|
| 1208 |
+ MpegEncContext * const s = &h->s; |
|
| 1209 |
+ int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1]; |
|
| 1210 |
+ if( IS_INTRA16x16( mb_type ) ) {
|
|
| 1211 |
+ //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 DC\n" ); |
|
| 1212 |
+ AV_ZERO128(h->mb_luma_dc[p]+0); |
|
| 1213 |
+ AV_ZERO128(h->mb_luma_dc[p]+8); |
|
| 1214 |
+ AV_ZERO128(h->mb_luma_dc[p]+16); |
|
| 1215 |
+ AV_ZERO128(h->mb_luma_dc[p]+24); |
|
| 1216 |
+ decode_cabac_residual_dc(h, h->mb_luma_dc[p], ctx_cat[0][p], LUMA_DC_BLOCK_INDEX+p, scan, 16); |
|
| 1217 |
+ |
|
| 1218 |
+ if( cbp&15 ) {
|
|
| 1219 |
+ qmul = h->dequant4_coeff[p][qscale]; |
|
| 1220 |
+ for( i4x4 = 0; i4x4 < 16; i4x4++ ) {
|
|
| 1221 |
+ const int index = 16*p + i4x4; |
|
| 1222 |
+ //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 AC:%d\n", index ); |
|
| 1223 |
+ decode_cabac_residual_nondc(h, h->mb + (16*index << pixel_shift), ctx_cat[1][p], index, scan + 1, qmul, 15); |
|
| 1224 |
+ } |
|
| 1225 |
+ } else {
|
|
| 1226 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[16*p]], 4, 4, 8, 0, 1); |
|
| 1227 |
+ } |
|
| 1228 |
+ } else {
|
|
| 1229 |
+ int cqm = (IS_INTRA( mb_type ) ? 0:3) + p; |
|
| 1230 |
+ for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
|
|
| 1231 |
+ if( cbp & (1<<i8x8) ) {
|
|
| 1232 |
+ if( IS_8x8DCT(mb_type) ) {
|
|
| 1233 |
+ const int index = 16*p + 4*i8x8; |
|
| 1234 |
+ decode_cabac_residual_nondc(h, h->mb + (16*index << pixel_shift), ctx_cat[3][p], index, |
|
| 1235 |
+ scan8x8, h->dequant8_coeff[cqm][qscale], 64); |
|
| 1236 |
+ } else {
|
|
| 1237 |
+ qmul = h->dequant4_coeff[cqm][qscale]; |
|
| 1238 |
+ for( i4x4 = 0; i4x4 < 4; i4x4++ ) {
|
|
| 1239 |
+ const int index = 16*p + 4*i8x8 + i4x4; |
|
| 1240 |
+ //av_log( s->avctx, AV_LOG_ERROR, "Luma4x4: %d\n", index ); |
|
| 1241 |
+//START_TIMER |
|
| 1242 |
+ decode_cabac_residual_nondc(h, h->mb + (16*index << pixel_shift), ctx_cat[2][p], index, scan, qmul, 16); |
|
| 1243 |
+//STOP_TIMER("decode_residual")
|
|
| 1244 |
+ } |
|
| 1245 |
+ } |
|
| 1246 |
+ } else {
|
|
| 1247 |
+ uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8+16*p] ]; |
|
| 1248 |
+ nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0; |
|
| 1249 |
+ } |
|
| 1250 |
+ } |
|
| 1251 |
+ } |
|
| 1252 |
+} |
|
| 1253 |
+ |
|
| 1199 | 1254 |
/** |
| 1200 | 1255 |
* decodes a macroblock |
| 1201 | 1256 |
* @return 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed |
| ... | ... |
@@ -1205,6 +1835,7 @@ int ff_h264_decode_mb_cabac(H264Context *h) {
|
| 1205 | 1205 |
int mb_xy; |
| 1206 | 1206 |
int mb_type, partition_count, cbp = 0; |
| 1207 | 1207 |
int dct8x8_allowed= h->pps.transform_8x8_mode; |
| 1208 |
+ int decode_chroma = h->sps.chroma_format_idc == 1 || h->sps.chroma_format_idc == 2; |
|
| 1208 | 1209 |
const int pixel_shift = h->pixel_shift; |
| 1209 | 1210 |
|
| 1210 | 1211 |
mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride; |
| ... | ... |
@@ -1314,7 +1945,8 @@ decode_intra_mb: |
| 1314 | 1314 |
h->slice_table[ mb_xy ]= h->slice_num; |
| 1315 | 1315 |
|
| 1316 | 1316 |
if(IS_INTRA_PCM(mb_type)) {
|
| 1317 |
- const int mb_size = (384*h->sps.bit_depth_luma) >> 3; |
|
| 1317 |
+ static const uint16_t mb_sizes[4] = {256,384,512,768};
|
|
| 1318 |
+ const int mb_size = mb_sizes[h->sps.chroma_format_idc]*h->sps.bit_depth_luma >> 3; |
|
| 1318 | 1319 |
const uint8_t *ptr; |
| 1319 | 1320 |
|
| 1320 | 1321 |
// We assume these blocks are very rare so we do not optimize it. |
| ... | ... |
@@ -1327,20 +1959,17 @@ decode_intra_mb: |
| 1327 | 1327 |
} |
| 1328 | 1328 |
|
| 1329 | 1329 |
// The pixels are stored in the same order as levels in h->mb array. |
| 1330 |
- memcpy(h->mb, ptr, 2*mb_size/3); ptr+=2*mb_size/3; |
|
| 1331 |
- if(CHROMA){
|
|
| 1332 |
- memcpy(h->mb+mb_size/3, ptr, mb_size/3); ptr+=mb_size/3; |
|
| 1333 |
- } |
|
| 1330 |
+ memcpy(h->mb, ptr, mb_size); ptr+=mb_size; |
|
| 1334 | 1331 |
|
| 1335 | 1332 |
ff_init_cabac_decoder(&h->cabac, ptr, h->cabac.bytestream_end - ptr); |
| 1336 | 1333 |
|
| 1337 | 1334 |
// All blocks are present |
| 1338 |
- h->cbp_table[mb_xy] = 0x1ef; |
|
| 1335 |
+ h->cbp_table[mb_xy] = 0xf7ef; |
|
| 1339 | 1336 |
h->chroma_pred_mode_table[mb_xy] = 0; |
| 1340 | 1337 |
// In deblocking, the quantizer is 0 |
| 1341 | 1338 |
s->current_picture.qscale_table[mb_xy]= 0; |
| 1342 | 1339 |
// All coeffs are present |
| 1343 |
- memset(h->non_zero_count[mb_xy], 16, 32); |
|
| 1340 |
+ memset(h->non_zero_count[mb_xy], 16, 48); |
|
| 1344 | 1341 |
s->current_picture.mb_type[mb_xy]= mb_type; |
| 1345 | 1342 |
h->last_qscale_diff = 0; |
| 1346 | 1343 |
return 0; |
| ... | ... |
@@ -1377,7 +2006,7 @@ decode_intra_mb: |
| 1377 | 1377 |
h->intra16x16_pred_mode= ff_h264_check_intra_pred_mode( h, h->intra16x16_pred_mode ); |
| 1378 | 1378 |
if( h->intra16x16_pred_mode < 0 ) return -1; |
| 1379 | 1379 |
} |
| 1380 |
- if(CHROMA){
|
|
| 1380 |
+ if(decode_chroma){
|
|
| 1381 | 1381 |
h->chroma_pred_mode_table[mb_xy] = |
| 1382 | 1382 |
pred_mode = decode_cabac_mb_chroma_pre_mode( h ); |
| 1383 | 1383 |
|
| ... | ... |
@@ -1606,7 +2235,7 @@ decode_intra_mb: |
| 1606 | 1606 |
|
| 1607 | 1607 |
if( !IS_INTRA16x16( mb_type ) ) {
|
| 1608 | 1608 |
cbp = decode_cabac_mb_cbp_luma( h ); |
| 1609 |
- if(CHROMA) |
|
| 1609 |
+ if(decode_chroma) |
|
| 1610 | 1610 |
cbp |= decode_cabac_mb_cbp_chroma( h ) << 4; |
| 1611 | 1611 |
} |
| 1612 | 1612 |
|
| ... | ... |
@@ -1615,6 +2244,28 @@ decode_intra_mb: |
| 1615 | 1615 |
if( dct8x8_allowed && (cbp&15) && !IS_INTRA( mb_type ) ) {
|
| 1616 | 1616 |
mb_type |= MB_TYPE_8x8DCT * get_cabac_noinline( &h->cabac, &h->cabac_state[399 + h->neighbor_transform_size] ); |
| 1617 | 1617 |
} |
| 1618 |
+ |
|
| 1619 |
+ /* It would be better to do this in fill_decode_caches, but we don't know |
|
| 1620 |
+ * the transform mode of the current macroblock there. */ |
|
| 1621 |
+ if (CHROMA444 && IS_8x8DCT(mb_type)){
|
|
| 1622 |
+ int i; |
|
| 1623 |
+ for (i = 0; i < 2; i++){
|
|
| 1624 |
+ if (h->left_type[i] && !IS_8x8DCT(h->left_type[i])){
|
|
| 1625 |
+ h->non_zero_count_cache[3+8* 1 + 2*8*i]= |
|
| 1626 |
+ h->non_zero_count_cache[3+8* 2 + 2*8*i]= |
|
| 1627 |
+ h->non_zero_count_cache[3+8* 6 + 2*8*i]= |
|
| 1628 |
+ h->non_zero_count_cache[3+8* 7 + 2*8*i]= |
|
| 1629 |
+ h->non_zero_count_cache[3+8*11 + 2*8*i]= |
|
| 1630 |
+ h->non_zero_count_cache[3+8*12 + 2*8*i]= IS_INTRA(mb_type) ? 64 : 0; |
|
| 1631 |
+ } |
|
| 1632 |
+ } |
|
| 1633 |
+ if (h->top_type && !IS_8x8DCT(h->top_type)){
|
|
| 1634 |
+ uint32_t top_empty = CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040; |
|
| 1635 |
+ AV_WN32A(&h->non_zero_count_cache[4+8* 0], top_empty); |
|
| 1636 |
+ AV_WN32A(&h->non_zero_count_cache[4+8* 5], top_empty); |
|
| 1637 |
+ AV_WN32A(&h->non_zero_count_cache[4+8*10], top_empty); |
|
| 1638 |
+ } |
|
| 1639 |
+ } |
|
| 1618 | 1640 |
s->current_picture.mb_type[mb_xy]= mb_type; |
| 1619 | 1641 |
|
| 1620 | 1642 |
if( cbp || IS_INTRA16x16( mb_type ) ) {
|
| ... | ... |
@@ -1659,76 +2310,38 @@ decode_intra_mb: |
| 1659 | 1659 |
}else |
| 1660 | 1660 |
h->last_qscale_diff=0; |
| 1661 | 1661 |
|
| 1662 |
- if( IS_INTRA16x16( mb_type ) ) {
|
|
| 1663 |
- int i; |
|
| 1664 |
- //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 DC\n" ); |
|
| 1665 |
- AV_ZERO128(h->mb_luma_dc+0); |
|
| 1666 |
- AV_ZERO128(h->mb_luma_dc+8); |
|
| 1667 |
- AV_ZERO128(h->mb_luma_dc+16); |
|
| 1668 |
- AV_ZERO128(h->mb_luma_dc+24); |
|
| 1669 |
- decode_cabac_residual_dc( h, h->mb_luma_dc, 0, LUMA_DC_BLOCK_INDEX, scan, 16); |
|
| 1670 |
- |
|
| 1671 |
- if( cbp&15 ) {
|
|
| 1672 |
- qmul = h->dequant4_coeff[0][s->qscale]; |
|
| 1673 |
- for( i = 0; i < 16; i++ ) {
|
|
| 1674 |
- //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 AC:%d\n", i ); |
|
| 1675 |
- decode_cabac_residual_nondc(h, h->mb + (16*i << pixel_shift), 1, i, scan + 1, qmul, 15); |
|
| 1676 |
- } |
|
| 1677 |
- } else {
|
|
| 1678 |
- fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1); |
|
| 1679 |
- } |
|
| 1662 |
+ decode_cabac_luma_residual(h, scan, scan8x8, pixel_shift, mb_type, cbp, 0); |
|
| 1663 |
+ if(CHROMA444){
|
|
| 1664 |
+ decode_cabac_luma_residual(h, scan, scan8x8, pixel_shift, mb_type, cbp, 1); |
|
| 1665 |
+ decode_cabac_luma_residual(h, scan, scan8x8, pixel_shift, mb_type, cbp, 2); |
|
| 1680 | 1666 |
} else {
|
| 1681 |
- int i8x8, i4x4; |
|
| 1682 |
- for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
|
|
| 1683 |
- if( cbp & (1<<i8x8) ) {
|
|
| 1684 |
- if( IS_8x8DCT(mb_type) ) {
|
|
| 1685 |
- decode_cabac_residual_nondc(h, h->mb + (64*i8x8 << pixel_shift), 5, 4*i8x8, |
|
| 1686 |
- scan8x8, h->dequant8_coeff[IS_INTRA( mb_type ) ? 0:1][s->qscale], 64); |
|
| 1687 |
- } else {
|
|
| 1688 |
- qmul = h->dequant4_coeff[IS_INTRA( mb_type ) ? 0:3][s->qscale]; |
|
| 1689 |
- for( i4x4 = 0; i4x4 < 4; i4x4++ ) {
|
|
| 1690 |
- const int index = 4*i8x8 + i4x4; |
|
| 1691 |
- //av_log( s->avctx, AV_LOG_ERROR, "Luma4x4: %d\n", index ); |
|
| 1692 |
-//START_TIMER |
|
| 1693 |
- decode_cabac_residual_nondc(h, h->mb + (16*index << pixel_shift), 2, index, scan, qmul, 16); |
|
| 1694 |
-//STOP_TIMER("decode_residual")
|
|
| 1695 |
- } |
|
| 1696 |
- } |
|
| 1697 |
- } else {
|
|
| 1698 |
- uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ]; |
|
| 1699 |
- nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0; |
|
| 1667 |
+ if( cbp&0x30 ){
|
|
| 1668 |
+ int c; |
|
| 1669 |
+ for( c = 0; c < 2; c++ ) {
|
|
| 1670 |
+ //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-DC\n",c ); |
|
| 1671 |
+ decode_cabac_residual_dc(h, h->mb + ((256 + 16*16*c) << pixel_shift), 3, CHROMA_DC_BLOCK_INDEX+c, chroma_dc_scan, 4); |
|
| 1700 | 1672 |
} |
| 1701 | 1673 |
} |
| 1702 |
- } |
|
| 1703 | 1674 |
|
| 1704 |
- if( cbp&0x30 ){
|
|
| 1705 |
- int c; |
|
| 1706 |
- for( c = 0; c < 2; c++ ) {
|
|
| 1707 |
- //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-DC\n",c ); |
|
| 1708 |
- decode_cabac_residual_dc(h, h->mb + ((256 + 16*4*c) << pixel_shift), 3, CHROMA_DC_BLOCK_INDEX+c, chroma_dc_scan, 4); |
|
| 1709 |
- } |
|
| 1710 |
- } |
|
| 1711 |
- |
|
| 1712 |
- if( cbp&0x20 ) {
|
|
| 1713 |
- int c, i; |
|
| 1714 |
- for( c = 0; c < 2; c++ ) {
|
|
| 1715 |
- qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[c]]; |
|
| 1716 |
- for( i = 0; i < 4; i++ ) {
|
|
| 1717 |
- const int index = 16 + 4 * c + i; |
|
| 1718 |
- //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16 ); |
|
| 1719 |
- decode_cabac_residual_nondc(h, h->mb + (16*index << pixel_shift), 4, index, scan + 1, qmul, 15); |
|
| 1675 |
+ if( cbp&0x20 ) {
|
|
| 1676 |
+ int c, i; |
|
| 1677 |
+ for( c = 0; c < 2; c++ ) {
|
|
| 1678 |
+ qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[c]]; |
|
| 1679 |
+ for( i = 0; i < 4; i++ ) {
|
|
| 1680 |
+ const int index = 16 + 16 * c + i; |
|
| 1681 |
+ //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16 ); |
|
| 1682 |
+ decode_cabac_residual_nondc(h, h->mb + (16*index << pixel_shift), 4, index, scan + 1, qmul, 15); |
|
| 1683 |
+ } |
|
| 1720 | 1684 |
} |
| 1685 |
+ } else {
|
|
| 1686 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1); |
|
| 1687 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1); |
|
| 1721 | 1688 |
} |
| 1722 |
- } else {
|
|
| 1723 |
- uint8_t * const nnz= &h->non_zero_count_cache[0]; |
|
| 1724 |
- nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] = |
