summaryrefslogtreecommitdiff
path: root/contrib/ffmpeg/libavcodec/i386/motion_est_mmx.c
diff options
context:
space:
mode:
Diffstat (limited to 'contrib/ffmpeg/libavcodec/i386/motion_est_mmx.c')
-rw-r--r--contrib/ffmpeg/libavcodec/i386/motion_est_mmx.c270
1 files changed, 161 insertions, 109 deletions
diff --git a/contrib/ffmpeg/libavcodec/i386/motion_est_mmx.c b/contrib/ffmpeg/libavcodec/i386/motion_est_mmx.c
index e33870e0f..888d891d9 100644
--- a/contrib/ffmpeg/libavcodec/i386/motion_est_mmx.c
+++ b/contrib/ffmpeg/libavcodec/i386/motion_est_mmx.c
@@ -3,6 +3,8 @@
* Copyright (c) 2001 Fabrice Bellard.
* Copyright (c) 2002-2004 Michael Niedermayer
*
+ * mostly by Michael Niedermayer <michaelni@gmx.at>
+ *
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
@@ -18,19 +20,17 @@
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- *
- * mostly by Michael Niedermayer <michaelni@gmx.at>
*/
-#include "../dsputil.h"
+#include "dsputil.h"
#include "x86_cpu.h"
-static const __attribute__ ((aligned(8))) uint64_t round_tab[3]={
+DECLARE_ASM_CONST(8, uint64_t, round_tab[3])={
0x0000000000000000ULL,
0x0001000100010001ULL,
0x0002000200020002ULL,
};
-static attribute_used __attribute__ ((aligned(8))) uint64_t bone= 0x0101010101010101LL;
+DECLARE_ASM_CONST(8, uint64_t, bone)= 0x0101010101010101LL;
static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
{
@@ -70,86 +70,127 @@ static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
static inline void sad8_1_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
{
- long len= -(stride*h);
asm volatile(
ASMALIGN(4)
"1: \n\t"
- "movq (%1, %%"REG_a"), %%mm0 \n\t"
- "movq (%2, %%"REG_a"), %%mm2 \n\t"
- "psadbw %%mm2, %%mm0 \n\t"
- "add %3, %%"REG_a" \n\t"
- "movq (%1, %%"REG_a"), %%mm1 \n\t"
- "movq (%2, %%"REG_a"), %%mm3 \n\t"
- "psadbw %%mm1, %%mm3 \n\t"
- "paddw %%mm3, %%mm0 \n\t"
+ "movq (%1), %%mm0 \n\t"
+ "movq (%1, %3), %%mm1 \n\t"
+ "psadbw (%2), %%mm0 \n\t"
+ "psadbw (%2, %3), %%mm1 \n\t"
"paddw %%mm0, %%mm6 \n\t"
- "add %3, %%"REG_a" \n\t"
- " js 1b \n\t"
- : "+a" (len)
- : "r" (blk1 - len), "r" (blk2 - len), "r" ((long)stride)
+ "paddw %%mm1, %%mm6 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((long)stride)
);
}
-static inline void sad8_2_mmx2(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h)
+static int sad16_sse2(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)
{
- long len= -(stride*h);
+ int ret;
asm volatile(
+ "pxor %%xmm6, %%xmm6 \n\t"
ASMALIGN(4)
"1: \n\t"
- "movq (%1, %%"REG_a"), %%mm0 \n\t"
- "movq (%2, %%"REG_a"), %%mm2 \n\t"
- "pavgb %%mm2, %%mm0 \n\t"
- "movq (%3, %%"REG_a"), %%mm2 \n\t"
- "psadbw %%mm2, %%mm0 \n\t"
- "add %4, %%"REG_a" \n\t"
- "movq (%1, %%"REG_a"), %%mm1 \n\t"
- "movq (%2, %%"REG_a"), %%mm3 \n\t"
- "pavgb %%mm1, %%mm3 \n\t"
- "movq (%3, %%"REG_a"), %%mm1 \n\t"
- "psadbw %%mm1, %%mm3 \n\t"
- "paddw %%mm3, %%mm0 \n\t"
+ "movdqu (%1), %%xmm0 \n\t"
+ "movdqu (%1, %3), %%xmm1 \n\t"
+ "psadbw (%2), %%xmm0 \n\t"
+ "psadbw (%2, %3), %%xmm1 \n\t"
+ "paddw %%xmm0, %%xmm6 \n\t"
+ "paddw %%xmm1, %%xmm6 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((long)stride)
+ );
+ asm volatile(
+ "movhlps %%xmm6, %%xmm0 \n\t"
+ "paddw %%xmm0, %%xmm6 \n\t"
+ "movd %%xmm6, %0 \n\t"
+ : "=r"(ret)
+ );
+ return ret;
+}
+
+static inline void sad8_x2a_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
+ asm volatile(
+ ASMALIGN(4)
+ "1: \n\t"
+ "movq (%1), %%mm0 \n\t"
+ "movq (%1, %3), %%mm1 \n\t"
+ "pavgb 1(%1), %%mm0 \n\t"
+ "pavgb 1(%1, %3), %%mm1 \n\t"
+ "psadbw (%2), %%mm0 \n\t"
+ "psadbw (%2, %3), %%mm1 \n\t"
"paddw %%mm0, %%mm6 \n\t"
- "add %4, %%"REG_a" \n\t"
- " js 1b \n\t"
- : "+a" (len)
- : "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" ((long)stride)
+ "paddw %%mm1, %%mm6 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((long)stride)
);
}
-static inline void sad8_4_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
-{ //FIXME reuse src
- long len= -(stride*h);
+static inline void sad8_y2a_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
asm volatile(
+ "movq (%1), %%mm0 \n\t"
+ "add %3, %1 \n\t"
ASMALIGN(4)
+ "1: \n\t"
+ "movq (%1), %%mm1 \n\t"
+ "movq (%1, %3), %%mm2 \n\t"
+ "pavgb %%mm1, %%mm0 \n\t"
+ "pavgb %%mm2, %%mm1 \n\t"
+ "psadbw (%2), %%mm0 \n\t"
+ "psadbw (%2, %3), %%mm1 \n\t"
+ "paddw %%mm0, %%mm6 \n\t"
+ "paddw %%mm1, %%mm6 \n\t"
+ "movq %%mm2, %%mm0 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((long)stride)
+ );
+}
+
+static inline void sad8_4_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
+ asm volatile(
"movq "MANGLE(bone)", %%mm5 \n\t"
+ "movq (%1), %%mm0 \n\t"
+ "pavgb 1(%1), %%mm0 \n\t"
+ "add %3, %1 \n\t"
+ ASMALIGN(4)
"1: \n\t"
- "movq (%1, %%"REG_a"), %%mm0 \n\t"
- "movq (%2, %%"REG_a"), %%mm2 \n\t"
- "movq 1(%1, %%"REG_a"), %%mm1 \n\t"
- "movq 1(%2, %%"REG_a"), %%mm3 \n\t"
- "pavgb %%mm2, %%mm0 \n\t"
- "pavgb %%mm1, %%mm3 \n\t"
- "psubusb %%mm5, %%mm3 \n\t"
- "pavgb %%mm3, %%mm0 \n\t"
- "movq (%3, %%"REG_a"), %%mm2 \n\t"
- "psadbw %%mm2, %%mm0 \n\t"
- "add %4, %%"REG_a" \n\t"
- "movq (%1, %%"REG_a"), %%mm1 \n\t"
- "movq (%2, %%"REG_a"), %%mm3 \n\t"
- "movq 1(%1, %%"REG_a"), %%mm2 \n\t"
- "movq 1(%2, %%"REG_a"), %%mm4 \n\t"
- "pavgb %%mm3, %%mm1 \n\t"
- "pavgb %%mm4, %%mm2 \n\t"
- "psubusb %%mm5, %%mm2 \n\t"
- "pavgb %%mm1, %%mm2 \n\t"
- "movq (%3, %%"REG_a"), %%mm1 \n\t"
- "psadbw %%mm1, %%mm2 \n\t"
- "paddw %%mm2, %%mm0 \n\t"
+ "movq (%1), %%mm1 \n\t"
+ "movq (%1,%3), %%mm2 \n\t"
+ "pavgb 1(%1), %%mm1 \n\t"
+ "pavgb 1(%1,%3), %%mm2 \n\t"
+ "psubusb %%mm5, %%mm1 \n\t"
+ "pavgb %%mm1, %%mm0 \n\t"
+ "pavgb %%mm2, %%mm1 \n\t"
+ "psadbw (%2), %%mm0 \n\t"
+ "psadbw (%2,%3), %%mm1 \n\t"
"paddw %%mm0, %%mm6 \n\t"
- "add %4, %%"REG_a" \n\t"
- " js 1b \n\t"
- : "+a" (len)
- : "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), "r" ((long)stride)
+ "paddw %%mm1, %%mm6 \n\t"
+ "movq %%mm2, %%mm0 \n\t"
+ "lea (%1,%3,2), %1 \n\t"
+ "lea (%2,%3,2), %2 \n\t"
+ "sub $2, %0 \n\t"
+ " jg 1b \n\t"
+ : "+r" (h), "+r" (blk1), "+r" (blk2)
+ : "r" ((long)stride)
);
}
@@ -195,45 +236,48 @@ static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
{
long len= -(stride*h);
asm volatile(
- ASMALIGN(4)
- "1: \n\t"
"movq (%1, %%"REG_a"), %%mm0 \n\t"
- "movq (%2, %%"REG_a"), %%mm1 \n\t"
- "movq %%mm0, %%mm4 \n\t"
- "movq %%mm1, %%mm2 \n\t"
- "punpcklbw %%mm7, %%mm0 \n\t"
- "punpcklbw %%mm7, %%mm1 \n\t"
- "punpckhbw %%mm7, %%mm4 \n\t"
- "punpckhbw %%mm7, %%mm2 \n\t"
- "paddw %%mm1, %%mm0 \n\t"
- "paddw %%mm2, %%mm4 \n\t"
"movq 1(%1, %%"REG_a"), %%mm2 \n\t"
- "movq 1(%2, %%"REG_a"), %%mm3 \n\t"
- "movq %%mm2, %%mm1 \n\t"
- "punpcklbw %%mm7, %%mm2 \n\t"
+ "movq %%mm0, %%mm1 \n\t"
+ "movq %%mm2, %%mm3 \n\t"
+ "punpcklbw %%mm7, %%mm0 \n\t"
"punpckhbw %%mm7, %%mm1 \n\t"
- "paddw %%mm0, %%mm2 \n\t"
- "paddw %%mm4, %%mm1 \n\t"
- "movq %%mm3, %%mm4 \n\t"
- "punpcklbw %%mm7, %%mm3 \n\t"
- "punpckhbw %%mm7, %%mm4 \n\t"
- "paddw %%mm3, %%mm2 \n\t"
- "paddw %%mm4, %%mm1 \n\t"
- "movq (%3, %%"REG_a"), %%mm3 \n\t"
- "movq (%3, %%"REG_a"), %%mm4 \n\t"
- "paddw %%mm5, %%mm2 \n\t"
+ "punpcklbw %%mm7, %%mm2 \n\t"
+ "punpckhbw %%mm7, %%mm3 \n\t"
+ "paddw %%mm2, %%mm0 \n\t"
+ "paddw %%mm3, %%mm1 \n\t"
+ ASMALIGN(4)
+ "1: \n\t"
+ "movq (%2, %%"REG_a"), %%mm2 \n\t"
+ "movq 1(%2, %%"REG_a"), %%mm4 \n\t"
+ "movq %%mm2, %%mm3 \n\t"
+ "movq %%mm4, %%mm5 \n\t"
+ "punpcklbw %%mm7, %%mm2 \n\t"
+ "punpckhbw %%mm7, %%mm3 \n\t"
+ "punpcklbw %%mm7, %%mm4 \n\t"
+ "punpckhbw %%mm7, %%mm5 \n\t"
+ "paddw %%mm4, %%mm2 \n\t"
+ "paddw %%mm5, %%mm3 \n\t"
+ "movq 16+"MANGLE(round_tab)", %%mm5 \n\t"
+ "paddw %%mm2, %%mm0 \n\t"
+ "paddw %%mm3, %%mm1 \n\t"
+ "paddw %%mm5, %%mm0 \n\t"
"paddw %%mm5, %%mm1 \n\t"
- "psrlw $2, %%mm2 \n\t"
+ "movq (%3, %%"REG_a"), %%mm4 \n\t"
+ "movq (%3, %%"REG_a"), %%mm5 \n\t"
+ "psrlw $2, %%mm0 \n\t"
"psrlw $2, %%mm1 \n\t"
- "packuswb %%mm1, %%mm2 \n\t"
- "psubusb %%mm2, %%mm3 \n\t"
- "psubusb %%mm4, %%mm2 \n\t"
- "por %%mm3, %%mm2 \n\t"
- "movq %%mm2, %%mm0 \n\t"
+ "packuswb %%mm1, %%mm0 \n\t"
+ "psubusb %%mm0, %%mm4 \n\t"
+ "psubusb %%mm5, %%mm0 \n\t"
+ "por %%mm4, %%mm0 \n\t"
+ "movq %%mm0, %%mm4 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
- "punpckhbw %%mm7, %%mm2 \n\t"
