diff -ru mpeg2enc.orig/MPEG.avi2yuv mpeg2enc/MPEG.avi2yuv --- mpeg2enc.orig/MPEG.avi2yuv Fri Apr 9 17:59:14 1999 +++ mpeg2enc/MPEG.avi2yuv Mon Apr 10 20:01:30 2000 @@ -1,5 +1,5 @@ recorded from video , 25 frames/sec -/video/Red_4 /* name of source files */ +/mnt/test /* name of source files */ - /* name of reconstructed images ("-": don't store) */ - /* name of intra quant matrix file ("-": default matrix) */ - /* name of non intra quant matrix file ("-": default matrix) */ diff -ru mpeg2enc.orig/Makefile mpeg2enc/Makefile --- mpeg2enc.orig/Makefile Fri Apr 9 18:15:43 1999 +++ mpeg2enc/Makefile Sat Apr 15 12:22:46 2000 @@ -36,7 +36,10 @@ # GNU gcc # CC = gcc -CFLAGS = -O3 -ffast-math -fomit-frame-pointer -funroll-loops -march=k6 -mcpu=k6 -Wall -I/home/uhecht/mpeg_encode/jpeg +#CFLAGS = -DHAVE_MMX -O3 -ffast-math -fomit-frame-pointer -funroll-loops -march=k6 -mcpu=k6 -Wall -I/home/uhecht/mpeg_encode/jpeg +#CFLAGS = -DAMD -DFDCTINT -DHAVE_MMX -O3 -ffast-math -fomit-frame-pointer -funroll-loops -malign-loops=2 -malign-jumps=2 -march=k6 -mcpu=k6 -Wall -I/home/uhecht/mpeg_encode/jpeg +CFLAGS = -DAMD -DFDCTINT -DHAVE_MMX -O3 -fomit-frame-pointer -funroll-loops -malign-loops=2 -malign-jumps=2 -march=k6 -mcpu=k6 -Wall -I/home/uhecht/mpeg_encode/jpeg +#CFLAGS = -Dinline=static -pg -DAMD -DFDCTINT -DHAVE_MMX -O2 -ffast-math -funroll-loops -malign-loops=2 -malign-jumps=2 -march=k6 -mcpu=k6 -Wall -I/home/uhecht/mpeg_encode/jpeg OBJ = mpeg2enc.o conform.o putseq.o putpic.o puthdr.o putmpg.o putvlc.o\ putbits.o motion.o predict.o readpic.o writepic.o transfrm.o fdctref.o\ Binary files mpeg2enc.orig/conform.o and mpeg2enc/conform.o differ Only in mpeg2enc: eins diff -ru mpeg2enc.orig/fdctref.c mpeg2enc/fdctref.c --- mpeg2enc.orig/fdctref.c Fri Apr 9 18:14:20 1999 +++ mpeg2enc/fdctref.c Sat Apr 15 06:19:35 2000 @@ -46,8 +46,16 @@ /* private data */ /* static double c[8][8]; */ /* transform coefficients */ +#ifdef FDCTINT +static int c[8][8]; +#ifdef AMD +static int d[8][8]; +#endif +#else static float c[8][8]; +#endif +#ifndef FDCTINT void init_fdct() { int i, j; @@ -64,7 +72,79 @@ c[i][j] = s * cos((PI/8.0)*i*(j+0.5)); } } +#else +#ifdef AMD +static int DIVIDE=65536; +void mmx_create_table(int* in, int* out) +{ + asm(" + .align 8 + movd (%%ecx),%%mm0; + pi2fd %%mm0, %%mm0; + movd DIVIDE, %%mm1; + pi2fd %%mm1, %%mm1; + pfrcp %%mm1, %%mm1; + pfmul %%mm1, %%mm0; + movd %%mm0, (%%ebx); + femms; + " + : : "b" (out), "c" (in)); +} +static int pointfive=0; +static int one=1; +static int two=2; + +void mmx_setpointfive(void) +{ + asm(" + .align 8 + movd one, %%mm0; + pi2fd %%mm0, %%mm0; + movd two, %%mm1; + pi2fd %%mm1, %%mm1; + pfrcp %%mm1, %%mm1; + pfmul %%mm1, %%mm0; + movd %%mm0, pointfive; + femms; + " : : ); +} +#endif + +void init_fdct() +{ + int i, j; + /* + double s; + */ + float s; +#ifdef AMD + int tmp; +#endif + for (i=0; i<8; i++) + { + s = (i==0) ? 0.3535533905932737622 /* sqrt(0.125) */ : 0.5; + + for (j=0; j<8; j++) + { +#ifndef AMD + c[i][j] = (s * cos((PI/8.0)*i*(j+0.5)))*65536.0; + //printf("%d ",c[i][j]); +#else + tmp=(s * cos((PI/8.0)*i*(j+0.5)))*65536.0; + printf("tmp %d\r",tmp); + mmx_create_table(&tmp,&d[i][j]); + //printf("%d\n",d[i][j]); +#endif + } + } +#ifdef AMD + mmx_setpointfive(); +#endif +} +#endif + +#ifndef FDCTINT void fdct(block) short *block; { @@ -107,3 +187,251 @@ */ } } +#else +#ifdef AMD +unsigned long long mul1=0x0001000100010001; + +inline void mmx_fdct(int* c, short* block, int* tmp) +{ +#ifdef STABLE + asm(" + .align 8 + movq mul1, %%mm7; + movq (%%ecx), %%mm6; // load 4 elements from block + movq %%mm6, %%mm5; // copy 4 unpack + pmulhw %%mm7, %%mm6; + pmullw %%mm7, %%mm5; + movq %%mm5, %%mm2; + punpcklwd %%mm6, %%mm5; // lower half + pi2fd %%mm5, %%mm4; // as float in mm4 + punpckhwd %%mm6, %%mm2; // upper half + pi2fd %%mm2, %%mm5; // as float in mm5 + movq (%%ebx), %%mm3; // lower half of table + movq 8(%%ebx), %%mm6; // upper half of table + pfmul %%mm4, %%mm3; + pfmul %%mm5, %%mm6; + pfacc %%mm3, %%mm6; + + movq 8(%%ecx), %%mm2; // load 4 elements from block + movq %%mm2, %%mm5; // copy 4 unpack + pmulhw %%mm7, %%mm2; + pmullw %%mm7, %%mm5; + movq %%mm5, %%mm3; + punpcklwd %%mm2, %%mm5; // lower half + pi2fd %%mm5, %%mm4; // as float in mm4 + punpckhwd %%mm2, %%mm3; // upper half + pi2fd %%mm3, %%mm5; // as float in mm5 + movq 16(%%ebx), %%mm3; // lower half of table + movq 24(%%ebx), %%mm2; // upper half of table + pfmul %%mm4, %%mm3; + pfmul %%mm5, %%mm2; + pfacc %%mm3, %%mm2; + pfadd %%mm2, %%mm6; + pf2id %%mm6, %%mm6; + movq %%mm6, %%mm5; + psrlq $32, %%mm5; + paddd %%mm5, %%mm6; + movd %%mm6, (%%edx); + femms; + " + : + : "b" (c), "c" (block), "d" (tmp) + : "st" ); +#else +#warning EXP + asm(" + .align 8 + movq mul1, %%mm7; + movq (%%ecx), %%mm6; // load 4 elements from block + movq %%mm6, %%mm5; // copy 4 unpack + pmulhw %%mm7, %%mm6; + pmullw %%mm7, %%mm5; + movq %%mm5, %%mm2; + punpcklwd %%mm6, %%mm5; // lower half + movq (%%ebx), %%mm3; // lower half of table + pi2fd %%mm5, %%mm4; // as float in mm4 + punpckhwd %%mm6, %%mm2; // upper half + pi2fd %%mm2, %%mm5; // as float in mm5 + movq 8(%%ebx), %%mm6; // upper half of table + pfmul %%mm4, %%mm3; + pfmul %%mm5, %%mm6; + movq 8(%%ecx), %%mm2; // load 4 elements from block + pfacc %%mm3, %%mm6; + + movq %%mm2, %%mm5; // copy 4 unpack + pmulhw %%mm7, %%mm2; + movq 16(%%ebx), %%mm3; // lower half of table + pmullw %%mm7, %%mm5; + movq %%mm5, %%mm1; + punpcklwd %%mm2, %%mm5; // lower half + pi2fd %%mm5, %%mm4; // as float in mm4 + punpckhwd %%mm2, %%mm1; // upper half + pi2fd %%mm1, %%mm5; // as float in mm5 + movq 24(%%ebx), %%mm2; // upper half of table + pfmul %%mm4, %%mm3; + pfmul %%mm5, %%mm2; + pfacc %%mm3, %%mm2; + pfadd %%mm2, %%mm6; + movq %%mm6, %%mm5; + psrlq $32, %%mm5; + pfadd %%mm5, %%mm6; + movd %%mm6, (%%edx); + //femms; + " + : + : "b" (c), "c" (block), "d" (tmp) + : "st" ); +#endif +} + +inline void mmx_fdct2(int* c, int* block,int* res) +{ +#ifdef STABLE2 + asm(" + .align 8 + movq (%%ecx), %%mm4; // load 2 elements from block + movq (%%ebx), %%mm3; // lower half of table + pfmul %%mm4, %%mm3; + movq 8(%%ecx), %%mm5; + movq 8(%%ebx), %%mm6; // upper half of table + pfmul %%mm5, %%mm6; + + movq 16(%%ecx), %%mm4; // load 2 elements from block + pfacc %%mm3, %%mm6; + movq 24(%%ecx), %%mm5; + movq 16(%%ebx), %%mm3; // lower half of table + movq 24(%%ebx), %%mm2; // upper half of table + pfmul %%mm4, %%mm3; + pfmul %%mm5, %%mm2; + pfacc %%mm3, %%mm2; + + pfadd %%mm2, %%mm6; + movq %%mm6, %%mm5; + psrlq $32, %%mm5; + pfadd %%mm5, %%mm6; + pfadd %%mm0, %%mm6; + pf2id %%mm6, %%mm6; + movd %%mm6, (%%edx); + //femms; + " + : + : "b" (c), "c" (block), "d" (res) + ); +#else +#warning EXP + asm(" + .align 8 + movq (%%ecx), %%mm4; // load 2 elements from block + movq (%%ebx), %%mm3; // lower half of table + pfmul %%mm4, %%mm3; + movq 8(%%ecx), %%mm5; + movq 8(%%ebx), %%mm6; // upper half of table + pfmul %%mm5, %%mm6; + + movq 16(%%ecx), %%mm4; // load 2 elements from block + pfacc %%mm3, %%mm6; + movq 16(%%ebx), %%mm3; // lower half of table + movq 24(%%ecx), %%mm5; + pfmul %%mm4, %%mm3; + movq 24(%%ebx), %%mm2; // upper half of table + pfmul %%mm5, %%mm2; + pfacc %%mm3, %%mm2; + + pfadd %%mm2, %%mm6; + movq %%mm6, %%mm5; + psrlq $32, %%mm5; + pfadd %%mm5, %%mm6; + pfadd %%mm0, %%mm6; + pf2id %%mm6, %%mm6; + movd %%mm6, (%%edx); + " + : + : "b" (c), "c" (block), "d" (res) + ); +#endif +} +#endif + +volatile static int res; + +void fdct(block) +short *block; +{ + int i, j; +#ifndef AMD + int k; + int s; +#endif + /* + double s; + double tmp[64]; + */ + int tmp[64]; + for (i=0; i<8; i++) + for (j=0; j<8; j++) + { +#ifndef AMD + s = 0; + + for (k=0; k<8; k++) + s += ((c[j][k])) * ((block[8*i+k])); +#else + mmx_fdct(d[j],&block[8*i],&tmp[8*j+i]); +#endif + +#ifndef AMD + tmp[8*j+i] = s; +#endif + } + +#ifdef AMD + asm(" + .align 8 + movd pointfive, %%mm0; + " : : ); +#endif + + for (j=0; j<8; j++) + for (i=0; i<8; i++) + { +#ifndef AMD + s = 0; + + for (k=0; k<8; k++) + s += ((c[i][k])>>8) * ((tmp[8*j+k])>>8); +#else + mmx_fdct2(d[i],&tmp[8*j],&res); + block[8*i+j]=res; +#endif + +#ifndef AMD + //block[8*i+j] = (int)floor(s+0.499999); + block[8*i+j] = (s+(0x7fff))>>16; + /* + * reason for adding 0.499999 instead of 0.5: + * s is quite often x.5 (at least for i and/or j = 0 or 4) + * and setting the rounding threshold exactly to 0.5 leads to an + * extremely high arithmetic implementation dependency of the result; + * s being between x.5 and x.500001 (which is now incorrectly rounded + * downwards instead of upwards) is assumed to occur less often + * (if at all) + */ +#else + asm("femms;"); +#endif + } +} +#endif + +#ifdef MAIN +main() +{ + volatile int vektor1[32]={1189904896,1178856448,-968627200,-957578752,-957578752,-968627200,1178856448,1189904896}; + volatile int vektor2[8]={42812,115897,-602414,115897,-425971,413432,1,1}; + volatile int i,j; + for(j=0;j<200000;j++) + fdct(vektor1); + printf("%d\n",i); +} +#endif + \ No newline at end of file Only in mpeg2enc.orig: fdctref.c,v Only in mpeg2enc: fdctref.c.int Binary files mpeg2enc.orig/fdctref.o and mpeg2enc/fdctref.o differ Only in mpeg2enc: fdctref.s Only in mpeg2enc: gmon.out Binary files mpeg2enc.orig/idct.o and mpeg2enc/idct.o differ diff -ru mpeg2enc.orig/motion.c mpeg2enc/motion.c --- mpeg2enc.orig/motion.c Sat Jul 20 00:47:49 1996 +++ mpeg2enc/motion.c Sun Apr 16 13:11:20 2000 @@ -1343,97 +1343,379 @@ * h: height of block (usually 8 or 16) * distlim: bail out if sum exceeds this value */ -static int dist1(blk1,blk2,lx,hx,hy,h,distlim) -unsigned char *blk1,*blk2; -int lx,hx,hy,h; -int distlim; + + +void inline mmx_start_block() { - unsigned char *p1,*p1a,*p2; - int i,j; - int s,v; + asm(" + .align 8 + pxor %%mm7, %%mm7; + pxor %%mm6, %%mm6; + " : : ); +} - s = 0; - p1 = blk1; - p2 = blk2; +void inline mmx_absdiff(unsigned char *a, unsigned char *b) +{ +#if STABLEABSDIFF +#warning STABLEABSDIFF + asm(" + .align 8 + movq (%%ebx), %%mm0; // Get first half of row1 + movq (%%ecx), %%mm1; // Get first half of row2 + movq %%mm0, %%mm2; // Make a copy of row1 for absdiff operation + movq 8(%%ebx), %%mm3; // Get second half of row1 + psubusb %%mm1, %%mm0; // Subtract the first halves one way + psubusb %%mm2, %%mm1; // Subtract the other way + movq 8(%%ecx), %%mm4; // Get second half of row2 + por %%mm1, %%mm0; // Merge first half results + movq %%mm3, %%mm5; // Copy for absdiff operation + movq %%mm0, %%mm1; // Keep a copy + psubusb %%mm4, %%mm3; // Subtract second halves one way + punpcklbw %%mm6, %%mm0; // Unpack to higher precision for accumulation + psubusb %%mm5, %%mm4; // Subtract the other way + psrlq $32, %%mm1; // Shift registeres for accumulation + por %%mm4, %%mm3; // merge results of 2nd half + punpcklbw %%mm6, %%mm1; // unpack to higher precision for accumulation + movq %%mm3, %%mm4; // keep a copy + punpcklbw %%mm6, %%mm3; // unpack to higher precision for accumulation + paddw %%mm0, %%mm7; // accumulate difference + psrlq $32, %%mm4; // shift results for accumulation + paddw %%mm1, %%mm7; // accumulate difference + punpcklbw %%mm6, %%mm4; // unpack to higher precision for accumulation + paddw %%mm3, %%mm7; // accumulate difference + paddw %%mm4, %%mm7; // accumulate difference + " + : + : "b" (a), "c" (b) + : "st" ); +#else + asm(" + .align 8 + movq (%%ebx), %%mm0; // Get first half + movq %%mm0, %%mm2; // Make a copy of + movq (%%ecx), %%mm1; // Get first half + psubusb %%mm1, %%mm0; // Subtract the + psubusb %%mm2, %%mm1; // Subtract the + movq 8(%%ebx), %%mm3; // Get second half + por %%mm1, %%mm0; // Merge first + movq 8(%%ecx), %%mm4; // Get second half of row2 + movq %%mm3, %%mm2; + psubusb %%mm4, %%mm3; + movq %%mm0, %%mm1; + psubusb %%mm2, %%mm4; + por %%mm4, %%mm3; + punpcklbw %%mm6, %%mm0; + movq %%mm3, %%mm4; + paddw %%mm0, %%mm7; + punpckhbw %%mm6, %%mm1; + punpcklbw %%mm6, %%mm3; + paddw %%mm1, %%mm7; + punpckhbw %%mm6, %%mm4; + paddw %%mm3, %%mm7; + paddw %%mm4, %%mm7; + " + : + : "b" (a), "c" (b) + : "st" ); +#endif +} - if (!hx && !hy) - for (j=0; j= distlim) - break; + return r; +} - p1+= lx; - p2+= lx; - } - else if (hx && !hy) - for (j=0; j>1) - p2[i]; - if (v>=0) - s+= v; - else - s-= v; - } - p1+= lx; - p2+= lx; - } - else if (!hx && hy) - { - p1a = p1 + lx; - for (j=0; j>1) - p2[i]; - if (v>=0) - s+= v; - else - s-= v; - } - p1 = p1a; - p1a+= lx; - p2+= lx; - } - } - else /* if (hx && hy) */ - { - p1a = p1 + lx; - for (j=0; j>2) - p2[i]; - if (v>=0) - s+= v; - else - s-= v; - } - p1 = p1a; - p1a+= lx; - p2+= lx; - } - } +static unsigned long MMX_AVGDIFF_1[] = {0x00010001, 0x00010001}; - return s; + +void inline mmx_avgdiff(unsigned char *p1, unsigned char *p2, unsigned char *p3) +{ +#if 0 + asm(" + .align 8 + movq (%%ebx), %%mm0; // Load 8 pixels from a + pxor %%mm4, %%mm4; // Zero out temp for unpacking a + movq %%mm0, %%mm2; // Make a copy of a for unpacking + movq (%%ecx), %%mm1; // Load 8 pixels from b + pxor %%mm3, %%mm3; // Zero out b's upper unpacked destination + punpcklbw %%mm4, %%mm2; // Unpack lower 4 pixels from a for addition + movq %%mm1, %%mm5; // Copy b for unpacking + punpckhbw %%mm4, %%mm0; // Unpack upper 4 pixels from a for addition + punpcklbw %%mm3, %%mm5; // Unpack lower 4 pixels from b for addition + paddw %%mm2, %%mm5; // Add lower a and lower b unpacked + punpckhbw %%mm3, %%mm1; // Unpack upper 4 pixels from b for addition + paddw %%mm0, %%mm1; // Add upper a and upper b unpacked + movq (%%edx), %%mm2; // Load c for difference + paddw MMX_AVGDIFF_1, %%mm5; // Add 1 to the result of lower a + b + pxor %%mm4, %%mm4; // Zero out temp for c unpacking + movq %%mm2, %%mm3; // Make a copy of c for unpacking + paddw MMX_AVGDIFF_1, %%mm1; // Add 1 to the result of upper a + b + punpcklbw %%mm4, %%mm3; // Unpack lower 4 pixels from c for subtraction + punpckhbw %%mm4, %%mm2; // Unpack upper 4 pixels from c + movq %%mm3, %%mm0; // Make a copy of lower c for absdiff + psraw $1, %%mm5; // Divide result of lower a + b by 2 + movq %%mm2, %%mm4; // Make a copy of upper c for absdiff + psraw $1, %%mm1; // Divide result of upper a + b by 2 + psubusw %%mm5, %%mm3; // Subtract lower pixels one way + psubusw %%mm1, %%mm2; // Subtract upper pixels one way + psubusw %%mm0, %%mm5; // Subtract lower pixels the other way + por %%mm5, %%mm3; // Or the result of the lower pixels + psubusw %%mm4, %%mm1; // Subtract upper pixels the other way + por %%mm1, %%mm2; // Or the result of the upper pixels + paddw %%mm3, %%mm7; // Accumulate lower pixels + paddw %%mm2, %%mm7; // Accumulate upper pixels + " + : + : "b" (p1), "c" (p2), "d" (p3)); +#else + asm(" + .align 8 + movq (%%ebx), %%mm0; // Load 8 pixels from a + movq (%%ecx), %%mm1; // Load 8 pixels from b + pavgusb %%mm1, %%mm0; // avg of a, b in mm0 + movq (%%edx), %%mm2; // Load c for difference + movq %%mm0, %%mm5; + psubusb %%mm2, %%mm0; // Subtract lower pixels one way + psubusb %%mm5, %%mm2; + por %%mm0, %%mm2; // Or the result of the lower pixels + punpcklbw %%mm2, %%mm3; + paddw %%mm3, %%mm7; // Accumulate lower pixels + punpckhbw %%mm2, %%mm4; + paddw %%mm4, %%mm7; // Accumulate upper pixels + " + : + : "b" (p1), "c" (p2), "d" (p3)); +#endif +} + +void inline mmx_avg4diff(unsigned char *p1, unsigned char *p2, unsigned char *p3) +{ + asm(" + .align 8 + movq (%%ebx), %%mm0; // Load 8 pixels from a + movq 1(%%ebx), %%mm1 + pavgusb %%mm1, %%mm0; // avg of a, b in mm0 + movq (%%ecx), %%mm2; // Load 8 pixels from b + movq 1(%%ecx), %%mm3; + pavgusb %%mm2, %%mm3; + movq (%%edx), %%mm2; // Load c for difference + pavgusb %%mm3, %%mm0; + movq %%mm0, %%mm5; + psubusb %%mm2, %%mm0; // Subtract lower pixels one way + psubusb %%mm5, %%mm2; + por %%mm0, %%mm2; // Or the result of the lower pixels + punpcklbw %%mm2, %%mm3; + paddw %%mm3, %%mm7; // Accumulate lower pixels + punpckhbw %%mm2, %%mm4; + paddw %%mm4, %%mm7; // Accumulate upper pixels + " + : + : "b" (p1), "c" (p2), "d" (p3)); +} + +static unsigned long MMX_ACCUM_AND[] = {0xffffffff, 0x00000000}; + +unsigned int inline mmx_accum_avgdiff() +{ + unsigned long long r = 0; + asm(" + .align 8 + pxor %%mm5, %%mm5; // Clear temp for unpacking + movq %%mm7, %%mm6; // Make a copy for unpacking + punpcklwd %%mm5, %%mm6; // Unpack lower 2 pixels for accumulation + punpckhwd %%mm5, %%mm7; // Unpack high 2 pixels for accumulation + paddw %%mm6, %%mm7; // Add 2 doublewords in each register + movq %%mm7, %%mm6; // Copy the result for a final add + pand MMX_ACCUM_AND, %%mm7; // And the result for accumulation + psrlq $32, %%mm6; // Shift the copy right for accumulation + paddd %%mm6, %%mm7; // Add the results + movq %%mm7, (%%ebx); // Store result + femms; + " + : : "b" (&r)); + + return (unsigned int)r; +} + +static int dist1(blk1, blk2, lx, hx, hy, h, distlim) +unsigned char *blk1,*blk2; +int lx,hx,hy,h; +int distlim; +{ + unsigned char *p1, *p1a, *p2; + int v, s; + int j; + + s = 0; + p1 = blk1; + p2 = blk2; + + if(!hx && !hy) + { +#ifdef HAVE_MMX + mmx_start_block(); +#endif + for(j = 0; j < h; j++) + { +#ifdef HAVE_MMX + mmx_absdiff(p1, p2); +#else + if((v = p1[0] - p2[0]) < 0) v = -v; s += v; + if((v = p1[1] - p2[1]) < 0) v = -v; s += v; + if((v = p1[2] - p2[2]) < 0) v = -v; s += v; + if((v = p1[3] - p2[3]) < 0) v = -v; s += v; + if((v = p1[4] - p2[4]) < 0) v = -v; s += v; + if((v = p1[5] - p2[5]) < 0) v = -v; s += v; + if((v = p1[6] - p2[6]) < 0) v = -v; s += v; + if((v = p1[7] - p2[7]) < 0) v = -v; s += v; + if((v = p1[8] - p2[8]) < 0) v = -v; s += v; + if((v = p1[9] - p2[9]) < 0) v = -v; s += v; + if((v = p1[10] - p2[10]) < 0) v = -v; s += v; + if((v = p1[11] - p2[11]) < 0) v = -v; s += v; + if((v = p1[12] - p2[12]) < 0) v = -v; s += v; + if((v = p1[13] - p2[13]) < 0) v = -v; s += v; + if((v = p1[14] - p2[14]) < 0) v = -v; s += v; + if((v = p1[15] - p2[15]) < 0) v = -v; s += v; + if(s >= distlim) break; +#endif + + p1 += lx; + p2 += lx; + } +#ifdef HAVE_MMX + s = mmx_accum_absdiff(); +#endif + } + else + if(hx && !hy) + { +#ifdef HAVE_MMX + mmx_start_block(); +#endif + for(j = 0; j < h; j++) + { +#ifdef HAVE_MMX + mmx_avgdiff(p1, &p1[1], p2); + mmx_avgdiff(&p1[8], &p1[9], &p2[8]); +#else + v = ((unsigned int)(p1[0] + p1[1] + 1) >> 1) - p2[0]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[1] + p1[2] + 1) >> 1) - p2[1]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[2] + p1[3] + 1) >> 1) - p2[2]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[3] + p1[4] + 1) >> 1) - p2[3]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[4] + p1[5] + 1) >> 1) - p2[4]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[5] + p1[6] + 1) >> 1) - p2[5]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[6] + p1[7] + 1) >> 1) - p2[6]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[7] + p1[8] + 1) >> 1) - p2[7]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[8] + p1[9] + 1) >> 1) - p2[8]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[9] + p1[10] + 1) >> 1) - p2[9]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[10] + p1[11] + 1) >> 1) - p2[10]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[11] + p1[12] + 1) >> 1) - p2[11]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[12] + p1[13] + 1) >> 1) - p2[12]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[13] + p1[14] + 1) >> 1) - p2[13]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[14] + p1[15] + 1) >> 1) - p2[14]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[15] + p1[16] + 1) >> 1) - p2[15]; if(v < 0) s -= v; else s += v; +#endif + + p1 += lx; + p2 += lx; + } +#ifdef HAVE_MMX + s = mmx_accum_avgdiff(); +#endif + } + else if(!hx && hy) + { +#ifdef HAVE_MMX + mmx_start_block(); +#endif + p1a = p1 + lx; + for(j = 0; j < h; j++) + { +#ifdef HAVE_MMX + mmx_avgdiff(p1, p1a, p2); + mmx_avgdiff(&p1[8], &p1a[8], &p2[8]); +#else + v = ((unsigned int)(p1[0] + p1a[0] + 1) >> 1) - p2[0]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[1] + p1a[1] + 1) >> 1) - p2[1]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[2] + p1a[2] + 1) >> 1) - p2[2]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[3] + p1a[3] + 1) >> 1) - p2[3]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[4] + p1a[4] + 1) >> 1) - p2[4]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[5] + p1a[5] + 1) >> 1) - p2[5]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[6] + p1a[6] + 1) >> 1) - p2[6]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[7] + p1a[7] + 1) >> 1) - p2[7]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[8] + p1a[8] + 1) >> 1) - p2[8]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[9] + p1a[9] + 1) >> 1) - p2[9]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[10] + p1a[10] + 1) >> 1) - p2[10]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[11] + p1a[11] + 1) >> 1) - p2[11]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[12] + p1a[12] + 1) >> 1) - p2[12]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[13] + p1a[13] + 1) >> 1) - p2[13]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[14] + p1a[14] + 1) >> 1) - p2[14]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[15] + p1a[15] + 1) >> 1) - p2[15]; if(v < 0) s -= v; else s += v; +#endif + + p1 = p1a; + p1a += lx; + p2 += lx; + } +#ifdef HAVE_MMX + s = mmx_accum_avgdiff(); +#endif + } + else /* if (hx && hy) */ + { +#ifdef HAVE_MMX + mmx_start_block(); +#endif + p1a = p1 + lx; + for(j = 0; j < h; j++) + { +#ifdef HAVE_MMX + mmx_avg4diff(p1,p1a,p2); + mmx_avg4diff(&p1[8],&p1a[8],&p2[8]); +#else + v = ((unsigned int)(p1[0] + p1[1] + p1a[0] + p1a[1] + 2) >> 2) - p2[0]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[1] + p1[2] + p1a[1] + p1a[3] + 2) >> 2) - p2[1]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[2] + p1[3] + p1a[2] + p1a[3] + 2) >> 2) - p2[2]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[3] + p1[4] + p1a[3] + p1a[4] + 2) >> 2) - p2[3]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[4] + p1[5] + p1a[4] + p1a[5] + 2) >> 2) - p2[4]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[5] + p1[6] + p1a[5] + p1a[6] + 2) >> 2) - p2[5]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[6] + p1[7] + p1a[6] + p1a[7] + 2) >> 2) - p2[6]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[7] + p1[8] + p1a[7] + p1a[8] + 2) >> 2) - p2[7]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[8] + p1[9] + p1a[8] + p1a[9] + 2) >> 2) - p2[8]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[9] + p1[10] + p1a[9] + p1a[10] + 2) >> 2) - p2[9]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[10] + p1[11] + p1a[10] + p1a[11] + 2) >> 2) - p2[10]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[11] + p1[12] + p1a[11] + p1a[12] + 2) >> 2) - p2[11]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[12] + p1[13] + p1a[12] + p1a[13] + 2) >> 2) - p2[12]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[13] + p1[14] + p1a[13] + p1a[14] + 2) >> 2) - p2[13]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[14] + p1[15] + p1a[14] + p1a[15] + 2) >> 2) - p2[14]; if(v < 0) s -= v; else s += v; + v = ((unsigned int)(p1[15] + p1[16] + p1a[15] + p1a[16] + 2) >> 2) - p2[15]; if(v < 0) s -= v; else s += v; +#endif + p1 = p1a; + p1a += lx; + p2+= lx; + } +#ifdef HAVE_MMX + s=mmx_accum_avgdiff(); +#endif + } + return s; } /* @@ -1444,71 +1726,121 @@ * hx,hy: flags for horizontal and/or vertical interpolation * h: height of block (usually 8 or 16) */ -static int dist2(blk1,blk2,lx,hx,hy,h) +static int dist2(blk1, blk2, lx, hx, hy, h) unsigned char *blk1,*blk2; int lx,hx,hy,h; { - unsigned char *p1,*p1a,*p2; - int i,j; - int s,v; - - s = 0; - p1 = blk1; - p2 = blk2; - if (!hx && !hy) - for (j=0; j>1) - p2[i]; - s+= v*v; - } - p1+= lx; - p2+= lx; - } - else if (!hx && hy) - { - p1a = p1 + lx; - for (j=0; j>1) - p2[i]; - s+= v*v; - } - p1 = p1a; - p1a+= lx; - p2+= lx; - } - } - else /* if (hx && hy) */ - { - p1a = p1 + lx; - for (j=0; j>2) - p2[i]; - s+= v*v; - } - p1 = p1a; - p1a+= lx; - p2+= lx; - } - } + register unsigned char *p1,*p1a,*p2; + register int v, s; + int j; + + s = 0; + p1 = blk1; + p2 = blk2; + if(!hx && !hy) + { + for(j = 0; j < h; j++) + { + v = p1[0] - p2[0]; s += v * v; + v = p1[1] - p2[1]; s += v * v; + v = p1[2] - p2[2]; s += v * v; + v = p1[3] - p2[3]; s += v * v; + v = p1[4] - p2[4]; s += v * v; + v = p1[5] - p2[5]; s += v * v; + v = p1[6] - p2[6]; s += v * v; + v = p1[7] - p2[7]; s += v * v; + v = p1[8] - p2[8]; s += v * v; + v = p1[9] - p2[9]; s += v * v; + v = p1[10] - p2[10]; s += v * v; + v = p1[11] - p2[11]; s += v * v; + v = p1[12] - p2[12]; s += v * v; + v = p1[13] - p2[13]; s += v * v; + v = p1[14] - p2[14]; s += v * v; + v = p1[15] - p2[15]; s += v * v; + p1 += lx; + p2 += lx; + } + } + else + if(hx && !hy) + { + for (j = 0; j < h; j++) + { + v = ((unsigned int)(p1[0] + p1[1] + 1) >> 1) - p2[0]; s += v * v; + v = ((unsigned int)(p1[1] + p1[2] + 1) >> 1) - p2[1]; s += v * v; + v = ((unsigned int)(p1[2] + p1[3] + 1) >> 1) - p2[2]; s += v * v; + v = ((unsigned int)(p1[3] + p1[4] + 1) >> 1) - p2[3]; s += v * v; + v = ((unsigned int)(p1[4] + p1[5] + 1) >> 1) - p2[4]; s += v * v; + v = ((unsigned int)(p1[5] + p1[6] + 1) >> 1) - p2[5]; s += v * v; + v = ((unsigned int)(p1[6] + p1[7] + 1) >> 1) - p2[6]; s += v * v; + v = ((unsigned int)(p1[7] + p1[8] + 1) >> 1) - p2[7]; s += v * v; + v = ((unsigned int)(p1[8] + p1[9] + 1) >> 1) - p2[8]; s += v * v; + v = ((unsigned int)(p1[9] + p1[10] + 1) >> 1) - p2[9]; s += v * v; + v = ((unsigned int)(p1[10]+ p1[11] + 1) >> 1) - p2[10]; s += v * v; + v = ((unsigned int)(p1[11]+ p1[12] + 1) >> 1) - p2[11]; s += v * v; + v = ((unsigned int)(p1[12]+ p1[13] + 1) >> 1) - p2[12]; s += v * v; + v = ((unsigned int)(p1[13]+ p1[14] + 1) >> 1) - p2[13]; s += v * v; + v = ((unsigned int)(p1[14]+ p1[15] + 1) >> 1) - p2[14]; s += v * v; + v = ((unsigned int)(p1[15]+ p1[16] + 1) >> 1) - p2[15]; s += v * v; + p1 += lx; + p2 += lx; + } + } + else + if(!hx && hy) + { + p1a = p1 + lx; + for(j = 0; j < h; j++) + { + v = ((unsigned int)(p1[0] + p1a[0] + 1) >> 1) - p2[0]; s += v * v; + v = ((unsigned int)(p1[1] + p1a[1] + 1) >> 1) - p2[1]; s += v * v; + v = ((unsigned int)(p1[2] + p1a[2] + 1) >> 1) - p2[2]; s += v * v; + v = ((unsigned int)(p1[3] + p1a[3] + 1) >> 1) - p2[3]; s += v * v; + v = ((unsigned int)(p1[4] + p1a[4] + 1) >> 