.align 16

Originally committed as revision 2800 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc

diff --git a/postproc/rgb2rgb.c b/postproc/rgb2rgb.c
index 4dc9857..2232343 100644 (file)
--- a/postproc/rgb2rgb.c
+++ b/postproc/rgb2rgb.c
@@ -585,6 +585,7 @@ void rgb32tobgr32(const uint8_t *src, uint8_t *dst, unsigned int src_size)
 #ifdef HAVE_MMX
        asm volatile (
                "xorl %%eax, %%eax              \n\t"
+               ".align 16                      \n\t"
                "1:                             \n\t"
                PREFETCH" 32(%0, %%eax)         \n\t"
                "movq (%0, %%eax), %%mm0        \n\t"
@@ -635,6 +636,7 @@ void yv12toyuy2(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, u
 //FIXME handle 2 lines a once (fewer prefetch, reuse some chrom, but very likely limited by mem anyway)
                asm volatile(
                        "xorl %%eax, %%eax              \n\t"
+                       ".align 16                      \n\t"
                        "1:                             \n\t"
                        PREFETCH" 32(%1, %%eax, 2)      \n\t"
                        PREFETCH" 32(%2, %%eax)         \n\t"
@@ -708,6 +710,7 @@ void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
                        "xorl %%eax, %%eax              \n\t"
                        "pcmpeqw %%mm7, %%mm7           \n\t"
                        "psrlw $8, %%mm7                \n\t" // FF,00,FF,00...
+                       ".align 16                      \n\t"
                        "1:                             \n\t"
                        PREFETCH" 64(%0, %%eax, 4)      \n\t"
                        "movq (%0, %%eax, 4), %%mm0     \n\t" // YUYV YUYV(0)
@@ -757,6 +760,7 @@ void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 
                asm volatile(
                        "xorl %%eax, %%eax              \n\t"
+                       ".align 16                      \n\t"
                        "1:                             \n\t"
                        PREFETCH" 64(%0, %%eax, 4)      \n\t"
                        "movq (%0, %%eax, 4), %%mm0     \n\t" // YUYV YUYV(0)

diff --git a/postproc/rgb2rgb_template.c b/postproc/rgb2rgb_template.c
index 4dc9857..2232343 100644 (file)
--- a/postproc/rgb2rgb_template.c
+++ b/postproc/rgb2rgb_template.c
@@ -585,6 +585,7 @@ void rgb32tobgr32(const uint8_t *src, uint8_t *dst, unsigned int src_size)
 #ifdef HAVE_MMX
        asm volatile (
                "xorl %%eax, %%eax              \n\t"
+               ".align 16                      \n\t"
                "1:                             \n\t"
                PREFETCH" 32(%0, %%eax)         \n\t"
                "movq (%0, %%eax), %%mm0        \n\t"
@@ -635,6 +636,7 @@ void yv12toyuy2(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, u
 //FIXME handle 2 lines a once (fewer prefetch, reuse some chrom, but very likely limited by mem anyway)
                asm volatile(
                        "xorl %%eax, %%eax              \n\t"
+                       ".align 16                      \n\t"
                        "1:                             \n\t"
                        PREFETCH" 32(%1, %%eax, 2)      \n\t"
                        PREFETCH" 32(%2, %%eax)         \n\t"
@@ -708,6 +710,7 @@ void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
                        "xorl %%eax, %%eax              \n\t"
                        "pcmpeqw %%mm7, %%mm7           \n\t"
                        "psrlw $8, %%mm7                \n\t" // FF,00,FF,00...
+                       ".align 16                      \n\t"
                        "1:                             \n\t"
                        PREFETCH" 64(%0, %%eax, 4)      \n\t"
                        "movq (%0, %%eax, 4), %%mm0     \n\t" // YUYV YUYV(0)
@@ -757,6 +760,7 @@ void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 
                asm volatile(
                        "xorl %%eax, %%eax              \n\t"
+                       ".align 16                      \n\t"
                        "1:                             \n\t"
                        PREFETCH" 64(%0, %%eax, 4)      \n\t"
                        "movq (%0, %%eax, 4), %%mm0     \n\t" // YUYV YUYV(0)

diff --git a/postproc/swscale.c b/postproc/swscale.c
index f40eccf..103c47d 100644 (file)
--- a/postproc/swscale.c
+++ b/postproc/swscale.c
@@ -143,6 +143,7 @@ static int canMMX2BeUsed=0;
                "punpcklwd %%mm5, %%mm5         \n\t"\
                "punpcklwd %%mm5, %%mm5         \n\t"\
                "xorl %%eax, %%eax              \n\t"\
+               ".align 16                      \n\t"\
                "1:                             \n\t"\
                "movq (%0, %%eax, 2), %%mm0     \n\t" /*buf0[eax]*/\
                "movq (%1, %%eax, 2), %%mm1     \n\t" /*buf1[eax]*/\
@@ -196,6 +197,7 @@ static int canMMX2BeUsed=0;
                "punpcklwd %%mm5, %%mm5         \n\t"\
                "movq %%mm5, asm_uvalpha1       \n\t"\
                "xorl %%eax, %%eax              \n\t"\
+               ".align 16                      \n\t"\
                "1:                             \n\t"\
                "movq (%2, %%eax), %%mm2        \n\t" /* uvbuf0[eax]*/\
                "movq (%3, %%eax), %%mm3        \n\t" /* uvbuf1[eax]*/\
@@ -260,6 +262,7 @@ static int canMMX2BeUsed=0;
 
 #define YSCALEYUV2RGB1 \
                "xorl %%eax, %%eax              \n\t"\
+               ".align 16                      \n\t"\
                "1:                             \n\t"\
                "movq (%2, %%eax), %%mm3        \n\t" /* uvbuf0[eax]*/\
                "movq 4096(%2, %%eax), %%mm4    \n\t" /* uvbuf0[eax+2048]*/\
@@ -308,6 +311,7 @@ static int canMMX2BeUsed=0;
 // do vertical chrominance interpolation
 #define YSCALEYUV2RGB1b \
                "xorl %%eax, %%eax              \n\t"\
+               ".align 16                      \n\t"\
                "1:                             \n\t"\
                "movq (%2, %%eax), %%mm2        \n\t" /* uvbuf0[eax]*/\
                "movq (%3, %%eax), %%mm3        \n\t" /* uvbuf1[eax]*/\
@@ -1306,6 +1310,7 @@ FUNNY_Y_CODE
                "xorl %%eax, %%eax              \n\t" // i
                "xorl %%ebx, %%ebx              \n\t" // xx
                "xorl %%ecx, %%ecx              \n\t" // 2*xalpha
+               ".align 16                      \n\t"
                "1:                             \n\t"
                "movzbl  (%0, %%ebx), %%edi     \n\t" //src[xx]
                "movzbl 1(%0, %%ebx), %%esi     \n\t" //src[xx+1]
@@ -1437,6 +1442,7 @@ FUNNYUVCODE
                "xorl %%eax, %%eax              \n\t" // i
                "xorl %%ebx, %%ebx              \n\t" // xx
                "xorl %%ecx, %%ecx              \n\t" // 2*xalpha
+               ".align 16                      \n\t"
                "1:                             \n\t"
                "movl %0, %%esi                 \n\t"
                "movzbl  (%%esi, %%ebx), %%edi  \n\t" //src[xx]

diff --git a/postproc/swscale_template.c b/postproc/swscale_template.c
index f40eccf..103c47d 100644 (file)
--- a/postproc/swscale_template.c
+++ b/postproc/swscale_template.c
@@ -143,6 +143,7 @@ static int canMMX2BeUsed=0;
                "punpcklwd %%mm5, %%mm5         \n\t"\
                "punpcklwd %%mm5, %%mm5         \n\t"\
                "xorl %%eax, %%eax              \n\t"\
+               ".align 16                      \n\t"\
                "1:                             \n\t"\
                "movq (%0, %%eax, 2), %%mm0     \n\t" /*buf0[eax]*/\
                "movq (%1, %%eax, 2), %%mm1     \n\t" /*buf1[eax]*/\
@@ -196,6 +197,7 @@ static int canMMX2BeUsed=0;
                "punpcklwd %%mm5, %%mm5         \n\t"\
                "movq %%mm5, asm_uvalpha1       \n\t"\
                "xorl %%eax, %%eax              \n\t"\
+               ".align 16                      \n\t"\
                "1:                             \n\t"\
                "movq (%2, %%eax), %%mm2        \n\t" /* uvbuf0[eax]*/\
                "movq (%3, %%eax), %%mm3        \n\t" /* uvbuf1[eax]*/\
@@ -260,6 +262,7 @@ static int canMMX2BeUsed=0;
 
 #define YSCALEYUV2RGB1 \
                "xorl %%eax, %%eax              \n\t"\
+               ".align 16                      \n\t"\
                "1:                             \n\t"\
                "movq (%2, %%eax), %%mm3        \n\t" /* uvbuf0[eax]*/\
                "movq 4096(%2, %%eax), %%mm4    \n\t" /* uvbuf0[eax+2048]*/\
@@ -308,6 +311,7 @@ static int canMMX2BeUsed=0;
 // do vertical chrominance interpolation
 #define YSCALEYUV2RGB1b \
                "xorl %%eax, %%eax              \n\t"\
+               ".align 16                      \n\t"\
                "1:                             \n\t"\
                "movq (%2, %%eax), %%mm2        \n\t" /* uvbuf0[eax]*/\
                "movq (%3, %%eax), %%mm3        \n\t" /* uvbuf1[eax]*/\
@@ -1306,6 +1310,7 @@ FUNNY_Y_CODE
                "xorl %%eax, %%eax              \n\t" // i
                "xorl %%ebx, %%ebx              \n\t" // xx
                "xorl %%ecx, %%ecx              \n\t" // 2*xalpha
+               ".align 16                      \n\t"
                "1:                             \n\t"
                "movzbl  (%0, %%ebx), %%edi     \n\t" //src[xx]
                "movzbl 1(%0, %%ebx), %%esi     \n\t" //src[xx+1]
@@ -1437,6 +1442,7 @@ FUNNYUVCODE
                "xorl %%eax, %%eax              \n\t" // i
                "xorl %%ebx, %%ebx              \n\t" // xx
                "xorl %%ecx, %%ecx              \n\t" // 2*xalpha
+               ".align 16                      \n\t"
                "1:                             \n\t"
                "movl %0, %%esi                 \n\t"
                "movzbl  (%%esi, %%ebx), %%edi  \n\t" //src[xx]