[FFmpeg-cvslog] libpostproc: mmx code uses stack below %esp, fix that
Yuriy Kaminskiy
git at videolan.org
Tue Sep 20 21:38:38 CEST 2011
ffmpeg | branch: master | Yuriy Kaminskiy <yumkam at mail.ru> | Tue Sep 20 21:32:35 2011 +0200| [2b2617da6b26d575c4fd9a6b2fb0906ece861210] | committer: Michael Niedermayer
libpostproc: mmx code uses stack below %esp, fix that
Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=2b2617da6b26d575c4fd9a6b2fb0906ece861210
---
libpostproc/postprocess_template.c | 61 ++++++++++++++++-------------------
1 files changed, 28 insertions(+), 33 deletions(-)
diff --git a/libpostproc/postprocess_template.c b/libpostproc/postprocess_template.c
index 918d005..21a8626 100644
--- a/libpostproc/postprocess_template.c
+++ b/libpostproc/postprocess_template.c
@@ -25,8 +25,6 @@
#include "libavutil/x86_cpu.h"
-#define ALIGN_MASK "$-8"
-
#undef REAL_PAVGB
#undef PAVGB
#undef PMINUB
@@ -767,11 +765,10 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
}
*/
#elif HAVE_MMX
+ DECLARE_ALIGNED(8, uint64_t, tmp)[4]; // make space for 4 8-byte vars
src+= stride*4;
__asm__ volatile(
"pxor %%mm7, %%mm7 \n\t"
- "lea -40(%%"REG_SP"), %%"REG_c" \n\t" // make space for 4 8-byte vars
- "and "ALIGN_MASK", %%"REG_c" \n\t" // align
// 0 1 2 3 4 5 6 7
// %0 %0+%1 %0+2%1 eax+2%1 %0+4%1 eax+4%1 edx+%1 edx+2%1
// %0 eax eax+%1 eax+2%1 %0+4%1 edx edx+%1 edx+2%1
@@ -813,8 +810,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2 - H3
"psubw %%mm2, %%mm0 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
- "movq %%mm0, (%%"REG_c") \n\t" // 2L0 - 5L1 + 5L2 - 2L3
- "movq %%mm1, 8(%%"REG_c") \n\t" // 2H0 - 5H1 + 5H2 - 2H3
+ "movq %%mm0, (%3) \n\t" // 2L0 - 5L1 + 5L2 - 2L3
+ "movq %%mm1, 8(%3) \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq (%%"REG_a", %1, 2), %%mm0 \n\t"
"movq %%mm0, %%mm1 \n\t"
@@ -823,8 +820,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"psubw %%mm0, %%mm2 \n\t" // L3 - L4
"psubw %%mm1, %%mm3 \n\t" // H3 - H4
- "movq %%mm2, 16(%%"REG_c") \n\t" // L3 - L4
- "movq %%mm3, 24(%%"REG_c") \n\t" // H3 - H4
+ "movq %%mm2, 16(%3) \n\t" // L3 - L4
+ "movq %%mm3, 24(%3) \n\t" // H3 - H4
"paddw %%mm4, %%mm4 \n\t" // 2L2
"paddw %%mm5, %%mm5 \n\t" // 2H2
"psubw %%mm2, %%mm4 \n\t" // 2L2 - L3 + L4
@@ -872,8 +869,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"psubw %%mm2, %%mm0 \n\t" // 2L4 - 5L5 + 5L6 - 2L7
"psubw %%mm3, %%mm1 \n\t" // 2H4 - 5H5 + 5H6 - 2H7
- "movq (%%"REG_c"), %%mm2 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
- "movq 8(%%"REG_c"), %%mm3 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
+ "movq (%3), %%mm2 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
+ "movq 8(%3), %%mm3 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
#if HAVE_MMX2
"movq %%mm7, %%mm6 \n\t" // 0
@@ -951,8 +948,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"psrlw $6, %%mm4 \n\t"
"psrlw $6, %%mm5 \n\t"
- "movq 16(%%"REG_c"), %%mm0 \n\t" // L3 - L4
- "movq 24(%%"REG_c"), %%mm1 \n\t" // H3 - H4
+ "movq 16(%3), %%mm0 \n\t" // L3 - L4
+ "movq 24(%3), %%mm1 \n\t" // H3 - H4
"pxor %%mm2, %%mm2 \n\t"
"pxor %%mm3, %%mm3 \n\t"
@@ -995,8 +992,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
"movq %%mm0, (%0, %1) \n\t"
: "+r" (src)
- : "r" ((x86_reg)stride), "m" (c->pQPb)
- : "%"REG_a, "%"REG_c
+ : "r" ((x86_reg)stride), "m" (c->pQPb), "r"(tmp)
+ : "%"REG_a
);
#else //HAVE_MMX2 || HAVE_AMD3DNOW
const int l1= stride;
@@ -1044,6 +1041,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
static inline void RENAME(dering)(uint8_t src[], int stride, PPContext *c)
{
#if HAVE_MMX2 || HAVE_AMD3DNOW
+ DECLARE_ALIGNED(8, uint64_t, tmp)[3];
__asm__ volatile(
"pxor %%mm6, %%mm6 \n\t"
"pcmpeqb %%mm7, %%mm7 \n\t"
@@ -1137,13 +1135,11 @@ FIND_MIN_MAX((%0, %1, 8))
"movd %%mm6, %%ecx \n\t"
"cmpb "MANGLE(deringThreshold)", %%cl \n\t"
" jb 1f \n\t"
- "lea -24(%%"REG_SP"), %%"REG_c" \n\t"
- "and "ALIGN_MASK", %%"REG_c" \n\t"
PAVGB(%%mm0, %%mm7) // a=(max + min)/2
"punpcklbw %%mm7, %%mm7 \n\t"
"punpcklbw %%mm7, %%mm7 \n\t"
"punpcklbw %%mm7, %%mm7 \n\t"
- "movq %%mm7, (%%"REG_c") \n\t"
+ "movq %%mm7, (%4) \n\t"
"movq (%0), %%mm0 \n\t" // L10
"movq %%mm0, %%mm1 \n\t" // L10
@@ -1207,8 +1203,8 @@ FIND_MIN_MAX((%0, %1, 8))
PAVGB(t0, lx) /* (src[-1] + src[+1])/2 */\
PAVGB(sx, lx) /* (src[-1] + 2src[0] + src[+1])/4 */\
PAVGB(lx, pplx) \
- "movq " #lx ", 8(%%"REG_c") \n\t"\
- "movq (%%"REG_c"), " #lx " \n\t"\
+ "movq " #lx ", 8(%4) \n\t"\
+ "movq (%4), " #lx " \n\t"\
"psubusb " #lx ", " #t1 " \n\t"\
"psubusb " #lx ", " #t0 " \n\t"\
"psubusb " #lx ", " #sx " \n\t"\
@@ -1235,7 +1231,7 @@ FIND_MIN_MAX((%0, %1, 8))
"pandn " #dst ", " #ppsx " \n\t"\
"por " #pplx ", " #ppsx " \n\t"\
"movq " #ppsx ", " #dst " \n\t"\
- "movq 8(%%"REG_c"), " #lx " \n\t"
+ "movq 8(%4), " #lx " \n\t"
#define DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1) \
REAL_DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1)
@@ -1265,7 +1261,7 @@ DERING_CORE((%%REGd, %1, 2),(%0, %1, 8) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,
DERING_CORE((%0, %1, 8) ,(%%REGd, %1, 4),%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
"1: \n\t"
- : : "r" (src), "r" ((x86_reg)stride), "m" (c->pQPb), "m"(c->pQPb2)
+ : : "r" (src), "r" ((x86_reg)stride), "m" (c->pQPb), "m"(c->pQPb2), "r"(tmp)
: "%"REG_a, "%"REG_d, "%"REG_c
);
#else //HAVE_MMX2 || HAVE_AMD3DNOW
@@ -2762,10 +2758,9 @@ static av_always_inline void RENAME(do_a_deblock)(uint8_t *src, int step, int st
if(eq_mask != -1LL){
uint8_t *temp_src= src;
+ DECLARE_ALIGNED(8, uint64_t, tmp)[4]; // make space for 4 8-byte vars
__asm__ volatile(
"pxor %%mm7, %%mm7 \n\t"
- "lea -40(%%"REG_SP"), %%"REG_c" \n\t" // make space for 4 8-byte vars
- "and "ALIGN_MASK", %%"REG_c" \n\t" // align
// 0 1 2 3 4 5 6 7 8 9
// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1 %1+8%1 ecx+4%1
@@ -2806,8 +2801,8 @@ static av_always_inline void RENAME(do_a_deblock)(uint8_t *src, int step, int st
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2 - H3
"psubw %%mm2, %%mm0 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
"psubw %%mm3, %%mm1 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
- "movq %%mm0, (%%"REG_c") \n\t" // 2L0 - 5L1 + 5L2 - 2L3
- "movq %%mm1, 8(%%"REG_c") \n\t" // 2H0 - 5H1 + 5H2 - 2H3
+ "movq %%mm0, (%4) \n\t" // 2L0 - 5L1 + 5L2 - 2L3
+ "movq %%mm1, 8(%4) \n\t" // 2H0 - 5H1 + 5H2 - 2H3
"movq (%%"REG_a", %1, 2), %%mm0 \n\t"
"movq %%mm0, %%mm1 \n\t"
@@ -2816,8 +2811,8 @@ static av_always_inline void RENAME(do_a_deblock)(uint8_t *src, int step, int st
"psubw %%mm0, %%mm2 \n\t" // L3 - L4
"psubw %%mm1, %%mm3 \n\t" // H3 - H4
- "movq %%mm2, 16(%%"REG_c") \n\t" // L3 - L4
- "movq %%mm3, 24(%%"REG_c") \n\t" // H3 - H4
+ "movq %%mm2, 16(%4) \n\t" // L3 - L4
+ "movq %%mm3, 24(%4) \n\t" // H3 - H4
"paddw %%mm4, %%mm4 \n\t" // 2L2
"paddw %%mm5, %%mm5 \n\t" // 2H2
"psubw %%mm2, %%mm4 \n\t" // 2L2 - L3 + L4
@@ -2865,8 +2860,8 @@ static av_always_inline void RENAME(do_a_deblock)(uint8_t *src, int step, int st
"psubw %%mm2, %%mm0 \n\t" // 2L4 - 5L5 + 5L6 - 2L7
"psubw %%mm3, %%mm1 \n\t" // 2H4 - 5H5 + 5H6 - 2H7
- "movq (%%"REG_c"), %%mm2 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
- "movq 8(%%"REG_c"), %%mm3 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
+ "movq (%4), %%mm2 \n\t" // 2L0 - 5L1 + 5L2 - 2L3
+ "movq 8(%4), %%mm3 \n\t" // 2H0 - 5H1 + 5H2 - 2H3
#if HAVE_MMX2
"movq %%mm7, %%mm6 \n\t" // 0
@@ -2944,8 +2939,8 @@ static av_always_inline void RENAME(do_a_deblock)(uint8_t *src, int step, int st
"psrlw $6, %%mm4 \n\t"
"psrlw $6, %%mm5 \n\t"
- "movq 16(%%"REG_c"), %%mm0 \n\t" // L3 - L4
- "movq 24(%%"REG_c"), %%mm1 \n\t" // H3 - H4
+ "movq 16(%4), %%mm0 \n\t" // L3 - L4
+ "movq 24(%4), %%mm1 \n\t" // H3 - H4
"pxor %%mm2, %%mm2 \n\t"
"pxor %%mm3, %%mm3 \n\t"
@@ -2990,8 +2985,8 @@ static av_always_inline void RENAME(do_a_deblock)(uint8_t *src, int step, int st
"movq %%mm0, (%0, %1) \n\t"
: "+r" (temp_src)
- : "r" ((x86_reg)step), "m" (c->pQPb), "m"(eq_mask)
- : "%"REG_a, "%"REG_c
+ : "r" ((x86_reg)step), "m" (c->pQPb), "m"(eq_mask), "r"(tmp)
+ : "%"REG_a
);
}
/*if(step==16){
More information about the ffmpeg-cvslog
mailing list