Skip to content

L
libswscale
Commit bf4c6c71 authored by diego's avatar diego
Browse files
Options

Replace asmalign.h hack by ASMALIGN cpp macros from config.h. 


git-svn-id: file:///var/local/repositories/mplayer/trunk/libswscale@19378 b3059339-0415-0410-9bf9-f77b7e298cf2
parent a327f3c1
Hide whitespace changes
Inline Side-by-side
Showing with 35 additions and 39 deletions
rgb2rgb_template.c
View file @ bf4c6c71
......@@ -12,8 +12,6 @@
#include <stddef.h>
#include <inttypes.h> /* for __WORDSIZE */
#include "asmalign.h"
#ifndef __WORDSIZE
// #warning You have misconfigured system and probably will lose performance!
#define __WORDSIZE MP_WORDSIZE
......@@ -343,7 +341,7 @@ static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, long src_
"movq %3, %%mm5 \n\t"
"movq %4, %%mm6 \n\t"
"movq %5, %%mm7 \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 32(%1) \n\t"
"movd (%1), %%mm0 \n\t"
......@@ -500,7 +498,7 @@ static inline void RENAME(rgb32to15)(const uint8_t *src, uint8_t *dst, long src_
"movq %3, %%mm5 \n\t"
"movq %4, %%mm6 \n\t"
"movq %5, %%mm7 \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 32(%1) \n\t"
"movd (%1), %%mm0 \n\t"
......@@ -1355,7 +1353,7 @@ static inline void RENAME(rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long s
/* TODO: unroll this loop */
asm volatile (
"xor %%"REG_a", %%"REG_a" \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 32(%0, %%"REG_a") \n\t"
"movq (%0, %%"REG_a"), %%mm0 \n\t"
......@@ -1405,7 +1403,7 @@ static inline void RENAME(rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long s
"movq "MANGLE(mask24r)", %%mm5 \n\t"
"movq "MANGLE(mask24g)", %%mm6 \n\t"
"movq "MANGLE(mask24b)", %%mm7 \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 32(%1, %%"REG_a") \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t" // BGR BGR BG
......@@ -1475,7 +1473,7 @@ static inline void RENAME(yuvPlanartoyuy2)(const uint8_t *ysrc, const uint8_t *u
//FIXME handle 2 lines a once (fewer prefetch, reuse some chrom, but very likely limited by mem anyway)
asm volatile(
"xor %%"REG_a", %%"REG_a" \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 32(%1, %%"REG_a", 2) \n\t"
PREFETCH" 32(%2, %%"REG_a") \n\t"
......@@ -1628,7 +1626,7 @@ static inline void RENAME(yuvPlanartouyvy)(const uint8_t *ysrc, const uint8_t *u
//FIXME handle 2 lines a once (fewer prefetch, reuse some chrom, but very likely limited by mem anyway)
asm volatile(
"xor %%"REG_a", %%"REG_a" \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 32(%1, %%"REG_a", 2) \n\t"
PREFETCH" 32(%2, %%"REG_a") \n\t"
......@@ -1752,7 +1750,7 @@ static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t
"xor %%"REG_a", %%"REG_a" \n\t"
"pcmpeqw %%mm7, %%mm7 \n\t"
"psrlw $8, %%mm7 \n\t" // FF,00,FF,00...
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 64(%0, %%"REG_a", 4) \n\t"
"movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0)
......@@ -1805,7 +1803,7 @@ static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t
asm volatile(
"xor %%"REG_a", %%"REG_a" \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 64(%0, %%"REG_a", 4) \n\t"
"movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0)
......@@ -1990,7 +1988,7 @@ static inline void RENAME(uyvytoyv12)(const uint8_t *src, uint8_t *ydst, uint8_t
"xorl %%eax, %%eax \n\t"
"pcmpeqw %%mm7, %%mm7 \n\t"
"psrlw $8, %%mm7 \n\t" // FF,00,FF,00...