- "paddw %%mm2, %%mm0 \n\t"
+ "punpckhbw %%mm7, %%mm4 \n\t"
"paddw %%mm0, %%mm6 \n\t"
+ "paddw %%mm4, %%mm6 \n\t"
+ "movq %%mm2, %%mm0 \n\t"
+ "movq %%mm3, %%mm1 \n\t"
"add %4, %%"REG_a" \n\t"
" js 1b \n\t"
: "+a" (len)
@@ -267,6 +311,15 @@ static inline int sum_mmx2(void)
return ret;
}
+static inline void sad8_x2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
+ sad8_2_mmx(blk1, blk1+1, blk2, stride, h);
+}
+static inline void sad8_y2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
+{
+ sad8_2_mmx(blk1, blk1+stride, blk2, stride, h);
+}
+
#define PIX_SAD(suf)\
static int sad8_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
@@ -288,7 +341,7 @@ static int sad8_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, in
:: "m"(round_tab[1]) \
);\
\
- sad8_2_ ## suf(blk1, blk1+1, blk2, stride, 8);\
+ sad8_x2a_ ## suf(blk1, blk2, stride, 8);\
\
return sum_ ## suf();\
}\
@@ -302,7 +355,7 @@ static int sad8_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, in
:: "m"(round_tab[1]) \
);\
\
- sad8_2_ ## suf(blk1, blk1+stride, blk2, stride, 8);\
+ sad8_y2a_ ## suf(blk1, blk2, stride, 8);\
\
return sum_ ## suf();\
}\
@@ -312,9 +365,7 @@ static int sad8_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, i
assert(h==8);\
asm volatile("pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
- "movq %0, %%mm5 \n\t"\
- :: "m"(round_tab[2]) \
- );\
+ ::);\
\
sad8_4_ ## suf(blk1, blk2, stride, 8);\
\
@@ -339,8 +390,8 @@ static int sad16_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, i
:: "m"(round_tab[1]) \
);\
\
- sad8_2_ ## suf(blk1 , blk1+1, blk2 , stride, h);\
- sad8_2_ ## suf(blk1+8, blk1+9, blk2+8, stride, h);\
+ sad8_x2a_ ## suf(blk1 , blk2 , stride, h);\
+ sad8_x2a_ ## suf(blk1+8, blk2+8, stride, h);\
\
return sum_ ## suf();\
}\
@@ -352,8 +403,8 @@ static int sad16_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, i
:: "m"(round_tab[1]) \
);\
\
- sad8_2_ ## suf(blk1 , blk1+stride, blk2 , stride, h);\
- sad8_2_ ## suf(blk1+8, blk1+stride+8,blk2+8, stride, h);\
+ sad8_y2a_ ## suf(blk1 , blk2 , stride, h);\
+ sad8_y2a_ ## suf(blk1+8, blk2+8, stride, h);\
\
return sum_ ## suf();\
}\
@@ -361,9 +412,7 @@ static int sad16_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride,
{\
asm volatile("pxor %%mm7, %%mm7 \n\t"\
"pxor %%mm6, %%mm6 \n\t"\
- "movq %0, %%mm5 \n\t"\
- :: "m"(round_tab[2]) \
- );\
+ ::);\
\
sad8_4_ ## suf(blk1 , blk2 , stride, h);\
sad8_4_ ## suf(blk1+8, blk2+8, stride, h);\
@@ -405,4 +454,7 @@ void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx)
c->pix_abs[1][3] = sad8_xy2_mmx2;
}
}
+ if ((mm_flags & MM_SSE2) && !(mm_flags & MM_3DNOW)) {
+ c->sad[0]= sad16_sse2;
+ }
}
generated by cgit v1.2.3 (git 2.25.1) at 2025-04-27 14:36:52 +0000