1) - p2[4]; s += v * v; + v = ((unsigned int)(p1[5] + p1a[5] + 1) >> 1) - p2[5]; s += v * v; + v = ((unsigned int)(p1[6] + p1a[6] + 1) >> 1) - p2[6]; s += v * v; + v = ((unsigned int)(p1[7] + p1a[7] + 1) >> 1) - p2[7]; s += v * v; + v = ((unsigned int)(p1[8] + p1a[8] + 1) >> 1) - p2[8]; s += v * v; + v = ((unsigned int)(p1[9] + p1a[9] + 1) >> 1) - p2[9]; s += v * v; + v = ((unsigned int)(p1[10]+ p1a[10]+ 1) >> 1) - p2[10]; s += v * v; + v = ((unsigned int)(p1[11]+ p1a[11]+ 1) >> 1) - p2[11]; s += v * v; + v = ((unsigned int)(p1[12]+ p1a[12]+ 1) >> 1) - p2[12]; s += v * v; + v = ((unsigned int)(p1[13]+ p1a[13]+ 1) >> 1) - p2[13]; s += v * v; + v = ((unsigned int)(p1[14]+ p1a[14]+ 1) >> 1) - p2[14]; s += v * v; + v = ((unsigned int)(p1[15]+ p1a[15]+ 1) >> 1) - p2[15]; s += v * v; + p1 = p1a; + p1a += lx; + p2 += lx; + } + } + else /* if (hx && hy) */ + { + p1a = p1 + lx; + for(j = 0; j < h; j++) + { + v = ((unsigned int)(p1[0] + p1[1] + p1a[0] + p1a[1] + 2) >> 2) - p2[0]; s += v * v; + v = ((unsigned int)(p1[1] + p1[2] + p1a[1] + p1a[2] + 2) >> 2) - p2[1]; s += v * v; + v = ((unsigned int)(p1[2] + p1[3] + p1a[2] + p1a[3] + 2) >> 2) - p2[2]; s += v * v; + v = ((unsigned int)(p1[3] + p1[4] + p1a[3] + p1a[4] + 2) >> 2) - p2[3]; s += v * v; + v = ((unsigned int)(p1[4] + p1[5] + p1a[4] + p1a[5] + 2) >> 2) - p2[4]; s += v * v; + v = ((unsigned int)(p1[5] + p1[6] + p1a[5] + p1a[6] + 2) >> 2) - p2[5]; s += v * v; + v = ((unsigned int)(p1[6] + p1[7] + p1a[6] + p1a[7] + 2) >> 2) - p2[6]; s += v * v; + v = ((unsigned int)(p1[7] + p1[8] + p1a[7] + p1a[8] + 2) >> 2) - p2[7]; s += v * v; + v = ((unsigned int)(p1[8] + p1[9] + p1a[8] + p1a[9] + 2) >> 2) - p2[8]; s += v * v; + v = ((unsigned int)(p1[9] + p1[10] + p1a[9] + p1a[10] + 2) >> 2) - p2[9]; s += v * v; + v = ((unsigned int)(p1[10] + p1[11] + p1a[10] + p1a[11] + 2) >> 2) - p2[10]; s += v * v; + v = ((unsigned int)(p1[11] + p1[12] + p1a[11] + p1a[12] + 2) >> 2) - p2[11]; s += v * v; + v = ((unsigned int)(p1[12] + p1[13] + p1a[12] + p1a[13] + 2) >> 2) - p2[12]; s += v * v; + v = ((unsigned int)(p1[13] + p1[14] + p1a[13] + p1a[14] + 2) >> 2) - p2[13]; s += v * v; + v = ((unsigned int)(p1[14] + p1[15] + p1a[14] + p1a[15] + 2) >> 2) - p2[14]; s += v * v; + v = ((unsigned int)(p1[15] + p1[16] + p1a[15] + p1a[16] + 2) >> 2) - p2[15]; s += v * v; + p1 = p1a; + p1a += lx; + p2 += lx; + } + } - return s; + return s; } /* @@ -1579,41 +1911,98 @@ unsigned char *pf,*pb,*p2; int lx,hxf,hyf,hxb,hyb,h; { - unsigned char *pfa,*pfb,*pfc,*pba,*pbb,*pbc; - int i,j; - int s,v; - - pfa = pf + hxf; - pfb = pf + lx*hyf; - pfc = pfb + hxf; - - pba = pb + hxb; - pbb = pb + lx*hyb; - pbc = pbb + hxb; - - s = 0; - - for (j=0; j>2) + - ((unsigned int)(*pb++ + *pba++ + *pbb++ + *pbc++ + 2)>>2) + 1)>>1) - - *p2++; - s+=v*v; - } - p2+= lx-16; - pf+= lx-16; - pfa+= lx-16; - pfb+= lx-16; - pfc+= lx-16; - pb+= lx-16; - pba+= lx-16; - pbb+= lx-16; - pbc+= lx-16; - } - - return s; + register unsigned char *pfa,*pfb,*pfc,*pba,*pbb,*pbc; + int j; + register int i, s, v; + + pfa = pf + hxf; + pfb = pf + lx*hyf; + pfc = pfb + hxf; + + pba = pb + hxb; + pbb = pb + lx*hyb; + pbc = pbb + hxb; + + s = 0; + + for(j = 0; j < h; j++) + { + v = ((((unsigned int)(pf[0] + pfa[0] + pfb[0] + pfc[0] + 2) >> 2) + + ((unsigned int)(pb[0] + pba[0] + pbb[0] + pbc[0] + 2) >> 2) + 1) >> 1) - + p2[0]; + s += v * v; + v = ((((unsigned int)(pf[1] + pfa[1] + pfb[1] + pfc[1] + 2) >> 2) + + ((unsigned int)(pb[1] + pba[1] + pbb[1] + pbc[1] + 2) >> 2) + 1) >> 1) - + p2[1]; + s += v * v; + v = ((((unsigned int)(pf[2] + pfa[2] + pfb[2] + pfc[2] + 2) >> 2) + + ((unsigned int)(pb[2] + pba[2] + pbb[2] + pbc[2] + 2) >> 2) + 1) >> 1) - + p2[2]; + s += v * v; + v = ((((unsigned int)(pf[3] + pfa[3] + pfb[3] + pfc[3] + 2) >> 2) + + ((unsigned int)(pb[3] + pba[3] + pbb[3] + pbc[3] + 2) >> 2) + 1) >> 1) - + p2[3]; + s += v * v; + v = ((((unsigned int)(pf[4] + pfa[4] + pfb[4] + pfc[4] + 2) >> 2) + + ((unsigned int)(pb[4] + pba[4] + pbb[4] + pbc[4] + 2) >> 2) + 1) >> 1) - + p2[4]; + s += v * v; + v = ((((unsigned int)(pf[5] + pfa[5] + pfb[5] + pfc[5] + 2) >> 2) + + ((unsigned int)(pb[5] + pba[5] + pbb[5] + pbc[5] + 2) >> 2) + 1) >> 1) - + p2[5]; + s += v * v; + v = ((((unsigned int)(pf[6] + pfa[6] + pfb[6] + pfc[6] + 2) >> 2) + + ((unsigned int)(pb[6] + pba[6] + pbb[6] + pbc[6] + 2) >> 2) + 1) >> 1) - + p2[6]; + s += v * v; + v = ((((unsigned int)(pf[7] + pfa[7] + pfb[7] + pfc[7] + 2) >> 2) + + ((unsigned int)(pb[7] + pba[7] + pbb[7] + pbc[7] + 2) >> 2) + 1) >> 1) - + p2[7]; + s += v * v; + v = ((((unsigned int)(pf[8] + pfa[8] + pfb[8] + pfc[8] + 2) >> 2) + + ((unsigned int)(pb[8] + pba[8] + pbb[8] + pbc[8] + 2) >> 2) + 1) >> 1) - + p2[8]; + s += v * v; + v = ((((unsigned int)(pf[9] + pfa[9] + pfb[9] + pfc[9] + 2) >> 2) + + ((unsigned int)(pb[9] + pba[9] + pbb[9] + pbc[9] + 2) >> 2) + 1) >> 1) - + p2[9]; + s += v * v; + v = ((((unsigned int)(pf[10] + pfa[10] + pfb[10] + pfc[10] + 2) >> 2) + + ((unsigned int)(pb[10] + pba[10] + pbb[10] + pbc[10] + 2) >> 2) + 1) >> 1) - + p2[10]; + s += v * v; + v = ((((unsigned int)(pf[11] + pfa[11] + pfb[11] + pfc[11] + 2) >> 2) + + ((unsigned int)(pb[11] + pba[11] + pbb[11] + pbc[11] + 2) >> 2) + 1) >> 1) - + p2[11]; + s += v * v; + v = ((((unsigned int)(pf[12] + pfa[12] + pfb[12] + pfc[12] + 2) >> 2) + + ((unsigned int)(pb[12] + pba[12] + pbb[12] + pbc[12] + 2) >> 2) + 1) >> 1) - + p2[12]; + s += v * v; + v = ((((unsigned int)(pf[13] + pfa[13] + pfb[13] + pfc[13] + 2) >> 2) + + ((unsigned int)(pb[13] + pba[13] + pbb[13] + pbc[13] + 2) >> 2) + 1) >> 1) - + p2[13]; + s += v * v; + v = ((((unsigned int)(pf[14] + pfa[14] + pfb[14] + pfc[14] + 2) >> 2) + + ((unsigned int)(pb[14] + pba[14] + pbb[14] + pbc[14] + 2) >> 2) + 1) >> 1) - + p2[14]; + s += v * v; + v = ((((unsigned int)(pf[15] + pfa[15] + pfb[15] + pfc[15] + 2) >> 2) + + ((unsigned int)(pb[15] + pba[15] + pbb[15] + pbc[15] + 2) >> 2) + 1) >> 1) - + p2[15]; + s += v * v; + + p2 += lx; + pf += lx; + pfa += lx; + pfb += lx; + pfc += lx; + pb += lx; + pba += lx; + pbb += lx; + pbc += lx; + } + return s; } /* @@ -1621,24 +2010,34 @@ * p: address of top left pel of block * lx: distance (in bytes) of vertically adjacent pels */ -static int variance(p,lx) +static int variance(p, lx) unsigned char *p; int lx; { - int i,j; - unsigned int v,s,s2; + int i, j; + register unsigned int v, s, s2; - s = s2 = 0; + s = s2 = 0; - for (j=0; j<16; j++) - { - for (i=0; i<16; i++) - { - v = *p++; - s+= v; - s2+= v*v; - } - p+= lx-16; - } - return s2 - (s*s)/256; + for (j=0; j<16; j++) + { + v = p[0]; s += v; s2 += v * v; + v = p[1]; s += v; s2 += v * v; + v = p[2]; s += v; s2 += v * v; + v = p[3]; s += v; s2 += v * v; + v = p[4]; s += v; s2 += v * v; + v = p[5]; s += v; s2 += v * v; + v = p[6]; s += v; s2 += v * v; + v = p[7]; s += v; s2 += v * v; + v = p[8]; s += v; s2 += v * v; + v = p[9]; s += v; s2 += v * v; + v = p[10]; s += v; s2 += v * v; + v = p[11]; s += v; s2 += v * v; + v = p[12]; s += v; s2 += v * v; + v = p[13]; s += v; s2 += v * v; + v = p[14]; s += v; s2 += v * v; + v = p[15]; s += v; s2 += v * v; + p += lx; + } + return s2 - (s*s)/256; } Binary files mpeg2enc.orig/motion.o and mpeg2enc/motion.o differ Binary files mpeg2enc.orig/mpeg2enc.o and mpeg2enc/mpeg2enc.o differ Binary files mpeg2enc.orig/mpeg2encode and mpeg2enc/mpeg2encode differ Only in mpeg2enc: mpeg2encode.ref Binary files mpeg2enc.orig/predict.o and mpeg2enc/predict.o differ Only in mpeg2enc: profrep1 Only in mpeg2enc: profrep2 Only in mpeg2enc: profrep3 Only in mpeg2enc: profrep4 Only in mpeg2enc: profrep5 Only in mpeg2enc: profrep6 Binary files mpeg2enc.orig/putbits.o and mpeg2enc/putbits.o differ Binary files mpeg2enc.orig/puthdr.o and mpeg2enc/puthdr.o differ Binary files mpeg2enc.orig/putmpg.o and mpeg2enc/putmpg.o differ Binary files mpeg2enc.orig/putpic.o and mpeg2enc/putpic.o differ Binary files mpeg2enc.orig/putseq.o and mpeg2enc/putseq.o differ Binary files mpeg2enc.orig/putvlc.o and mpeg2enc/putvlc.o differ Binary files mpeg2enc.orig/quantize.o and mpeg2enc/quantize.o differ Only in mpeg2enc: r.diff Only in mpeg2enc: ratectl.c.mmx Only in mpeg2enc: ratectl.c.orig Only in mpeg2enc: ratectl.c.rej Binary files mpeg2enc.orig/ratectl.o and mpeg2enc/ratectl.o differ Binary files mpeg2enc.orig/readavi.o and mpeg2enc/readavi.o differ Binary files mpeg2enc.orig/readpic.o and mpeg2enc/readpic.o differ Binary files mpeg2enc.orig/stats.o and mpeg2enc/stats.o differ Only in mpeg2enc.orig: test.m1v Only in mpeg2enc: test.m1v.ref Only in mpeg2enc.orig: test.mpg Only in mpeg2enc: testabsdiff Only in mpeg2enc: testabsdiff.c Binary files mpeg2enc.orig/transfrm.o and mpeg2enc/transfrm.o differ Only in mpeg2enc: x Only in mpeg2enc: xx Only in mpeg2enc: xxx Only in mpeg2enc: zwei