RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

postproc: Made QP, nonBQP, and pQPb arrays 

Also pulled QP initialization out of inner loop, which removed some redundent code.

Added some dummy fields to PPContext to allow current code to work while
changing the rest of the postprocessing code to support the arrays.

I also increased alignment requirements for some fields in the PPContext struct to
support future avx2 code.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Tucker DiNapoli 2015-04-22 16:27:27 -04:00 committed by Michael Niedermayer
parent 6264b6227c
commit cbe27006ce
2 changed files with 46 additions and 45 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
libpostproc/postprocess_internal.h
View File

@ -143,8 +143,11 @@ typedef struct PPContext{
DECLARE_ALIGNED(8, uint64_t, pQPb); DECLARE_ALIGNED(8, uint64_t, pQPb);
DECLARE_ALIGNED(8, uint64_t, pQPb2); DECLARE_ALIGNED(8, uint64_t, pQPb2);
DECLARE_ALIGNED(8, uint64_t, mmxDcOffset)[64]; DECLARE_ALIGNED(32, uint64_t, pQPb_block)[4];
DECLARE_ALIGNED(8, uint64_t, mmxDcThreshold)[64]; DECLARE_ALIGNED(32, uint64_t, pQPb2_block)[4];
DECLARE_ALIGNED(32, uint64_t, mmxDcOffset)[64];
DECLARE_ALIGNED(32, uint64_t, mmxDcThreshold)[64];
QP_STORE_T *stdQPTable; ///< used to fix MPEG2 style qscale QP_STORE_T *stdQPTable; ///< used to fix MPEG2 style qscale
QP_STORE_T *nonBQPTable; QP_STORE_T *nonBQPTable;
@ -153,6 +156,9 @@ typedef struct PPContext{
int QP; int QP;
int nonBQP; int nonBQP;
DECLARE_ALIGNED(32, int, QP_block)[4];
DECLARE_ALIGNED(32, int, nonBQP_block)[4];
int frameNum; int frameNum;
int cpuCaps; int cpuCaps;

81
libpostproc/postprocess_template.c
View File

@ -3479,7 +3479,7 @@ static void RENAME(postProcess)(const uint8_t src[], int srcStride, uint8_t dst[
#endif #endif
const int8_t *QPptr= &QPs[(y>>qpVShift)*QPStride]; const int8_t *QPptr= &QPs[(y>>qpVShift)*QPStride];
int8_t *nonBQPptr= &c.nonBQPTable[(y>>qpVShift)*FFABS(QPStride)]; int8_t *nonBQPptr= &c.nonBQPTable[(y>>qpVShift)*FFABS(QPStride)];
int QP=0; int QP=0, nonBQP=0;
/* can we mess with a 8x16 block from srcBlock/dstBlock downwards and 1 line upwards /* can we mess with a 8x16 block from srcBlock/dstBlock downwards and 1 line upwards
if not than use a temporary buffer */ if not than use a temporary buffer */
if(y+15 >= height){ if(y+15 >= height){
@ -3512,6 +3512,29 @@ static void RENAME(postProcess)(const uint8_t src[], int srcStride, uint8_t dst[
int endx = FFMIN(width, x+32); int endx = FFMIN(width, x+32);
uint8_t *dstBlockStart = dstBlock; uint8_t *dstBlockStart = dstBlock;
const uint8_t *srcBlockStart = srcBlock; const uint8_t *srcBlockStart = srcBlock;
int qp_index = 0;
for(qp_index=0; qp_index < (endx-startx)/BLOCK_SIZE; qp_index++){
QP = QPptr[(x+qp_index*BLOCK_SIZE)>>qpHShift];
nonBQP = nonBQPptr[(x+qp_index*BLOCK_SIZE)>>qpHShift];
if(!isColor){
QP= (QP* QPCorrecture + 256*128)>>16;
nonBQP= (nonBQP* QPCorrecture + 256*128)>>16;
yHistogram[(srcBlock+qp_index*8)[srcStride*12 + 4]]++;
}
c.QP_block[qp_index] = QP;
c.nonBQP_block[qp_index] = nonBQP;
#if TEMPLATE_PP_MMX
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"packuswb %%mm7, %%mm7 \n\t" // 0, 0, 0, QP, 0, 0, 0, QP
"packuswb %%mm7, %%mm7 \n\t" // 0,QP, 0, QP, 0,QP, 0, QP
"packuswb %%mm7, %%mm7 \n\t" // QP,..., QP
"movq %%mm7, %0 \n\t"
: "=m" (c.pQPb_block[qp_index])
: "r" (QP)
);
#endif
}
for(; x < endx; x+=BLOCK_SIZE){ for(; x < endx; x+=BLOCK_SIZE){
RENAME(prefetchnta)(srcBlock + (((x>>2)&6) + copyAhead)*srcStride + 32); RENAME(prefetchnta)(srcBlock + (((x>>2)&6) + copyAhead)*srcStride + 32);
RENAME(prefetchnta)(srcBlock + (((x>>2)&6) + copyAhead+1)*srcStride + 32); RENAME(prefetchnta)(srcBlock + (((x>>2)&6) + copyAhead+1)*srcStride + 32);
@ -3543,27 +3566,15 @@ static void RENAME(postProcess)(const uint8_t src[], int srcStride, uint8_t dst[
dstBlock = dstBlockStart; dstBlock = dstBlockStart;
srcBlock = srcBlockStart; srcBlock = srcBlockStart;
for(x = startx; x < endx; x+=BLOCK_SIZE){ for(x = startx, qp_index = 0; x < endx; x+=BLOCK_SIZE, qp_index++){
const int stride= dstStride; const int stride= dstStride;
QP = QPptr[x>>qpHShift]; //temporary while changing QP stuff to make things continue to work
c.nonBQP = nonBQPptr[x>>qpHShift]; //eventually QP,nonBQP,etc will be arrays and this will be unnecessary
if(!isColor){ c.QP = c.QP_block[qp_index];
QP= (QP* QPCorrecture + 256*128)>>16; c.nonBQP = c.nonBQP_block[qp_index];
c.nonBQP= (c.nonBQP* QPCorrecture + 256*128)>>16; c.pQPb = c.pQPb_block[qp_index];
yHistogram[srcBlock[srcStride*12 + 4]]++; c.pQPb2 = c.pQPb2_block[qp_index];
}
c.QP= QP;
#if TEMPLATE_PP_MMX
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"packuswb %%mm7, %%mm7 \n\t" // 0, 0, 0, QP, 0, 0, 0, QP
"packuswb %%mm7, %%mm7 \n\t" // 0,QP, 0, QP, 0,QP, 0, QP
"packuswb %%mm7, %%mm7 \n\t" // QP,..., QP
"movq %%mm7, %0 \n\t"
: "=m" (c.pQPb)
: "r" (QP)
);
#endif
/* only deblock if we have 2 blocks */ /* only deblock if we have 2 blocks */
if(y + 8 < height){ if(y + 8 < height){
if(mode & V_X1_FILTER) if(mode & V_X1_FILTER)
@ -3587,30 +3598,14 @@ static void RENAME(postProcess)(const uint8_t src[], int srcStride, uint8_t dst[
dstBlock = dstBlockStart; dstBlock = dstBlockStart;
srcBlock = srcBlockStart; srcBlock = srcBlockStart;
for(x = startx; x < endx; x+=BLOCK_SIZE){ for(x = startx, qp_index=0; x < endx; x+=BLOCK_SIZE, qp_index++){
const int stride= dstStride; const int stride= dstStride;
av_unused uint8_t *tmpXchg; av_unused uint8_t *tmpXchg;
c.QP = c.QP_block[qp_index];
if(isColor){ c.nonBQP = c.nonBQP_block[qp_index];
QP= QPptr[x>>qpHShift]; c.pQPb = c.pQPb_block[qp_index];
c.nonBQP= nonBQPptr[x>>qpHShift]; c.pQPb2 = c.pQPb2_block[qp_index];
}else{
QP= QPptr[x>>4];
QP= (QP* QPCorrecture + 256*128)>>16;
c.nonBQP= nonBQPptr[x>>4];
c.nonBQP= (c.nonBQP* QPCorrecture + 256*128)>>16;
}
c.QP= QP;
#if TEMPLATE_PP_MMX #if TEMPLATE_PP_MMX
__asm__ volatile(
"movd %1, %%mm7 \n\t"
"packuswb %%mm7, %%mm7 \n\t" // 0, 0, 0, QP, 0, 0, 0, QP
"packuswb %%mm7, %%mm7 \n\t" // 0,QP, 0, QP, 0,QP, 0, QP
"packuswb %%mm7, %%mm7 \n\t" // QP,..., QP
"movq %%mm7, %0 \n\t"
: "=m" (c.pQPb)
: "r" (QP)
);
RENAME(transpose1)(tempBlock1, tempBlock2, dstBlock, dstStride); RENAME(transpose1)(tempBlock1, tempBlock2, dstBlock, dstStride);
#endif #endif
/* check if we have a previous block to deblock it with dstBlock */ /* check if we have a previous block to deblock it with dstBlock */
@ -3632,7 +3627,7 @@ static void RENAME(postProcess)(const uint8_t src[], int srcStride, uint8_t dst[
#else #else
if(mode & H_X1_FILTER) if(mode & H_X1_FILTER)
horizX1Filter(dstBlock-4, stride, QP); horizX1Filter(dstBlock-4, stride, c.QP);
else if(mode & H_DEBLOCK){ else if(mode & H_DEBLOCK){
#if TEMPLATE_PP_ALTIVEC #if TEMPLATE_PP_ALTIVEC
DECLARE_ALIGNED(16, unsigned char, tempBlock)[272]; DECLARE_ALIGNED(16, unsigned char, tempBlock)[272];