[SCM] libav/experimental: and another +2% speedup
siretart at users.alioth.debian.org
siretart at users.alioth.debian.org
Sun Jun 30 15:30:15 UTC 2013
The following commit has been merged in the experimental branch:
commit ac0b0b2f6d2a7a528a04353e1d8e1f42166525c2
Author: Michael Niedermayer <michaelni at gmx.at>
Date: Tue Oct 23 12:05:34 2001 +0000
and another +2% speedup
Originally committed as revision 2417 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc
diff --git a/postproc/postprocess.c b/postproc/postprocess.c
index b32d3fe..5ed2477 100644
--- a/postproc/postprocess.c
+++ b/postproc/postprocess.c
@@ -336,7 +336,7 @@ static inline int isVertMinMaxOk(uint8_t src[], int stride, int QP)
: "=r" (isOk)
: "r" (src), "r" (stride)
);
- return isOk ? 1 : 0;
+ return isOk;
#else
int isOk2= 1;
@@ -1303,40 +1303,60 @@ static inline int isHorizDCAndCopy2Temp(uint8_t src[], int stride)
#ifdef HAVE_MMX
asm volatile (
// "int $3 \n\t"
- "pushl %1\n\t"
+ "leal (%1, %2), %%ecx \n\t"
+ "leal (%%ecx, %2, 4), %%ebx \n\t"
+// 0 1 2 3 4 5 6 7 8 9
+// %1 ecx ecx+%2 ecx+2%2 %1+4%2 ebx ebx+%2 ebx+2%2 %1+8%2 ebx+4%2
"movq b7E, %%mm7 \n\t" // mm7 = 0x7F
"movq b7C, %%mm6 \n\t" // mm6 = 0x7D
- "leal tempBlock, %%eax \n\t"
"pxor %%mm0, %%mm0 \n\t"
+ "movl %1, %%eax \n\t"
+ "andl $0x1F, %%eax \n\t"
+ "cmpl $24, %%eax \n\t"
+ "leal tempBlock, %%eax \n\t"
+ "jb 1f \n\t"
-#define HDC_CHECK_AND_CPY(i) \
- "movq -4(%1), %%mm2 \n\t"\
- "psrlq $32, %%mm2 \n\t"\
- "punpckldq 4(%1), %%mm2 \n\t" /* (%1) */\
+#define HDC_CHECK_AND_CPY(src, dst) \
+ "movd " #src ", %%mm2 \n\t"\
+ "punpckldq 4" #src ", %%mm2 \n\t" /* (%1) */\
"movq %%mm2, %%mm1 \n\t"\
"psrlq $8, %%mm2 \n\t"\
"psubb %%mm1, %%mm2 \n\t"\
"paddb %%mm7, %%mm2 \n\t"\
"pcmpgtb %%mm6, %%mm2 \n\t"\
"paddb %%mm2, %%mm0 \n\t"\
- "movq %%mm1," #i "(%%eax) \n\t"
-
- HDC_CHECK_AND_CPY(0)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(8)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(16)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(24)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(32)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(40)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(48)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(56)
-
+ "movq %%mm1," #dst "(%%eax) \n\t"
+
+ HDC_CHECK_AND_CPY((%1),0)
+ HDC_CHECK_AND_CPY((%%ecx),8)
+ HDC_CHECK_AND_CPY((%%ecx, %2),16)
+ HDC_CHECK_AND_CPY((%%ecx, %2, 2),24)
+ HDC_CHECK_AND_CPY((%1, %2, 4),32)
+ HDC_CHECK_AND_CPY((%%ebx),40)
+ HDC_CHECK_AND_CPY((%%ebx, %2),48)
+ HDC_CHECK_AND_CPY((%%ebx, %2, 2),56)
+ "jmp 2f \n\t"
+ "1: \n\t"
+// src does not cross a 32 byte cache line so dont waste time with alignment
+#define HDC_CHECK_AND_CPY2(src, dst) \
+ "movq " #src ", %%mm2 \n\t"\
+ "movq " #src ", %%mm1 \n\t"\
+ "psrlq $8, %%mm2 \n\t"\
+ "psubb %%mm1, %%mm2 \n\t"\
+ "paddb %%mm7, %%mm2 \n\t"\
+ "pcmpgtb %%mm6, %%mm2 \n\t"\
+ "paddb %%mm2, %%mm0 \n\t"\
+ "movq %%mm1," #dst "(%%eax) \n\t"
+
+ HDC_CHECK_AND_CPY2((%1),0)
+ HDC_CHECK_AND_CPY2((%%ecx),8)
+ HDC_CHECK_AND_CPY2((%%ecx, %2),16)
+ HDC_CHECK_AND_CPY2((%%ecx, %2, 2),24)
+ HDC_CHECK_AND_CPY2((%1, %2, 4),32)
+ HDC_CHECK_AND_CPY2((%%ebx),40)
+ HDC_CHECK_AND_CPY2((%%ebx, %2),48)
+ HDC_CHECK_AND_CPY2((%%ebx, %2, 2),56)
+ "2: \n\t"
"psllq $8, %%mm0 \n\t" // remove dummy value
"movq %%mm0, %%mm1 \n\t"
"psrlw $8, %%mm0 \n\t"
@@ -1347,14 +1367,13 @@ asm volatile (
"movq %%mm0, %%mm1 \n\t"
"psrlq $32, %%mm0 \n\t"
"paddb %%mm1, %%mm0 \n\t"
- "popl %1\n\t"
"movd %%mm0, %0 \n\t"
: "=r" (numEq)
: "r" (src), "r" (stride)
- : "%eax"
+ : "%eax", "%ebx", "%ecx"
);
// printf("%d\n", numEq);
- numEq= (256 - (numEq & 0xFF)) &0xFF;
+ numEq= (256 - numEq) &0xFF;
#else
int y;
for(y=0; y<BLOCK_SIZE; y++)
diff --git a/postproc/postprocess_template.c b/postproc/postprocess_template.c
index b32d3fe..5ed2477 100644
--- a/postproc/postprocess_template.c
+++ b/postproc/postprocess_template.c
@@ -336,7 +336,7 @@ static inline int isVertMinMaxOk(uint8_t src[], int stride, int QP)
: "=r" (isOk)
: "r" (src), "r" (stride)
);
- return isOk ? 1 : 0;
+ return isOk;
#else
int isOk2= 1;
@@ -1303,40 +1303,60 @@ static inline int isHorizDCAndCopy2Temp(uint8_t src[], int stride)
#ifdef HAVE_MMX
asm volatile (
// "int $3 \n\t"
- "pushl %1\n\t"
+ "leal (%1, %2), %%ecx \n\t"
+ "leal (%%ecx, %2, 4), %%ebx \n\t"
+// 0 1 2 3 4 5 6 7 8 9
+// %1 ecx ecx+%2 ecx+2%2 %1+4%2 ebx ebx+%2 ebx+2%2 %1+8%2 ebx+4%2
"movq b7E, %%mm7 \n\t" // mm7 = 0x7F
"movq b7C, %%mm6 \n\t" // mm6 = 0x7D
- "leal tempBlock, %%eax \n\t"
"pxor %%mm0, %%mm0 \n\t"
+ "movl %1, %%eax \n\t"
+ "andl $0x1F, %%eax \n\t"
+ "cmpl $24, %%eax \n\t"
+ "leal tempBlock, %%eax \n\t"
+ "jb 1f \n\t"
-#define HDC_CHECK_AND_CPY(i) \
- "movq -4(%1), %%mm2 \n\t"\
- "psrlq $32, %%mm2 \n\t"\
- "punpckldq 4(%1), %%mm2 \n\t" /* (%1) */\
+#define HDC_CHECK_AND_CPY(src, dst) \
+ "movd " #src ", %%mm2 \n\t"\
+ "punpckldq 4" #src ", %%mm2 \n\t" /* (%1) */\
"movq %%mm2, %%mm1 \n\t"\
"psrlq $8, %%mm2 \n\t"\
"psubb %%mm1, %%mm2 \n\t"\
"paddb %%mm7, %%mm2 \n\t"\
"pcmpgtb %%mm6, %%mm2 \n\t"\
"paddb %%mm2, %%mm0 \n\t"\
- "movq %%mm1," #i "(%%eax) \n\t"
-
- HDC_CHECK_AND_CPY(0)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(8)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(16)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(24)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(32)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(40)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(48)
- "addl %2, %1 \n\t"
- HDC_CHECK_AND_CPY(56)
-
+ "movq %%mm1," #dst "(%%eax) \n\t"
+
+ HDC_CHECK_AND_CPY((%1),0)
+ HDC_CHECK_AND_CPY((%%ecx),8)
+ HDC_CHECK_AND_CPY((%%ecx, %2),16)
+ HDC_CHECK_AND_CPY((%%ecx, %2, 2),24)
+ HDC_CHECK_AND_CPY((%1, %2, 4),32)
+ HDC_CHECK_AND_CPY((%%ebx),40)
+ HDC_CHECK_AND_CPY((%%ebx, %2),48)
+ HDC_CHECK_AND_CPY((%%ebx, %2, 2),56)
+ "jmp 2f \n\t"
+ "1: \n\t"
+// src does not cross a 32 byte cache line so dont waste time with alignment
+#define HDC_CHECK_AND_CPY2(src, dst) \
+ "movq " #src ", %%mm2 \n\t"\
+ "movq " #src ", %%mm1 \n\t"\
+ "psrlq $8, %%mm2 \n\t"\
+ "psubb %%mm1, %%mm2 \n\t"\
+ "paddb %%mm7, %%mm2 \n\t"\
+ "pcmpgtb %%mm6, %%mm2 \n\t"\
+ "paddb %%mm2, %%mm0 \n\t"\
+ "movq %%mm1," #dst "(%%eax) \n\t"
+
+ HDC_CHECK_AND_CPY2((%1),0)
+ HDC_CHECK_AND_CPY2((%%ecx),8)
+ HDC_CHECK_AND_CPY2((%%ecx, %2),16)
+ HDC_CHECK_AND_CPY2((%%ecx, %2, 2),24)
+ HDC_CHECK_AND_CPY2((%1, %2, 4),32)
+ HDC_CHECK_AND_CPY2((%%ebx),40)
+ HDC_CHECK_AND_CPY2((%%ebx, %2),48)
+ HDC_CHECK_AND_CPY2((%%ebx, %2, 2),56)
+ "2: \n\t"
"psllq $8, %%mm0 \n\t" // remove dummy value
"movq %%mm0, %%mm1 \n\t"
"psrlw $8, %%mm0 \n\t"
@@ -1347,14 +1367,13 @@ asm volatile (
"movq %%mm0, %%mm1 \n\t"
"psrlq $32, %%mm0 \n\t"
"paddb %%mm1, %%mm0 \n\t"
- "popl %1\n\t"
"movd %%mm0, %0 \n\t"
: "=r" (numEq)
: "r" (src), "r" (stride)
- : "%eax"
+ : "%eax", "%ebx", "%ecx"
);
// printf("%d\n", numEq);
- numEq= (256 - (numEq & 0xFF)) &0xFF;
+ numEq= (256 - numEq) &0xFF;
#else
int y;
for(y=0; y<BLOCK_SIZE; y++)
--
Libav/FFmpeg packaging
More information about the pkg-multimedia-commits
mailing list