* @author Michael Niedermayer <michaelni@gmx.at>
*/
+#include "libavutil/attributes.h"
+#include "dsputil.h"
#include "h264pred.h"
#define BIT_DEPTH 8
#include "h264pred_template.c"
#undef BIT_DEPTH
-static void pred4x4_vertical_vp8_c(uint8_t *src, const uint8_t *topright, int stride){
- const int lt= src[-1-1*stride];
+static void pred4x4_vertical_vp8_c(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride)
+{
+ const unsigned lt = src[-1-1*stride];
LOAD_TOP_EDGE
LOAD_TOP_RIGHT_EDGE
uint32_t v = PACK_4U8((lt + 2*t0 + t1 + 2) >> 2,
AV_WN32A(src+3*stride, v);
}
-static void pred4x4_horizontal_vp8_c(uint8_t *src, const uint8_t *topright, int stride){
- const int lt= src[-1-1*stride];
+static void pred4x4_horizontal_vp8_c(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride)
+{
+ const unsigned lt = src[-1-1*stride];
LOAD_LEFT_EDGE
AV_WN32A(src+0*stride, ((lt + 2*l0 + l1 + 2) >> 2)*0x01010101);
AV_WN32A(src+3*stride, ((l2 + 2*l3 + l3 + 2) >> 2)*0x01010101);
}
-static void pred4x4_down_left_svq3_c(uint8_t *src, const uint8_t *topright, int stride){
+static void pred4x4_down_left_svq3_c(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride)
+{
LOAD_TOP_EDGE
LOAD_LEFT_EDGE
src[3+3*stride]=(l3 + t3)>>1;
}
-static void pred4x4_down_left_rv40_c(uint8_t *src, const uint8_t *topright, int stride){
+static void pred4x4_down_left_rv40_c(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride)
+{
LOAD_TOP_EDGE
LOAD_TOP_RIGHT_EDGE
LOAD_LEFT_EDGE
src[3+3*stride]=(t6 + t7 + 1 + l6 + l7 + 1)>>2;
}
-static void pred4x4_down_left_rv40_nodown_c(uint8_t *src, const uint8_t *topright, int stride){
+static void pred4x4_down_left_rv40_nodown_c(uint8_t *src,
+ const uint8_t *topright,
+ ptrdiff_t stride)
+{
LOAD_TOP_EDGE
LOAD_TOP_RIGHT_EDGE
LOAD_LEFT_EDGE
src[3+3*stride]=(t6 + t7 + 1 + 2*l3 + 1)>>2;
}
-static void pred4x4_vertical_left_rv40(uint8_t *src, const uint8_t *topright, int stride,
- const int l0, const int l1, const int l2, const int l3, const int l4){
+static void pred4x4_vertical_left_rv40(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride,
+ const int l0, const int l1, const int l2,
+ const int l3, const int l4)
+{
LOAD_TOP_EDGE
LOAD_TOP_RIGHT_EDGE
src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
}
-static void pred4x4_vertical_left_rv40_c(uint8_t *src, const uint8_t *topright, int stride){
+static void pred4x4_vertical_left_rv40_c(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride)
+{
LOAD_LEFT_EDGE
LOAD_DOWN_LEFT_EDGE
pred4x4_vertical_left_rv40(src, topright, stride, l0, l1, l2, l3, l4);
}
-static void pred4x4_vertical_left_rv40_nodown_c(uint8_t *src, const uint8_t *topright, int stride){
+static void pred4x4_vertical_left_rv40_nodown_c(uint8_t *src,
+ const uint8_t *topright,
+ ptrdiff_t stride)
+{
LOAD_LEFT_EDGE
pred4x4_vertical_left_rv40(src, topright, stride, l0, l1, l2, l3, l3);
}
-static void pred4x4_vertical_left_vp8_c(uint8_t *src, const uint8_t *topright, int stride){
+static void pred4x4_vertical_left_vp8_c(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride)
+{
LOAD_TOP_EDGE
LOAD_TOP_RIGHT_EDGE
src[3+3*stride]=(t5 + 2*t6 + t7 + 2)>>2;
}
-static void pred4x4_horizontal_up_rv40_c(uint8_t *src, const uint8_t *topright, int stride){
+static void pred4x4_horizontal_up_rv40_c(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride)
+{
LOAD_LEFT_EDGE
LOAD_DOWN_LEFT_EDGE
LOAD_TOP_EDGE
src[3+3*stride]=(l4 + 2*l5 + l6 + 2)>>2;
}
-static void pred4x4_horizontal_up_rv40_nodown_c(uint8_t *src, const uint8_t *topright, int stride){
+static void pred4x4_horizontal_up_rv40_nodown_c(uint8_t *src,
+ const uint8_t *topright,
+ ptrdiff_t stride)
+{
LOAD_LEFT_EDGE
LOAD_TOP_EDGE
LOAD_TOP_RIGHT_EDGE
src[3+3*stride]=l3;
}
-static void pred4x4_tm_vp8_c(uint8_t *src, const uint8_t *topright, int stride){
- uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
+static void pred4x4_tm_vp8_c(uint8_t *src, const uint8_t *topright,
+ ptrdiff_t stride)
+{
+ const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
uint8_t *top = src-stride;
int y;
for (y = 0; y < 4; y++) {
- uint8_t *cm_in = cm + src[-1];
+ const uint8_t *cm_in = cm + src[-1];
src[0] = cm_in[top[0]];
src[1] = cm_in[top[1]];
src[2] = cm_in[top[2]];
}
}
-static void pred16x16_plane_svq3_c(uint8_t *src, int stride){
+static void pred16x16_plane_svq3_c(uint8_t *src, ptrdiff_t stride)
+{
pred16x16_plane_compat_8_c(src, stride, 1, 0);
}
-static void pred16x16_plane_rv40_c(uint8_t *src, int stride){
+static void pred16x16_plane_rv40_c(uint8_t *src, ptrdiff_t stride)
+{
pred16x16_plane_compat_8_c(src, stride, 0, 1);
}
-static void pred16x16_tm_vp8_c(uint8_t *src, int stride){
- uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
+static void pred16x16_tm_vp8_c(uint8_t *src, ptrdiff_t stride)
+{
+ const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
uint8_t *top = src-stride;
int y;
for (y = 0; y < 16; y++) {
- uint8_t *cm_in = cm + src[-1];
+ const uint8_t *cm_in = cm + src[-1];
src[0] = cm_in[top[0]];
src[1] = cm_in[top[1]];
src[2] = cm_in[top[2]];
}
}
-static void pred8x8_left_dc_rv40_c(uint8_t *src, int stride){
+static void pred8x8_left_dc_rv40_c(uint8_t *src, ptrdiff_t stride)
+{
int i;
- int dc0;
+ unsigned dc0;
dc0=0;
for(i=0;i<8; i++)
}
}
-static void pred8x8_top_dc_rv40_c(uint8_t *src, int stride){
+static void pred8x8_top_dc_rv40_c(uint8_t *src, ptrdiff_t stride)
+{
int i;
- int dc0;
+ unsigned dc0;
dc0=0;
for(i=0;i<8; i++)
}
}
-static void pred8x8_dc_rv40_c(uint8_t *src, int stride){
+static void pred8x8_dc_rv40_c(uint8_t *src, ptrdiff_t stride)
+{
int i;
- int dc0=0;
+ unsigned dc0 = 0;
for(i=0;i<4; i++){
dc0+= src[-1+i*stride] + src[i-stride];
}
}
-static void pred8x8_tm_vp8_c(uint8_t *src, int stride){
- uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
+static void pred8x8_tm_vp8_c(uint8_t *src, ptrdiff_t stride)
+{
+ const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP - src[-1-stride];
uint8_t *top = src-stride;
int y;
for (y = 0; y < 8; y++) {
- uint8_t *cm_in = cm + src[-1];
+ const uint8_t *cm_in = cm + src[-1];
src[0] = cm_in[top[0]];
src[1] = cm_in[top[1]];
src[2] = cm_in[top[2]];
/**
* Set the intra prediction function pointers.
*/
-void ff_h264_pred_init(H264PredContext *h, int codec_id, const int bit_depth){
-// MpegEncContext * const s = &h->s;
-
+av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id,
+ const int bit_depth,
+ const int chroma_format_idc)
+{
#undef FUNC
#undef FUNCC
#define FUNC(a, depth) a ## _ ## depth
#define FUNCD(a) a ## _c
#define H264_PRED(depth) \
- if(codec_id != CODEC_ID_RV40){\
- if(codec_id == CODEC_ID_VP8) {\
+ if(codec_id != AV_CODEC_ID_RV40){\
+ if(codec_id == AV_CODEC_ID_VP8) {\
h->pred4x4[VERT_PRED ]= FUNCD(pred4x4_vertical_vp8);\
h->pred4x4[HOR_PRED ]= FUNCD(pred4x4_horizontal_vp8);\
} else {\
h->pred4x4[HOR_PRED ]= FUNCC(pred4x4_horizontal , depth);\
}\
h->pred4x4[DC_PRED ]= FUNCC(pred4x4_dc , depth);\
- if(codec_id == CODEC_ID_SVQ3)\
+ if(codec_id == AV_CODEC_ID_SVQ3)\
h->pred4x4[DIAG_DOWN_LEFT_PRED ]= FUNCD(pred4x4_down_left_svq3);\
else\
h->pred4x4[DIAG_DOWN_LEFT_PRED ]= FUNCC(pred4x4_down_left , depth);\
h->pred4x4[DIAG_DOWN_RIGHT_PRED]= FUNCC(pred4x4_down_right , depth);\
h->pred4x4[VERT_RIGHT_PRED ]= FUNCC(pred4x4_vertical_right , depth);\
h->pred4x4[HOR_DOWN_PRED ]= FUNCC(pred4x4_horizontal_down , depth);\
- if (codec_id == CODEC_ID_VP8) {\
+ if (codec_id == AV_CODEC_ID_VP8) {\
h->pred4x4[VERT_LEFT_PRED ]= FUNCD(pred4x4_vertical_left_vp8);\
} else\
h->pred4x4[VERT_LEFT_PRED ]= FUNCC(pred4x4_vertical_left , depth);\
h->pred4x4[HOR_UP_PRED ]= FUNCC(pred4x4_horizontal_up , depth);\
- if(codec_id != CODEC_ID_VP8) {\
+ if(codec_id != AV_CODEC_ID_VP8) {\
h->pred4x4[LEFT_DC_PRED ]= FUNCC(pred4x4_left_dc , depth);\
h->pred4x4[TOP_DC_PRED ]= FUNCC(pred4x4_top_dc , depth);\
h->pred4x4[DC_128_PRED ]= FUNCC(pred4x4_128_dc , depth);\
h->pred8x8l[TOP_DC_PRED ]= FUNCC(pred8x8l_top_dc , depth);\
h->pred8x8l[DC_128_PRED ]= FUNCC(pred8x8l_128_dc , depth);\
\
- h->pred8x8[VERT_PRED8x8 ]= FUNCC(pred8x8_vertical , depth);\
- h->pred8x8[HOR_PRED8x8 ]= FUNCC(pred8x8_horizontal , depth);\
- if (codec_id != CODEC_ID_VP8) {\
- h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x8_plane , depth);\
+ if (chroma_format_idc == 1) {\
+ h->pred8x8[VERT_PRED8x8 ]= FUNCC(pred8x8_vertical , depth);\
+ h->pred8x8[HOR_PRED8x8 ]= FUNCC(pred8x8_horizontal , depth);\
+ } else {\
+ h->pred8x8[VERT_PRED8x8 ]= FUNCC(pred8x16_vertical , depth);\
+ h->pred8x8[HOR_PRED8x8 ]= FUNCC(pred8x16_horizontal , depth);\
+ }\
+ if (codec_id != AV_CODEC_ID_VP8) {\
+ if (chroma_format_idc == 1) {\
+ h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x8_plane , depth);\
+ } else {\
+ h->pred8x8[PLANE_PRED8x8]= FUNCC(pred8x16_plane , depth);\
+ }\
} else\
h->pred8x8[PLANE_PRED8x8]= FUNCD(pred8x8_tm_vp8);\
- if(codec_id != CODEC_ID_RV40 && codec_id != CODEC_ID_VP8){\
- h->pred8x8[DC_PRED8x8 ]= FUNCC(pred8x8_dc , depth);\
- h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x8_left_dc , depth);\
- h->pred8x8[TOP_DC_PRED8x8 ]= FUNCC(pred8x8_top_dc , depth);\
- h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l0t, depth);\
- h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0lt, depth);\
- h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l00, depth);\
- h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0l0, depth);\
+ if(codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP8){\
+ if (chroma_format_idc == 1) {\
+ h->pred8x8[DC_PRED8x8 ]= FUNCC(pred8x8_dc , depth);\