|
| 1725 |
- nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0; |
|
| 1726 | 1689 |
} |
| 1727 | 1690 |
} else {
|
| 1728 |
- uint8_t * const nnz= &h->non_zero_count_cache[0]; |
|
| 1729 |
- fill_rectangle(&nnz[scan8[0]], 4, 4, 8, 0, 1); |
|
| 1730 |
- nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] = |
|
| 1731 |
- nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0; |
|
| 1691 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[ 0]], 4, 4, 8, 0, 1); |
|
| 1692 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1); |
|
| 1693 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1); |
|
| 1732 | 1694 |
h->last_qscale_diff = 0; |
| 1733 | 1695 |
} |
| 1734 | 1696 |
|
| ... | ... |
@@ -371,12 +371,12 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in |
| 371 | 371 |
|
| 372 | 372 |
//FIXME put trailing_onex into the context |
| 373 | 373 |
|
| 374 |
- if(n >= CHROMA_DC_BLOCK_INDEX){
|
|
| 374 |
+ if(max_coeff <= 8){
|
|
| 375 | 375 |
coeff_token= get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1); |
| 376 | 376 |
total_coeff= coeff_token>>2; |
| 377 | 377 |
}else{
|
| 378 |
- if(n == LUMA_DC_BLOCK_INDEX){
|
|
| 379 |
- total_coeff= pred_non_zero_count(h, 0); |
|
| 378 |
+ if(n >= LUMA_DC_BLOCK_INDEX){
|
|
| 379 |
+ total_coeff= pred_non_zero_count(h, (n - LUMA_DC_BLOCK_INDEX)*16); |
|
| 380 | 380 |
coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2); |
| 381 | 381 |
total_coeff= coeff_token>>2; |
| 382 | 382 |
}else{
|
| ... | ... |
@@ -482,7 +482,8 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in |
| 482 | 482 |
if(total_coeff == max_coeff) |
| 483 | 483 |
zeros_left=0; |
| 484 | 484 |
else{
|
| 485 |
- if(n >= CHROMA_DC_BLOCK_INDEX) |
|
| 485 |
+ /* FIXME: we don't actually support 4:2:2 yet. */ |
|
| 486 |
+ if(max_coeff <= 8) |
|
| 486 | 487 |
zeros_left= get_vlc2(gb, (chroma_dc_total_zeros_vlc-1)[ total_coeff ].table, CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1); |
| 487 | 488 |
else |
| 488 | 489 |
zeros_left= get_vlc2(gb, (total_zeros_vlc-1)[ total_coeff ].table, TOTAL_ZEROS_VLC_BITS, 1); |
| ... | ... |
@@ -536,12 +537,80 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in |
| 536 | 536 |
return 0; |
| 537 | 537 |
} |
| 538 | 538 |
|
| 539 |
+static av_always_inline int decode_luma_residual(H264Context *h, GetBitContext *gb, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p){
|
|
| 540 |
+ int i4x4, i8x8; |
|
| 541 |
+ MpegEncContext * const s = &h->s; |
|
| 542 |
+ int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1]; |
|
| 543 |
+ if(IS_INTRA16x16(mb_type)){
|
|
| 544 |
+ AV_ZERO128(h->mb_luma_dc[p]+0); |
|
| 545 |
+ AV_ZERO128(h->mb_luma_dc[p]+8); |
|
| 546 |
+ AV_ZERO128(h->mb_luma_dc[p]+16); |
|
| 547 |
+ AV_ZERO128(h->mb_luma_dc[p]+24); |
|
| 548 |
+ if( decode_residual(h, h->intra_gb_ptr, h->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX+p, scan, NULL, 16) < 0){
|
|
| 549 |
+ return -1; //FIXME continue if partitioned and other return -1 too |
|
| 550 |
+ } |
|
| 551 |
+ |
|
| 552 |
+ assert((cbp&15) == 0 || (cbp&15) == 15); |
|
| 553 |
+ |
|
| 554 |
+ if(cbp&15){
|
|
| 555 |
+ for(i8x8=0; i8x8<4; i8x8++){
|
|
| 556 |
+ for(i4x4=0; i4x4<4; i4x4++){
|
|
| 557 |
+ const int index= i4x4 + 4*i8x8 + p*16; |
|
| 558 |
+ if( decode_residual(h, h->intra_gb_ptr, h->mb + (16*index << pixel_shift), |
|
| 559 |
+ index, scan + 1, h->dequant4_coeff[p][qscale], 15) < 0 ){
|
|
| 560 |
+ return -1; |
|
| 561 |
+ } |
|
| 562 |
+ } |
|
| 563 |
+ } |
|
| 564 |
+ return 0xf; |
|
| 565 |
+ }else{
|
|
| 566 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[p*16]], 4, 4, 8, 0, 1); |
|
| 567 |
+ return 0; |
|
| 568 |
+ } |
|
| 569 |
+ }else{
|
|
| 570 |
+ int cqm = (IS_INTRA( mb_type ) ? 0:3)+p; |
|
| 571 |
+ /* For CAVLC 4:4:4, we need to keep track of the luma 8x8 CBP for deblocking nnz purposes. */ |
|
| 572 |
+ int new_cbp = 0; |
|
| 573 |
+ for(i8x8=0; i8x8<4; i8x8++){
|
|
| 574 |
+ if(cbp & (1<<i8x8)){
|
|
| 575 |
+ if(IS_8x8DCT(mb_type)){
|
|
| 576 |
+ DCTELEM *buf = &h->mb[64*i8x8+256*p << pixel_shift]; |
|
| 577 |
+ uint8_t *nnz; |
|
| 578 |
+ for(i4x4=0; i4x4<4; i4x4++){
|
|
| 579 |
+ const int index= i4x4 + 4*i8x8 + p*16; |
|
| 580 |
+ if( decode_residual(h, gb, buf, index, scan8x8+16*i4x4, |
|
| 581 |
+ h->dequant8_coeff[cqm][qscale], 16) < 0 ) |
|
| 582 |
+ return -1; |
|
| 583 |
+ } |
|
| 584 |
+ nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ]; |
|
| 585 |
+ nnz[0] += nnz[1] + nnz[8] + nnz[9]; |
|
| 586 |
+ new_cbp |= !!nnz[0] << i8x8; |
|
| 587 |
+ }else{
|
|
| 588 |
+ for(i4x4=0; i4x4<4; i4x4++){
|
|
| 589 |
+ const int index= i4x4 + 4*i8x8 + p*16; |
|
| 590 |
+ if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, |
|
| 591 |
+ scan, h->dequant4_coeff[cqm][qscale], 16) < 0 ){
|
|
| 592 |
+ return -1; |
|
| 593 |
+ } |
|
| 594 |
+ new_cbp |= h->non_zero_count_cache[ scan8[index] ] << i8x8; |
|
| 595 |
+ } |
|
| 596 |
+ } |
|
| 597 |
+ }else{
|
|
| 598 |
+ uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ]; |
|
| 599 |
+ nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0; |
|
| 600 |
+ } |
|
| 601 |
+ } |
|
| 602 |
+ return new_cbp; |
|
| 603 |
+ } |
|
| 604 |
+} |
|
| 605 |
+ |
|
| 539 | 606 |
int ff_h264_decode_mb_cavlc(H264Context *h){
|
| 540 | 607 |
MpegEncContext * const s = &h->s; |
| 541 | 608 |
int mb_xy; |
| 542 | 609 |
int partition_count; |
| 543 | 610 |
unsigned int mb_type, cbp; |
| 544 | 611 |
int dct8x8_allowed= h->pps.transform_8x8_mode; |
| 612 |
+ int decode_chroma = h->sps.chroma_format_idc == 1 || h->sps.chroma_format_idc == 2; |
|
| 545 | 613 |
const int pixel_shift = h->pixel_shift; |
| 546 | 614 |
|
| 547 | 615 |
mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride; |
| ... | ... |
@@ -608,19 +677,21 @@ decode_intra_mb: |
| 608 | 608 |
|
| 609 | 609 |
if(IS_INTRA_PCM(mb_type)){
|
| 610 | 610 |
unsigned int x; |
| 611 |
+ static const uint16_t mb_sizes[4] = {256,384,512,768};
|
|
| 612 |
+ const int mb_size = mb_sizes[h->sps.chroma_format_idc]*h->sps.bit_depth_luma >> 3; |
|
| 611 | 613 |
|
| 612 | 614 |
// We assume these blocks are very rare so we do not optimize it. |
| 613 | 615 |
align_get_bits(&s->gb); |
| 614 | 616 |
|
| 615 | 617 |
// The pixels are stored in the same order as levels in h->mb array. |
| 616 |
- for(x=0; x < (CHROMA ? 384 : 256)*h->sps.bit_depth_luma/8; x++){
|
|
| 618 |
+ for(x=0; x < mb_size; x++){
|
|
| 617 | 619 |
((uint8_t*)h->mb)[x]= get_bits(&s->gb, 8); |
| 618 | 620 |
} |
| 619 | 621 |
|
| 620 | 622 |
// In deblocking, the quantizer is 0 |
| 621 | 623 |
s->current_picture.qscale_table[mb_xy]= 0; |
| 622 | 624 |
// All coeffs are present |
| 623 |
- memset(h->non_zero_count[mb_xy], 16, 32); |
|
| 625 |
+ memset(h->non_zero_count[mb_xy], 16, 48); |
|
| 624 | 626 |
|
| 625 | 627 |
s->current_picture.mb_type[mb_xy]= mb_type; |
| 626 | 628 |
return 0; |
| ... | ... |
@@ -668,7 +739,7 @@ decode_intra_mb: |
| 668 | 668 |
if(h->intra16x16_pred_mode < 0) |
| 669 | 669 |
return -1; |
| 670 | 670 |
} |
| 671 |
- if(CHROMA){
|
|
| 671 |
+ if(decode_chroma){
|
|
| 672 | 672 |
pred_mode= ff_h264_check_intra_pred_mode(h, get_ue_golomb_31(&s->gb)); |
| 673 | 673 |
if(pred_mode < 0) |
| 674 | 674 |
return -1; |
| ... | ... |
@@ -896,15 +967,19 @@ decode_intra_mb: |
| 896 | 896 |
|
| 897 | 897 |
if(!IS_INTRA16x16(mb_type)){
|
| 898 | 898 |
cbp= get_ue_golomb(&s->gb); |
| 899 |
- if(cbp > 47){
|
|
| 900 |
- av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y); |
|
| 901 |
- return -1; |
|
| 902 |
- } |
|
| 903 | 899 |
|
| 904 |
- if(CHROMA){
|
|
| 900 |
+ if(decode_chroma){
|
|
| 901 |
+ if(cbp > 47){
|
|
| 902 |
+ av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y); |
|
| 903 |
+ return -1; |
|
| 904 |
+ } |
|
| 905 | 905 |
if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp[cbp]; |
| 906 | 906 |
else cbp= golomb_to_inter_cbp [cbp]; |
| 907 | 907 |
}else{
|
| 908 |
+ if(cbp > 15){
|
|
| 909 |
+ av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y); |
|
| 910 |
+ return -1; |
|
| 911 |
+ } |
|
| 908 | 912 |
if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp_gray[cbp]; |
| 909 | 913 |
else cbp= golomb_to_inter_cbp_gray[cbp]; |
| 910 | 914 |
} |
| ... | ... |
@@ -918,8 +993,9 @@ decode_intra_mb: |
| 918 | 918 |
s->current_picture.mb_type[mb_xy]= mb_type; |
| 919 | 919 |
|
| 920 | 920 |
if(cbp || IS_INTRA16x16(mb_type)){
|
| 921 |
- int i8x8, i4x4, chroma_idx; |
|
| 921 |
+ int i4x4, chroma_idx; |
|
| 922 | 922 |
int dquant; |
| 923 |
+ int ret; |
|
| 923 | 924 |
GetBitContext *gb= IS_INTRA(mb_type) ? h->intra_gb_ptr : h->inter_gb_ptr; |
| 924 | 925 |
const uint8_t *scan, *scan8x8; |
| 925 | 926 |
const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8); |
| ... | ... |
@@ -947,85 +1023,45 @@ decode_intra_mb: |
| 947 | 947 |
|
| 948 | 948 |
h->chroma_qp[0]= get_chroma_qp(h, 0, s->qscale); |
| 949 | 949 |
h->chroma_qp[1]= get_chroma_qp(h, 1, s->qscale); |
| 950 |
- if(IS_INTRA16x16(mb_type)){
|
|
| 951 |
- AV_ZERO128(h->mb_luma_dc+0); |
|
| 952 |
- AV_ZERO128(h->mb_luma_dc+8); |
|
| 953 |
- AV_ZERO128(h->mb_luma_dc+16); |
|
| 954 |
- AV_ZERO128(h->mb_luma_dc+24); |
|
| 955 |
- if( decode_residual(h, h->intra_gb_ptr, h->mb_luma_dc, LUMA_DC_BLOCK_INDEX, scan, h->dequant4_coeff[0][s->qscale], 16) < 0){
|
|
| 956 |
- return -1; //FIXME continue if partitioned and other return -1 too |
|
| 957 |
- } |
|
| 958 | 950 |
|
| 959 |
- assert((cbp&15) == 0 || (cbp&15) == 15); |
|
| 951 |
+ if( (ret = decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 0)) < 0 ){
|
|
| 952 |
+ return -1; |
|
| 953 |
+ } |
|
| 954 |
+ h->cbp_table[mb_xy] |= ret << 12; |
|
| 955 |
+ if(CHROMA444){
|
|
| 956 |
+ if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 1) < 0 ){
|
|
| 957 |
+ return -1; |
|
| 958 |
+ } |
|
| 959 |
+ if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 2) < 0 ){
|
|
| 960 |
+ return -1; |
|
| 961 |
+ } |
|
| 962 |
+ } else {
|
|
| 963 |
+ if(cbp&0x30){
|
|
| 964 |
+ for(chroma_idx=0; chroma_idx<2; chroma_idx++) |
|
| 965 |
+ if( decode_residual(h, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift), CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma_dc_scan, NULL, 4) < 0){
|
|
| 966 |
+ return -1; |
|
| 967 |
+ } |
|
| 968 |
+ } |
|
| 960 | 969 |
|
| 961 |
- if(cbp&15){
|
|
| 962 |
- for(i8x8=0; i8x8<4; i8x8++){
|
|
| 970 |
+ if(cbp&0x20){
|
|
| 971 |
+ for(chroma_idx=0; chroma_idx<2; chroma_idx++){
|
|
| 972 |
+ const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]]; |
|
| 963 | 973 |
for(i4x4=0; i4x4<4; i4x4++){
|
| 964 |
- const int index= i4x4 + 4*i8x8; |
|
| 965 |
- if( decode_residual(h, h->intra_gb_ptr, h->mb + (16*index << pixel_shift), index, scan + 1, h->dequant4_coeff[0][s->qscale], 15) < 0 ){
|
|
| 974 |
+ const int index= 16 + 16*chroma_idx + i4x4; |
|
| 975 |
+ if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){
|
|
| 966 | 976 |
return -1; |
| 967 | 977 |
} |
| 968 | 978 |
} |
| 969 | 979 |
} |
| 970 | 980 |
}else{
|
| 971 |
- fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1); |
|
| 981 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1); |
|
| 982 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1); |
|
| 972 | 983 |
} |
| 973 |
- }else{
|
|
| 974 |
- for(i8x8=0; i8x8<4; i8x8++){
|
|
| 975 |
- if(cbp & (1<<i8x8)){
|
|
| 976 |
- if(IS_8x8DCT(mb_type)){
|
|
| 977 |
- DCTELEM *buf = &h->mb[64*i8x8 << pixel_shift]; |
|
| 978 |
- uint8_t *nnz; |
|
| 979 |
- for(i4x4=0; i4x4<4; i4x4++){
|
|
| 980 |
- if( decode_residual(h, gb, buf, i4x4+4*i8x8, scan8x8+16*i4x4, |
|
| 981 |
- h->dequant8_coeff[IS_INTRA( mb_type ) ? 0:1][s->qscale], 16) <0 ) |
|
| 982 |
- return -1; |
|
| 983 |
- } |
|
| 984 |
- nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ]; |
|
| 985 |
- nnz[0] += nnz[1] + nnz[8] + nnz[9]; |
|
| 986 |
- }else{
|
|
| 987 |
- for(i4x4=0; i4x4<4; i4x4++){
|
|
| 988 |
- const int index= i4x4 + 4*i8x8; |
|
| 989 |
- |
|
| 990 |
- if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan, h->dequant4_coeff[IS_INTRA( mb_type ) ? 0:3][s->qscale], 16) <0 ){
|
|
| 991 |
- return -1; |
|
| 992 |
- } |
|
| 993 |
- } |
|
| 994 |
- } |
|
| 995 |
- }else{
|
|
| 996 |
- uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ]; |
|
| 997 |
- nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0; |
|
| 998 |
- } |
|
| 999 |
- } |
|
| 1000 |
- } |
|
| 1001 |
- |
|
| 1002 |
- if(cbp&0x30){
|
|
| 1003 |
- for(chroma_idx=0; chroma_idx<2; chroma_idx++) |
|
| 1004 |
- if( decode_residual(h, gb, h->mb + ((256 + 16*4*chroma_idx) << pixel_shift), CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma_dc_scan, NULL, 4) < 0){
|
|
| 1005 |
- return -1; |
|
| 1006 |
- } |
|
| 1007 |
- } |
|
| 1008 |
- |
|
| 1009 |
- if(cbp&0x20){
|
|
| 1010 |
- for(chroma_idx=0; chroma_idx<2; chroma_idx++){
|
|
| 1011 |
- const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]]; |
|
| 1012 |
- for(i4x4=0; i4x4<4; i4x4++){
|
|
| 1013 |
- const int index= 16 + 4*chroma_idx + i4x4; |
|
| 1014 |
- if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){
|
|
| 1015 |
- return -1; |
|
| 1016 |
- } |
|
| 1017 |
- } |
|
| 1018 |
- } |
|
| 1019 |
- }else{
|
|
| 1020 |
- uint8_t * const nnz= &h->non_zero_count_cache[0]; |
|
| 1021 |
- nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] = |
|
| 1022 |
- nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0; |
|
| 1023 | 984 |
} |
| 1024 | 985 |
}else{
|
| 1025 |
- uint8_t * const nnz= &h->non_zero_count_cache[0]; |
|
| 1026 |
- fill_rectangle(&nnz[scan8[0]], 4, 4, 8, 0, 1); |
|
| 1027 |
- nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] = |
|
| 1028 |
- nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0; |
|
| 986 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[ 0]], 4, 4, 8, 0, 1); |
|
| 987 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1); |
|
| 988 |
+ fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1); |
|
| 1029 | 989 |
} |
| 1030 | 990 |
s->current_picture.qscale_table[mb_xy]= s->qscale; |
| 1031 | 991 |
write_back_non_zero_count(h); |
| ... | ... |
@@ -217,10 +217,11 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, |
| 217 | 217 |
int mb_xy; |
| 218 | 218 |
int mb_type, left_type; |
| 219 | 219 |
int qp, qp0, qp1, qpc, qpc0, qpc1, qp_thresh; |
| 220 |
+ int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY)); |
|
| 220 | 221 |
|
| 221 | 222 |
mb_xy = h->mb_xy; |
| 222 | 223 |
|
| 223 |
- if(!h->top_type || !h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff) {
|
|
| 224 |
+ if(!h->top_type || !h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff || CHROMA444) {
|
|
| 224 | 225 |
ff_h264_filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize); |
| 225 | 226 |
return; |
| 226 | 227 |
} |
| ... | ... |
@@ -262,16 +263,18 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, |
| 262 | 262 |
filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, h); |
| 263 | 263 |
filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, h); |
| 264 | 264 |
} |
| 265 |
- if(left_type){
|
|
| 266 |
- filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, h); |
|
| 267 |
- filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, h); |
|
| 265 |
+ if(chroma){
|
|
| 266 |
+ if(left_type){
|
|
| 267 |
+ filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, h); |
|
| 268 |
+ filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, h); |
|
| 269 |
+ } |
|
| 270 |
+ filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, h); |
|
| 271 |
+ filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, h); |
|
| 272 |
+ filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, h); |
|
| 273 |
+ filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, h); |
|
| 274 |
+ filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, h); |
|
| 275 |
+ filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, h); |
|
| 268 | 276 |
} |
| 269 |
- filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, h); |
|
| 270 |
- filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, h); |
|
| 271 |
- filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, h); |
|
| 272 |
- filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, h); |
|
| 273 |
- filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, h); |
|
| 274 |
- filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, h); |
|
| 275 | 277 |
return; |
| 276 | 278 |
} else {
|
| 277 | 279 |
LOCAL_ALIGNED_8(int16_t, bS, [2], [4][4]); |
| ... | ... |
@@ -298,7 +301,7 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, |
| 298 | 298 |
#define FILTER(hv,dir,edge)\ |
| 299 | 299 |
if(AV_RN64A(bS[dir][edge])) { \
|
| 300 | 300 |
filter_mb_edge##hv( &img_y[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qp : qp##dir, h );\ |
| 301 |
- if(!(edge&1)) {\
|
|
| 301 |
+ if(chroma && !(edge&1)) {\
|
|
| 302 | 302 |
filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h );\ |
| 303 | 303 |
filter_mb_edgec##hv( &img_cr[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h );\ |
| 304 | 304 |
}\ |
| ... | ... |
@@ -353,9 +356,10 @@ static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){
|
| 353 | 353 |
return v; |
| 354 | 354 |
} |
| 355 | 355 |
|
| 356 |
-static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int dir) {
|
|
| 356 |
+static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int chroma, int chroma444, int dir) {
|
|
| 357 | 357 |
MpegEncContext * const s = &h->s; |
| 358 | 358 |
int edge; |
| 359 |
+ int chroma_qp_avg[2]; |
|
| 359 | 360 |
const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy; |
| 360 | 361 |
const int mbm_type = dir == 0 ? h->left_type[0] : h->top_type; |
| 361 | 362 |
|
| ... | ... |
@@ -394,7 +398,7 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u |
| 394 | 394 |
bS[2]= 1+((h->cbp_table[mbn_xy] & 8)||h->non_zero_count_cache[scan8[0]+2]); |
| 395 | 395 |
bS[3]= 1+((h->cbp_table[mbn_xy] & 8)||h->non_zero_count_cache[scan8[0]+3]); |
| 396 | 396 |
}else{
|
| 397 |
- const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 4+3*8; |
|
| 397 |
+ const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 3*4; |
|
| 398 | 398 |
int i; |
| 399 | 399 |
for( i = 0; i < 4; i++ ) {
|
| 400 | 400 |
bS[i] = 1 + !!(h->non_zero_count_cache[scan8[0]+i] | mbn_nnz[i]); |
| ... | ... |
@@ -407,10 +411,17 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u |
| 407 | 407 |
tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize); |
| 408 | 408 |
{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
|
| 409 | 409 |
filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, h ); |
| 410 |
- filter_mb_edgech( &img_cb[j*uvlinesize], tmp_uvlinesize, bS, |
|
| 411 |
- ( h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1, h); |
|
| 412 |
- filter_mb_edgech( &img_cr[j*uvlinesize], tmp_uvlinesize, bS, |
|
| 413 |
- ( h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1, h); |
|
| 410 |
+ chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1; |
|
| 411 |
+ chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1; |
|
| 412 |
+ if (chroma) {
|
|
| 413 |
+ if (chroma444) {
|
|
| 414 |
+ filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], h); |
|
| 415 |
+ filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h); |
|
| 416 |
+ } else {
|
|
| 417 |
+ filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], h); |
|
| 418 |
+ filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h); |
|
| 419 |
+ } |
|
| 420 |
+ } |
|
| 414 | 421 |
} |
| 415 | 422 |
}else{
|
| 416 | 423 |
DECLARE_ALIGNED(8, int16_t, bS)[4]; |
| ... | ... |
@@ -465,23 +476,29 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u |
| 465 | 465 |
//tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], s->current_picture.qscale_table[mbn_xy]); |
| 466 | 466 |
tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize); |
| 467 | 467 |
//{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
|
| 468 |
+ chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1; |
|
| 469 |
+ chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1; |
|
| 468 | 470 |
if( dir == 0 ) {
|
| 469 | 471 |
filter_mb_edgev( &img_y[0], linesize, bS, qp, h ); |
| 470 |
- {
|
|
| 471 |
- int qp= ( h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1; |
|
| 472 |
- filter_mb_edgecv( &img_cb[0], uvlinesize, bS, qp, h); |
|
| 473 |
- if(h->pps.chroma_qp_diff) |
|
| 474 |
- qp= ( h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1; |
|
| 475 |
- filter_mb_edgecv( &img_cr[0], uvlinesize, bS, qp, h); |
|
| 472 |
+ if (chroma) {
|
|
| 473 |
+ if (chroma444) {
|
|
| 474 |
+ filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h); |
|
| 475 |
+ filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h); |
|
| 476 |
+ } else {
|
|
| 477 |
+ filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h); |
|
| 478 |
+ filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h); |
|
| 479 |
+ } |
|
| 476 | 480 |
} |
| 477 | 481 |
} else {
|
| 478 | 482 |
filter_mb_edgeh( &img_y[0], linesize, bS, qp, h ); |
| 479 |
- {
|
|
| 480 |
- int qp= ( h->chroma_qp[0] + get_chroma_qp( h, 0, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1; |
|
| 481 |
- filter_mb_edgech( &img_cb[0], uvlinesize, bS, qp, h); |
|
| 482 |
- if(h->pps.chroma_qp_diff) |
|
| 483 |
- qp= ( h->chroma_qp[1] + get_chroma_qp( h, 1, s->current_picture.qscale_table[mbm_xy] ) + 1 ) >> 1; |
|
| 484 |
- filter_mb_edgech( &img_cr[0], uvlinesize, bS, qp, h); |
|
| 483 |
+ if (chroma) {
|
|
| 484 |
+ if (chroma444) {
|
|
| 485 |
+ filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h); |
|
| 486 |
+ filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h); |
|
| 487 |
+ } else {
|
|
| 488 |
+ filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h); |
|
| 489 |
+ filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h); |
|
| 490 |
+ } |
|
| 485 | 491 |
} |
| 486 | 492 |
} |
| 487 | 493 |
} |
| ... | ... |
@@ -545,15 +562,25 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u |
| 545 | 545 |
//{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
|
| 546 | 546 |
if( dir == 0 ) {
|
| 547 | 547 |
filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, h ); |
| 548 |
- if( (edge&1) == 0 ) {
|
|
| 549 |
- filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h); |
|
| 550 |
- filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h); |
|
| 548 |
+ if (chroma) {
|
|
| 549 |
+ if (chroma444) {
|
|
| 550 |
+ filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h); |
|
| 551 |
+ filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h); |
|
| 552 |
+ } else if( (edge&1) == 0 ) {
|
|
| 553 |
+ filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h); |
|
| 554 |
+ filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h); |
|
| 555 |
+ } |
|
| 551 | 556 |
} |
| 552 | 557 |
} else {
|
| 553 | 558 |
filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, h ); |
| 554 |
- if( (edge&1) == 0 ) {
|
|
| 555 |
- filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h); |
|
| 556 |
- filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h); |
|
| 559 |
+ if (chroma) {
|
|
| 560 |
+ if (chroma444) {
|
|
| 561 |
+ filter_mb_edgeh ( &img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h); |
|
| 562 |
+ filter_mb_edgeh ( &img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h); |
|
| 563 |
+ } else if( (edge&1) == 0 ) {
|
|
| 564 |
+ filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h); |
|
| 565 |
+ filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h); |
|
| 566 |
+ } |
|
| 557 | 567 |
} |
| 558 | 568 |
} |
| 559 | 569 |
} |
| ... | ... |
@@ -566,6 +593,7 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint |
| 566 | 566 |
const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4; |
| 567 | 567 |
int first_vertical_edge_done = 0; |
| 568 | 568 |
av_unused int dir; |
| 569 |
+ int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY)); |
|
| 569 | 570 |
|
| 570 | 571 |
if (FRAME_MBAFF |
| 571 | 572 |
// and current and left pair do not have the same interlaced type |
| ... | ... |
@@ -589,11 +617,11 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint |
| 589 | 589 |
} else {
|
| 590 | 590 |
static const uint8_t offset[2][2][8]={
|
| 591 | 591 |
{
|
| 592 |
- {7+8*0, 7+8*0, 7+8*0, 7+8*0, 7+8*1, 7+8*1, 7+8*1, 7+8*1},
|
|
| 593 |
- {7+8*2, 7+8*2, 7+8*2, 7+8*2, 7+8*3, 7+8*3, 7+8*3, 7+8*3},
|
|
| 592 |
+ {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1},
|
|
| 593 |
+ {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3},
|
|
| 594 | 594 |
},{
|
| 595 |
- {7+8*0, 7+8*1, 7+8*2, 7+8*3, 7+8*0, 7+8*1, 7+8*2, 7+8*3},
|
|
| 596 |
- {7+8*0, 7+8*1, 7+8*2, 7+8*3, 7+8*0, 7+8*1, 7+8*2, 7+8*3},
|
|
| 595 |
+ {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
|
|
| 596 |
+ {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
|
|
| 597 | 597 |
} |
| 598 | 598 |
}; |
| 599 | 599 |
const uint8_t *off= offset[MB_FIELD][mb_y&1]; |
| ... | ... |
@@ -634,25 +662,29 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint |
| 634 | 634 |
if(MB_FIELD){
|
| 635 | 635 |
filter_mb_mbaff_edgev ( h, img_y , linesize, bS , 1, qp [0] ); |
| 636 | 636 |
filter_mb_mbaff_edgev ( h, img_y + 8* linesize, linesize, bS+4, 1, qp [1] ); |
| 637 |
- filter_mb_mbaff_edgecv( h, img_cb, uvlinesize, bS , 1, bqp[0] ); |
|
| 638 |
- filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1] ); |
|
| 639 |
- filter_mb_mbaff_edgecv( h, img_cr, uvlinesize, bS , 1, rqp[0] ); |
|
| 640 |
- filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1] ); |
|
| 637 |
+ if (chroma){
|
|
| 638 |
+ filter_mb_mbaff_edgecv( h, img_cb, uvlinesize, bS , 1, bqp[0] ); |
|
| 639 |
+ filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1] ); |
|
| 640 |
+ filter_mb_mbaff_edgecv( h, img_cr, uvlinesize, bS , 1, rqp[0] ); |
|
| 641 |
+ filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1] ); |
|
| 642 |
+ } |
|
| 641 | 643 |
}else{
|
| 642 | 644 |
filter_mb_mbaff_edgev ( h, img_y , 2* linesize, bS , 2, qp [0] ); |
| 643 | 645 |
filter_mb_mbaff_edgev ( h, img_y + linesize, 2* linesize, bS+1, 2, qp [1] ); |
| 644 |
- filter_mb_mbaff_edgecv( h, img_cb, 2*uvlinesize, bS , 2, bqp[0] ); |
|
| 645 |
- filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1] ); |
|
| 646 |
- filter_mb_mbaff_edgecv( h, img_cr, 2*uvlinesize, bS , 2, rqp[0] ); |
|
| 647 |
- filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1] ); |
|
| 646 |
+ if (chroma){
|
|
| 647 |
+ filter_mb_mbaff_edgecv( h, img_cb, 2*uvlinesize, bS , 2, bqp[0] ); |
|
| 648 |
+ filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1] ); |
|
| 649 |
+ filter_mb_mbaff_edgecv( h, img_cr, 2*uvlinesize, bS , 2, rqp[0] ); |
|
| 650 |
+ filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1] ); |
|
| 651 |
+ } |
|
| 648 | 652 |
} |
| 649 | 653 |
} |
| 650 | 654 |
|
| 651 | 655 |
#if CONFIG_SMALL |
| 652 | 656 |
for( dir = 0; dir < 2; dir++ ) |
| 653 |
- filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, dir); |
|
| 657 |
+ filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, chroma, CHROMA444, dir); |
|
| 654 | 658 |
#else |
| 655 |
- filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, 0); |
|
| 656 |
- filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, 1); |
|
| 659 |
+ filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, chroma, CHROMA444, 0); |
|
| 660 |
+ filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, chroma, CHROMA444, 1); |
|
| 657 | 661 |
#endif |
| 658 | 662 |
} |
| ... | ... |
@@ -269,7 +269,7 @@ static void decode_scaling_matrices(H264Context *h, SPS *sps, PPS *pps, int is_s |
| 269 | 269 |
fallback_sps ? sps->scaling_matrix4[0] : default_scaling4[0], |
| 270 | 270 |
fallback_sps ? sps->scaling_matrix4[3] : default_scaling4[1], |
| 271 | 271 |
fallback_sps ? sps->scaling_matrix8[0] : default_scaling8[0], |
| 272 |
- fallback_sps ? sps->scaling_matrix8[1] : default_scaling8[1] |
|
| 272 |
+ fallback_sps ? sps->scaling_matrix8[3] : default_scaling8[1] |
|
| 273 | 273 |
}; |
| 274 | 274 |
if(get_bits1(&s->gb)){
|
| 275 | 275 |
sps->scaling_matrix_present |= is_sps; |
| ... | ... |
@@ -281,7 +281,15 @@ static void decode_scaling_matrices(H264Context *h, SPS *sps, PPS *pps, int is_s |
| 281 | 281 |
decode_scaling_list(h,scaling_matrix4[5],16,default_scaling4[1],scaling_matrix4[4]); // Inter, Cb |
| 282 | 282 |
if(is_sps || pps->transform_8x8_mode){
|
| 283 | 283 |
decode_scaling_list(h,scaling_matrix8[0],64,default_scaling8[0],fallback[2]); // Intra, Y |
| 284 |
- decode_scaling_list(h,scaling_matrix8[1],64,default_scaling8[1],fallback[3]); // Inter, Y |
|
| 284 |
+ if(h->sps.chroma_format_idc == 3){
|
|
| 285 |
+ decode_scaling_list(h,scaling_matrix8[1],64,default_scaling8[0],scaling_matrix8[0]); // Intra, Cr |
|
| 286 |
+ decode_scaling_list(h,scaling_matrix8[2],64,default_scaling8[0],scaling_matrix8[1]); // Intra, Cb |
|
| 287 |
+ } |
|
| 288 |
+ decode_scaling_list(h,scaling_matrix8[3],64,default_scaling8[1],fallback[3]); // Inter, Y |
|
| 289 |
+ if(h->sps.chroma_format_idc == 3){
|
|
| 290 |
+ decode_scaling_list(h,scaling_matrix8[4],64,default_scaling8[1],scaling_matrix8[3]); // Inter, Cr |
|
| 291 |
+ decode_scaling_list(h,scaling_matrix8[5],64,default_scaling8[1],scaling_matrix8[4]); // Inter, Cb |
|
| 292 |
+ } |
|
| 285 | 293 |
} |
| 286 | 294 |
} |
| 287 | 295 |
} |
| ... | ... |
@@ -395,7 +403,7 @@ int ff_h264_decode_seq_parameter_set(H264Context *h){
|
| 395 | 395 |
if(sps->crop_left || sps->crop_top){
|
| 396 | 396 |
av_log(h->s.avctx, AV_LOG_ERROR, "insane cropping not completely supported, this could look slightly wrong ...\n"); |
| 397 | 397 |
} |
| 398 |
- if(sps->crop_right >= 8 || sps->crop_bottom >= 8){
|
|
| 398 |
+ if(sps->crop_right >= (8<<CHROMA444) || sps->crop_bottom >= (8<<CHROMA444)){
|
|
| 399 | 399 |
av_log(h->s.avctx, AV_LOG_ERROR, "brainfart cropping not supported, this could look slightly wrong ...\n"); |
| 400 | 400 |
} |
| 401 | 401 |
}else{
|
| ... | ... |
@@ -66,10 +66,10 @@ typedef struct H264DSPContext{
|
| 66 | 66 |
void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride); |
| 67 | 67 |
void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride); |
| 68 | 68 |
|
| 69 |
- void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]); |
|
| 70 |
- void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]); |
|
| 71 |
- void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]); |
|
| 72 |
- void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]); |
|
| 69 |
+ void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[15*8]); |
|
| 70 |
+ void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[15*8]); |
|
| 71 |
+ void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[15*8]); |
|
| 72 |
+ void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[15*8]); |
|
| 73 | 73 |
void (*h264_luma_dc_dequant_idct)(DCTELEM *output, DCTELEM *input/*align 16*/, int qmul); |
| 74 | 74 |
void (*h264_chroma_dc_dequant_idct)(DCTELEM *block, int qmul); |
| 75 | 75 |
}H264DSPContext; |
| ... | ... |
@@ -30,15 +30,19 @@ |
| 30 | 30 |
#ifndef AVCODEC_H264IDCT_INTERNAL_H |
| 31 | 31 |
#define AVCODEC_H264IDCT_INTERNAL_H |
| 32 | 32 |
//FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split |
| 33 |
-static const uint8_t scan8[16 + 2*4]={
|
|
| 34 |
- 4+1*8, 5+1*8, 4+2*8, 5+2*8, |
|
| 35 |
- 6+1*8, 7+1*8, 6+2*8, 7+2*8, |
|
| 36 |
- 4+3*8, 5+3*8, 4+4*8, 5+4*8, |
|
| 37 |
- 6+3*8, 7+3*8, 6+4*8, 7+4*8, |
|
| 38 |
- 1+1*8, 2+1*8, |
|
| 39 |
- 1+2*8, 2+2*8, |
|
| 40 |
- 1+4*8, 2+4*8, |
|
| 41 |
- 1+5*8, 2+5*8, |
|
| 33 |
+static const uint8_t scan8[16*3]={
|
|
| 34 |
+ 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8, |
|
| 35 |
+ 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8, |
|
| 36 |
+ 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8, |
|
| 37 |
+ 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8, |
|
| 38 |
+ 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8, |
|
| 39 |
+ 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8, |
|
| 40 |
+ 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8, |
|
| 41 |
+ 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8, |
|
| 42 |
+ 4+11*8, 5+11*8, 4+12*8, 5+12*8, |
|
| 43 |
+ 6+11*8, 7+11*8, 6+12*8, 7+12*8, |
|
| 44 |
+ 4+13*8, 5+13*8, 4+14*8, 5+14*8, |
|
| 45 |
+ 6+13*8, 7+13*8, 6+14*8, 7+14*8 |
|
| 42 | 46 |
}; |
| 43 | 47 |
#endif |
| 44 | 48 |
|
| ... | ... |
@@ -190,7 +194,7 @@ void FUNCC(ff_h264_idct8_dc_add)(uint8_t *p_dst, DCTELEM *block, int stride){
|
| 190 | 190 |
} |
| 191 | 191 |
} |
| 192 | 192 |
|
| 193 |
-void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
|
|
| 193 |
+void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
|
|
| 194 | 194 |
int i; |
| 195 | 195 |
for(i=0; i<16; i++){
|
| 196 | 196 |
int nnz = nnzc[ scan8[i] ]; |
| ... | ... |
@@ -201,7 +205,7 @@ void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *b |
| 201 | 201 |
} |
| 202 | 202 |
} |
| 203 | 203 |
|
| 204 |
-void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
|
|
| 204 |
+void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
|
|
| 205 | 205 |
int i; |
| 206 | 206 |
for(i=0; i<16; i++){
|
| 207 | 207 |
if(nnzc[ scan8[i] ]) FUNCC(idct_internal )(dst + block_offset[i], block + i*16*sizeof(pixel), stride, 4, 6, 1); |
| ... | ... |
@@ -209,7 +213,7 @@ void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTEL |
| 209 | 209 |
} |
| 210 | 210 |
} |
| 211 | 211 |
|
| 212 |
-void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
|
|
| 212 |
+void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
|
|
| 213 | 213 |
int i; |
| 214 | 214 |
for(i=0; i<16; i+=4){
|
| 215 | 215 |
int nnz = nnzc[ scan8[i] ]; |
| ... | ... |
@@ -220,13 +224,15 @@ void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *b |
| 220 | 220 |
} |
| 221 | 221 |
} |
| 222 | 222 |
|
| 223 |
-void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
|
|
| 224 |
- int i; |
|
| 225 |
- for(i=16; i<16+8; i++){
|
|
| 226 |
- if(nnzc[ scan8[i] ]) |
|
| 227 |
- FUNCC(ff_h264_idct_add )(dest[(i&4)>>2] + block_offset[i], block + i*16*sizeof(pixel), stride); |
|
| 228 |
- else if(((dctcoef*)block)[i*16]) |
|
| 229 |
- FUNCC(ff_h264_idct_dc_add)(dest[(i&4)>>2] + block_offset[i], block + i*16*sizeof(pixel), stride); |
|
| 223 |
+void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
|
|
| 224 |
+ int i, j; |
|
| 225 |
+ for(j=1; j<3; j++){
|
|
| 226 |
+ for(i=j*16; i<j*16+4; i++){
|
|
| 227 |
+ if(nnzc[ scan8[i] ]) |
|
| 228 |
+ FUNCC(ff_h264_idct_add )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride); |
|
| 229 |
+ else if(((dctcoef*)block)[i*16]) |
|
| 230 |
+ FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride); |
|
| 231 |
+ } |
|
| 230 | 232 |
} |
| 231 | 233 |
} |
| 232 | 234 |
/** |
| ... | ... |
@@ -1180,12 +1180,17 @@ void MPV_frame_end(MpegEncContext *s) |
| 1180 | 1180 |
&& s->current_picture.reference |
| 1181 | 1181 |
&& !s->intra_only |
| 1182 | 1182 |
&& !(s->flags&CODEC_FLAG_EMU_EDGE)) {
|
| 1183 |
- int edges = EDGE_BOTTOM | EDGE_TOP, h = s->v_edge_pos; |
|
| 1184 |
- |
|
| 1185 |
- s->dsp.draw_edges(s->current_picture.data[0], s->linesize , s->h_edge_pos , h , EDGE_WIDTH , edges); |
|
| 1186 |
- s->dsp.draw_edges(s->current_picture.data[1], s->uvlinesize, s->h_edge_pos>>1, h>>1, EDGE_WIDTH/2, edges); |
|
| 1187 |
- s->dsp.draw_edges(s->current_picture.data[2], s->uvlinesize, s->h_edge_pos>>1, h>>1, EDGE_WIDTH/2, edges); |
|
| 1188 |
- |
|
| 1183 |
+ int hshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_w; |
|
| 1184 |
+ int vshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_h; |
|
| 1185 |
+ s->dsp.draw_edges(s->current_picture.data[0], s->linesize , |
|
| 1186 |
+ s->h_edge_pos , s->v_edge_pos, |
|
| 1187 |
+ EDGE_WIDTH , EDGE_WIDTH , EDGE_TOP | EDGE_BOTTOM); |
|
| 1188 |
+ s->dsp.draw_edges(s->current_picture.data[1], s->uvlinesize, |
|
| 1189 |
+ s->h_edge_pos>>hshift, s->v_edge_pos>>vshift, |
|
| 1190 |
+ EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, EDGE_TOP | EDGE_BOTTOM); |
|
| 1191 |
+ s->dsp.draw_edges(s->current_picture.data[2], s->uvlinesize, |
|
| 1192 |
+ s->h_edge_pos>>hshift, s->v_edge_pos>>vshift, |
|
| 1193 |
+ EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, EDGE_TOP | EDGE_BOTTOM); |
|
| 1189 | 1194 |
} |
| 1190 | 1195 |
|
| 1191 | 1196 |
emms_c(); |
| ... | ... |
@@ -2289,14 +2294,19 @@ void ff_draw_horiz_band(MpegEncContext *s, int y, int h){
|
| 2289 | 2289 |
&& !s->intra_only |
| 2290 | 2290 |
&& !(s->flags&CODEC_FLAG_EMU_EDGE)) {
|
| 2291 | 2291 |
int sides = 0, edge_h; |
| 2292 |
+ int hshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_w; |
|
| 2293 |
+ int vshift = av_pix_fmt_descriptors[s->avctx->pix_fmt].log2_chroma_h; |
|
| 2292 | 2294 |
if (y==0) sides |= EDGE_TOP; |
| 2293 | 2295 |
if (y + h >= s->v_edge_pos) sides |= EDGE_BOTTOM; |
| 2294 | 2296 |
|
| 2295 | 2297 |
edge_h= FFMIN(h, s->v_edge_pos - y); |
| 2296 | 2298 |
|
| 2297 |
- s->dsp.draw_edges(s->current_picture_ptr->data[0] + y *s->linesize , s->linesize , s->h_edge_pos , edge_h , EDGE_WIDTH , sides); |
|
| 2298 |
- s->dsp.draw_edges(s->current_picture_ptr->data[1] + (y>>1)*s->uvlinesize, s->uvlinesize, s->h_edge_pos>>1, edge_h>>1, EDGE_WIDTH/2, sides); |
|
| 2299 |
- s->dsp.draw_edges(s->current_picture_ptr->data[2] + (y>>1)*s->uvlinesize, s->uvlinesize, s->h_edge_pos>>1, edge_h>>1, EDGE_WIDTH/2, sides); |
|
| 2299 |
+ s->dsp.draw_edges(s->current_picture_ptr->data[0] + y *s->linesize , s->linesize, |
|
| 2300 |
+ s->h_edge_pos , edge_h , EDGE_WIDTH , EDGE_WIDTH , sides); |
|
| 2301 |
+ s->dsp.draw_edges(s->current_picture_ptr->data[1] + (y>>vshift)*s->uvlinesize, s->uvlinesize, |
|
| 2302 |
+ s->h_edge_pos>>hshift, edge_h>>hshift, EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, sides); |
|
| 2303 |
+ s->dsp.draw_edges(s->current_picture_ptr->data[2] + (y>>vshift)*s->uvlinesize, s->uvlinesize, |
|
| 2304 |
+ s->h_edge_pos>>hshift, edge_h>>hshift, EDGE_WIDTH>>hshift, EDGE_WIDTH>>vshift, sides); |
|
| 2300 | 2305 |
} |
| 2301 | 2306 |
|
| 2302 | 2307 |
h= FFMIN(h, s->avctx->height - y); |
| ... | ... |
@@ -527,7 +527,7 @@ static void ff_h264_idct8_dc_add_altivec(uint8_t *dst, DCTELEM *block, int strid |
| 527 | 527 |
h264_idct_dc_add_internal(dst, block, stride, 8); |
| 528 | 528 |
} |
| 529 | 529 |
|
| 530 |
-static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
|
|
| 530 |
+static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
|
|
| 531 | 531 |
int i; |
| 532 | 532 |
for(i=0; i<16; i++){
|
| 533 | 533 |
int nnz = nnzc[ scan8[i] ]; |
| ... | ... |
@@ -538,7 +538,7 @@ static void ff_h264_idct_add16_altivec(uint8_t *dst, const int *block_offset, DC |
| 538 | 538 |
} |
| 539 | 539 |
} |
| 540 | 540 |
|
| 541 |
-static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
|
|
| 541 |
+static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
|
|
| 542 | 542 |
int i; |
| 543 | 543 |
for(i=0; i<16; i++){
|
| 544 | 544 |
if(nnzc[ scan8[i] ]) ff_h264_idct_add_altivec(dst + block_offset[i], block + i*16, stride); |
| ... | ... |
@@ -546,7 +546,7 @@ static void ff_h264_idct_add16intra_altivec(uint8_t *dst, const int *block_offse |
| 546 | 546 |
} |
| 547 | 547 |
} |
| 548 | 548 |
|
| 549 |
-static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
|
|
| 549 |
+static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
|
|
| 550 | 550 |
int i; |
| 551 | 551 |
for(i=0; i<16; i+=4){
|
| 552 | 552 |
int nnz = nnzc[ scan8[i] ]; |
| ... | ... |
@@ -557,13 +557,15 @@ static void ff_h264_idct8_add4_altivec(uint8_t *dst, const int *block_offset, DC |
| 557 | 557 |
} |
| 558 | 558 |
} |
| 559 | 559 |
|
| 560 |
-static void ff_h264_idct_add8_altivec(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
|
|
| 561 |
- int i; |
|
| 562 |
- for(i=16; i<16+8; i++){
|
|
| 563 |
- if(nnzc[ scan8[i] ]) |
|
| 564 |
- ff_h264_idct_add_altivec(dest[(i&4)>>2] + block_offset[i], block + i*16, stride); |
|
| 565 |
- else if(block[i*16]) |
|
| 566 |
- h264_idct_dc_add_altivec(dest[(i&4)>>2] + block_offset[i], block + i*16, stride); |
|
| 560 |
+static void ff_h264_idct_add8_altivec(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
|
|
| 561 |
+ int i, j; |
|
| 562 |
+ for (j = 1; j < 3; j++) {
|
|
| 563 |
+ for(i = j * 16; i < j * 16 + 4; i++){
|
|
| 564 |
+ if(nnzc[ scan8[i] ]) |
|
| 565 |
+ ff_h264_idct_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride); |
|
| 566 |
+ else if(block[i*16]) |
|
| 567 |
+ h264_idct_dc_add_altivec(dest[j-1] + block_offset[i], block + i*16, stride); |
|
| 568 |
+ } |
|
| 567 | 569 |
} |
| 568 | 570 |
} |
| 569 | 571 |
|
| ... | ... |
@@ -1976,9 +1976,15 @@ static int frame_start(SnowContext *s){
|
| 1976 | 1976 |
int h= s->avctx->height; |
| 1977 | 1977 |
|
| 1978 | 1978 |
if(s->current_picture.data[0]){
|
| 1979 |
- s->dsp.draw_edges(s->current_picture.data[0], s->current_picture.linesize[0], w , h , EDGE_WIDTH , EDGE_TOP|EDGE_BOTTOM); |
|
| 1980 |
- s->dsp.draw_edges(s->current_picture.data[1], s->current_picture.linesize[1], w>>1, h>>1, EDGE_WIDTH/2, EDGE_TOP|EDGE_BOTTOM); |
|
| 1981 |
- s->dsp.draw_edges(s->current_picture.data[2], s->current_picture.linesize[2], w>>1, h>>1, EDGE_WIDTH/2, EDGE_TOP|EDGE_BOTTOM); |
|
| 1979 |
+ s->dsp.draw_edges(s->current_picture.data[0], |
|
| 1980 |
+ s->current_picture.linesize[0], w , h , |
|
| 1981 |
+ EDGE_WIDTH , EDGE_WIDTH , EDGE_TOP | EDGE_BOTTOM); |
|
| 1982 |
+ s->dsp.draw_edges(s->current_picture.data[1], |
|
| 1983 |
+ s->current_picture.linesize[1], w>>1, h>>1, |
|
| 1984 |
+ EDGE_WIDTH/2, EDGE_WIDTH/2, EDGE_TOP | EDGE_BOTTOM); |
|
| 1985 |
+ s->dsp.draw_edges(s->current_picture.data[2], |
|
| 1986 |
+ s->current_picture.linesize[2], w>>1, h>>1, |
|
| 1987 |
+ EDGE_WIDTH/2, EDGE_WIDTH/2, EDGE_TOP | EDGE_BOTTOM); |
|
| 1982 | 1988 |
} |
| 1983 | 1989 |
|
| 1984 | 1990 |
release_buffer(s->avctx); |
| ... | ... |
@@ -635,8 +635,9 @@ static int svq3_decode_mb(SVQ3Context *svq3, unsigned int mb_type) |
| 635 | 635 |
memset(h->intra4x4_pred_mode+h->mb2br_xy[mb_xy], DC_PRED, 8); |
| 636 | 636 |
} |
| 637 | 637 |
if (!IS_SKIP(mb_type) || s->pict_type == AV_PICTURE_TYPE_B) {
|
| 638 |
- memset(h->non_zero_count_cache + 8, 0, 4*9*sizeof(uint8_t)); |
|
| 639 |
- s->dsp.clear_blocks(h->mb); |
|
| 638 |
+ memset(h->non_zero_count_cache + 8, 0, 14*8*sizeof(uint8_t)); |
|
| 639 |
+ s->dsp.clear_blocks(h->mb+ 0); |
|
| 640 |
+ s->dsp.clear_blocks(h->mb+384); |
|
| 640 | 641 |
} |
| 641 | 642 |
|
| 642 | 643 |
if (!IS_INTRA16x16(mb_type) && (!IS_SKIP(mb_type) || s->pict_type == AV_PICTURE_TYPE_B)) {
|
| ... | ... |
@@ -656,8 +657,8 @@ static int svq3_decode_mb(SVQ3Context *svq3, unsigned int mb_type) |
| 656 | 656 |
} |
| 657 | 657 |
} |
| 658 | 658 |
if (IS_INTRA16x16(mb_type)) {
|
| 659 |
- AV_ZERO128(h->mb_luma_dc+0); |
|
| 660 |
- AV_ZERO128(h->mb_luma_dc+8); |
|
| 659 |
+ AV_ZERO128(h->mb_luma_dc[0]+0); |
|
| 660 |
+ AV_ZERO128(h->mb_luma_dc[0]+8); |
|
| 661 | 661 |
if (svq3_decode_block(&s->gb, h->mb_luma_dc, 0, 1)){
|
| 662 | 662 |
av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding intra luma dc\n"); |
| 663 | 663 |
return -1; |
| ... | ... |
@@ -683,20 +684,23 @@ static int svq3_decode_mb(SVQ3Context *svq3, unsigned int mb_type) |
| 683 | 683 |
} |
| 684 | 684 |
|
| 685 | 685 |
if ((cbp & 0x30)) {
|
| 686 |
- for (i = 0; i < 2; ++i) {
|
|
| 687 |
- if (svq3_decode_block(&s->gb, &h->mb[16*(16 + 4*i)], 0, 3)){
|
|
| 686 |
+ for (i = 1; i < 3; ++i) {
|
|
| 687 |
+ if (svq3_decode_block(&s->gb, &h->mb[16*16*i], 0, 3)){
|
|
| 688 | 688 |
av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding chroma dc block\n"); |
| 689 | 689 |
return -1; |
| 690 | 690 |
} |
| 691 | 691 |
} |
| 692 | 692 |
|
| 693 | 693 |
if ((cbp & 0x20)) {
|
| 694 |
- for (i = 0; i < 8; i++) {
|
|
| 695 |
- h->non_zero_count_cache[ scan8[16+i] ] = 1; |
|
| 696 |
- |
|
| 697 |
- if (svq3_decode_block(&s->gb, &h->mb[16*(16 + i)], 1, 1)){
|
|
| 698 |
- av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding chroma ac block\n"); |
|
| 699 |
- return -1; |
|
| 694 |
+ for (i = 1; i < 3; i++) {
|
|
| 695 |
+ for (j = 0; j < 4; j++) {
|
|
| 696 |
+ k = 16*i + j; |
|
| 697 |
+ h->non_zero_count_cache[ scan8[k] ] = 1; |
|
| 698 |
+ |
|
| 699 |
+ if (svq3_decode_block(&s->gb, &h->mb[16*k], 1, 1)){
|
|
| 700 |
+ av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding chroma ac block\n"); |
|
| 701 |
+ return -1; |
|
| 702 |
+ } |
|
| 700 | 703 |
} |
| 701 | 704 |
} |
| 702 | 705 |
} |
| ... | ... |
@@ -762,7 +762,7 @@ static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale){
|
| 762 | 762 |
|
| 763 | 763 |
/* draw the edges of width 'w' of an image of size width, height |
| 764 | 764 |
this mmx version can only handle w==8 || w==16 */ |
| 765 |
-static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w, int sides) |
|
| 765 |
+static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides) |
|
| 766 | 766 |
{
|
| 767 | 767 |
uint8_t *ptr, *last_line; |
| 768 | 768 |
int i; |
| ... | ... |
@@ -817,7 +817,7 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w, |
| 817 | 817 |
|
| 818 | 818 |
/* top and bottom (and hopefully also the corners) */ |
| 819 | 819 |
if (sides&EDGE_TOP) {
|
| 820 |
- for(i = 0; i < w; i += 4) {
|
|
| 820 |
+ for(i = 0; i < h; i += 4) {
|
|
| 821 | 821 |
ptr= buf - (i + 1) * wrap - w; |
| 822 | 822 |
__asm__ volatile( |
| 823 | 823 |
"1: \n\t" |
| ... | ... |
@@ -36,7 +36,7 @@ |
| 36 | 36 |
#if ARCH_X86 && HAVE_7REGS && HAVE_EBX_AVAILABLE && !defined(BROKEN_RELOCATIONS) |
| 37 | 37 |
static int decode_significance_x86(CABACContext *c, int max_coeff, |
| 38 | 38 |
uint8_t *significant_coeff_ctx_base, |
| 39 |
- int *index){
|
|
| 39 |
+ int *index, x86_reg last_off){
|
|
| 40 | 40 |
void *end= significant_coeff_ctx_base + max_coeff - 1; |
| 41 | 41 |
int minusstart= -(int)significant_coeff_ctx_base; |
| 42 | 42 |
int minusindex= 4-(int)index; |
| ... | ... |
@@ -52,10 +52,12 @@ static int decode_significance_x86(CABACContext *c, int max_coeff, |
| 52 | 52 |
|
| 53 | 53 |
"test $1, %%edx \n\t" |
| 54 | 54 |
" jz 3f \n\t" |
| 55 |
+ "add %7, %1 \n\t" |
|
| 55 | 56 |
|
| 56 |
- BRANCHLESS_GET_CABAC("%%edx", "%3", "61(%1)", "%%ebx",
|
|
| 57 |
+ BRANCHLESS_GET_CABAC("%%edx", "%3", "(%1)", "%%ebx",
|
|
| 57 | 58 |
"%%bx", "%%esi", "%%eax", "%%al") |
| 58 | 59 |
|
| 60 |
+ "sub %7, %1 \n\t" |
|
| 59 | 61 |
"mov %2, %%"REG_a" \n\t" |
| 60 | 62 |
"movl %4, %%ecx \n\t" |
| 61 | 63 |
"add %1, %%"REG_c" \n\t" |
| ... | ... |
@@ -82,7 +84,7 @@ static int decode_significance_x86(CABACContext *c, int max_coeff, |
| 82 | 82 |
"movl %%esi, "RANGE "(%3) \n\t" |
| 83 | 83 |
"movl %%ebx, "LOW "(%3) \n\t" |
| 84 | 84 |
:"=&a"(coeff_count), "+r"(significant_coeff_ctx_base), "+m"(index) |
| 85 |
- :"r"(c), "m"(minusstart), "m"(end), "m"(minusindex) |
|
| 85 |
+ :"r"(c), "m"(minusstart), "m"(end), "m"(minusindex), "m"(last_off) |
|
| 86 | 86 |
: "%"REG_c, "%ebx", "%edx", "%esi", "memory" |
| 87 | 87 |
); |
| 88 | 88 |
return coeff_count; |
| ... | ... |
@@ -90,7 +92,7 @@ static int decode_significance_x86(CABACContext *c, int max_coeff, |
| 90 | 90 |
|
| 91 | 91 |
static int decode_significance_8x8_x86(CABACContext *c, |
| 92 | 92 |
uint8_t *significant_coeff_ctx_base, |
| 93 |
- int *index, const uint8_t *sig_off){
|
|
| 93 |
+ int *index, x86_reg last_off, const uint8_t *sig_off){
|
|
| 94 | 94 |
int minusindex= 4-(int)index; |
| 95 | 95 |
int coeff_count; |
| 96 | 96 |
x86_reg last=0; |
| ... | ... |
@@ -114,8 +116,9 @@ static int decode_significance_8x8_x86(CABACContext *c, |
| 114 | 114 |
|
| 115 | 115 |
"movzbl "MANGLE(last_coeff_flag_offset_8x8)"(%%edi), %%edi\n\t" |
| 116 | 116 |
"add %5, %%"REG_D" \n\t" |
| 117 |
+ "add %7, %%"REG_D" \n\t" |
|
| 117 | 118 |
|
| 118 |
- BRANCHLESS_GET_CABAC("%%edx", "%3", "15(%%"REG_D")", "%%ebx",
|
|
| 119 |
+ BRANCHLESS_GET_CABAC("%%edx", "%3", "(%%"REG_D")", "%%ebx",
|
|
| 119 | 120 |
"%%bx", "%%esi", "%%eax", "%%al") |
| 120 | 121 |
|
| 121 | 122 |
"mov %2, %%"REG_a" \n\t" |
| ... | ... |
@@ -142,7 +145,7 @@ static int decode_significance_8x8_x86(CABACContext *c, |
| 142 | 142 |
"movl %%esi, "RANGE "(%3) \n\t" |
| 143 | 143 |
"movl %%ebx, "LOW "(%3) \n\t" |
| 144 | 144 |
:"=&a"(coeff_count),"+m"(last), "+m"(index) |
| 145 |
- :"r"(c), "m"(minusindex), "m"(significant_coeff_ctx_base), "m"(sig_off) |
|
| 145 |
+ :"r"(c), "m"(minusindex), "m"(significant_coeff_ctx_base), "m"(sig_off), "m"(last_off) |
|
| 146 | 146 |
: "%"REG_c, "%ebx", "%edx", "%esi", "%"REG_D, "memory" |
| 147 | 147 |
); |
| 148 | 148 |
return coeff_count; |
| ... | ... |
@@ -32,14 +32,18 @@ |
| 32 | 32 |
SECTION_RODATA |
| 33 | 33 |
|
| 34 | 34 |
; FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split |
| 35 |
-scan8_mem: db 4+1*8, 5+1*8, 4+2*8, 5+2*8 |
|
| 36 |
- db 6+1*8, 7+1*8, 6+2*8, 7+2*8 |
|
| 37 |
- db 4+3*8, 5+3*8, 4+4*8, 5+4*8 |
|
| 38 |
- db 6+3*8, 7+3*8, 6+4*8, 7+4*8 |
|
| 39 |
- db 1+1*8, 2+1*8 |
|
| 40 |
- db 1+2*8, 2+2*8 |
|
| 41 |
- db 1+4*8, 2+4*8 |
|
| 42 |
- db 1+5*8, 2+5*8 |
|
| 35 |
+scan8_mem: db 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8 |
|
| 36 |
+ db 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8 |
|
| 37 |
+ db 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8 |
|
| 38 |
+ db 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8 |
|
| 39 |
+ db 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8 |
|
| 40 |
+ db 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8 |
|
| 41 |
+ db 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8 |
|
| 42 |
+ db 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8 |
|
| 43 |
+ db 4+11*8, 5+11*8, 4+12*8, 5+12*8 |
|
| 44 |
+ db 6+11*8, 7+11*8, 6+12*8, 7+12*8 |
|
| 45 |
+ db 4+13*8, 5+13*8, 4+14*8, 5+14*8 |
|
| 46 |
+ db 6+13*8, 7+13*8, 6+14*8, 7+14*8 |
|
| 43 | 47 |
%ifdef PIC |
| 44 | 48 |
%define scan8 r11 |
| 45 | 49 |
%else |
| ... | ... |
@@ -617,6 +621,8 @@ cglobal h264_idct_add8_8_mmx, 5, 7, 0 |
| 617 | 617 |
mov r10, r0 |
| 618 | 618 |
%endif |
| 619 | 619 |
call h264_idct_add8_mmx_plane |
| 620 |
+ mov r5, 32 |
|
| 621 |
+ add r2, 384 |
|
| 620 | 622 |
%ifdef ARCH_X86_64 |
| 621 | 623 |
add r10, gprsize |
| 622 | 624 |
%else |
| ... | ... |
@@ -678,6 +684,8 @@ cglobal h264_idct_add8_8_mmx2, 5, 7, 0 |
| 678 | 678 |
lea r11, [scan8_mem] |
| 679 | 679 |
%endif |
| 680 | 680 |
call h264_idct_add8_mmx2_plane |
| 681 |
+ mov r5, 32 |
|
| 682 |
+ add r2, 384 |
|
| 681 | 683 |
%ifdef ARCH_X86_64 |
| 682 | 684 |
add r10, gprsize |
| 683 | 685 |
%else |
| ... | ... |
@@ -810,12 +818,12 @@ cglobal h264_idct_add16intra_8_sse2, 5, 7, 8 |
| 810 | 810 |
test r0, r0 |
| 811 | 811 |
jz .try%1dc |
| 812 | 812 |
%ifdef ARCH_X86_64 |
| 813 |
- mov r0d, dword [r1+%1*8+64] |
|
| 813 |
+ mov r0d, dword [r1+(%1&1)*8+64*(1+(%1>>1))] |
|
| 814 | 814 |
add r0, [r10] |
| 815 | 815 |
%else |
| 816 | 816 |
mov r0, r0m |
| 817 | 817 |
mov r0, [r0] |
| 818 |
- add r0, dword [r1+%1*8+64] |
|
| 818 |
+ add r0, dword [r1+(%1&1)*8+64*(1+(%1>>1))] |
|
| 819 | 819 |
%endif |
| 820 | 820 |
call x264_add8x4_idct_sse2 |
| 821 | 821 |
jmp .cycle%1end |
| ... | ... |
@@ -824,16 +832,18 @@ cglobal h264_idct_add16intra_8_sse2, 5, 7, 8 |
| 824 | 824 |
or r0w, word [r2+32] |
| 825 | 825 |
jz .cycle%1end |
| 826 | 826 |
%ifdef ARCH_X86_64 |
| 827 |
- mov r0d, dword [r1+%1*8+64] |
|
| 827 |
+ mov r0d, dword [r1+(%1&1)*8+64*(1+(%1>>1))] |
|
| 828 | 828 |
add r0, [r10] |
| 829 | 829 |
%else |
| 830 | 830 |
mov r0, r0m |
| 831 | 831 |
mov r0, [r0] |
| 832 |
- add r0, dword [r1+%1*8+64] |
|
| 832 |
+ add r0, dword [r1+(%1&1)*8+64*(1+(%1>>1))] |
|
| 833 | 833 |
%endif |
| 834 | 834 |
call h264_idct_dc_add8_mmx2 |
| 835 | 835 |
.cycle%1end |
| 836 |
-%if %1 < 3 |
|
| 836 |
+%if %1 == 1 |
|
| 837 |
+ add r2, 384+64 |
|
| 838 |
+%elif %1 < 3 |
|
| 837 | 839 |
add r2, 64 |
| 838 | 840 |
%endif |
| 839 | 841 |
%endmacro |
| ... | ... |
@@ -845,15 +855,15 @@ cglobal h264_idct_add8_8_sse2, 5, 7, 8 |
| 845 | 845 |
%ifdef ARCH_X86_64 |
| 846 | 846 |
mov r10, r0 |
| 847 | 847 |
%endif |
| 848 |
- add8_sse2_cycle 0, 0x09 |
|
| 849 |
- add8_sse2_cycle 1, 0x11 |
|
| 848 |
+ add8_sse2_cycle 0, 0x34 |
|
| 849 |
+ add8_sse2_cycle 1, 0x3c |
|
| 850 | 850 |
%ifdef ARCH_X86_64 |
| 851 | 851 |
add r10, gprsize |
| 852 | 852 |
%else |
| 853 | 853 |
add r0mp, gprsize |
| 854 | 854 |
%endif |
| 855 |
- add8_sse2_cycle 2, 0x21 |
|
| 856 |
- add8_sse2_cycle 3, 0x29 |
|
| 855 |
+ add8_sse2_cycle 2, 0x5c |
|
| 856 |
+ add8_sse2_cycle 3, 0x64 |
|
| 857 | 857 |
RET |
| 858 | 858 |
|
| 859 | 859 |
;void ff_h264_luma_dc_dequant_idct_mmx(DCTELEM *output, DCTELEM *input, int qmul) |
| ... | ... |
@@ -29,14 +29,18 @@ SECTION_RODATA |
| 29 | 29 |
|
| 30 | 30 |
pw_pixel_max: times 8 dw ((1 << 10)-1) |
| 31 | 31 |
pd_32: times 4 dd 32 |
| 32 |
-scan8_mem: db 4+1*8, 5+1*8, 4+2*8, 5+2*8 |
|
| 33 |
- db 6+1*8, 7+1*8, 6+2*8, 7+2*8 |
|
| 34 |
- db 4+3*8, 5+3*8, 4+4*8, 5+4*8 |
|
| 35 |
- db 6+3*8, 7+3*8, 6+4*8, 7+4*8 |
|
| 36 |
- db 1+1*8, 2+1*8 |
|
| 37 |
- db 1+2*8, 2+2*8 |
|
| 38 |
- db 1+4*8, 2+4*8 |
|
| 39 |
- db 1+5*8, 2+5*8 |
|
| 32 |
+scan8_mem: db 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8 |
|
| 33 |
+ db 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8 |
|
| 34 |
+ db 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8 |
|
| 35 |
+ db 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8 |
|
| 36 |
+ db 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8 |
|
| 37 |
+ db 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8 |
|
| 38 |
+ db 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8 |
|
| 39 |
+ db 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8 |
|
| 40 |
+ db 4+11*8, 5+11*8, 4+12*8, 5+12*8 |
|
| 41 |
+ db 6+11*8, 7+11*8, 6+12*8, 7+12*8 |
|
| 42 |
+ db 4+13*8, 5+13*8, 4+14*8, 5+14*8 |
|
| 43 |
+ db 6+13*8, 7+13*8, 6+14*8, 7+14*8 |
|
| 40 | 44 |
|
| 41 | 45 |
%ifdef PIC |
| 42 | 46 |
%define scan8 r11 |
| ... | ... |
@@ -306,7 +310,7 @@ INIT_AVX |
| 306 | 306 |
IDCT_ADD16INTRA_10 avx |
| 307 | 307 |
%endif |
| 308 | 308 |
|
| 309 |
-%assign last_block 24 |
|
| 309 |
+%assign last_block 36 |
|
| 310 | 310 |
;----------------------------------------------------------------------------- |
| 311 | 311 |
; h264_idct_add8(pixel **dst, const int *block_offset, dctcoef *block, int stride, const uint8_t nnzc[6*8]) |
| 312 | 312 |
;----------------------------------------------------------------------------- |
| ... | ... |
@@ -317,21 +321,22 @@ cglobal h264_idct_add8_10_%1,5,7 |
| 317 | 317 |
%endif |
| 318 | 318 |
add r2, 1024 |
| 319 | 319 |
mov r0, [r0] |
| 320 |
- ADD16_OP_INTRA %1, 16, 1+1*8 |
|
| 321 |
- ADD16_OP_INTRA %1, 18, 1+2*8 |
|
| 320 |
+ ADD16_OP_INTRA %1, 16, 4+ 6*8 |
|
| 321 |
+ ADD16_OP_INTRA %1, 18, 4+ 7*8 |
|
| 322 |
+ add r2, 1024-128*2 |
|
| 322 | 323 |
%ifdef ARCH_X86_64 |
| 323 | 324 |
mov r0, [r10+gprsize] |
| 324 | 325 |
%else |
| 325 | 326 |
mov r0, r0m |
| 326 | 327 |
mov r0, [r0+gprsize] |
| 327 | 328 |
%endif |
| 328 |
- ADD16_OP_INTRA %1, 20, 1+4*8 |
|
| 329 |
- ADD16_OP_INTRA %1, 22, 1+5*8 |
|
| 329 |
+ ADD16_OP_INTRA %1, 32, 4+11*8 |
|
| 330 |
+ ADD16_OP_INTRA %1, 34, 4+12*8 |
|
| 330 | 331 |
REP_RET |
| 331 | 332 |
AC %1, 16 |
| 332 | 333 |
AC %1, 18 |
| 333 |
- AC %1, 20 |
|
| 334 |
- AC %1, 22 |
|
| 334 |
+ AC %1, 32 |
|
| 335 |
+ AC %1, 34 |
|
| 335 | 336 |
|
| 336 | 337 |
%endmacro ; IDCT_ADD8 |
| 337 | 338 |
|
| ... | ... |
@@ -506,6 +506,13 @@ static void yuv2nv12X_c(SwsContext *c, const int16_t *lumFilter, |
| 506 | 506 |
} |
| 507 | 507 |
} |
| 508 | 508 |
|
| 509 |
+#define output_pixel(pos, val) \ |
|
| 510 |
+ if (target == PIX_FMT_GRAY16BE) { \
|
|
| 511 |
+ AV_WB16(pos, val); \ |
|
| 512 |
+ } else { \
|
|
| 513 |
+ AV_WL16(pos, val); \ |
|
| 514 |
+ } |
|
| 515 |
+ |
|
| 509 | 516 |
static av_always_inline void |
| 510 | 517 |
yuv2gray16_X_c_template(SwsContext *c, const int16_t *lumFilter, |
| 511 | 518 |
const int16_t **lumSrc, int lumFilterSize, |
| ... | ... |
@@ -516,12 +523,6 @@ yuv2gray16_X_c_template(SwsContext *c, const int16_t *lumFilter, |
| 516 | 516 |
{
|
| 517 | 517 |
int i; |
| 518 | 518 |
|
| 519 |
-#define output_pixel(pos, val) \ |
|
| 520 |
- if (target == PIX_FMT_GRAY16BE) { \
|
|
| 521 |
- AV_WB16(pos, val); \ |
|
| 522 |
- } else { \
|
|
| 523 |
- AV_WL16(pos, val); \ |
|
| 524 |
- } |
|
| 525 | 519 |
for (i = 0; i < (dstW >> 1); i++) {
|
| 526 | 520 |
int j; |
| 527 | 521 |
int Y1 = 1 << 18; |
| ... | ... |
@@ -583,10 +584,11 @@ yuv2gray16_1_c_template(SwsContext *c, const uint16_t *buf0, |
| 583 | 583 |
output_pixel(&dest[2 * i2 + 0], Y1); |
| 584 | 584 |
output_pixel(&dest[2 * i2 + 2], Y2); |
| 585 | 585 |
} |
| 586 |
-#undef output_pixel |
|
| 587 | 586 |
} |
| 588 | 587 |
|
| 589 |
-#define YUV2PACKEDWRAPPER(name, ext, fmt) \ |
|
| 588 |
+#undef output_pixel |
|
| 589 |
+ |
|
| 590 |
+#define YUV2PACKEDWRAPPER(name, base, ext, fmt) \ |
|
| 590 | 591 |
static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \ |
| 591 | 592 |
const int16_t **lumSrc, int lumFilterSize, \ |
| 592 | 593 |
const int16_t *chrFilter, const int16_t **chrUSrc, \ |
| ... | ... |
@@ -594,7 +596,7 @@ static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \ |
| 594 | 594 |
const int16_t **alpSrc, uint8_t *dest, int dstW, \ |
| 595 | 595 |
int y) \ |
| 596 | 596 |
{ \
|
| 597 |
- name ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \ |
|
| 597 |
+ name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \ |
|
| 598 | 598 |
chrFilter, chrUSrc, chrVSrc, chrFilterSize, \ |
| 599 | 599 |
alpSrc, dest, dstW, y, fmt); \ |
| 600 | 600 |
} \ |
| ... | ... |
@@ -606,7 +608,7 @@ static void name ## ext ## _2_c(SwsContext *c, const uint16_t *buf0, \ |
| 606 | 606 |
const uint16_t *abuf1, uint8_t *dest, int dstW, \ |
| 607 | 607 |
int yalpha, int uvalpha, int y) \ |
| 608 | 608 |
{ \
|
| 609 |
- name ## _2_c_template(c, buf0, buf1, ubuf0, ubuf1, \ |
|
| 609 |
+ name ## base ## _2_c_template(c, buf0, buf1, ubuf0, ubuf1, \ |
|
| 610 | 610 |
vbuf0, vbuf1, abuf0, abuf1, \ |
| 611 | 611 |
dest, dstW, yalpha, uvalpha, y, fmt); \ |
| 612 | 612 |
} \ |
| ... | ... |
@@ -618,13 +620,20 @@ static void name ## ext ## _1_c(SwsContext *c, const uint16_t *buf0, \ |
| 618 | 618 |
int uvalpha, enum PixelFormat dstFormat, \ |
| 619 | 619 |
int flags, int y) \ |
| 620 | 620 |
{ \
|
| 621 |
- name ## _1_c_template(c, buf0, ubuf0, ubuf1, vbuf0, \ |
|
| 621 |
+ name ## base ## _1_c_template(c, buf0, ubuf0, ubuf1, vbuf0, \ |
|
| 622 | 622 |
vbuf1, abuf0, dest, dstW, uvalpha, \ |
| 623 | 623 |
dstFormat, flags, y, fmt); \ |
| 624 | 624 |
} |
| 625 | 625 |
|
| 626 |
-YUV2PACKEDWRAPPER(yuv2gray16, LE, PIX_FMT_GRAY16LE); |
|
| 627 |
-YUV2PACKEDWRAPPER(yuv2gray16, BE, PIX_FMT_GRAY16BE); |
|
| 626 |
+YUV2PACKEDWRAPPER(yuv2gray16,, LE, PIX_FMT_GRAY16LE); |
|
| 627 |
+YUV2PACKEDWRAPPER(yuv2gray16,, BE, PIX_FMT_GRAY16BE); |
|
| 628 |
+ |
|
| 629 |
+#define output_pixel(pos, acc) \ |
|
| 630 |
+ if (target == PIX_FMT_MONOBLACK) { \
|
|
| 631 |
+ pos = acc; \ |
|
| 632 |
+ } else { \
|
|
| 633 |
+ pos = ~acc; \ |
|
| 634 |
+ } |
|
| 628 | 635 |
|
| 629 | 636 |
static av_always_inline void |
| 630 | 637 |
yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter, |
| ... | ... |
@@ -639,12 +648,6 @@ yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter, |
| 639 | 639 |
int i; |
| 640 | 640 |
int acc = 0; |
| 641 | 641 |
|
| 642 |
-#define output_pixel(pos, acc) \ |
|
| 643 |
- if (target == PIX_FMT_MONOBLACK) { \
|
|
| 644 |
- pos = acc; \ |
|
| 645 |
- } else { \
|
|
| 646 |
- pos = ~acc; \ |
|
| 647 |
- } |
|
| 648 | 642 |
for (i = 0; i < dstW - 1; i += 2) {
|
| 649 | 643 |
int j; |
| 650 | 644 |
int Y1 = 1 << 18; |
| ... | ... |
@@ -718,21 +721,12 @@ yuv2mono_1_c_template(SwsContext *c, const uint16_t *buf0, |
| 718 | 718 |
acc += acc + g[(buf0[i + 7] >> 7) + d128[7]]; |
| 719 | 719 |
output_pixel(*dest++, acc); |
| 720 | 720 |
} |
| 721 |
-#undef output_pixel |
|
| 722 | 721 |
} |
| 723 | 722 |
|
| 724 |
-YUV2PACKEDWRAPPER(yuv2mono, white, PIX_FMT_MONOWHITE); |
|
| 725 |
-YUV2PACKEDWRAPPER(yuv2mono, black, PIX_FMT_MONOBLACK); |
|
| 723 |
+#undef output_pixel |
|
| 726 | 724 |
|
| 727 |
-static av_always_inline void |
|
| 728 |
-yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter, |
|
| 729 |
- const int16_t **lumSrc, int lumFilterSize, |
|
| 730 |
- const int16_t *chrFilter, const int16_t **chrUSrc, |
|
| 731 |
- const int16_t **chrVSrc, int chrFilterSize, |
|
| 732 |
- const int16_t **alpSrc, uint8_t *dest, int dstW, |
|
| 733 |
- int y, enum PixelFormat target) |
|
| 734 |
-{
|
|
| 735 |
- int i; |
|
| 725 |
+YUV2PACKEDWRAPPER(yuv2mono,, white, PIX_FMT_MONOWHITE); |
|
| 726 |
+YUV2PACKEDWRAPPER(yuv2mono,, black, PIX_FMT_MONOBLACK); |
|
| 736 | 727 |
|
| 737 | 728 |
#define output_pixels(pos, Y1, U, Y2, V) \ |
| 738 | 729 |
if (target == PIX_FMT_YUYV422) { \
|
| ... | ... |
@@ -747,6 +741,16 @@ yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter, |
| 747 | 747 |
dest[pos + 3] = Y2; \ |
| 748 | 748 |
} |
| 749 | 749 |
|
| 750 |
+static av_always_inline void |
|
| 751 |
+yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter, |
|
| 752 |
+ const int16_t **lumSrc, int lumFilterSize, |
|
| 753 |
+ const int16_t *chrFilter, const int16_t **chrUSrc, |
|
| 754 |
+ const int16_t **chrVSrc, int chrFilterSize, |
|
| 755 |
+ const int16_t **alpSrc, uint8_t *dest, int dstW, |
|
| 756 |
+ int y, enum PixelFormat target) |
|
| 757 |
+{
|
|
| 758 |
+ int i; |
|
| 759 |
+ |
|
| 750 | 760 |
for (i = 0; i < (dstW >> 1); i++) {
|
| 751 | 761 |
int j; |
| 752 | 762 |
int Y1 = 1 << 18; |
| ... | ... |
@@ -828,11 +832,156 @@ yuv2422_1_c_template(SwsContext *c, const uint16_t *buf0, |
| 828 | 828 |
output_pixels(i * 4, Y1, U, Y2, V); |
| 829 | 829 |
} |
| 830 | 830 |
} |
| 831 |
+} |
|
| 832 |
+ |
|
| 831 | 833 |
#undef output_pixels |
| 834 |
+ |
|
| 835 |
+YUV2PACKEDWRAPPER(yuv2, 422, yuyv422, PIX_FMT_YUYV422); |
|
| 836 |
+YUV2PACKEDWRAPPER(yuv2, 422, uyvy422, PIX_FMT_UYVY422); |
|
| 837 |
+ |
|
| 838 |
+#define r_b ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? r : b) |
|
| 839 |
+#define b_r ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? b : r) |
|
| 840 |
+ |
|
| 841 |
+static av_always_inline void |
|
| 842 |
+yuv2rgb48_X_c_template(SwsContext *c, const int16_t *lumFilter, |
|
| 843 |
+ const int16_t **lumSrc, int lumFilterSize, |
|
| 844 |
+ const int16_t *chrFilter, const int16_t **chrUSrc, |
|
| 845 |
+ const int16_t **chrVSrc, int chrFilterSize, |
|
| 846 |
+ const int16_t **alpSrc, uint8_t *dest, int dstW, |
|
| 847 |
+ int y, enum PixelFormat target) |
|
| 848 |
+{
|
|
| 849 |
+ int i; |
|
| 850 |
+ |
|
| 851 |
+ for (i = 0; i < (dstW >> 1); i++) {
|
|
| 852 |
+ int j; |
|
| 853 |
+ int Y1 = 1 << 18; |
|
| 854 |
+ int Y2 = 1 << 18; |
|
| 855 |
+ int U = 1 << 18; |
|
| 856 |
+ int V = 1 << 18; |
|
| 857 |
+ const uint8_t *r, *g, *b; |
|
| 858 |
+ |
|
| 859 |
+ for (j = 0; j < lumFilterSize; j++) {
|
|
| 860 |
+ Y1 += lumSrc[j][i * 2] * lumFilter[j]; |
|
| 861 |
+ Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j]; |
|
| 862 |
+ } |
|
| 863 |
+ for (j = 0; j < chrFilterSize; j++) {
|
|
| 864 |
+ U += chrUSrc[j][i] * chrFilter[j]; |
|
| 865 |
+ V += chrVSrc[j][i] * chrFilter[j]; |
|
| 866 |
+ } |
|
| 867 |
+ Y1 >>= 19; |
|
| 868 |
+ Y2 >>= 19; |
|
| 869 |
+ U >>= 19; |
|
| 870 |
+ V >>= 19; |
|
| 871 |
+ if ((Y1 | Y2 | U | V) & 0x100) {
|
|
| 872 |
+ Y1 = av_clip_uint8(Y1); |
|
| 873 |
+ Y2 = av_clip_uint8(Y2); |
|
| 874 |
+ U = av_clip_uint8(U); |
|
| 875 |
+ V = av_clip_uint8(V); |
|
| 876 |
+ } |
|
| 877 |
+ |
|
| 878 |
+ /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/ |
|
| 879 |
+ r = (const uint8_t *) c->table_rV[V]; |
|
| 880 |
+ g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]); |
|
| 881 |
+ b = (const uint8_t *) c->table_bU[U]; |
|
| 882 |
+ |
|
| 883 |
+ dest[ 0] = dest[ 1] = r_b[Y1]; |
|
| 884 |
+ dest[ 2] = dest[ 3] = g[Y1]; |
|
| 885 |
+ dest[ 4] = dest[ 5] = b_r[Y1]; |
|
| 886 |
+ dest[ 6] = dest[ 7] = r_b[Y2]; |
|
| 887 |
+ dest[ 8] = dest[ 9] = g[Y2]; |
|
| 888 |
+ dest[10] = dest[11] = b_r[Y2]; |
|
| 889 |
+ dest += 12; |
|
| 890 |
+ } |
|
| 891 |
+} |
|
| 892 |
+ |
|
| 893 |
+static av_always_inline void |
|
| 894 |
+yuv2rgb48_2_c_template(SwsContext *c, const uint16_t *buf0, |
|
| 895 |
+ const uint16_t *buf1, const uint16_t *ubuf0, |
|
| 896 |
+ const uint16_t *ubuf1, const uint16_t *vbuf0, |
|
| 897 |
+ const uint16_t *vbuf1, const uint16_t *abuf0, |
|
| 898 |
+ const uint16_t *abuf1, uint8_t *dest, int dstW, |
|
| 899 |
+ int yalpha, int uvalpha, int y, |
|
| 900 |
+ enum PixelFormat target) |
|
| 901 |
+{
|
|
| 902 |
+ int yalpha1 = 4095 - yalpha; |
|
| 903 |
+ int uvalpha1 = 4095 - uvalpha; |
|
| 904 |
+ int i; |
|
| 905 |
+ |
|
| 906 |
+ for (i = 0; i < (dstW >> 1); i++) {
|
|
| 907 |
+ int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19; |
|
| 908 |
+ int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19; |
|
| 909 |
+ int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19; |
|
| 910 |
+ int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19; |
|
| 911 |
+ const uint8_t *r = (const uint8_t *) c->table_rV[V], |
|
| 912 |
+ *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]), |
|
| 913 |
+ *b = (const uint8_t *) c->table_bU[U]; |
|
| 914 |
+ |
|
| 915 |
+ dest[ 0] = dest[ 1] = r_b[Y1]; |
|
| 916 |
+ dest[ 2] = dest[ 3] = g[Y1]; |
|
| 917 |
+ dest[ 4] = dest[ 5] = b_r[Y1]; |
|
| 918 |
+ dest[ 6] = dest[ 7] = r_b[Y2]; |
|
| 919 |
+ dest[ 8] = dest[ 9] = g[Y2]; |
|
| 920 |
+ dest[10] = dest[11] = b_r[Y2]; |
|
| 921 |
+ dest += 12; |
|
| 922 |
+ } |
|
| 923 |
+} |
|
| 924 |
+ |
|
| 925 |
+static av_always_inline void |
|
| 926 |
+yuv2rgb48_1_c_template(SwsContext *c, const uint16_t *buf0, |
|
| 927 |
+ const uint16_t *ubuf0, const uint16_t *ubuf1, |
|
| 928 |
+ const uint16_t *vbuf0, const uint16_t *vbuf1, |
|
| 929 |
+ const uint16_t *abuf0, uint8_t *dest, int dstW, |
|
| 930 |
+ int uvalpha, enum PixelFormat dstFormat, |
|
| 931 |
+ int flags, int y, enum PixelFormat target) |
|
| 932 |
+{
|
|
| 933 |
+ int i; |
|
| 934 |
+ |
|
| 935 |
+ if (uvalpha < 2048) {
|
|
| 936 |
+ for (i = 0; i < (dstW >> 1); i++) {
|
|
| 937 |
+ int Y1 = buf0[i * 2] >> 7; |
|
| 938 |
+ int Y2 = buf0[i * 2 + 1] >> 7; |
|
| 939 |
+ int U = ubuf1[i] >> 7; |
|
| 940 |
+ int V = vbuf1[i] >> 7; |
|
| 941 |
+ const uint8_t *r = (const uint8_t *) c->table_rV[V], |
|
| 942 |
+ *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]), |
|
| 943 |
+ *b = (const uint8_t *) c->table_bU[U]; |
|
| 944 |
+ |
|
| 945 |
+ dest[ 0] = dest[ 1] = r_b[Y1]; |
|
| 946 |
+ dest[ 2] = dest[ 3] = g[Y1]; |
|
| 947 |
+ dest[ 4] = dest[ 5] = b_r[Y1]; |
|
| 948 |
+ dest[ 6] = dest[ 7] = r_b[Y2]; |
|
| 949 |
+ dest[ 8] = dest[ 9] = g[Y2]; |
|
| 950 |
+ dest[10] = dest[11] = b_r[Y2]; |
|
| 951 |
+ dest += 12; |
|
| 952 |
+ } |
|
| 953 |
+ } else {
|
|
| 954 |
+ for (i = 0; i < (dstW >> 1); i++) {
|
|
| 955 |
+ int Y1 = buf0[i * 2] >> 7; |
|
| 956 |
+ int Y2 = buf0[i * 2 + 1] >> 7; |
|
| 957 |
+ int U = (ubuf0[i] + ubuf1[i]) >> 8; |
|
| 958 |
+ int V = (vbuf0[i] + vbuf1[i]) >> 8; |
|
| 959 |
+ const uint8_t *r = (const uint8_t *) c->table_rV[V], |
|
| 960 |
+ *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]), |
|
| 961 |
+ *b = (const uint8_t *) c->table_bU[U]; |
|
| 962 |
+ |
|
| 963 |
+ dest[ 0] = dest[ 1] = r_b[Y1]; |
|
| 964 |
+ dest[ 2] = dest[ 3] = g[Y1]; |
|
| 965 |
+ dest[ 4] = dest[ 5] = b_r[Y1]; |
|
| 966 |
+ dest[ 6] = dest[ 7] = r_b[Y2]; |
|
| 967 |
+ dest[ 8] = dest[ 9] = g[Y2]; |
|
| 968 |
+ dest[10] = dest[11] = b_r[Y2]; |
|
| 969 |
+ dest += 12; |
|
| 970 |
+ } |
|
| 971 |
+ } |
|
| 832 | 972 |
} |
| 833 | 973 |
|
| 834 |
-YUV2PACKEDWRAPPER(yuv2422, yuyv, PIX_FMT_YUYV422); |
|
| 835 |
-YUV2PACKEDWRAPPER(yuv2422, uyvy, PIX_FMT_UYVY422); |
|
| 974 |
+#undef r_b |
|
| 975 |
+#undef b_r |
|
| 976 |
+ |
|
| 977 |
+YUV2PACKEDWRAPPER(yuv2, rgb48, rgb48be, PIX_FMT_RGB48BE); |
|
| 978 |
+//YUV2PACKEDWRAPPER(yuv2, rgb48, rgb48le, PIX_FMT_RGB48LE); |
|
| 979 |
+YUV2PACKEDWRAPPER(yuv2, rgb48, bgr48be, PIX_FMT_BGR48BE); |
|
| 980 |
+//YUV2PACKEDWRAPPER(yuv2, rgb48, bgr48le, PIX_FMT_BGR48LE); |
|
| 836 | 981 |
|
| 837 | 982 |
#define YSCALE_YUV_2_RGBX_C(type,alpha) \ |
| 838 | 983 |
for (i=0; i<(dstW>>1); i++) {\
|
| ... | ... |
@@ -974,36 +1123,6 @@ YUV2PACKEDWRAPPER(yuv2422, uyvy, PIX_FMT_UYVY422); |
| 974 | 974 |
|
| 975 | 975 |
#define YSCALE_YUV_2_ANYRGB_C(func)\ |
| 976 | 976 |
switch(c->dstFormat) {\
|
| 977 |
- case PIX_FMT_RGB48BE:\ |
|
| 978 |
- case PIX_FMT_RGB48LE:\ |
|
| 979 |
- func(uint8_t,0)\ |
|
| 980 |
- ((uint8_t*)dest)[ 0]= r[Y1];\ |
|
| 981 |
- ((uint8_t*)dest)[ 1]= r[Y1];\ |
|
| 982 |
- ((uint8_t*)dest)[ 2]= g[Y1];\ |
|
| 983 |
- ((uint8_t*)dest)[ 3]= g[Y1];\ |
|
| 984 |
- ((uint8_t*)dest)[ 4]= b[Y1];\ |
|
| 985 |
- ((uint8_t*)dest)[ 5]= b[Y1];\ |
|
| 986 |
- ((uint8_t*)dest)[ 6]= r[Y2];\ |
|
| 987 |
- ((uint8_t*)dest)[ 7]= r[Y2];\ |
|
| 988 |
- ((uint8_t*)dest)[ 8]= g[Y2];\ |
|
| 989 |
- ((uint8_t*)dest)[ 9]= g[Y2];\ |
|
| 990 |
- ((uint8_t*)dest)[10]= b[Y2];\ |
|
| 991 |
- ((uint8_t*)dest)[11]= b[Y2];\ |
|
| 992 |
- dest+=12;\ |
|
| 993 |
- }\ |
|
| 994 |
- break;\ |
|
| 995 |
- case PIX_FMT_BGR48BE:\ |
|
| 996 |
- case PIX_FMT_BGR48LE:\ |
|
| 997 |
- func(uint8_t,0)\ |
|
| 998 |
- ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\ |
|
| 999 |
- ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\ |
|
| 1000 |
- ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\ |
|
| 1001 |
- ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\ |
|
| 1002 |
- ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\ |
|
| 1003 |
- ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\ |
|
| 1004 |
- dest+=12;\ |
|
| 1005 |
- }\ |
|
| 1006 |
- break;\ |
|
| 1007 | 977 |
case PIX_FMT_RGBA:\ |
| 1008 | 978 |
case PIX_FMT_BGRA:\ |
| 1009 | 979 |
if (CONFIG_SMALL) {\
|
| ... | ... |
@@ -1294,19 +1413,21 @@ static av_always_inline void fillPlane(uint8_t* plane, int stride, |
| 1294 | 1294 |
} |
| 1295 | 1295 |
} |
| 1296 | 1296 |
|
| 1297 |
+#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos)) |
|
| 1298 |
+ |
|
| 1299 |
+#define r ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? b_r : r_b) |
|
| 1300 |
+#define b ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? r_b : b_r) |
|
| 1301 |
+ |
|
| 1297 | 1302 |
static av_always_inline void |
| 1298 | 1303 |
rgb48ToY_c_template(int16_t *dst, const uint16_t *src, int width, |
| 1299 | 1304 |
enum PixelFormat origin) |
| 1300 | 1305 |
{
|
| 1301 | 1306 |
int i; |
| 1302 | 1307 |
for (i = 0; i < width; i++) {
|
| 1303 |
-#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos)) |
|
| 1304 |
- int a = input_pixel(&src[i*3+0]); |
|
| 1305 |
- int g = input_pixel(&src[i*3+1]); |
|
| 1306 |
- int c = input_pixel(&src[i*3+2]); |
|
| 1308 |
+ int r_b = input_pixel(&src[i*3+0]); |
|
| 1309 |
+ int g = input_pixel(&src[i*3+1]); |
|
| 1310 |
+ int b_r = input_pixel(&src[i*3+2]); |
|
| 1307 | 1311 |
|
| 1308 |
-#define r ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? c : a) |
|
| 1309 |
-#define b ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? a : c) |
|
| 1310 | 1312 |
dst[i] = (RY*r + GY*g + BY*b + (32<<(RGB2YUV_SHIFT-1+8)) + (1<<(RGB2YUV_SHIFT-7+8))) >> (RGB2YUV_SHIFT-6+8); |
| 1311 | 1313 |
} |
| 1312 | 1314 |
} |
| ... | ... |
@@ -1319,9 +1440,9 @@ rgb48ToUV_c_template(int16_t *dstU, int16_t *dstV, |
| 1319 | 1319 |
int i; |
| 1320 | 1320 |
assert(src1==src2); |
| 1321 | 1321 |
for (i = 0; i < width; i++) {
|
| 1322 |
- int a = input_pixel(&src1[3*i + 0]); |
|
| 1323 |
- int g = input_pixel(&src1[3*i + 1]); |
|
| 1324 |
- int c = input_pixel(&src1[3*i + 2]); |
|
| 1322 |
+ int r_b = input_pixel(&src1[i*3+0]); |
|
| 1323 |
+ int g = input_pixel(&src1[i*3+1]); |
|
| 1324 |
+ int b_r = input_pixel(&src1[i*3+2]); |
|
| 1325 | 1325 |
|
| 1326 | 1326 |
dstU[i] = (RU*r + GU*g + BU*b + (256<<(RGB2YUV_SHIFT-1+8)) + (1<<(RGB2YUV_SHIFT-7+8))) >> (RGB2YUV_SHIFT-6+8); |
| 1327 | 1327 |
dstV[i] = (RV*r + GV*g + BV*b + (256<<(RGB2YUV_SHIFT-1+8)) + (1<<(RGB2YUV_SHIFT-7+8))) >> (RGB2YUV_SHIFT-6+8); |
| ... | ... |
@@ -1336,17 +1457,18 @@ rgb48ToUV_half_c_template(int16_t *dstU, int16_t *dstV, |
| 1336 | 1336 |
int i; |
| 1337 | 1337 |
assert(src1==src2); |
| 1338 | 1338 |
for (i = 0; i < width; i++) {
|
| 1339 |
- int a = (input_pixel(&src1[6*i + 0])) + (input_pixel(&src1[6*i + 3])); |
|
| 1340 |
- int g = (input_pixel(&src1[6*i + 1])) + (input_pixel(&src1[6*i + 4])); |
|
| 1341 |
- int c = (input_pixel(&src1[6*i + 2])) + (input_pixel(&src1[6*i + 5])); |
|
| 1339 |
+ int r_b = (input_pixel(&src1[6*i + 0])) + (input_pixel(&src1[6*i + 3])); |
|
| 1340 |
+ int g = (input_pixel(&src1[6*i + 1])) + (input_pixel(&src1[6*i + 4])); |
|
| 1341 |
+ int b_r = (input_pixel(&src1[6*i + 2])) + (input_pixel(&src1[6*i + 5])); |
|
| 1342 | 1342 |
|
| 1343 | 1343 |
dstU[i]= (RU*r + GU*g + BU*b + (256U<<(RGB2YUV_SHIFT+8)) + (1<<(RGB2YUV_SHIFT-6+8))) >> (RGB2YUV_SHIFT-5+8); |
| 1344 | 1344 |
dstV[i]= (RV*r + GV*g + BV*b + (256U<<(RGB2YUV_SHIFT+8)) + (1<<(RGB2YUV_SHIFT-6+8))) >> (RGB2YUV_SHIFT-5+8); |
| 1345 | 1345 |
} |
| 1346 |
+} |
|
| 1347 |
+ |
|
| 1346 | 1348 |
#undef r |
| 1347 | 1349 |
#undef b |
| 1348 | 1350 |
#undef input_pixel |
| 1349 |
-} |
|
| 1350 | 1351 |
|
| 1351 | 1352 |
#define rgb48funcs(pattern, BE_LE, origin) \ |
| 1352 | 1353 |
static void pattern ## 48 ## BE_LE ## ToY_c(uint8_t *dst, const uint8_t *src, \ |
| ... | ... |
@@ -1374,6 +1496,10 @@ rgb48funcs(rgb, BE, PIX_FMT_RGB48BE); |
| 1374 | 1374 |
rgb48funcs(bgr, LE, PIX_FMT_BGR48LE); |
| 1375 | 1375 |
rgb48funcs(bgr, BE, PIX_FMT_BGR48BE); |
| 1376 | 1376 |
|
| 1377 |
+#define input_pixel(i) ((origin == PIX_FMT_RGBA || origin == PIX_FMT_BGRA || \ |
|
| 1378 |
+ origin == PIX_FMT_ARGB || origin == PIX_FMT_ABGR) ? AV_RN32A(&src[(i)*4]) : \ |
|
| 1379 |
+ (isBE(origin) ? AV_RB16(&src[(i)*2]) : AV_RL16(&src[(i)*2]))) |
|
| 1380 |
+ |
|
| 1377 | 1381 |
static av_always_inline void |
| 1378 | 1382 |
rgb16_32ToY_c_template(int16_t *dst, const uint8_t *src, |
| 1379 | 1383 |
int width, enum PixelFormat origin, |
| ... | ... |
@@ -1386,9 +1512,6 @@ rgb16_32ToY_c_template(int16_t *dst, const uint8_t *src, |
| 1386 | 1386 |
int i; |
| 1387 | 1387 |
|
| 1388 | 1388 |
for (i = 0; i < width; i++) {
|
| 1389 |
-#define input_pixel(i) ((origin == PIX_FMT_RGBA || origin == PIX_FMT_BGRA || \ |
|
| 1390 |
- origin == PIX_FMT_ARGB || origin == PIX_FMT_ABGR) ? AV_RN32A(&src[(i)*4]) : \ |
|
| 1391 |
- (isBE(origin) ? AV_RB16(&src[(i)*2]) : AV_RL16(&src[(i)*2]))) |
|
| 1392 | 1389 |
int px = input_pixel(i) >> shp; |
| 1393 | 1390 |
int b = (px & maskb) >> shb; |
| 1394 | 1391 |
int g = (px & maskg) >> shg; |
| ... | ... |
@@ -1454,9 +1577,10 @@ rgb16_32ToUV_half_c_template(int16_t *dstU, int16_t *dstV, |
| 1454 | 1454 |
dstU[i] = (ru * r + gu * g + bu * b + (unsigned)rnd) >> ((S)-6+1); |
| 1455 | 1455 |
dstV[i] = (rv * r + gv * g + bv * b + (unsigned)rnd) >> ((S)-6+1); |
| 1456 | 1456 |
} |
| 1457 |
-#undef input_pixel |
|
| 1458 | 1457 |
} |
| 1459 | 1458 |
|
| 1459 |
+#undef input_pixel |
|
| 1460 |
+ |
|
| 1460 | 1461 |
#define rgb16_32_wrapper(fmt, name, shr, shg, shb, shp, maskr, \ |
| 1461 | 1462 |
maskg, maskb, rsh, gsh, bsh, S) \ |
| 1462 | 1463 |
static void name ## ToY_c(uint8_t *dst, const uint8_t *src, \ |
| ... | ... |
@@ -1637,6 +1761,22 @@ static av_always_inline void nvXXtoUV_c(uint8_t *dst1, uint8_t *dst2, |
| 1637 | 1637 |
} |
| 1638 | 1638 |
} |
| 1639 | 1639 |
|
| 1640 |
+static void nv12ToUV_c(uint8_t *dstU, uint8_t *dstV, |
|
| 1641 |
+ const uint8_t *src1, const uint8_t *src2, |
|
| 1642 |
+ int width, uint32_t *unused) |
|
| 1643 |
+{
|
|
| 1644 |
+ nvXXtoUV_c(dstU, dstV, src1, width); |
|
| 1645 |
+} |
|
| 1646 |
+ |
|
| 1647 |
+static void nv21ToUV_c(uint8_t *dstU, uint8_t *dstV, |
|
| 1648 |
+ const uint8_t *src1, const uint8_t *src2, |
|
| 1649 |
+ int width, uint32_t *unused) |
|
| 1650 |
+{
|
|
| 1651 |
+ nvXXtoUV_c(dstV, dstU, src1, width); |
|
| 1652 |
+} |
|
| 1653 |
+ |
|
| 1654 |
+#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos)) |
|
| 1655 |
+ |
|
| 1640 | 1656 |
// FIXME Maybe dither instead. |
| 1641 | 1657 |
static av_always_inline void |
| 1642 | 1658 |
yuv9_OR_10ToUV_c_template(uint8_t *dstU, uint8_t *dstV, |
| ... | ... |
@@ -1647,7 +1787,6 @@ yuv9_OR_10ToUV_c_template(uint8_t *dstU, uint8_t *dstV, |
| 1647 | 1647 |
const uint16_t *srcU = (const uint16_t *) _srcU; |
| 1648 | 1648 |
const uint16_t *srcV = (const uint16_t *) _srcV; |
| 1649 | 1649 |
|
| 1650 |
-#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos)) |
|
| 1651 | 1650 |
for (i = 0; i < width; i++) {
|
| 1652 | 1651 |
dstU[i] = input_pixel(&srcU[i]) >> (depth - 8); |
| 1653 | 1652 |
dstV[i] = input_pixel(&srcV[i]) >> (depth - 8); |
| ... | ... |
@@ -1663,9 +1802,10 @@ yuv9_or_10ToY_c_template(uint8_t *dstY, const uint8_t *_srcY, |
| 1663 | 1663 |
|
| 1664 | 1664 |
for (i = 0; i < width; i++) |
| 1665 | 1665 |
dstY[i] = input_pixel(&srcY[i]) >> (depth - 8); |
| 1666 |
-#undef input_pixel |
|
| 1667 | 1666 |
} |
| 1668 | 1667 |
|
| 1668 |
+#undef input_pixel |
|
| 1669 |
+ |
|
| 1669 | 1670 |
#define YUV_NBPS(depth, BE_LE, origin) \ |
| 1670 | 1671 |
static void BE_LE ## depth ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \ |
| 1671 | 1672 |
const uint8_t *srcU, const uint8_t *srcV, \ |
| ... | ... |
@@ -1684,20 +1824,6 @@ YUV_NBPS( 9, BE, PIX_FMT_YUV420P9BE); |
| 1684 | 1684 |
YUV_NBPS(10, LE, PIX_FMT_YUV420P10LE); |
| 1685 | 1685 |
YUV_NBPS(10, BE, PIX_FMT_YUV420P10BE); |
| 1686 | 1686 |
|
| 1687 |
-static inline void nv12ToUV_c(uint8_t *dstU, uint8_t *dstV, |
|
| 1688 |
- const uint8_t *src1, const uint8_t *src2, |
|
| 1689 |
- int width, uint32_t *unused) |
|
| 1690 |
-{
|
|
| 1691 |
- nvXXtoUV_c(dstU, dstV, src1, width); |
|
| 1692 |
-} |
|
| 1693 |
- |
|
| 1694 |
-static inline void nv21ToUV_c(uint8_t *dstU, uint8_t *dstV, |
|
| 1695 |
- const uint8_t *src1, const uint8_t *src2, |
|
| 1696 |
- int width, uint32_t *unused) |
|
| 1697 |
-{
|
|
| 1698 |
- nvXXtoUV_c(dstV, dstU, src1, width); |
|
| 1699 |
-} |
|
| 1700 |
- |
|
| 1701 | 1687 |
static void bgr24ToY_c(int16_t *dst, const uint8_t *src, |
| 1702 | 1688 |
int width, uint32_t *unused) |
| 1703 | 1689 |
{
|
| ... | ... |
@@ -2003,14 +2129,34 @@ find_c_packed_planar_out_funcs(SwsContext *c, |
| 2003 | 2003 |
*yuv2packedX = yuv2monoblack_X_c; |
| 2004 | 2004 |
break; |
| 2005 | 2005 |
case PIX_FMT_YUYV422: |
| 2006 |
- *yuv2packed1 = yuv2422yuyv_1_c; |
|
| 2007 |
- *yuv2packed2 = yuv2422yuyv_2_c; |
|
| 2008 |
- *yuv2packedX = yuv2422yuyv_X_c; |
|
| 2006 |
+ *yuv2packed1 = yuv2yuyv422_1_c; |
|
| 2007 |
+ *yuv2packed2 = yuv2yuyv422_2_c; |
|
| 2008 |
+ *yuv2packedX = yuv2yuyv422_X_c; |
|
| 2009 | 2009 |
break; |
| 2010 | 2010 |
case PIX_FMT_UYVY422: |
| 2011 |
- *yuv2packed1 = yuv2422uyvy_1_c; |
|
| 2012 |
- *yuv2packed2 = yuv2422uyvy_2_c; |
|
| 2013 |
- *yuv2packedX = yuv2422uyvy_X_c; |
|
| 2011 |
+ *yuv2packed1 = yuv2uyvy422_1_c; |
|
| 2012 |
+ *yuv2packed2 = yuv2uyvy422_2_c; |
|
| 2013 |
+ *yuv2packedX = yuv2uyvy422_X_c; |
|
| 2014 |
+ break; |
|
| 2015 |
+ case PIX_FMT_RGB48LE: |
|
| 2016 |
+ //*yuv2packed1 = yuv2rgb48le_1_c; |
|
| 2017 |
+ //*yuv2packed2 = yuv2rgb48le_2_c; |
|
| 2018 |
+ //*yuv2packedX = yuv2rgb48le_X_c; |
|
| 2019 |
+ //break; |
|
| 2020 |
+ case PIX_FMT_RGB48BE: |
|
| 2021 |
+ *yuv2packed1 = yuv2rgb48be_1_c; |
|
| 2022 |
+ *yuv2packed2 = yuv2rgb48be_2_c; |
|
| 2023 |
+ *yuv2packedX = yuv2rgb48be_X_c; |
|
| 2024 |
+ break; |
|
| 2025 |
+ case PIX_FMT_BGR48LE: |
|
| 2026 |
+ //*yuv2packed1 = yuv2bgr48le_1_c; |
|
| 2027 |
+ //*yuv2packed2 = yuv2bgr48le_2_c; |
|
| 2028 |
+ //*yuv2packedX = yuv2bgr48le_X_c; |
|
| 2029 |
+ //break; |
|
| 2030 |
+ case PIX_FMT_BGR48BE: |
|
| 2031 |
+ *yuv2packed1 = yuv2bgr48be_1_c; |
|
| 2032 |
+ *yuv2packed2 = yuv2bgr48be_2_c; |
|
| 2033 |
+ *yuv2packedX = yuv2bgr48be_X_c; |
|
| 2014 | 2034 |
break; |
| 2015 | 2035 |
default: |
| 2016 | 2036 |
*yuv2packed1 = yuv2packed1_c; |