Mercurial > libavcodec.hg
comparison i386/motion_est_mmx.c @ 3576:f7125bf10892 libavcodec
Support for MacIntel, last part: balign directives
Determines whether .align's arg is power-of-two or not, then defines ASMALIGN appropriately in config.h. Changes all .baligns to ASMALIGNs.
Patch by John Dalgliesh % johnd AH defyne P org %
Original thread:
Date: Aug 11, 2006 8:00 AM
Subject: Re: [Ffmpeg-devel] Mac OS X Intel last part: balign directives
| author | gpoirier |
|---|---|
| date | Sat, 12 Aug 2006 16:37:31 +0000 |
| parents | e0927bc44a10 |
| children | c8c591fe26f8 |
comparison
equal
deleted
inserted
replaced
| 3575:4baed6f6577b | 3576:f7125bf10892 |
|---|---|
| 32 | 32 |
| 33 static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) | 33 static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
| 34 { | 34 { |
| 35 long len= -(stride*h); | 35 long len= -(stride*h); |
| 36 asm volatile( | 36 asm volatile( |
| 37 ".balign 16 \n\t" | 37 ASMALIGN(4) |
| 38 "1: \n\t" | 38 "1: \n\t" |
| 39 "movq (%1, %%"REG_a"), %%mm0 \n\t" | 39 "movq (%1, %%"REG_a"), %%mm0 \n\t" |
| 40 "movq (%2, %%"REG_a"), %%mm2 \n\t" | 40 "movq (%2, %%"REG_a"), %%mm2 \n\t" |
| 41 "movq (%2, %%"REG_a"), %%mm4 \n\t" | 41 "movq (%2, %%"REG_a"), %%mm4 \n\t" |
| 42 "add %3, %%"REG_a" \n\t" | 42 "add %3, %%"REG_a" \n\t" |
| 68 | 68 |
| 69 static inline void sad8_1_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h) | 69 static inline void sad8_1_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
| 70 { | 70 { |
| 71 long len= -(stride*h); | 71 long len= -(stride*h); |
| 72 asm volatile( | 72 asm volatile( |
| 73 ".balign 16 \n\t" | 73 ASMALIGN(4) |
| 74 "1: \n\t" | 74 "1: \n\t" |
| 75 "movq (%1, %%"REG_a"), %%mm0 \n\t" | 75 "movq (%1, %%"REG_a"), %%mm0 \n\t" |
| 76 "movq (%2, %%"REG_a"), %%mm2 \n\t" | 76 "movq (%2, %%"REG_a"), %%mm2 \n\t" |
| 77 "psadbw %%mm2, %%mm0 \n\t" | 77 "psadbw %%mm2, %%mm0 \n\t" |
| 78 "add %3, %%"REG_a" \n\t" | 78 "add %3, %%"REG_a" \n\t" |
| 90 | 90 |
| 91 static inline void sad8_2_mmx2(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h) | 91 static inline void sad8_2_mmx2(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h) |
| 92 { | 92 { |
| 93 long len= -(stride*h); | 93 long len= -(stride*h); |
| 94 asm volatile( | 94 asm volatile( |
| 95 ".balign 16 \n\t" | 95 ASMALIGN(4) |
| 96 "1: \n\t" | 96 "1: \n\t" |
| 97 "movq (%1, %%"REG_a"), %%mm0 \n\t" | 97 "movq (%1, %%"REG_a"), %%mm0 \n\t" |
| 98 "movq (%2, %%"REG_a"), %%mm2 \n\t" | 98 "movq (%2, %%"REG_a"), %%mm2 \n\t" |
| 99 "pavgb %%mm2, %%mm0 \n\t" | 99 "pavgb %%mm2, %%mm0 \n\t" |
| 100 "movq (%3, %%"REG_a"), %%mm2 \n\t" | 100 "movq (%3, %%"REG_a"), %%mm2 \n\t" |
| 116 | 116 |
| 117 static inline void sad8_4_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h) | 117 static inline void sad8_4_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
| 118 { //FIXME reuse src | 118 { //FIXME reuse src |
| 119 long len= -(stride*h); | 119 long len= -(stride*h); |
| 120 asm volatile( | 120 asm volatile( |
| 121 ".balign 16 \n\t" | 121 ASMALIGN(4) |
| 122 "movq "MANGLE(bone)", %%mm5 \n\t" | 122 "movq "MANGLE(bone)", %%mm5 \n\t" |
| 123 "1: \n\t" | 123 "1: \n\t" |
| 124 "movq (%1, %%"REG_a"), %%mm0 \n\t" | 124 "movq (%1, %%"REG_a"), %%mm0 \n\t" |
| 125 "movq (%2, %%"REG_a"), %%mm2 \n\t" | 125 "movq (%2, %%"REG_a"), %%mm2 \n\t" |
| 126 "movq 1(%1, %%"REG_a"), %%mm1 \n\t" | 126 "movq 1(%1, %%"REG_a"), %%mm1 \n\t" |
| 153 | 153 |
| 154 static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h) | 154 static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h) |
| 155 { | 155 { |
| 156 long len= -(stride*h); | 156 long len= -(stride*h); |
| 157 asm volatile( | 157 asm volatile( |
| 158 ".balign 16 \n\t" | 158 ASMALIGN(4) |
| 159 "1: \n\t" | 159 "1: \n\t" |
| 160 "movq (%1, %%"REG_a"), %%mm0 \n\t" | 160 "movq (%1, %%"REG_a"), %%mm0 \n\t" |
| 161 "movq (%2, %%"REG_a"), %%mm1 \n\t" | 161 "movq (%2, %%"REG_a"), %%mm1 \n\t" |
| 162 "movq (%1, %%"REG_a"), %%mm2 \n\t" | 162 "movq (%1, %%"REG_a"), %%mm2 \n\t" |
| 163 "movq (%2, %%"REG_a"), %%mm3 \n\t" | 163 "movq (%2, %%"REG_a"), %%mm3 \n\t" |
| 191 | 191 |
| 192 static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) | 192 static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) |
| 193 { | 193 { |
| 194 long len= -(stride*h); | 194 long len= -(stride*h); |
| 195 asm volatile( | 195 asm volatile( |
| 196 ".balign 16 \n\t" | 196 ASMALIGN(4) |
| 197 "1: \n\t" | 197 "1: \n\t" |
| 198 "movq (%1, %%"REG_a"), %%mm0 \n\t" | 198 "movq (%1, %%"REG_a"), %%mm0 \n\t" |
| 199 "movq (%2, %%"REG_a"), %%mm1 \n\t" | 199 "movq (%2, %%"REG_a"), %%mm1 \n\t" |
| 200 "movq %%mm0, %%mm4 \n\t" | 200 "movq %%mm0, %%mm4 \n\t" |
| 201 "movq %%mm1, %%mm2 \n\t" | 201 "movq %%mm1, %%mm2 \n\t" |