+ h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x8_left_dc , depth);\
+ h->pred8x8[TOP_DC_PRED8x8 ]= FUNCC(pred8x8_top_dc , depth);\
+ h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l0t, depth);\
+ h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0lt, depth);\
+ h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_l00, depth);\
+ h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= FUNC(pred8x8_mad_cow_dc_0l0, depth);\
+ } else {\
+ h->pred8x8[DC_PRED8x8 ]= FUNCC(pred8x16_dc , depth);\
+ h->pred8x8[LEFT_DC_PRED8x8]= FUNCC(pred8x16_left_dc , depth);\
+ h->pred8x8[TOP_DC_PRED8x8 ]= FUNCC(pred8x16_top_dc , depth);\
+ h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8 ]= FUNC(pred8x16_mad_cow_dc_l0t, depth);\
+ h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8 ]= FUNC(pred8x16_mad_cow_dc_0lt, depth);\
+ h->pred8x8[ALZHEIMER_DC_L00_PRED8x8 ]= FUNC(pred8x16_mad_cow_dc_l00, depth);\
+ h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8 ]= FUNC(pred8x16_mad_cow_dc_0l0, depth);\
+ }\
}else{\
h->pred8x8[DC_PRED8x8 ]= FUNCD(pred8x8_dc_rv40);\
h->pred8x8[LEFT_DC_PRED8x8]= FUNCD(pred8x8_left_dc_rv40);\
h->pred8x8[TOP_DC_PRED8x8 ]= FUNCD(pred8x8_top_dc_rv40);\
- if (codec_id == CODEC_ID_VP8) {\
+ if (codec_id == AV_CODEC_ID_VP8) {\
h->pred8x8[DC_127_PRED8x8]= FUNCC(pred8x8_127_dc , depth);\
h->pred8x8[DC_129_PRED8x8]= FUNCC(pred8x8_129_dc , depth);\
}\
}\
- h->pred8x8[DC_128_PRED8x8 ]= FUNCC(pred8x8_128_dc , depth);\
+ if (chroma_format_idc == 1) {\
+ h->pred8x8[DC_128_PRED8x8 ]= FUNCC(pred8x8_128_dc , depth);\
+ } else {\
+ h->pred8x8[DC_128_PRED8x8 ]= FUNCC(pred8x16_128_dc , depth);\
+ }\
\
h->pred16x16[DC_PRED8x8 ]= FUNCC(pred16x16_dc , depth);\
h->pred16x16[VERT_PRED8x8 ]= FUNCC(pred16x16_vertical , depth);\
h->pred16x16[HOR_PRED8x8 ]= FUNCC(pred16x16_horizontal , depth);\
switch(codec_id){\
- case CODEC_ID_SVQ3:\
+ case AV_CODEC_ID_SVQ3:\
h->pred16x16[PLANE_PRED8x8 ]= FUNCD(pred16x16_plane_svq3);\
break;\
- case CODEC_ID_RV40:\
+ case AV_CODEC_ID_RV40:\
h->pred16x16[PLANE_PRED8x8 ]= FUNCD(pred16x16_plane_rv40);\
break;\
- case CODEC_ID_VP8:\
+ case AV_CODEC_ID_VP8:\
h->pred16x16[PLANE_PRED8x8 ]= FUNCD(pred16x16_tm_vp8);\
h->pred16x16[DC_127_PRED8x8]= FUNCC(pred16x16_127_dc , depth);\
h->pred16x16[DC_129_PRED8x8]= FUNCC(pred16x16_129_dc , depth);\
h->pred4x4_add [ HOR_PRED ]= FUNCC(pred4x4_horizontal_add , depth);\
h->pred8x8l_add [VERT_PRED ]= FUNCC(pred8x8l_vertical_add , depth);\
h->pred8x8l_add [ HOR_PRED ]= FUNCC(pred8x8l_horizontal_add , depth);\
+ if (chroma_format_idc == 1) {\
h->pred8x8_add [VERT_PRED8x8]= FUNCC(pred8x8_vertical_add , depth);\
h->pred8x8_add [ HOR_PRED8x8]= FUNCC(pred8x8_horizontal_add , depth);\
+ } else {\
+ h->pred8x8_add [VERT_PRED8x8]= FUNCC(pred8x16_vertical_add , depth);\
+ h->pred8x8_add [ HOR_PRED8x8]= FUNCC(pred8x16_horizontal_add , depth);\
+ }\
h->pred16x16_add[VERT_PRED8x8]= FUNCC(pred16x16_vertical_add , depth);\
h->pred16x16_add[ HOR_PRED8x8]= FUNCC(pred16x16_horizontal_add , depth);\
break;
}
- if (ARCH_ARM) ff_h264_pred_init_arm(h, codec_id, bit_depth);
- if (HAVE_MMX) ff_h264_pred_init_x86(h, codec_id, bit_depth);
+ if (ARCH_ARM) ff_h264_pred_init_arm(h, codec_id, bit_depth, chroma_format_idc);
+ if (ARCH_X86) ff_h264_pred_init_x86(h, codec_id, bit_depth, chroma_format_idc);
}