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 64(%0, %%eax, 4) \n\t"
"movq (%0, %%eax, 4), %%mm0 \n\t" // UYVY UYVY(0)
......@@ -2043,7 +2041,7 @@ static inline void RENAME(uyvytoyv12)(const uint8_t *src, uint8_t *ydst, uint8_t
asm volatile(
"xorl %%eax, %%eax \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 64(%0, %%eax, 4) \n\t"
"movq (%0, %%eax, 4), %%mm0 \n\t" // YUYV YUYV(0)
......@@ -2121,7 +2119,7 @@ static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_
"movq "MANGLE(w1111)", %%mm5 \n\t"
"pxor %%mm7, %%mm7 \n\t"
"lea (%%"REG_a", %%"REG_a", 2), %%"REG_b"\n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 64(%0, %%"REG_b") \n\t"
"movd (%0, %%"REG_b"), %%mm0 \n\t"
......@@ -2195,7 +2193,7 @@ static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_
"pxor %%mm7, %%mm7 \n\t"
"lea (%%"REG_a", %%"REG_a", 2), %%"REG_b"\n\t"
"add %%"REG_b", %%"REG_b" \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 64(%0, %%"REG_b") \n\t"
PREFETCH" 64(%1, %%"REG_b") \n\t"
......
swscale_template.c
View file @ bf4c6c71
......@@ -16,8 +16,6 @@
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "asmalign.h"
#undef REAL_MOVNTQ
#undef MOVNTQ
#undef PAVGB
......@@ -74,7 +72,7 @@
"movq %%mm3, %%mm4 \n\t"\
"lea " offset "(%0), %%"REG_d" \n\t"\
"mov (%%"REG_d"), %%"REG_S" \n\t"\
ASMALIGN16 /* FIXME Unroll? */\
ASMALIGN(4) /* FIXME Unroll? */\
"1: \n\t"\
"movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
"movq " #x "(%%"REG_S", %%"REG_a", 2), %%mm2\n\t" /* srcData */\
......@@ -112,7 +110,7 @@
"pxor %%mm6, %%mm6 \n\t"\
"pxor %%mm7, %%mm7 \n\t"\
"mov (%%"REG_d"), %%"REG_S" \n\t"\
ASMALIGN16 \
ASMALIGN(4) \
"1: \n\t"\
"movq " #x "(%%"REG_S", %%"REG_a", 2), %%mm0\n\t" /* srcData */\
"movq 8+" #x "(%%"REG_S", %%"REG_a", 2), %%mm2\n\t" /* srcData */\
......@@ -167,7 +165,7 @@
#define YSCALEYUV2YV121 \
"mov %2, %%"REG_a" \n\t"\
ASMALIGN16 /* FIXME Unroll? */\
ASMALIGN(4) /* FIXME Unroll? */\
"1: \n\t"\
"movq (%0, %%"REG_a", 2), %%mm0 \n\t"\
"movq 8(%0, %%"REG_a", 2), %%mm1\n\t"\
......@@ -188,14 +186,14 @@
#define YSCALEYUV2PACKEDX \
asm volatile(\
"xor %%"REG_a", %%"REG_a" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"nop \n\t"\
"1: \n\t"\
"lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\
"mov (%%"REG_d"), %%"REG_S" \n\t"\
"movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\
"movq %%mm3, %%mm4 \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"2: \n\t"\
"movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
"movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\
......@@ -213,7 +211,7 @@
"mov (%%"REG_d"), %%"REG_S" \n\t"\
"movq "VROUNDER_OFFSET"(%0), %%mm1\n\t"\
"movq %%mm1, %%mm7 \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"2: \n\t"\
"movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
"movq (%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y1srcData */\
......@@ -237,7 +235,7 @@
#define YSCALEYUV2PACKEDX_ACCURATE \
asm volatile(\
"xor %%"REG_a", %%"REG_a" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"nop \n\t"\
"1: \n\t"\
"lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"\n\t"\
......@@ -246,7 +244,7 @@
"pxor %%mm5, %%mm5 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
"pxor %%mm7, %%mm7 \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"2: \n\t"\
"movq (%%"REG_S", %%"REG_a"), %%mm0 \n\t" /* UsrcData */\
"movq 4096(%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* VsrcData */\
......@@ -290,7 +288,7 @@
"pxor %%mm5, %%mm5 \n\t"\
"pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"2: \n\t"\
"movq (%%"REG_S", %%"REG_a", 2), %%mm0 \n\t" /* Y1srcData */\
"movq 8(%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y2srcData */\
......@@ -374,7 +372,7 @@
"punpcklwd %%mm5, %%mm5 \n\t"\
"punpcklwd %%mm5, %%mm5 \n\t"\
"xor %%"REG_a", %%"REG_a" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"1: \n\t"\
"movq (%0, %%"REG_a", 2), %%mm0 \n\t" /*buf0[eax]*/\
"movq (%1, %%"REG_a", 2), %%mm1 \n\t" /*buf1[eax]*/\
......@@ -427,7 +425,7 @@
"movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c")\n\t"\
"movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c")\n\t"\
"xor "#index", "#index" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"1: \n\t"\
"movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
"movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
......@@ -459,7 +457,7 @@
#define REAL_YSCALEYUV2RGB(index, c) \
"xor "#index", "#index" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"1: \n\t"\
"movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
"movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
......@@ -525,7 +523,7 @@
#define REAL_YSCALEYUV2PACKED1(index, c) \
"xor "#index", "#index" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"1: \n\t"\
"movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
"movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
......@@ -540,7 +538,7 @@
#define REAL_YSCALEYUV2RGB1(index, c) \
"xor "#index", "#index" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"1: \n\t"\
"movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
"movq 4096(%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
......@@ -589,7 +587,7 @@
#define REAL_YSCALEYUV2PACKED1b(index, c) \
"xor "#index", "#index" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"1: \n\t"\
"movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
"movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
......@@ -608,7 +606,7 @@
// do vertical chrominance interpolation
#define REAL_YSCALEYUV2RGB1b(index, c) \
"xor "#index", "#index" \n\t"\
ASMALIGN16\
ASMALIGN(4)\
"1: \n\t"\
"movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
"movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
......@@ -1868,7 +1866,7 @@ static inline void RENAME(bgr24ToY)(uint8_t *dst, uint8_t *src, long width)
"movq "MANGLE(w1111)", %%mm5 \n\t"
"pxor %%mm7, %%mm7 \n\t"
"lea (%%"REG_a", %%"REG_a", 2), %%"REG_b"\n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 64(%0, %%"REG_b") \n\t"
"movd (%0, %%"REG_b"), %%mm0 \n\t"
......@@ -1954,7 +1952,7 @@ static inline void RENAME(bgr24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1
"pxor %%mm7, %%mm7 \n\t"
"lea (%%"REG_a", %%"REG_a", 2), %%"REG_b" \n\t"
"add %%"REG_b", %%"REG_b" \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
PREFETCH" 64(%0, %%"REG_b") \n\t"
PREFETCH" 64(%1, %%"REG_b") \n\t"
......@@ -2261,7 +2259,7 @@ static inline void RENAME(hScale)(int16_t *dst, int dstW, uint8_t *src, int srcW
"movq "MANGLE(w02)", %%mm6 \n\t"
"push %%"REG_BP" \n\t" // we use 7 regs here ...
"mov %%"REG_a", %%"REG_BP" \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
"movzwl (%2, %%"REG_BP"), %%eax \n\t"
"movzwl 2(%2, %%"REG_BP"), %%ebx\n\t"
......@@ -2299,7 +2297,7 @@ static inline void RENAME(hScale)(int16_t *dst, int dstW, uint8_t *src, int srcW
"movq "MANGLE(w02)", %%mm6 \n\t"
"push %%"REG_BP" \n\t" // we use 7 regs here ...
"mov %%"REG_a", %%"REG_BP" \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
"movzwl (%2, %%"REG_BP"), %%eax \n\t"
"movzwl 2(%2, %%"REG_BP"), %%ebx\n\t"
......@@ -2348,7 +2346,7 @@ static inline void RENAME(hScale)(int16_t *dst, int dstW, uint8_t *src, int srcW
asm volatile(
"pxor %%mm7, %%mm7 \n\t"
"movq "MANGLE(w02)", %%mm6 \n\t"
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
"mov %2, %%"REG_c" \n\t"
"movzwl (%%"REG_c", %0), %%eax \n\t"
......@@ -2532,7 +2530,7 @@ FUNNY_Y_CODE
"xor %%"REG_a", %%"REG_a" \n\t" // i
"xor %%"REG_b", %%"REG_b" \n\t" // xx
"xorl %%ecx, %%ecx \n\t" // 2*xalpha
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
"movzbl (%0, %%"REG_b"), %%edi \n\t" //src[xx]
"movzbl 1(%0, %%"REG_b"), %%esi \n\t" //src[xx+1]
......@@ -2729,7 +2727,7 @@ FUNNY_UV_CODE
"xor %%"REG_a", %%"REG_a" \n\t" // i
"xor %%"REG_b", %%"REG_b" \n\t" // xx
"xorl %%ecx, %%ecx \n\t" // 2*xalpha
ASMALIGN16
ASMALIGN(4)
"1: \n\t"
"mov %0, %%"REG_S" \n\t"
"movzbl (%%"REG_S", %%"REG_b"), %%edi \n\t" //src[xx]
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment