From 40c9e400fdaf1e9bb49eb0d0f0e437025b0ab6e8 Mon Sep 17 00:00:00 2001 From: Miguel Freitas Date: Sun, 5 Feb 2006 14:11:16 +0000 Subject: ffmpeg sync CVS patchset: 7872 CVS date: 2006/02/05 14:11:16 --- src/libffmpeg/libavcodec/libpostproc/postprocess.c | 1570 ++++++++++---------- 1 file changed, 785 insertions(+), 785 deletions(-) (limited to 'src/libffmpeg/libavcodec/libpostproc/postprocess.c') diff --git a/src/libffmpeg/libavcodec/libpostproc/postprocess.c b/src/libffmpeg/libavcodec/libpostproc/postprocess.c index 9f3e522ed..63d65fece 100644 --- a/src/libffmpeg/libavcodec/libpostproc/postprocess.c +++ b/src/libffmpeg/libavcodec/libpostproc/postprocess.c @@ -15,35 +15,35 @@ You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software - Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ /** * @file postprocess.c * postprocessing. */ - + /* - C MMX MMX2 3DNow AltiVec -isVertDC Ec Ec Ec -isVertMinMaxOk Ec Ec Ec -doVertLowPass E e e Ec -doVertDefFilter Ec Ec e e Ec -isHorizDC Ec Ec Ec -isHorizMinMaxOk a E Ec -doHorizLowPass E e e Ec -doHorizDefFilter Ec Ec e e Ec -do_a_deblock Ec E Ec E -deRing E e e* Ecp -Vertical RKAlgo1 E a a -Horizontal RKAlgo1 a a -Vertical X1# a E E -Horizontal X1# a E E -LinIpolDeinterlace e E E* -CubicIpolDeinterlace a e e* -LinBlendDeinterlace e E E* -MedianDeinterlace# E Ec Ec -TempDeNoiser# E e e Ec + C MMX MMX2 3DNow AltiVec +isVertDC Ec Ec Ec +isVertMinMaxOk Ec Ec Ec +doVertLowPass E e e Ec +doVertDefFilter Ec Ec e e Ec +isHorizDC Ec Ec Ec +isHorizMinMaxOk a E Ec +doHorizLowPass E e e Ec +doHorizDefFilter Ec Ec e e Ec +do_a_deblock Ec E Ec E +deRing E e e* Ecp +Vertical RKAlgo1 E a a +Horizontal RKAlgo1 a a +Vertical X1# a E E +Horizontal X1# a E E +LinIpolDeinterlace e E E* +CubicIpolDeinterlace a e e* +LinBlendDeinterlace e E E* +MedianDeinterlace# E Ec Ec +TempDeNoiser# E e e Ec * i dont have a 3dnow CPU -> its untested, but noone said it doesnt work so it seems to work # more or less selfinvented filters so the exactness isnt too meaningfull @@ -61,7 +61,7 @@ unroll stuff if instructions depend too much on the prior one move YScale thing to the end instead of fixing QP write a faster and higher quality deblocking filter :) make the mainloop more flexible (variable number of blocks at once - (the if/else stuff per block is slowing things down) + (the if/else stuff per block is slowing things down) compare the quality & speed of all filters split this huge file optimize c versions @@ -120,14 +120,14 @@ try to unroll inner for(x=0 ... loop to avoid these damn if(x ... checks #endif #if defined(ARCH_X86) || defined(ARCH_X86_64) -static uint64_t __attribute__((aligned(8))) attribute_used w05= 0x0005000500050005LL; -static uint64_t __attribute__((aligned(8))) attribute_used w04= 0x0004000400040004LL; -static uint64_t __attribute__((aligned(8))) attribute_used w20= 0x0020002000200020LL; -static uint64_t __attribute__((aligned(8))) attribute_used b00= 0x0000000000000000LL; -static uint64_t __attribute__((aligned(8))) attribute_used b01= 0x0101010101010101LL; -static uint64_t __attribute__((aligned(8))) attribute_used b02= 0x0202020202020202LL; -static uint64_t __attribute__((aligned(8))) attribute_used b08= 0x0808080808080808LL; -static uint64_t __attribute__((aligned(8))) attribute_used b80= 0x8080808080808080LL; +static uint64_t __attribute__((aligned(8))) attribute_used w05= 0x0005000500050005LL; +static uint64_t __attribute__((aligned(8))) attribute_used w04= 0x0004000400040004LL; +static uint64_t __attribute__((aligned(8))) attribute_used w20= 0x0020002000200020LL; +static uint64_t __attribute__((aligned(8))) attribute_used b00= 0x0000000000000000LL; +static uint64_t __attribute__((aligned(8))) attribute_used b01= 0x0101010101010101LL; +static uint64_t __attribute__((aligned(8))) attribute_used b02= 0x0202020202020202LL; +static uint64_t __attribute__((aligned(8))) attribute_used b08= 0x0808080808080808LL; +static uint64_t __attribute__((aligned(8))) attribute_used b80= 0x8080808080808080LL; #endif static uint8_t clip_table[3*256]; @@ -140,65 +140,65 @@ static const int attribute_used deringThreshold= 20; static struct PPFilter filters[]= { - {"hb", "hdeblock", 1, 1, 3, H_DEBLOCK}, - {"vb", "vdeblock", 1, 2, 4, V_DEBLOCK}, -/* {"hr", "rkhdeblock", 1, 1, 3, H_RK1_FILTER}, - {"vr", "rkvdeblock", 1, 2, 4, V_RK1_FILTER},*/ - {"h1", "x1hdeblock", 1, 1, 3, H_X1_FILTER}, - {"v1", "x1vdeblock", 1, 2, 4, V_X1_FILTER}, - {"ha", "ahdeblock", 1, 1, 3, H_A_DEBLOCK}, - {"va", "avdeblock", 1, 2, 4, V_A_DEBLOCK}, - {"dr", "dering", 1, 5, 6, DERING}, - {"al", "autolevels", 0, 1, 2, LEVEL_FIX}, - {"lb", "linblenddeint", 1, 1, 4, LINEAR_BLEND_DEINT_FILTER}, - {"li", "linipoldeint", 1, 1, 4, LINEAR_IPOL_DEINT_FILTER}, - {"ci", "cubicipoldeint", 1, 1, 4, CUBIC_IPOL_DEINT_FILTER}, - {"md", "mediandeint", 1, 1, 4, MEDIAN_DEINT_FILTER}, - {"fd", "ffmpegdeint", 1, 1, 4, FFMPEG_DEINT_FILTER}, - {"l5", "lowpass5", 1, 1, 4, LOWPASS5_DEINT_FILTER}, - {"tn", "tmpnoise", 1, 7, 8, TEMP_NOISE_FILTER}, - {"fq", "forcequant", 1, 0, 0, FORCE_QUANT}, - {NULL, NULL,0,0,0,0} //End Marker + {"hb", "hdeblock", 1, 1, 3, H_DEBLOCK}, + {"vb", "vdeblock", 1, 2, 4, V_DEBLOCK}, +/* {"hr", "rkhdeblock", 1, 1, 3, H_RK1_FILTER}, + {"vr", "rkvdeblock", 1, 2, 4, V_RK1_FILTER},*/ + {"h1", "x1hdeblock", 1, 1, 3, H_X1_FILTER}, + {"v1", "x1vdeblock", 1, 2, 4, V_X1_FILTER}, + {"ha", "ahdeblock", 1, 1, 3, H_A_DEBLOCK}, + {"va", "avdeblock", 1, 2, 4, V_A_DEBLOCK}, + {"dr", "dering", 1, 5, 6, DERING}, + {"al", "autolevels", 0, 1, 2, LEVEL_FIX}, + {"lb", "linblenddeint", 1, 1, 4, LINEAR_BLEND_DEINT_FILTER}, + {"li", "linipoldeint", 1, 1, 4, LINEAR_IPOL_DEINT_FILTER}, + {"ci", "cubicipoldeint", 1, 1, 4, CUBIC_IPOL_DEINT_FILTER}, + {"md", "mediandeint", 1, 1, 4, MEDIAN_DEINT_FILTER}, + {"fd", "ffmpegdeint", 1, 1, 4, FFMPEG_DEINT_FILTER}, + {"l5", "lowpass5", 1, 1, 4, LOWPASS5_DEINT_FILTER}, + {"tn", "tmpnoise", 1, 7, 8, TEMP_NOISE_FILTER}, + {"fq", "forcequant", 1, 0, 0, FORCE_QUANT}, + {NULL, NULL,0,0,0,0} //End Marker }; static char *replaceTable[]= { - "default", "hdeblock:a,vdeblock:a,dering:a", - "de", "hdeblock:a,vdeblock:a,dering:a", - "fast", "x1hdeblock:a,x1vdeblock:a,dering:a", - "fa", "x1hdeblock:a,x1vdeblock:a,dering:a", - "ac", "ha:a:128:7,va:a,dering:a", - NULL //End Marker + "default", "hdeblock:a,vdeblock:a,dering:a", + "de", "hdeblock:a,vdeblock:a,dering:a", + "fast", "x1hdeblock:a,x1vdeblock:a,dering:a", + "fa", "x1hdeblock:a,x1vdeblock:a,dering:a", + "ac", "ha:a:128:7,va:a,dering:a", + NULL //End Marker }; #if defined(ARCH_X86) || defined(ARCH_X86_64) static inline void prefetchnta(void *p) { - asm volatile( "prefetchnta (%0)\n\t" - : : "r" (p) - ); + asm volatile( "prefetchnta (%0)\n\t" + : : "r" (p) + ); } static inline void prefetcht0(void *p) { - asm volatile( "prefetcht0 (%0)\n\t" - : : "r" (p) - ); + asm volatile( "prefetcht0 (%0)\n\t" + : : "r" (p) + ); } static inline void prefetcht1(void *p) { - asm volatile( "prefetcht1 (%0)\n\t" - : : "r" (p) - ); + asm volatile( "prefetcht1 (%0)\n\t" + : : "r" (p) + ); } static inline void prefetcht2(void *p) { - asm volatile( "prefetcht2 (%0)\n\t" - : : "r" (p) - ); + asm volatile( "prefetcht2 (%0)\n\t" + : : "r" (p) + ); } #endif @@ -209,171 +209,171 @@ static inline void prefetcht2(void *p) */ static inline int isHorizDC_C(uint8_t src[], int stride, PPContext *c) { - int numEq= 0; - int y; - const int dcOffset= ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1; - const int dcThreshold= dcOffset*2 + 1; - - for(y=0; y c->ppMode.flatnessThreshold; + int numEq= 0; + int y; + const int dcOffset= ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1; + const int dcThreshold= dcOffset*2 + 1; + + for(y=0; y c->ppMode.flatnessThreshold; } /** * Check if the middle 8x8 Block in the given 8x16 block is flat */ static inline int isVertDC_C(uint8_t src[], int stride, PPContext *c){ - int numEq= 0; - int y; - const int dcOffset= ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1; - const int dcThreshold= dcOffset*2 + 1; - - src+= stride*4; // src points to begin of the 8x8 Block - for(y=0; y c->ppMode.flatnessThreshold; + int numEq= 0; + int y; + const int dcOffset= ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1; + const int dcThreshold= dcOffset*2 + 1; + + src+= stride*4; // src points to begin of the 8x8 Block + for(y=0; y c->ppMode.flatnessThreshold; } static inline int isHorizMinMaxOk_C(uint8_t src[], int stride, int QP) { - int i; + int i; #if 1 - for(i=0; i<2; i++){ - if((unsigned)(src[0] - src[5] + 2*QP) > 4*QP) return 0; - src += stride; - if((unsigned)(src[2] - src[7] + 2*QP) > 4*QP) return 0; - src += stride; - if((unsigned)(src[4] - src[1] + 2*QP) > 4*QP) return 0; - src += stride; - if((unsigned)(src[6] - src[3] + 2*QP) > 4*QP) return 0; - src += stride; - } -#else - for(i=0; i<8; i++){ - if((unsigned)(src[0] - src[7] + 2*QP) > 4*QP) return 0; - src += stride; - } + for(i=0; i<2; i++){ + if((unsigned)(src[0] - src[5] + 2*QP) > 4*QP) return 0; + src += stride; + if((unsigned)(src[2] - src[7] + 2*QP) > 4*QP) return 0; + src += stride; + if((unsigned)(src[4] - src[1] + 2*QP) > 4*QP) return 0; + src += stride; + if((unsigned)(src[6] - src[3] + 2*QP) > 4*QP) return 0; + src += stride; + } +#else + for(i=0; i<8; i++){ + if((unsigned)(src[0] - src[7] + 2*QP) > 4*QP) return 0; + src += stride; + } #endif - return 1; + return 1; } static inline int isVertMinMaxOk_C(uint8_t src[], int stride, int QP) { #if 1 #if 1 - int x; - src+= stride*4; - for(x=0; x 4*QP) return 0; - if((unsigned)(src[1+x + 2*stride] - src[1+x + 7*stride] + 2*QP) > 4*QP) return 0; - if((unsigned)(src[2+x + 4*stride] - src[2+x + 1*stride] + 2*QP) > 4*QP) return 0; - if((unsigned)(src[3+x + 6*stride] - src[3+x + 3*stride] + 2*QP) > 4*QP) return 0; - } + int x; + src+= stride*4; + for(x=0; x 4*QP) return 0; + if((unsigned)(src[1+x + 2*stride] - src[1+x + 7*stride] + 2*QP) > 4*QP) return 0; + if((unsigned)(src[2+x + 4*stride] - src[2+x + 1*stride] + 2*QP) > 4*QP) return 0; + if((unsigned)(src[3+x + 6*stride] - src[3+x + 3*stride] + 2*QP) > 4*QP) return 0; + } #else - int x; - src+= stride*3; - for(x=0; x 4*QP) return 0; - } + int x; + src+= stride*3; + for(x=0; x 4*QP) return 0; + } #endif - return 1; + return 1; #else - int x; - src+= stride*4; - for(x=0; xmax) max=v; - if(v 2*QP) return 0; - } - return 1; + int x; + src+= stride*4; + for(x=0; xmax) max=v; + if(v 2*QP) return 0; + } + return 1; #endif } static inline int horizClassify_C(uint8_t src[], int stride, PPContext *c){ - if( isHorizDC_C(src, stride, c) ){ - if( isHorizMinMaxOk_C(src, stride, c->QP) ) - return 1; - else - return 0; - }else{ - return 2; - } + if( isHorizDC_C(src, stride, c) ){ + if( isHorizMinMaxOk_C(src, stride, c->QP) ) + return 1; + else + return 0; + }else{ + return 2; + } } static inline int vertClassify_C(uint8_t src[], int stride, PPContext *c){ - if( isVertDC_C(src, stride, c) ){ - if( isVertMinMaxOk_C(src, stride, c->QP) ) - return 1; - else - return 0; - }else{ - return 2; - } + if( isVertDC_C(src, stride, c) ){ + if( isVertMinMaxOk_C(src, stride, c->QP) ) + return 1; + else + return 0; + }else{ + return 2; + } } static inline void doHorizDefFilter_C(uint8_t dst[], int stride, PPContext *c) { - int y; - for(y=0; yQP) - { - const int q=(dst[3] - dst[4])/2; - const int leftEnergy= 5*(dst[2] - dst[1]) + 2*(dst[0] - dst[3]); - const int rightEnergy= 5*(dst[6] - dst[5]) + 2*(dst[4] - dst[7]); - - int d= ABS(middleEnergy) - MIN( ABS(leftEnergy), ABS(rightEnergy) ); - d= MAX(d, 0); - - d= (5*d + 32) >> 6; - d*= SIGN(-middleEnergy); - - if(q>0) - { - d= d<0 ? 0 : d; - d= d>q ? q : d; - } - else - { - d= d>0 ? 0 : d; - d= dQP) + { + const int q=(dst[3] - dst[4])/2; + const int leftEnergy= 5*(dst[2] - dst[1]) + 2*(dst[0] - dst[3]); + const int rightEnergy= 5*(dst[6] - dst[5]) + 2*(dst[4] - dst[7]); + + int d= ABS(middleEnergy) - MIN( ABS(leftEnergy), ABS(rightEnergy) ); + d= MAX(d, 0); + + d= (5*d + 32) >> 6; + d*= SIGN(-middleEnergy); + + if(q>0) + { + d= d<0 ? 0 : d; + d= d>q ? q : d; + } + else + { + d= d>0 ? 0 : d; + d= dQP ? dst[-1] : dst[0]; - const int last= ABS(dst[8] - dst[7]) < c->QP ? dst[8] : dst[7]; - - int sums[10]; - sums[0] = 4*first + dst[0] + dst[1] + dst[2] + 4; - sums[1] = sums[0] - first + dst[3]; - sums[2] = sums[1] - first + dst[4]; - sums[3] = sums[2] - first + dst[5]; - sums[4] = sums[3] - first + dst[6]; - sums[5] = sums[4] - dst[0] + dst[7]; - sums[6] = sums[5] - dst[1] + last; - sums[7] = sums[6] - dst[2] + last; - sums[8] = sums[7] - dst[3] + last; - sums[9] = sums[8] - dst[4] + last; - - dst[0]= (sums[0] + sums[2] + 2*dst[0])>>4; - dst[1]= (sums[1] + sums[3] + 2*dst[1])>>4; - dst[2]= (sums[2] + sums[4] + 2*dst[2])>>4; - dst[3]= (sums[3] + sums[5] + 2*dst[3])>>4; - dst[4]= (sums[4] + sums[6] + 2*dst[4])>>4; - dst[5]= (sums[5] + sums[7] + 2*dst[5])>>4; - dst[6]= (sums[6] + sums[8] + 2*dst[6])>>4; - dst[7]= (sums[7] + sums[9] + 2*dst[7])>>4; - - dst+= stride; - } + int y; + for(y=0; yQP ? dst[-1] : dst[0]; + const int last= ABS(dst[8] - dst[7]) < c->QP ? dst[8] : dst[7]; + + int sums[10]; + sums[0] = 4*first + dst[0] + dst[1] + dst[2] + 4; + sums[1] = sums[0] - first + dst[3]; + sums[2] = sums[1] - first + dst[4]; + sums[3] = sums[2] - first + dst[5]; + sums[4] = sums[3] - first + dst[6]; + sums[5] = sums[4] - dst[0] + dst[7]; + sums[6] = sums[5] - dst[1] + last; + sums[7] = sums[6] - dst[2] + last; + sums[8] = sums[7] - dst[3] + last; + sums[9] = sums[8] - dst[4] + last; + + dst[0]= (sums[0] + sums[2] + 2*dst[0])>>4; + dst[1]= (sums[1] + sums[3] + 2*dst[1])>>4; + dst[2]= (sums[2] + sums[4] + 2*dst[2])>>4; + dst[3]= (sums[3] + sums[5] + 2*dst[3])>>4; + dst[4]= (sums[4] + sums[6] + 2*dst[4])>>4; + dst[5]= (sums[5] + sums[7] + 2*dst[5])>>4; + dst[6]= (sums[6] + sums[8] + 2*dst[6])>>4; + dst[7]= (sums[7] + sums[9] + 2*dst[7])>>4; + + dst+= stride; + } } /** @@ -423,161 +423,161 @@ static inline void doHorizLowPass_C(uint8_t dst[], int stride, PPContext *c) */ static inline void horizX1Filter(uint8_t *src, int stride, int QP) { - int y; - static uint64_t *lut= NULL; - if(lut==NULL) - { - int i; - lut= (uint64_t*)memalign(8, 256*8); - for(i=0; i<256; i++) - { - int v= i < 128 ? 2*i : 2*(i-256); + int y; + static uint64_t *lut= NULL; + if(lut==NULL) + { + int i; + lut= (uint64_t*)memalign(8, 256*8); + for(i=0; i<256; i++) + { + int v= i < 128 ? 2*i : 2*(i-256); /* //Simulate 112242211 9-Tap filter - uint64_t a= (v/16) & 0xFF; - uint64_t b= (v/8) & 0xFF; - uint64_t c= (v/4) & 0xFF; - uint64_t d= (3*v/8) & 0xFF; + uint64_t a= (v/16) & 0xFF; + uint64_t b= (v/8) & 0xFF; + uint64_t c= (v/4) & 0xFF; + uint64_t d= (3*v/8) & 0xFF; */ //Simulate piecewise linear interpolation - uint64_t a= (v/16) & 0xFF; - uint64_t b= (v*3/16) & 0xFF; - uint64_t c= (v*5/16) & 0xFF; - uint64_t d= (7*v/16) & 0xFF; - uint64_t A= (0x100 - a)&0xFF; - uint64_t B= (0x100 - b)&0xFF; - uint64_t C= (0x100 - c)&0xFF; - uint64_t D= (0x100 - c)&0xFF; - - lut[i] = (a<<56) | (b<<48) | (c<<40) | (d<<32) | - (D<<24) | (C<<16) | (B<<8) | (A); - //lut[i] = (v<<32) | (v<<24); - } - } - - for(y=0; yQP; - const int dcOffset= ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1; - const int dcThreshold= dcOffset*2 + 1; + int y; + const int QP= c->QP; + const int dcOffset= ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1; + const int dcThreshold= dcOffset*2 + 1; //START_TIMER - src+= step*4; // src points to begin of the 8x8 Block - for(y=0; y<8; y++){ - int numEq= 0; - - if(((unsigned)(src[-1*step] - src[0*step] + dcOffset)) < dcThreshold) numEq++; - if(((unsigned)(src[ 0*step] - src[1*step] + dcOffset)) < dcThreshold) numEq++; - if(((unsigned)(src[ 1*step] - src[2*step] + dcOffset)) < dcThreshold) numEq++; - if(((unsigned)(src[ 2*step] - src[3*step] + dcOffset)) < dcThreshold) numEq++; - if(((unsigned)(src[ 3*step] - src[4*step] + dcOffset)) < dcThreshold) numEq++; - if(((unsigned)(src[ 4*step] - src[5*step] + dcOffset)) < dcThreshold) numEq++; - if(((unsigned)(src[ 5*step] - src[6*step] + dcOffset)) < dcThreshold) numEq++; - if(((unsigned)(src[ 6*step] - src[7*step] + dcOffset)) < dcThreshold) numEq++; - if(((unsigned)(src[ 7*step] - src[8*step] + dcOffset)) < dcThreshold) numEq++; - if(numEq > c->ppMode.flatnessThreshold){ - int min, max, x; - - if(src[0] > src[step]){ - max= src[0]; - min= src[step]; - }else{ - max= src[step]; - min= src[0]; - } - for(x=2; x<8; x+=2){ - if(src[x*step] > src[(x+1)*step]){ - if(src[x *step] > max) max= src[ x *step]; - if(src[(x+1)*step] < min) min= src[(x+1)*step]; - }else{ - if(src[(x+1)*step] > max) max= src[(x+1)*step]; - if(src[ x *step] < min) min= src[ x *step]; - } - } - if(max-min < 2*QP){ - const int first= ABS(src[-1*step] - src[0]) < QP ? src[-1*step] : src[0]; - const int last= ABS(src[8*step] - src[7*step]) < QP ? src[8*step] : src[7*step]; - - int sums[10]; - sums[0] = 4*first + src[0*step] + src[1*step] + src[2*step] + 4; - sums[1] = sums[0] - first + src[3*step]; - sums[2] = sums[1] - first + src[4*step]; - sums[3] = sums[2] - first + src[5*step]; - sums[4] = sums[3] - first + src[6*step]; - sums[5] = sums[4] - src[0*step] + src[7*step]; - sums[6] = sums[5] - src[1*step] + last; - sums[7] = sums[6] - src[2*step] + last; - sums[8] = sums[7] - src[3*step] + last; - sums[9] = sums[8] - src[4*step] + last; - - src[0*step]= (sums[0] + sums[2] + 2*src[0*step])>>4; - src[1*step]= (sums[1] + sums[3] + 2*src[1*step])>>4; - src[2*step]= (sums[2] + sums[4] + 2*src[2*step])>>4; - src[3*step]= (sums[3] + sums[5] + 2*src[3*step])>>4; - src[4*step]= (sums[4] + sums[6] + 2*src[4*step])>>4; - src[5*step]= (sums[5] + sums[7] + 2*src[5*step])>>4; - src[6*step]= (sums[6] + sums[8] + 2*src[6*step])>>4; - src[7*step]= (sums[7] + sums[9] + 2*src[7*step])>>4; - } - }else{ - const int middleEnergy= 5*(src[4*step] - src[3*step]) + 2*(src[2*step] - src[5*step]); - - if(ABS(middleEnergy) < 8*QP) - { - const int q=(src[3*step] - src[4*step])/2; - const int leftEnergy= 5*(src[2*step] - src[1*step]) + 2*(src[0*step] - src[3*step]); - const int rightEnergy= 5*(src[6*step] - src[5*step]) + 2*(src[4*step] - src[7*step]); - - int d= ABS(middleEnergy) - MIN( ABS(leftEnergy), ABS(rightEnergy) ); - d= MAX(d, 0); - - d= (5*d + 32) >> 6; - d*= SIGN(-middleEnergy); - - if(q>0) - { - d= d<0 ? 0 : d; - d= d>q ? q : d; - } - else - { - d= d>0 ? 0 : d; - d= d c->ppMode.flatnessThreshold){ + int min, max, x; + + if(src[0] > src[step]){ + max= src[0]; + min= src[step]; + }else{ + max= src[step]; + min= src[0]; + } + for(x=2; x<8; x+=2){ + if(src[x*step] > src[(x+1)*step]){ + if(src[x *step] > max) max= src[ x *step]; + if(src[(x+1)*step] < min) min= src[(x+1)*step]; + }else{ + if(src[(x+1)*step] > max) max= src[(x+1)*step]; + if(src[ x *step] < min) min= src[ x *step]; + } + } + if(max-min < 2*QP){ + const int first= ABS(src[-1*step] - src[0]) < QP ? src[-1*step] : src[0]; + const int last= ABS(src[8*step] - src[7*step]) < QP ? src[8*step] : src[7*step]; + + int sums[10]; + sums[0] = 4*first + src[0*step] + src[1*step] + src[2*step] + 4; + sums[1] = sums[0] - first + src[3*step]; + sums[2] = sums[1] - first + src[4*step]; + sums[3] = sums[2] - first + src[5*step]; + sums[4] = sums[3] - first + src[6*step]; + sums[5] = sums[4] - src[0*step] + src[7*step]; + sums[6] = sums[5] - src[1*step] + last; + sums[7] = sums[6] - src[2*step] + last; + sums[8] = sums[7] - src[3*step] + last; + sums[9] = sums[8] - src[4*step] + last; + + src[0*step]= (sums[0] + sums[2] + 2*src[0*step])>>4; + src[1*step]= (sums[1] + sums[3] + 2*src[1*step])>>4; + src[2*step]= (sums[2] + sums[4] + 2*src[2*step])>>4; + src[3*step]= (sums[3] + sums[5] + 2*src[3*step])>>4; + src[4*step]= (sums[4] + sums[6] + 2*src[4*step])>>4; + src[5*step]= (sums[5] + sums[7] + 2*src[5*step])>>4; + src[6*step]= (sums[6] + sums[8] + 2*src[6*step])>>4; + src[7*step]= (sums[7] + sums[9] + 2*src[7*step])>>4; + } + }else{ + const int middleEnergy= 5*(src[4*step] - src[3*step]) + 2*(src[2*step] - src[5*step]); + + if(ABS(middleEnergy) < 8*QP) + { + const int q=(src[3*step] - src[4*step])/2; + const int leftEnergy= 5*(src[2*step] - src[1*step]) + 2*(src[0*step] - src[3*step]); + const int rightEnergy= 5*(src[6*step] - src[5*step]) + 2*(src[4*step] - src[7*step]); + + int d= ABS(middleEnergy) - MIN( ABS(leftEnergy), ABS(rightEnergy) ); + d= MAX(d, 0); + + d= (5*d + 32) >> 6; + d*= SIGN(-middleEnergy); + + if(q>0) + { + d= d<0 ? 0 : d; + d= d>q ? q : d; + } + else + { + d= d>0 ? 0 : d; + d= dppMode= *ppMode; //FIXME + PPContext *c= (PPContext *)vc; + PPMode *ppMode= (PPMode *)vm; + c->ppMode= *ppMode; //FIXME - // useing ifs here as they are faster than function pointers allthough the - // difference wouldnt be messureable here but its much better because - // someone might exchange the cpu whithout restarting mplayer ;) + // useing ifs here as they are faster than function pointers allthough the + // difference wouldnt be messureable here but its much better because + // someone might exchange the cpu whithout restarting mplayer ;) #ifdef RUNTIME_CPUDETECT #if defined(ARCH_X86) || defined(ARCH_X86_64) - // ordered per speed fasterst first - if(c->cpuCaps & PP_CPU_CAPS_MMX2) - postProcess_MMX2(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); - else if(c->cpuCaps & PP_CPU_CAPS_3DNOW) - postProcess_3DNow(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); - else if(c->cpuCaps & PP_CPU_CAPS_MMX) - postProcess_MMX(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); - else - postProcess_C(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + // ordered per speed fasterst first + if(c->cpuCaps & PP_CPU_CAPS_MMX2) + postProcess_MMX2(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + else if(c->cpuCaps & PP_CPU_CAPS_3DNOW) + postProcess_3DNow(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + else if(c->cpuCaps & PP_CPU_CAPS_MMX) + postProcess_MMX(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + else + postProcess_C(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); #else #ifdef ARCH_POWERPC #ifdef HAVE_ALTIVEC if(c->cpuCaps & PP_CPU_CAPS_ALTIVEC) - postProcess_altivec(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + postProcess_altivec(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); else #endif #endif - postProcess_C(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + postProcess_C(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); #endif #else //RUNTIME_CPUDETECT #ifdef HAVE_MMX2 - postProcess_MMX2(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + postProcess_MMX2(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); #elif defined (HAVE_3DNOW) - postProcess_3DNow(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + postProcess_3DNow(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); #elif defined (HAVE_MMX) - postProcess_MMX(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + postProcess_MMX(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); #elif defined (HAVE_ALTIVEC) - postProcess_altivec(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + postProcess_altivec(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); #else - postProcess_C(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); + postProcess_C(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c); #endif #endif //!RUNTIME_CPUDETECT } //static void postProcess(uint8_t src[], int srcStride, uint8_t dst[], int dstStride, int width, int height, -// QP_STORE_T QPs[], int QPStride, int isColor, struct PPMode *ppMode); +// QP_STORE_T QPs[], int QPStride, int isColor, struct PPMode *ppMode); /* -pp Command line Help */ char *pp_help= "Available postprocessing filters:\n" -"Filters Options\n" -"short long name short long option Description\n" -"* * a autoq CPU power dependent enabler\n" -" c chrom chrominance filtering enabled\n" -" y nochrom chrominance filtering disabled\n" -" n noluma luma filtering disabled\n" -"hb hdeblock (2 threshold) horizontal deblocking filter\n" -" 1. difference factor: default=32, higher -> more deblocking\n" -" 2. flatness threshold: default=39, lower -> more deblocking\n" -" the h & v deblocking filters share these\n" -" so you can't set different thresholds for h / v\n" -"vb vdeblock (2 threshold) vertical deblocking filter\n" -"ha hadeblock (2 threshold) horizontal deblocking filter\n" -"va vadeblock (2 threshold) vertical deblocking filter\n" -"h1 x1hdeblock experimental h deblock filter 1\n" -"v1 x1vdeblock experimental v deblock filter 1\n" -"dr dering deringing filter\n" -"al autolevels automatic brightness / contrast\n" -" f fullyrange stretch luminance to (0..255)\n" -"lb linblenddeint linear blend deinterlacer\n" -"li linipoldeint linear interpolating deinterlace\n" -"ci cubicipoldeint cubic interpolating deinterlacer\n" -"md mediandeint median deinterlacer\n" -"fd ffmpegdeint ffmpeg deinterlacer\n" -"l5 lowpass5 FIR lowpass deinterlacer\n" -"de default hb:a,vb:a,dr:a\n" -"fa fast h1:a,v1:a,dr:a\n" -"ac ha:a:128:7,va:a,dr:a\n" -"tn tmpnoise (3 threshold) temporal noise reducer\n" -" 1. <= 2. <= 3. larger -> stronger filtering\n" -"fq forceQuant force quantizer\n" +"Filters Options\n" +"short long name short long option Description\n" +"* * a autoq CPU power dependent enabler\n" +" c chrom chrominance filtering enabled\n" +" y nochrom chrominance filtering disabled\n" +" n noluma luma filtering disabled\n" +"hb hdeblock (2 threshold) horizontal deblocking filter\n" +" 1. difference factor: default=32, higher -> more deblocking\n" +" 2. flatness threshold: default=39, lower -> more deblocking\n" +" the h & v deblocking filters share these\n" +" so you can't set different thresholds for h / v\n" +"vb vdeblock (2 threshold) vertical deblocking filter\n" +"ha hadeblock (2 threshold) horizontal deblocking filter\n" +"va vadeblock (2 threshold) vertical deblocking filter\n" +"h1 x1hdeblock experimental h deblock filter 1\n" +"v1 x1vdeblock experimental v deblock filter 1\n" +"dr dering deringing filter\n" +"al autolevels automatic brightness / contrast\n" +" f fullyrange stretch luminance to (0..255)\n" +"lb linblenddeint linear blend deinterlacer\n" +"li linipoldeint linear interpolating deinterlace\n" +"ci cubicipoldeint cubic interpolating deinterlacer\n" +"md mediandeint median deinterlacer\n" +"fd ffmpegdeint ffmpeg deinterlacer\n" +"l5 lowpass5 FIR lowpass deinterlacer\n" +"de default hb:a,vb:a,dr:a\n" +"fa fast h1:a,v1:a,dr:a\n" +"ac ha:a:128:7,va:a,dr:a\n" +"tn tmpnoise (3 threshold) temporal noise reducer\n" +" 1. <= 2. <= 3. larger -> stronger filtering\n" +"fq forceQuant force quantizer\n" "Usage:\n" "[: