diff options
Diffstat (limited to 'src/libffmpeg/libavcodec/i386')
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/cputest.c | 34 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c | 255 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/dsputil_mmx.c | 1924 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/dsputil_mmx_avg.h | 1288 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/dsputil_mmx_rnd.h | 828 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/fdct_mmx.c | 290 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/fft_sse.c | 26 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/h264dsp_mmx.c | 508 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/idct_mmx.c | 516 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c | 19 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/mmx.h | 434 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/motion_est_mmx.c | 384 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c | 922 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c | 392 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/simple_idct_mmx.c | 2076 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/vp3dsp_mmx.c | 4 | ||||
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/vp3dsp_sse2.c | 32 |
17 files changed, 5075 insertions, 4857 deletions
diff --git a/src/libffmpeg/libavcodec/i386/cputest.c b/src/libffmpeg/libavcodec/i386/cputest.c index 593e0550d..64656c65a 100644 --- a/src/libffmpeg/libavcodec/i386/cputest.c +++ b/src/libffmpeg/libavcodec/i386/cputest.c @@ -15,7 +15,7 @@ /* ebx saving is necessary for PIC. gcc seems unable to see it alone */ #define cpuid(index,eax,ebx,ecx,edx)\ __asm __volatile\ - ("mov %%"REG_b", %%"REG_S"\n\t"\ + ("mov %%"REG_b", %%"REG_S"\n\t"\ "cpuid\n\t"\ "xchg %%"REG_b", %%"REG_S\ : "=a" (eax), "=S" (ebx),\ @@ -29,28 +29,28 @@ int mm_support(void) int eax, ebx, ecx, edx; int max_std_level, max_ext_level, std_caps=0, ext_caps=0; long a, c; - + __asm__ __volatile__ ( /* See if CPUID instruction is supported ... */ /* ... Get copies of EFLAGS into eax and ecx */ "pushf\n\t" "pop %0\n\t" "mov %0, %1\n\t" - + /* ... Toggle the ID bit in one copy and store */ /* to the EFLAGS reg */ "xor $0x200000, %0\n\t" "push %0\n\t" "popf\n\t" - + /* ... Get the (hopefully modified) EFLAGS */ "pushf\n\t" "pop %0\n\t" : "=a" (a), "=c" (c) : - : "cc" + : "cc" ); - + if (a == c) return 0; /* CPUID not supported */ @@ -60,9 +60,9 @@ int mm_support(void) cpuid(1, eax, ebx, ecx, std_caps); if (std_caps & (1<<23)) rval |= MM_MMX; - if (std_caps & (1<<25)) + if (std_caps & (1<<25)) rval |= MM_MMXEXT | MM_SSE; - if (std_caps & (1<<26)) + if (std_caps & (1<<26)) rval |= MM_SSE2; } @@ -89,8 +89,8 @@ int mm_support(void) edx == 0x48727561 && ecx == 0x736c7561) { /* "CentaurHauls" */ /* VIA C3 */ - if(ext_caps & (1<<24)) - rval |= MM_MMXEXT; + if(ext_caps & (1<<24)) + rval |= MM_MMXEXT; } else if (ebx == 0x69727943 && edx == 0x736e4978 && ecx == 0x64616574) { @@ -103,18 +103,18 @@ int mm_support(void) According to the table, the only CPU which supports level 2 is also the only one which supports extended CPUID levels. */ - if (eax < 2) + if (eax < 2) return rval; if (ext_caps & (1<<24)) rval |= MM_MMXEXT; } #if 0 - av_log(NULL, AV_LOG_DEBUG, "%s%s%s%s%s%s\n", - (rval&MM_MMX) ? "MMX ":"", - (rval&MM_MMXEXT) ? "MMX2 ":"", - (rval&MM_SSE) ? "SSE ":"", - (rval&MM_SSE2) ? "SSE2 ":"", - (rval&MM_3DNOW) ? "3DNow ":"", + av_log(NULL, AV_LOG_DEBUG, "%s%s%s%s%s%s\n", + (rval&MM_MMX) ? "MMX ":"", + (rval&MM_MMXEXT) ? "MMX2 ":"", + (rval&MM_SSE) ? "SSE ":"", + (rval&MM_SSE2) ? "SSE2 ":"", + (rval&MM_3DNOW) ? "3DNow ":"", (rval&MM_3DNOWEXT) ? "3DNowExt ":""); #endif return rval; diff --git a/src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c b/src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c index 4cd4d52d8..d52938ccf 100644 --- a/src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c +++ b/src/libffmpeg/libavcodec/i386/dsputil_h264_template_mmx.c @@ -1,5 +1,6 @@ /* - * Copyright (c) 2005 Zoltan Hidvegi <hzoli -a- hzoli -d- com> + * Copyright (c) 2005 Zoltan Hidvegi <hzoli -a- hzoli -d- com>, + * Loren Merritt * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public @@ -13,27 +14,143 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ /** * MMX optimized version of (put|avg)_h264_chroma_mc8. - * H264_CHROMA_MC8_TMPL must be defined to the desired function name and - * H264_CHROMA_OP must be defined to empty for put and pavgb/pavgusb for avg. + * H264_CHROMA_MC8_TMPL must be defined to the desired function name + * H264_CHROMA_OP must be defined to empty for put and pavgb/pavgusb for avg + * H264_CHROMA_MC8_MV0 must be defined to a (put|avg)_pixels8 function */ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y) { - uint64_t AA __align8; - uint64_t DD __align8; - unsigned long srcos = (long)src & 7; - uint64_t sh1 __align8 = srcos * 8; - uint64_t sh2 __align8 = 56 - sh1; + DECLARE_ALIGNED_8(uint64_t, AA); + DECLARE_ALIGNED_8(uint64_t, DD); int i; + if(y==0 && x==0) { + /* no filter needed */ + H264_CHROMA_MC8_MV0(dst, src, stride, h); + return; + } + assert(x<8 && y<8 && x>=0 && y>=0); - asm volatile("movd %1, %%mm4\n\t" - "movd %2, %%mm6\n\t" + if(y==0) + { + /* horizontal filter only */ + asm volatile("movd %0, %%mm5\n\t" + "punpcklwd %%mm5, %%mm5\n\t" + "punpckldq %%mm5, %%mm5\n\t" /* mm5 = B = x */ + "movq %1, %%mm4\n\t" + "pxor %%mm7, %%mm7\n\t" + "psubw %%mm5, %%mm4\n\t" /* mm4 = A = 8-x */ + : : "rm" (x), "m" (ff_pw_8)); + + for(i=0; i<h; i++) { + asm volatile( + /* mm0 = src[0..7], mm1 = src[1..8] */ + "movq %0, %%mm0\n\t" + "movq %1, %%mm1\n\t" + : : "m" (src[0]), "m" (src[1])); + + asm volatile( + /* [mm2,mm3] = A * src[0..7] */ + "movq %%mm0, %%mm2\n\t" + "punpcklbw %%mm7, %%mm2\n\t" + "pmullw %%mm4, %%mm2\n\t" + "movq %%mm0, %%mm3\n\t" + "punpckhbw %%mm7, %%mm3\n\t" + "pmullw %%mm4, %%mm3\n\t" + + /* [mm2,mm3] += B * src[1..8] */ + "movq %%mm1, %%mm0\n\t" + "punpcklbw %%mm7, %%mm0\n\t" + "pmullw %%mm5, %%mm0\n\t" + "punpckhbw %%mm7, %%mm1\n\t" + "pmullw %%mm5, %%mm1\n\t" + "paddw %%mm0, %%mm2\n\t" + "paddw %%mm1, %%mm3\n\t" + + /* dst[0..7] = pack(([mm2,mm3] + 32) >> 6) */ + "paddw %1, %%mm2\n\t" + "paddw %1, %%mm3\n\t" + "psrlw $3, %%mm2\n\t" + "psrlw $3, %%mm3\n\t" + "packuswb %%mm3, %%mm2\n\t" + H264_CHROMA_OP(%0, %%mm2) + "movq %%mm2, %0\n\t" + : "=m" (dst[0]) : "m" (ff_pw_4)); + + src += stride; + dst += stride; + } + return; + } + + if(x==0) + { + /* vertical filter only */ + asm volatile("movd %0, %%mm6\n\t" + "punpcklwd %%mm6, %%mm6\n\t" + "punpckldq %%mm6, %%mm6\n\t" /* mm6 = C = y */ + "movq %1, %%mm4\n\t" + "pxor %%mm7, %%mm7\n\t" + "psubw %%mm6, %%mm4\n\t" /* mm4 = A = 8-y */ + : : "rm" (y), "m" (ff_pw_8)); + + asm volatile( + /* mm0 = src[0..7] */ + "movq %0, %%mm0\n\t" + : : "m" (src[0])); + + for(i=0; i<h; i++) { + asm volatile( + /* [mm2,mm3] = A * src[0..7] */ + "movq %mm0, %mm2\n\t" + "punpcklbw %mm7, %mm2\n\t" + "pmullw %mm4, %mm2\n\t" + "movq %mm0, %mm3\n\t" + "punpckhbw %mm7, %mm3\n\t" + "pmullw %mm4, %mm3\n\t"); + + src += stride; + asm volatile( + /* mm0 = src[0..7] */ + "movq %0, %%mm0\n\t" + : : "m" (src[0])); + + asm volatile( + /* [mm2,mm3] += C * src[0..7] */ + "movq %mm0, %mm1\n\t" + "punpcklbw %mm7, %mm1\n\t" + "pmullw %mm6, %mm1\n\t" + "paddw %mm1, %mm2\n\t" + "movq %mm0, %mm5\n\t" + "punpckhbw %mm7, %mm5\n\t" + "pmullw %mm6, %mm5\n\t" + "paddw %mm5, %mm3\n\t"); + + asm volatile( + /* dst[0..7] = pack(([mm2,mm3] + 32) >> 6) */ + "paddw %1, %%mm2\n\t" + "paddw %1, %%mm3\n\t" + "psrlw $3, %%mm2\n\t" + "psrlw $3, %%mm3\n\t" + "packuswb %%mm3, %%mm2\n\t" + H264_CHROMA_OP(%0, %%mm2) + "movq %%mm2, %0\n\t" + : "=m" (dst[0]) : "m" (ff_pw_4)); + + dst += stride; + } + return; + } + + /* general case, bilinear */ + asm volatile("movd %2, %%mm4\n\t" + "movd %3, %%mm6\n\t" "punpcklwd %%mm4, %%mm4\n\t" "punpcklwd %%mm6, %%mm6\n\t" "punpckldq %%mm4, %%mm4\n\t" /* mm4 = x words */ @@ -44,29 +161,20 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1* "psllw $3, %%mm6\n\t" "movq %%mm5, %%mm7\n\t" "paddw %%mm6, %%mm7\n\t" - "movq %%mm4, %0\n\t" /* DD = x * y */ + "movq %%mm4, %1\n\t" /* DD = x * y */ "psubw %%mm4, %%mm5\n\t" /* mm5 = B = 8x - xy */ "psubw %%mm4, %%mm6\n\t" /* mm6 = C = 8y - xy */ - "paddw %3, %%mm4\n\t" + "paddw %4, %%mm4\n\t" "psubw %%mm7, %%mm4\n\t" /* mm4 = A = xy - (8x+8y) + 64 */ "pxor %%mm7, %%mm7\n\t" - : "=m" (DD) : "rm" (x), "rm" (y), "m" (ff_pw_64)); - - asm volatile("movq %%mm4, %0" : "=m" (AA)); + "movq %%mm4, %0\n\t" + : "=m" (AA), "=m" (DD) : "rm" (x), "rm" (y), "m" (ff_pw_64)); - src -= srcos; asm volatile( /* mm0 = src[0..7], mm1 = src[1..8] */ - "movq %0, %%mm1\n\t" - "movq %1, %%mm0\n\t" - "psrlq %2, %%mm1\n\t" - "psllq %3, %%mm0\n\t" - "movq %%mm0, %%mm4\n\t" - "psllq $8, %%mm0\n\t" - "por %%mm1, %%mm0\n\t" - "psrlq $8, %%mm1\n\t" - "por %%mm4, %%mm1\n\t" - : : "m" (src[0]), "m" (src[8]), "m" (sh1), "m" (sh2)); + "movq %0, %%mm0\n\t" + "movq %1, %%mm1\n\t" + : : "m" (src[0]), "m" (src[1])); for(i=0; i<h; i++) { asm volatile( @@ -91,16 +199,9 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1* src += stride; asm volatile( /* mm0 = src[0..7], mm1 = src[1..8] */ - "movq %0, %%mm1\n\t" - "movq %1, %%mm0\n\t" - "psrlq %2, %%mm1\n\t" - "psllq %3, %%mm0\n\t" - "movq %%mm0, %%mm4\n\t" - "psllq $8, %%mm0\n\t" - "por %%mm1, %%mm0\n\t" - "psrlq $8, %%mm1\n\t" - "por %%mm4, %%mm1\n\t" - : : "m" (src[0]), "m" (src[8]), "m" (sh1), "m" (sh2)); + "movq %0, %%mm0\n\t" + "movq %1, %%mm1\n\t" + : : "m" (src[0]), "m" (src[1])); asm volatile( /* [mm2,mm3] += C * src[0..7] */ @@ -138,3 +239,83 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1* dst+= stride; } } + +static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y) +{ + DECLARE_ALIGNED_8(uint64_t, AA); + DECLARE_ALIGNED_8(uint64_t, DD); + int i; + + /* no special case for mv=(0,0) in 4x*, since it's much less common than in 8x*. + * could still save a few cycles, but maybe not worth the complexity. */ + + assert(x<8 && y<8 && x>=0 && y>=0); + + asm volatile("movd %2, %%mm4\n\t" + "movd %3, %%mm6\n\t" + "punpcklwd %%mm4, %%mm4\n\t" + "punpcklwd %%mm6, %%mm6\n\t" + "punpckldq %%mm4, %%mm4\n\t" /* mm4 = x words */ + "punpckldq %%mm6, %%mm6\n\t" /* mm6 = y words */ + "movq %%mm4, %%mm5\n\t" + "pmullw %%mm6, %%mm4\n\t" /* mm4 = x * y */ + "psllw $3, %%mm5\n\t" + "psllw $3, %%mm6\n\t" + "movq %%mm5, %%mm7\n\t" + "paddw %%mm6, %%mm7\n\t" + "movq %%mm4, %1\n\t" /* DD = x * y */ + "psubw %%mm4, %%mm5\n\t" /* mm5 = B = 8x - xy */ + "psubw %%mm4, %%mm6\n\t" /* mm6 = C = 8y - xy */ + "paddw %4, %%mm4\n\t" + "psubw %%mm7, %%mm4\n\t" /* mm4 = A = xy - (8x+8y) + 64 */ + "pxor %%mm7, %%mm7\n\t" + "movq %%mm4, %0\n\t" + : "=m" (AA), "=m" (DD) : "rm" (x), "rm" (y), "m" (ff_pw_64)); + + asm volatile( + /* mm0 = src[0..3], mm1 = src[1..4] */ + "movd %0, %%mm0\n\t" + "movd %1, %%mm1\n\t" + "punpcklbw %%mm7, %%mm0\n\t" + "punpcklbw %%mm7, %%mm1\n\t" + : : "m" (src[0]), "m" (src[1])); + + for(i=0; i<h; i++) { + asm volatile( + /* mm2 = A * src[0..3] + B * src[1..4] */ + "movq %%mm0, %%mm2\n\t" + "pmullw %0, %%mm2\n\t" + "pmullw %%mm5, %%mm1\n\t" + "paddw %%mm1, %%mm2\n\t" + : : "m" (AA)); + + src += stride; + asm volatile( + /* mm0 = src[0..3], mm1 = src[1..4] */ + "movd %0, %%mm0\n\t" + "movd %1, %%mm1\n\t" + "punpcklbw %%mm7, %%mm0\n\t" + "punpcklbw %%mm7, %%mm1\n\t" + : : "m" (src[0]), "m" (src[1])); + + asm volatile( + /* mm2 += C * src[0..3] + D * src[1..4] */ + "movq %%mm0, %%mm3\n\t" + "movq %%mm1, %%mm4\n\t" + "pmullw %%mm6, %%mm3\n\t" + "pmullw %0, %%mm4\n\t" + "paddw %%mm3, %%mm2\n\t" + "paddw %%mm4, %%mm2\n\t" + : : "m" (DD)); + + asm volatile( + /* dst[0..3] = pack((mm2 + 32) >> 6) */ + "paddw %1, %%mm2\n\t" + "psrlw $6, %%mm2\n\t" + "packuswb %%mm7, %%mm2\n\t" + H264_CHROMA_OP4(%0, %%mm2, %%mm3) + "movd %%mm2, %0\n\t" + : "=m" (dst[0]) : "m" (ff_pw_32)); + dst += stride; + } +} diff --git a/src/libffmpeg/libavcodec/i386/dsputil_mmx.c b/src/libffmpeg/libavcodec/i386/dsputil_mmx.c index c1dd2176a..7d69859a6 100644 --- a/src/libffmpeg/libavcodec/i386/dsputil_mmx.c +++ b/src/libffmpeg/libavcodec/i386/dsputil_mmx.c @@ -15,7 +15,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA * * MMX optimization by Nick Kurshev <nickols_k@mail.ru> */ @@ -43,6 +43,7 @@ static const uint64_t ff_pw_20 attribute_used __attribute__ ((aligned(8))) = 0x0 static const uint64_t ff_pw_3 attribute_used __attribute__ ((aligned(8))) = 0x0003000300030003ULL; static const uint64_t ff_pw_4 attribute_used __attribute__ ((aligned(8))) = 0x0004000400040004ULL; static const uint64_t ff_pw_5 attribute_used __attribute__ ((aligned(8))) = 0x0005000500050005ULL; +static const uint64_t ff_pw_8 attribute_used __attribute__ ((aligned(8))) = 0x0008000800080008ULL; static const uint64_t ff_pw_16 attribute_used __attribute__ ((aligned(8))) = 0x0010001000100010ULL; static const uint64_t ff_pw_32 attribute_used __attribute__ ((aligned(8))) = 0x0020002000200020ULL; static const uint64_t ff_pw_64 attribute_used __attribute__ ((aligned(8))) = 0x0040004000400040ULL; @@ -88,56 +89,56 @@ static const uint64_t ff_pb_FC attribute_used __attribute__ ((aligned(8))) = 0xF // first argument is unmodifed and second is trashed // regfe is supposed to contain 0xfefefefefefefefe #define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \ - "movq " #rega ", " #regr " \n\t"\ - "pand " #regb ", " #regr " \n\t"\ - "pxor " #rega ", " #regb " \n\t"\ - "pand " #regfe "," #regb " \n\t"\ - "psrlq $1, " #regb " \n\t"\ - "paddb " #regb ", " #regr " \n\t" + "movq " #rega ", " #regr " \n\t"\ + "pand " #regb ", " #regr " \n\t"\ + "pxor " #rega ", " #regb " \n\t"\ + "pand " #regfe "," #regb " \n\t"\ + "psrlq $1, " #regb " \n\t"\ + "paddb " #regb ", " #regr " \n\t" #define PAVGB_MMX(rega, regb, regr, regfe) \ - "movq " #rega ", " #regr " \n\t"\ - "por " #regb ", " #regr " \n\t"\ - "pxor " #rega ", " #regb " \n\t"\ - "pand " #regfe "," #regb " \n\t"\ - "psrlq $1, " #regb " \n\t"\ - "psubb " #regb ", " #regr " \n\t" + "movq " #rega ", " #regr " \n\t"\ + "por " #regb ", " #regr " \n\t"\ + "pxor " #rega ", " #regb " \n\t"\ + "pand " #regfe "," #regb " \n\t"\ + "psrlq $1, " #regb " \n\t"\ + "psubb " #regb ", " #regr " \n\t" // mm6 is supposed to contain 0xfefefefefefefefe #define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \ - "movq " #rega ", " #regr " \n\t"\ - "movq " #regc ", " #regp " \n\t"\ - "pand " #regb ", " #regr " \n\t"\ - "pand " #regd ", " #regp " \n\t"\ - "pxor " #rega ", " #regb " \n\t"\ - "pxor " #regc ", " #regd " \n\t"\ - "pand %%mm6, " #regb " \n\t"\ - "pand %%mm6, " #regd " \n\t"\ - "psrlq $1, " #regb " \n\t"\ - "psrlq $1, " #regd " \n\t"\ - "paddb " #regb ", " #regr " \n\t"\ - "paddb " #regd ", " #regp " \n\t" + "movq " #rega ", " #regr " \n\t"\ + "movq " #regc ", " #regp " \n\t"\ + "pand " #regb ", " #regr " \n\t"\ + "pand " #regd ", " #regp " \n\t"\ + "pxor " #rega ", " #regb " \n\t"\ + "pxor " #regc ", " #regd " \n\t"\ + "pand %%mm6, " #regb " \n\t"\ + "pand %%mm6, " #regd " \n\t"\ + "psrlq $1, " #regb " \n\t"\ + "psrlq $1, " #regd " \n\t"\ + "paddb " #regb ", " #regr " \n\t"\ + "paddb " #regd ", " #regp " \n\t" #define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \ - "movq " #rega ", " #regr " \n\t"\ - "movq " #regc ", " #regp " \n\t"\ - "por " #regb ", " #regr " \n\t"\ - "por " #regd ", " #regp " \n\t"\ - "pxor " #rega ", " #regb " \n\t"\ - "pxor " #regc ", " #regd " \n\t"\ - "pand %%mm6, " #regb " \n\t"\ - "pand %%mm6, " #regd " \n\t"\ - "psrlq $1, " #regd " \n\t"\ - "psrlq $1, " #regb " \n\t"\ - "psubb " #regb ", " #regr " \n\t"\ - "psubb " #regd ", " #regp " \n\t" + "movq " #rega ", " #regr " \n\t"\ + "movq " #regc ", " #regp " \n\t"\ + "por " #regb ", " #regr " \n\t"\ + "por " #regd ", " #regp " \n\t"\ + "pxor " #rega ", " #regb " \n\t"\ + "pxor " #regc ", " #regd " \n\t"\ + "pand %%mm6, " #regb " \n\t"\ + "pand %%mm6, " #regd " \n\t"\ + "psrlq $1, " #regd " \n\t"\ + "psrlq $1, " #regb " \n\t"\ + "psubb " #regb ", " #regr " \n\t"\ + "psubb " #regd ", " #regp " \n\t" /***********************************/ /* MMX no rounding */ #define DEF(x, y) x ## _no_rnd_ ## y ##_mmx #define SET_RND MOVQ_WONE -#define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX_NO_RND(a, b, c, d, e, f) -#define PAVGB(a, b, c, e) PAVGB_MMX_NO_RND(a, b, c, e) +#define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX_NO_RND(a, b, c, d, e, f) +#define PAVGB(a, b, c, e) PAVGB_MMX_NO_RND(a, b, c, e) #include "dsputil_mmx_rnd.h" @@ -150,8 +151,8 @@ static const uint64_t ff_pb_FC attribute_used __attribute__ ((aligned(8))) = 0xF #define DEF(x, y) x ## _ ## y ##_mmx #define SET_RND MOVQ_WTWO -#define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX(a, b, c, d, e, f) -#define PAVGB(a, b, c, e) PAVGB_MMX(a, b, c, e) +#define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX(a, b, c, d, e, f) +#define PAVGB(a, b, c, e) PAVGB_MMX(a, b, c, e) #include "dsputil_mmx_rnd.h" @@ -192,25 +193,25 @@ static const uint64_t ff_pb_FC attribute_used __attribute__ ((aligned(8))) = 0xF static void get_pixels_mmx(DCTELEM *block, const uint8_t *pixels, int line_size) { asm volatile( - "mov $-128, %%"REG_a" \n\t" - "pxor %%mm7, %%mm7 \n\t" - ".balign 16 \n\t" - "1: \n\t" - "movq (%0), %%mm0 \n\t" - "movq (%0, %2), %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpckhbw %%mm7, %%mm3 \n\t" - "movq %%mm0, (%1, %%"REG_a")\n\t" - "movq %%mm1, 8(%1, %%"REG_a")\n\t" - "movq %%mm2, 16(%1, %%"REG_a")\n\t" - "movq %%mm3, 24(%1, %%"REG_a")\n\t" - "add %3, %0 \n\t" - "add $32, %%"REG_a" \n\t" - "js 1b \n\t" + "mov $-128, %%"REG_a" \n\t" + "pxor %%mm7, %%mm7 \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%0), %%mm0 \n\t" + "movq (%0, %2), %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "movq %%mm0, (%1, %%"REG_a") \n\t" + "movq %%mm1, 8(%1, %%"REG_a") \n\t" + "movq %%mm2, 16(%1, %%"REG_a") \n\t" + "movq %%mm3, 24(%1, %%"REG_a") \n\t" + "add %3, %0 \n\t" + "add $32, %%"REG_a" \n\t" + "js 1b \n\t" : "+r" (pixels) : "r" (block+64), "r" ((long)line_size), "r" ((long)line_size*2) : "%"REG_a @@ -220,26 +221,26 @@ static void get_pixels_mmx(DCTELEM *block, const uint8_t *pixels, int line_size) static inline void diff_pixels_mmx(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride) { asm volatile( - "pxor %%mm7, %%mm7 \n\t" - "mov $-128, %%"REG_a" \n\t" - ".balign 16 \n\t" - "1: \n\t" - "movq (%0), %%mm0 \n\t" - "movq (%1), %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpckhbw %%mm7, %%mm3 \n\t" - "psubw %%mm2, %%mm0 \n\t" - "psubw %%mm3, %%mm1 \n\t" - "movq %%mm0, (%2, %%"REG_a")\n\t" - "movq %%mm1, 8(%2, %%"REG_a")\n\t" - "add %3, %0 \n\t" - "add %3, %1 \n\t" - "add $16, %%"REG_a" \n\t" - "jnz 1b \n\t" + "pxor %%mm7, %%mm7 \n\t" + "mov $-128, %%"REG_a" \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%0), %%mm0 \n\t" + "movq (%1), %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "psubw %%mm2, %%mm0 \n\t" + "psubw %%mm3, %%mm1 \n\t" + "movq %%mm0, (%2, %%"REG_a") \n\t" + "movq %%mm1, 8(%2, %%"REG_a") \n\t" + "add %3, %0 \n\t" + "add %3, %1 \n\t" + "add $16, %%"REG_a" \n\t" + "jnz 1b \n\t" : "+r" (s1), "+r" (s2) : "r" (block+64), "r" ((long)stride) : "%"REG_a @@ -256,25 +257,25 @@ void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size p = block; pix = pixels; /* unrolled loop */ - __asm __volatile( - "movq %3, %%mm0\n\t" - "movq 8%3, %%mm1\n\t" - "movq 16%3, %%mm2\n\t" - "movq 24%3, %%mm3\n\t" - "movq 32%3, %%mm4\n\t" - "movq 40%3, %%mm5\n\t" - "movq 48%3, %%mm6\n\t" - "movq 56%3, %%mm7\n\t" - "packuswb %%mm1, %%mm0\n\t" - "packuswb %%mm3, %%mm2\n\t" - "packuswb %%mm5, %%mm4\n\t" - "packuswb %%mm7, %%mm6\n\t" - "movq %%mm0, (%0)\n\t" - "movq %%mm2, (%0, %1)\n\t" - "movq %%mm4, (%0, %1, 2)\n\t" - "movq %%mm6, (%0, %2)\n\t" - ::"r" (pix), "r" ((long)line_size), "r" ((long)line_size*3), "m"(*p) - :"memory"); + __asm __volatile( + "movq %3, %%mm0 \n\t" + "movq 8%3, %%mm1 \n\t" + "movq 16%3, %%mm2 \n\t" + "movq 24%3, %%mm3 \n\t" + "movq 32%3, %%mm4 \n\t" + "movq 40%3, %%mm5 \n\t" + "movq 48%3, %%mm6 \n\t" + "movq 56%3, %%mm7 \n\t" + "packuswb %%mm1, %%mm0 \n\t" + "packuswb %%mm3, %%mm2 \n\t" + "packuswb %%mm5, %%mm4 \n\t" + "packuswb %%mm7, %%mm6 \n\t" + "movq %%mm0, (%0) \n\t" + "movq %%mm2, (%0, %1) \n\t" + "movq %%mm4, (%0, %1, 2) \n\t" + "movq %%mm6, (%0, %2) \n\t" + ::"r" (pix), "r" ((long)line_size), "r" ((long)line_size*3), "m"(*p) + :"memory"); pix += line_size*4; p += 32; @@ -282,27 +283,27 @@ void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size // compiler would generate some very strange code // thus using "r" __asm __volatile( - "movq (%3), %%mm0\n\t" - "movq 8(%3), %%mm1\n\t" - "movq 16(%3), %%mm2\n\t" - "movq 24(%3), %%mm3\n\t" - "movq 32(%3), %%mm4\n\t" - "movq 40(%3), %%mm5\n\t" - "movq 48(%3), %%mm6\n\t" - "movq 56(%3), %%mm7\n\t" - "packuswb %%mm1, %%mm0\n\t" - "packuswb %%mm3, %%mm2\n\t" - "packuswb %%mm5, %%mm4\n\t" - "packuswb %%mm7, %%mm6\n\t" - "movq %%mm0, (%0)\n\t" - "movq %%mm2, (%0, %1)\n\t" - "movq %%mm4, (%0, %1, 2)\n\t" - "movq %%mm6, (%0, %2)\n\t" - ::"r" (pix), "r" ((long)line_size), "r" ((long)line_size*3), "r"(p) - :"memory"); + "movq (%3), %%mm0 \n\t" + "movq 8(%3), %%mm1 \n\t" + "movq 16(%3), %%mm2 \n\t" + "movq 24(%3), %%mm3 \n\t" + "movq 32(%3), %%mm4 \n\t" + "movq 40(%3), %%mm5 \n\t" + "movq 48(%3), %%mm6 \n\t" + "movq 56(%3), %%mm7 \n\t" + "packuswb %%mm1, %%mm0 \n\t" + "packuswb %%mm3, %%mm2 \n\t" + "packuswb %%mm5, %%mm4 \n\t" + "packuswb %%mm7, %%mm6 \n\t" + "movq %%mm0, (%0) \n\t" + "movq %%mm2, (%0, %1) \n\t" + "movq %%mm4, (%0, %1, 2) \n\t" + "movq %%mm6, (%0, %2) \n\t" + ::"r" (pix), "r" ((long)line_size), "r" ((long)line_size*3), "r"(p) + :"memory"); } -static const unsigned char __align8 vector128[8] = +static DECLARE_ALIGNED_8(const unsigned char, vector128[8]) = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size) @@ -332,30 +333,30 @@ void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size MOVQ_ZERO(mm7); i = 4; do { - __asm __volatile( - "movq (%2), %%mm0\n\t" - "movq 8(%2), %%mm1\n\t" - "movq 16(%2), %%mm2\n\t" - "movq 24(%2), %%mm3\n\t" - "movq %0, %%mm4\n\t" - "movq %1, %%mm6\n\t" - "movq %%mm4, %%mm5\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpckhbw %%mm7, %%mm5\n\t" - "paddsw %%mm4, %%mm0\n\t" - "paddsw %%mm5, %%mm1\n\t" - "movq %%mm6, %%mm5\n\t" - "punpcklbw %%mm7, %%mm6\n\t" - "punpckhbw %%mm7, %%mm5\n\t" - "paddsw %%mm6, %%mm2\n\t" - "paddsw %%mm5, %%mm3\n\t" - "packuswb %%mm1, %%mm0\n\t" - "packuswb %%mm3, %%mm2\n\t" - "movq %%mm0, %0\n\t" - "movq %%mm2, %1\n\t" - :"+m"(*pix), "+m"(*(pix+line_size)) - :"r"(p) - :"memory"); + __asm __volatile( + "movq (%2), %%mm0 \n\t" + "movq 8(%2), %%mm1 \n\t" + "movq 16(%2), %%mm2 \n\t" + "movq 24(%2), %%mm3 \n\t" + "movq %0, %%mm4 \n\t" + "movq %1, %%mm6 \n\t" + "movq %%mm4, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpckhbw %%mm7, %%mm5 \n\t" + "paddsw %%mm4, %%mm0 \n\t" + "paddsw %%mm5, %%mm1 \n\t" + "movq %%mm6, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm6 \n\t" + "punpckhbw %%mm7, %%mm5 \n\t" + "paddsw %%mm6, %%mm2 \n\t" + "paddsw %%mm5, %%mm3 \n\t" + "packuswb %%mm1, %%mm0 \n\t" + "packuswb %%mm3, %%mm2 \n\t" + "movq %%mm0, %0 \n\t" + "movq %%mm2, %1 \n\t" + :"+m"(*pix), "+m"(*(pix+line_size)) + :"r"(p) + :"memory"); pix += line_size*2; p += 16; } while (--i); @@ -364,101 +365,101 @@ void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size static void put_pixels4_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movd (%1), %%mm0 \n\t" - "movd (%1, %3), %%mm1 \n\t" - "movd %%mm0, (%2) \n\t" - "movd %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movd (%1), %%mm0 \n\t" - "movd (%1, %3), %%mm1 \n\t" - "movd %%mm0, (%2) \n\t" - "movd %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - : "+g"(h), "+r" (pixels), "+r" (block) - : "r"((long)line_size) - : "%"REG_a, "memory" - ); + "lea (%3, %3), %%"REG_a" \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movd (%1), %%mm0 \n\t" + "movd (%1, %3), %%mm1 \n\t" + "movd %%mm0, (%2) \n\t" + "movd %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movd (%1), %%mm0 \n\t" + "movd (%1, %3), %%mm1 \n\t" + "movd %%mm0, (%2) \n\t" + "movd %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + : "+g"(h), "+r" (pixels), "+r" (block) + : "r"((long)line_size) + : "%"REG_a, "memory" + ); } static void put_pixels8_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - : "+g"(h), "+r" (pixels), "+r" (block) - : "r"((long)line_size) - : "%"REG_a, "memory" - ); + "lea (%3, %3), %%"REG_a" \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + : "+g"(h), "+r" (pixels), "+r" (block) + : "r"((long)line_size) + : "%"REG_a, "memory" + ); } static void put_pixels16_mmx(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm4 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq 8(%1, %3), %%mm5 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm4, 8(%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "movq %%mm5, 8(%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm4 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq 8(%1, %3), %%mm5 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm4, 8(%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "movq %%mm5, 8(%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - : "+g"(h), "+r" (pixels), "+r" (block) - : "r"((long)line_size) - : "%"REG_a, "memory" - ); + "lea (%3, %3), %%"REG_a" \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm4 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq 8(%1, %3), %%mm5 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm4, 8(%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "movq %%mm5, 8(%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm4 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq 8(%1, %3), %%mm5 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm4, 8(%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "movq %%mm5, 8(%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + : "+g"(h), "+r" (pixels), "+r" (block) + : "r"((long)line_size) + : "%"REG_a, "memory" + ); } static void clear_blocks_mmx(DCTELEM *blocks) { __asm __volatile( - "pxor %%mm7, %%mm7 \n\t" - "mov $-128*6, %%"REG_a" \n\t" - "1: \n\t" - "movq %%mm7, (%0, %%"REG_a") \n\t" - "movq %%mm7, 8(%0, %%"REG_a") \n\t" - "movq %%mm7, 16(%0, %%"REG_a") \n\t" - "movq %%mm7, 24(%0, %%"REG_a") \n\t" - "add $32, %%"REG_a" \n\t" - " js 1b \n\t" + "pxor %%mm7, %%mm7 \n\t" + "mov $-128*6, %%"REG_a" \n\t" + "1: \n\t" + "movq %%mm7, (%0, %%"REG_a") \n\t" + "movq %%mm7, 8(%0, %%"REG_a") \n\t" + "movq %%mm7, 16(%0, %%"REG_a") \n\t" + "movq %%mm7, 24(%0, %%"REG_a") \n\t" + "add $32, %%"REG_a" \n\t" + " js 1b \n\t" : : "r" (((uint8_t *)blocks)+128*6) : "%"REG_a ); @@ -471,31 +472,31 @@ static int pix_sum16_mmx(uint8_t * pix, int line_size){ long index= -line_size*h; __asm __volatile( - "pxor %%mm7, %%mm7 \n\t" - "pxor %%mm6, %%mm6 \n\t" - "1: \n\t" - "movq (%2, %1), %%mm0 \n\t" - "movq (%2, %1), %%mm1 \n\t" - "movq 8(%2, %1), %%mm2 \n\t" - "movq 8(%2, %1), %%mm3 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpckhbw %%mm7, %%mm3 \n\t" - "paddw %%mm0, %%mm1 \n\t" - "paddw %%mm2, %%mm3 \n\t" - "paddw %%mm1, %%mm3 \n\t" - "paddw %%mm3, %%mm6 \n\t" - "add %3, %1 \n\t" - " js 1b \n\t" - "movq %%mm6, %%mm5 \n\t" - "psrlq $32, %%mm6 \n\t" - "paddw %%mm5, %%mm6 \n\t" - "movq %%mm6, %%mm5 \n\t" - "psrlq $16, %%mm6 \n\t" - "paddw %%mm5, %%mm6 \n\t" - "movd %%mm6, %0 \n\t" - "andl $0xFFFF, %0 \n\t" + "pxor %%mm7, %%mm7 \n\t" + "pxor %%mm6, %%mm6 \n\t" + "1: \n\t" + "movq (%2, %1), %%mm0 \n\t" + "movq (%2, %1), %%mm1 \n\t" + "movq 8(%2, %1), %%mm2 \n\t" + "movq 8(%2, %1), %%mm3 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "paddw %%mm0, %%mm1 \n\t" + "paddw %%mm2, %%mm3 \n\t" + "paddw %%mm1, %%mm3 \n\t" + "paddw %%mm3, %%mm6 \n\t" + "add %3, %1 \n\t" + " js 1b \n\t" + "movq %%mm6, %%mm5 \n\t" + "psrlq $32, %%mm6 \n\t" + "paddw %%mm5, %%mm6 \n\t" + "movq %%mm6, %%mm5 \n\t" + "psrlq $16, %%mm6 \n\t" + "paddw %%mm5, %%mm6 \n\t" + "movd %%mm6, %0 \n\t" + "andl $0xFFFF, %0 \n\t" : "=&r" (sum), "+r" (index) : "r" (pix - index), "r" ((long)line_size) ); @@ -507,18 +508,18 @@ static int pix_sum16_mmx(uint8_t * pix, int line_size){ static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){ long i=0; asm volatile( - "1: \n\t" - "movq (%1, %0), %%mm0 \n\t" - "movq (%2, %0), %%mm1 \n\t" - "paddb %%mm0, %%mm1 \n\t" - "movq %%mm1, (%2, %0) \n\t" - "movq 8(%1, %0), %%mm0 \n\t" - "movq 8(%2, %0), %%mm1 \n\t" - "paddb %%mm0, %%mm1 \n\t" - "movq %%mm1, 8(%2, %0) \n\t" - "add $16, %0 \n\t" - "cmp %3, %0 \n\t" - " jb 1b \n\t" + "1: \n\t" + "movq (%1, %0), %%mm0 \n\t" + "movq (%2, %0), %%mm1 \n\t" + "paddb %%mm0, %%mm1 \n\t" + "movq %%mm1, (%2, %0) \n\t" + "movq 8(%1, %0), %%mm0 \n\t" + "movq 8(%2, %0), %%mm1 \n\t" + "paddb %%mm0, %%mm1 \n\t" + "movq %%mm1, 8(%2, %0) \n\t" + "add $16, %0 \n\t" + "cmp %3, %0 \n\t" + " jb 1b \n\t" : "+r" (i) : "r"(src), "r"(dst), "r"((long)w-15) ); @@ -527,87 +528,87 @@ static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){ } #define H263_LOOP_FILTER \ - "pxor %%mm7, %%mm7 \n\t"\ - "movq %0, %%mm0 \n\t"\ - "movq %0, %%mm1 \n\t"\ - "movq %3, %%mm2 \n\t"\ - "movq %3, %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpckhbw %%mm7, %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpckhbw %%mm7, %%mm3 \n\t"\ - "psubw %%mm2, %%mm0 \n\t"\ - "psubw %%mm3, %%mm1 \n\t"\ - "movq %1, %%mm2 \n\t"\ - "movq %1, %%mm3 \n\t"\ - "movq %2, %%mm4 \n\t"\ - "movq %2, %%mm5 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpckhbw %%mm7, %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm4 \n\t"\ - "punpckhbw %%mm7, %%mm5 \n\t"\ - "psubw %%mm2, %%mm4 \n\t"\ - "psubw %%mm3, %%mm5 \n\t"\ - "psllw $2, %%mm4 \n\t"\ - "psllw $2, %%mm5 \n\t"\ - "paddw %%mm0, %%mm4 \n\t"\ - "paddw %%mm1, %%mm5 \n\t"\ - "pxor %%mm6, %%mm6 \n\t"\ - "pcmpgtw %%mm4, %%mm6 \n\t"\ - "pcmpgtw %%mm5, %%mm7 \n\t"\ - "pxor %%mm6, %%mm4 \n\t"\ - "pxor %%mm7, %%mm5 \n\t"\ - "psubw %%mm6, %%mm4 \n\t"\ - "psubw %%mm7, %%mm5 \n\t"\ - "psrlw $3, %%mm4 \n\t"\ - "psrlw $3, %%mm5 \n\t"\ - "packuswb %%mm5, %%mm4 \n\t"\ - "packsswb %%mm7, %%mm6 \n\t"\ - "pxor %%mm7, %%mm7 \n\t"\ - "movd %4, %%mm2 \n\t"\ - "punpcklbw %%mm2, %%mm2 \n\t"\ - "punpcklbw %%mm2, %%mm2 \n\t"\ - "punpcklbw %%mm2, %%mm2 \n\t"\ - "psubusb %%mm4, %%mm2 \n\t"\ - "movq %%mm2, %%mm3 \n\t"\ - "psubusb %%mm4, %%mm3 \n\t"\ - "psubb %%mm3, %%mm2 \n\t"\ - "movq %1, %%mm3 \n\t"\ - "movq %2, %%mm4 \n\t"\ - "pxor %%mm6, %%mm3 \n\t"\ - "pxor %%mm6, %%mm4 \n\t"\ - "paddusb %%mm2, %%mm3 \n\t"\ - "psubusb %%mm2, %%mm4 \n\t"\ - "pxor %%mm6, %%mm3 \n\t"\ - "pxor %%mm6, %%mm4 \n\t"\ - "paddusb %%mm2, %%mm2 \n\t"\ - "packsswb %%mm1, %%mm0 \n\t"\ - "pcmpgtb %%mm0, %%mm7 \n\t"\ - "pxor %%mm7, %%mm0 \n\t"\ - "psubb %%mm7, %%mm0 \n\t"\ - "movq %%mm0, %%mm1 \n\t"\ - "psubusb %%mm2, %%mm0 \n\t"\ - "psubb %%mm0, %%mm1 \n\t"\ - "pand %5, %%mm1 \n\t"\ - "psrlw $2, %%mm1 \n\t"\ - "pxor %%mm7, %%mm1 \n\t"\ - "psubb %%mm7, %%mm1 \n\t"\ - "movq %0, %%mm5 \n\t"\ - "movq %3, %%mm6 \n\t"\ - "psubb %%mm1, %%mm5 \n\t"\ - "paddb %%mm1, %%mm6 \n\t" + "pxor %%mm7, %%mm7 \n\t"\ + "movq %0, %%mm0 \n\t"\ + "movq %0, %%mm1 \n\t"\ + "movq %3, %%mm2 \n\t"\ + "movq %3, %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpckhbw %%mm7, %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpckhbw %%mm7, %%mm3 \n\t"\ + "psubw %%mm2, %%mm0 \n\t"\ + "psubw %%mm3, %%mm1 \n\t"\ + "movq %1, %%mm2 \n\t"\ + "movq %1, %%mm3 \n\t"\ + "movq %2, %%mm4 \n\t"\ + "movq %2, %%mm5 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpckhbw %%mm7, %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm4 \n\t"\ + "punpckhbw %%mm7, %%mm5 \n\t"\ + "psubw %%mm2, %%mm4 \n\t"\ + "psubw %%mm3, %%mm5 \n\t"\ + "psllw $2, %%mm4 \n\t"\ + "psllw $2, %%mm5 \n\t"\ + "paddw %%mm0, %%mm4 \n\t"\ + "paddw %%mm1, %%mm5 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "pcmpgtw %%mm4, %%mm6 \n\t"\ + "pcmpgtw %%mm5, %%mm7 \n\t"\ + "pxor %%mm6, %%mm4 \n\t"\ + "pxor %%mm7, %%mm5 \n\t"\ + "psubw %%mm6, %%mm4 \n\t"\ + "psubw %%mm7, %%mm5 \n\t"\ + "psrlw $3, %%mm4 \n\t"\ + "psrlw $3, %%mm5 \n\t"\ + "packuswb %%mm5, %%mm4 \n\t"\ + "packsswb %%mm7, %%mm6 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "movd %4, %%mm2 \n\t"\ + "punpcklbw %%mm2, %%mm2 \n\t"\ + "punpcklbw %%mm2, %%mm2 \n\t"\ + "punpcklbw %%mm2, %%mm2 \n\t"\ + "psubusb %%mm4, %%mm2 \n\t"\ + "movq %%mm2, %%mm3 \n\t"\ + "psubusb %%mm4, %%mm3 \n\t"\ + "psubb %%mm3, %%mm2 \n\t"\ + "movq %1, %%mm3 \n\t"\ + "movq %2, %%mm4 \n\t"\ + "pxor %%mm6, %%mm3 \n\t"\ + "pxor %%mm6, %%mm4 \n\t"\ + "paddusb %%mm2, %%mm3 \n\t"\ + "psubusb %%mm2, %%mm4 \n\t"\ + "pxor %%mm6, %%mm3 \n\t"\ + "pxor %%mm6, %%mm4 \n\t"\ + "paddusb %%mm2, %%mm2 \n\t"\ + "packsswb %%mm1, %%mm0 \n\t"\ + "pcmpgtb %%mm0, %%mm7 \n\t"\ + "pxor %%mm7, %%mm0 \n\t"\ + "psubb %%mm7, %%mm0 \n\t"\ + "movq %%mm0, %%mm1 \n\t"\ + "psubusb %%mm2, %%mm0 \n\t"\ + "psubb %%mm0, %%mm1 \n\t"\ + "pand %5, %%mm1 \n\t"\ + "psrlw $2, %%mm1 \n\t"\ + "pxor %%mm7, %%mm1 \n\t"\ + "psubb %%mm7, %%mm1 \n\t"\ + "movq %0, %%mm5 \n\t"\ + "movq %3, %%mm6 \n\t"\ + "psubb %%mm1, %%mm5 \n\t"\ + "paddb %%mm1, %%mm6 \n\t" static void h263_v_loop_filter_mmx(uint8_t *src, int stride, int qscale){ const int strength= ff_h263_loop_filter_strength[qscale]; asm volatile( - + H263_LOOP_FILTER - - "movq %%mm3, %1 \n\t" - "movq %%mm4, %2 \n\t" - "movq %%mm5, %0 \n\t" - "movq %%mm6, %3 \n\t" + + "movq %%mm3, %1 \n\t" + "movq %%mm4, %2 \n\t" + "movq %%mm5, %0 \n\t" + "movq %%mm6, %3 \n\t" : "+m" (*(uint64_t*)(src - 2*stride)), "+m" (*(uint64_t*)(src - 1*stride)), "+m" (*(uint64_t*)(src + 0*stride)), @@ -617,32 +618,31 @@ static void h263_v_loop_filter_mmx(uint8_t *src, int stride, int qscale){ } static inline void transpose4x4(uint8_t *dst, uint8_t *src, int dst_stride, int src_stride){ - void *dst_reg = dst, *src_reg = src; - asm volatile( //FIXME could save 1 instruction if done as 8x4 ... - "movd (%1), %%mm0 \n\t" - "movd (%1,%5), %%mm1 \n\t" - "lea (%1, %5, 2), %1 \n\t" - "movd (%1), %%mm2 \n\t" - "movd (%1,%5), %%mm3 \n\t" - "punpcklbw %%mm1, %%mm0 \n\t" - "punpcklbw %%mm3, %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "punpcklwd %%mm2, %%mm0 \n\t" - "punpckhwd %%mm2, %%mm1 \n\t" - "movd %%mm0, (%0) \n\t" - "punpckhdq %%mm0, %%mm0 \n\t" - "movd %%mm0, (%0,%4) \n\t" - "lea (%0, %4, 2), %0 \n\t" - "movd %%mm1, (%0) \n\t" - "punpckhdq %%mm1, %%mm1 \n\t" - "movd %%mm1, (%0,%4) \n\t" - : "=&r" (dst_reg), - "=&r" (src_reg) - : "0" (dst_reg), - "1" (src_reg), - "r" (dst_stride), - "r" (src_stride) + "movd %4, %%mm0 \n\t" + "movd %5, %%mm1 \n\t" + "movd %6, %%mm2 \n\t" + "movd %7, %%mm3 \n\t" + "punpcklbw %%mm1, %%mm0 \n\t" + "punpcklbw %%mm3, %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "punpcklwd %%mm2, %%mm0 \n\t" + "punpckhwd %%mm2, %%mm1 \n\t" + "movd %%mm0, %0 \n\t" + "punpckhdq %%mm0, %%mm0 \n\t" + "movd %%mm0, %1 \n\t" + "movd %%mm1, %2 \n\t" + "punpckhdq %%mm1, %%mm1 \n\t" + "movd %%mm1, %3 \n\t" + + : "=m" (*(uint32_t*)(dst + 0*dst_stride)), + "=m" (*(uint32_t*)(dst + 1*dst_stride)), + "=m" (*(uint32_t*)(dst + 2*dst_stride)), + "=m" (*(uint32_t*)(dst + 3*dst_stride)) + : "m" (*(uint32_t*)(src + 0*src_stride)), + "m" (*(uint32_t*)(src + 1*src_stride)), + "m" (*(uint32_t*)(src + 2*src_stride)), + "m" (*(uint32_t*)(src + 3*src_stride)) ); } @@ -650,14 +650,14 @@ static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale){ const int strength= ff_h263_loop_filter_strength[qscale]; uint64_t temp[4] __attribute__ ((aligned(8))); uint8_t *btemp= (uint8_t*)temp; - + src -= 2; transpose4x4(btemp , src , 8, stride); transpose4x4(btemp+4, src + 4*stride, 8, stride); asm volatile( H263_LOOP_FILTER // 5 3 4 6 - + : "+m" (temp[0]), "+m" (temp[1]), "+m" (temp[2]), @@ -666,30 +666,30 @@ static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale){ ); asm volatile( - "movq %%mm5, %%mm1 \n\t" - "movq %%mm4, %%mm0 \n\t" - "punpcklbw %%mm3, %%mm5 \n\t" - "punpcklbw %%mm6, %%mm4 \n\t" - "punpckhbw %%mm3, %%mm1 \n\t" - "punpckhbw %%mm6, %%mm0 \n\t" - "movq %%mm5, %%mm3 \n\t" - "movq %%mm1, %%mm6 \n\t" - "punpcklwd %%mm4, %%mm5 \n\t" - "punpcklwd %%mm0, %%mm1 \n\t" - "punpckhwd %%mm4, %%mm3 \n\t" - "punpckhwd %%mm0, %%mm6 \n\t" - "movd %%mm5, (%0) \n\t" - "punpckhdq %%mm5, %%mm5 \n\t" - "movd %%mm5, (%0,%2) \n\t" - "movd %%mm3, (%0,%2,2) \n\t" - "punpckhdq %%mm3, %%mm3 \n\t" - "movd %%mm3, (%0,%3) \n\t" - "movd %%mm1, (%1) \n\t" - "punpckhdq %%mm1, %%mm1 \n\t" - "movd %%mm1, (%1,%2) \n\t" - "movd %%mm6, (%1,%2,2) \n\t" - "punpckhdq %%mm6, %%mm6 \n\t" - "movd %%mm6, (%1,%3) \n\t" + "movq %%mm5, %%mm1 \n\t" + "movq %%mm4, %%mm0 \n\t" + "punpcklbw %%mm3, %%mm5 \n\t" + "punpcklbw %%mm6, %%mm4 \n\t" + "punpckhbw %%mm3, %%mm1 \n\t" + "punpckhbw %%mm6, %%mm0 \n\t" + "movq %%mm5, %%mm3 \n\t" + "movq %%mm1, %%mm6 \n\t" + "punpcklwd %%mm4, %%mm5 \n\t" + "punpcklwd %%mm0, %%mm1 \n\t" + "punpckhwd %%mm4, %%mm3 \n\t" + "punpckhwd %%mm0, %%mm6 \n\t" + "movd %%mm5, (%0) \n\t" + "punpckhdq %%mm5, %%mm5 \n\t" + "movd %%mm5, (%0,%2) \n\t" + "movd %%mm3, (%0,%2,2) \n\t" + "punpckhdq %%mm3, %%mm3 \n\t" + "movd %%mm3, (%0,%3) \n\t" + "movd %%mm1, (%1) \n\t" + "punpckhdq %%mm1, %%mm1 \n\t" + "movd %%mm1, (%1,%2) \n\t" + "movd %%mm6, (%1,%2,2) \n\t" + "punpckhdq %%mm6, %%mm6 \n\t" + "movd %%mm6, (%1,%3) \n\t" :: "r" (src), "r" (src + 4*stride), "r" ((long) stride ), @@ -705,26 +705,26 @@ static int pix_norm1_mmx(uint8_t *pix, int line_size) { "pxor %%mm0,%%mm0\n" "pxor %%mm7,%%mm7\n" "1:\n" - "movq (%0),%%mm2\n" /* mm2 = pix[0-7] */ - "movq 8(%0),%%mm3\n" /* mm3 = pix[8-15] */ + "movq (%0),%%mm2\n" /* mm2 = pix[0-7] */ + "movq 8(%0),%%mm3\n" /* mm3 = pix[8-15] */ - "movq %%mm2,%%mm1\n" /* mm1 = mm2 = pix[0-7] */ + "movq %%mm2,%%mm1\n" /* mm1 = mm2 = pix[0-7] */ - "punpckhbw %%mm0,%%mm1\n" /* mm1 = [pix4-7] */ - "punpcklbw %%mm0,%%mm2\n" /* mm2 = [pix0-3] */ + "punpckhbw %%mm0,%%mm1\n" /* mm1 = [pix4-7] */ + "punpcklbw %%mm0,%%mm2\n" /* mm2 = [pix0-3] */ - "movq %%mm3,%%mm4\n" /* mm4 = mm3 = pix[8-15] */ - "punpckhbw %%mm0,%%mm3\n" /* mm3 = [pix12-15] */ - "punpcklbw %%mm0,%%mm4\n" /* mm4 = [pix8-11] */ + "movq %%mm3,%%mm4\n" /* mm4 = mm3 = pix[8-15] */ + "punpckhbw %%mm0,%%mm3\n" /* mm3 = [pix12-15] */ + "punpcklbw %%mm0,%%mm4\n" /* mm4 = [pix8-11] */ - "pmaddwd %%mm1,%%mm1\n" /* mm1 = (pix0^2+pix1^2,pix2^2+pix3^2) */ - "pmaddwd %%mm2,%%mm2\n" /* mm2 = (pix4^2+pix5^2,pix6^2+pix7^2) */ + "pmaddwd %%mm1,%%mm1\n" /* mm1 = (pix0^2+pix1^2,pix2^2+pix3^2) */ + "pmaddwd %%mm2,%%mm2\n" /* mm2 = (pix4^2+pix5^2,pix6^2+pix7^2) */ "pmaddwd %%mm3,%%mm3\n" "pmaddwd %%mm4,%%mm4\n" - "paddd %%mm1,%%mm2\n" /* mm2 = (pix0^2+pix1^2+pix4^2+pix5^2, - pix2^2+pix3^2+pix6^2+pix7^2) */ + "paddd %%mm1,%%mm2\n" /* mm2 = (pix0^2+pix1^2+pix4^2+pix5^2, + pix2^2+pix3^2+pix6^2+pix7^2) */ "paddd %%mm3,%%mm4\n" "paddd %%mm2,%%mm7\n" @@ -734,7 +734,7 @@ static int pix_norm1_mmx(uint8_t *pix, int line_size) { "jnz 1b\n" "movq %%mm7,%%mm1\n" - "psrlq $32, %%mm7\n" /* shift hi dword to lo */ + "psrlq $32, %%mm7\n" /* shift hi dword to lo */ "paddd %%mm7,%%mm1\n" "movd %%mm1,%1\n" : "+r" (pix), "=r"(tmp) : "r" ((long)line_size) : "%ecx" ); @@ -746,13 +746,13 @@ static int sse8_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int asm volatile ( "movl %4,%%ecx\n" "shr $1,%%ecx\n" - "pxor %%mm0,%%mm0\n" /* mm0 = 0 */ - "pxor %%mm7,%%mm7\n" /* mm7 holds the sum */ + "pxor %%mm0,%%mm0\n" /* mm0 = 0 */ + "pxor %%mm7,%%mm7\n" /* mm7 holds the sum */ "1:\n" - "movq (%0),%%mm1\n" /* mm1 = pix1[0][0-7] */ - "movq (%1),%%mm2\n" /* mm2 = pix2[0][0-7] */ - "movq (%0,%3),%%mm3\n" /* mm3 = pix1[1][0-7] */ - "movq (%1,%3),%%mm4\n" /* mm4 = pix2[1][0-7] */ + "movq (%0),%%mm1\n" /* mm1 = pix1[0][0-7] */ + "movq (%1),%%mm2\n" /* mm2 = pix2[0][0-7] */ + "movq (%0,%3),%%mm3\n" /* mm3 = pix1[1][0-7] */ + "movq (%1,%3),%%mm4\n" /* mm4 = pix2[1][0-7] */ /* todo: mm1-mm2, mm3-mm4 */ /* algo: substract mm1 from mm2 with saturation and vice versa */ @@ -773,16 +773,16 @@ static int sse8_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int "punpckhbw %%mm0,%%mm2\n" "punpckhbw %%mm0,%%mm4\n" - "punpcklbw %%mm0,%%mm1\n" /* mm1 now spread over (mm1,mm2) */ - "punpcklbw %%mm0,%%mm3\n" /* mm4 now spread over (mm3,mm4) */ + "punpcklbw %%mm0,%%mm1\n" /* mm1 now spread over (mm1,mm2) */ + "punpcklbw %%mm0,%%mm3\n" /* mm4 now spread over (mm3,mm4) */ "pmaddwd %%mm2,%%mm2\n" "pmaddwd %%mm4,%%mm4\n" "pmaddwd %%mm1,%%mm1\n" "pmaddwd %%mm3,%%mm3\n" - "lea (%0,%3,2), %0\n" /* pix1 += 2*line_size */ - "lea (%1,%3,2), %1\n" /* pix2 += 2*line_size */ + "lea (%0,%3,2), %0\n" /* pix1 += 2*line_size */ + "lea (%1,%3,2), %1\n" /* pix2 += 2*line_size */ "paddd %%mm2,%%mm1\n" "paddd %%mm4,%%mm3\n" @@ -793,10 +793,10 @@ static int sse8_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int "jnz 1b\n" "movq %%mm7,%%mm1\n" - "psrlq $32, %%mm7\n" /* shift hi dword to lo */ + "psrlq $32, %%mm7\n" /* shift hi dword to lo */ "paddd %%mm7,%%mm1\n" "movd %%mm1,%2\n" - : "+r" (pix1), "+r" (pix2), "=r"(tmp) + : "+r" (pix1), "+r" (pix2), "=r"(tmp) : "r" ((long)line_size) , "m" (h) : "%ecx"); return tmp; @@ -806,13 +806,13 @@ static int sse16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int int tmp; asm volatile ( "movl %4,%%ecx\n" - "pxor %%mm0,%%mm0\n" /* mm0 = 0 */ - "pxor %%mm7,%%mm7\n" /* mm7 holds the sum */ + "pxor %%mm0,%%mm0\n" /* mm0 = 0 */ + "pxor %%mm7,%%mm7\n" /* mm7 holds the sum */ "1:\n" - "movq (%0),%%mm1\n" /* mm1 = pix1[0-7] */ - "movq (%1),%%mm2\n" /* mm2 = pix2[0-7] */ - "movq 8(%0),%%mm3\n" /* mm3 = pix1[8-15] */ - "movq 8(%1),%%mm4\n" /* mm4 = pix2[8-15] */ + "movq (%0),%%mm1\n" /* mm1 = pix1[0-7] */ + "movq (%1),%%mm2\n" /* mm2 = pix2[0-7] */ + "movq 8(%0),%%mm3\n" /* mm3 = pix1[8-15] */ + "movq 8(%1),%%mm4\n" /* mm4 = pix2[8-15] */ /* todo: mm1-mm2, mm3-mm4 */ /* algo: substract mm1 from mm2 with saturation and vice versa */ @@ -833,8 +833,8 @@ static int sse16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int "punpckhbw %%mm0,%%mm2\n" "punpckhbw %%mm0,%%mm4\n" - "punpcklbw %%mm0,%%mm1\n" /* mm1 now spread over (mm1,mm2) */ - "punpcklbw %%mm0,%%mm3\n" /* mm4 now spread over (mm3,mm4) */ + "punpcklbw %%mm0,%%mm1\n" /* mm1 now spread over (mm1,mm2) */ + "punpcklbw %%mm0,%%mm3\n" /* mm4 now spread over (mm3,mm4) */ "pmaddwd %%mm2,%%mm2\n" "pmaddwd %%mm4,%%mm4\n" @@ -853,10 +853,10 @@ static int sse16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int "jnz 1b\n" "movq %%mm7,%%mm1\n" - "psrlq $32, %%mm7\n" /* shift hi dword to lo */ + "psrlq $32, %%mm7\n" /* shift hi dword to lo */ "paddd %%mm7,%%mm1\n" "movd %%mm1,%2\n" - : "+r" (pix1), "+r" (pix2), "=r"(tmp) + : "+r" (pix1), "+r" (pix2), "=r"(tmp) : "r" ((long)line_size) , "m" (h) : "%ecx"); return tmp; @@ -866,13 +866,13 @@ static int sse16_sse2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in int tmp; asm volatile ( "shr $1,%2\n" - "pxor %%xmm0,%%xmm0\n" /* mm0 = 0 */ - "pxor %%xmm7,%%xmm7\n" /* mm7 holds the sum */ + "pxor %%xmm0,%%xmm0\n" /* mm0 = 0 */ + "pxor %%xmm7,%%xmm7\n" /* mm7 holds the sum */ "1:\n" - "movdqu (%0),%%xmm1\n" /* mm1 = pix1[0][0-15] */ - "movdqu (%1),%%xmm2\n" /* mm2 = pix2[0][0-15] */ - "movdqu (%0,%4),%%xmm3\n" /* mm3 = pix1[1][0-15] */ - "movdqu (%1,%4),%%xmm4\n" /* mm4 = pix2[1][0-15] */ + "movdqu (%0),%%xmm1\n" /* mm1 = pix1[0][0-15] */ + "movdqu (%1),%%xmm2\n" /* mm2 = pix2[0][0-15] */ + "movdqu (%0,%4),%%xmm3\n" /* mm3 = pix1[1][0-15] */ + "movdqu (%1,%4),%%xmm4\n" /* mm4 = pix2[1][0-15] */ /* todo: mm1-mm2, mm3-mm4 */ /* algo: substract mm1 from mm2 with saturation and vice versa */ @@ -893,16 +893,16 @@ static int sse16_sse2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in "punpckhbw %%xmm0,%%xmm2\n" "punpckhbw %%xmm0,%%xmm4\n" - "punpcklbw %%xmm0,%%xmm1\n" /* mm1 now spread over (mm1,mm2) */ - "punpcklbw %%xmm0,%%xmm3\n" /* mm4 now spread over (mm3,mm4) */ + "punpcklbw %%xmm0,%%xmm1\n" /* mm1 now spread over (mm1,mm2) */ + "punpcklbw %%xmm0,%%xmm3\n" /* mm4 now spread over (mm3,mm4) */ "pmaddwd %%xmm2,%%xmm2\n" "pmaddwd %%xmm4,%%xmm4\n" "pmaddwd %%xmm1,%%xmm1\n" "pmaddwd %%xmm3,%%xmm3\n" - "lea (%0,%4,2), %0\n" /* pix1 += 2*line_size */ - "lea (%1,%4,2), %1\n" /* pix2 += 2*line_size */ + "lea (%0,%4,2), %0\n" /* pix1 += 2*line_size */ + "lea (%1,%4,2), %1\n" /* pix2 += 2*line_size */ "paddd %%xmm2,%%xmm1\n" "paddd %%xmm4,%%xmm3\n" @@ -913,13 +913,13 @@ static int sse16_sse2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in "jnz 1b\n" "movdqa %%xmm7,%%xmm1\n" - "psrldq $8, %%xmm7\n" /* shift hi qword to lo */ + "psrldq $8, %%xmm7\n" /* shift hi qword to lo */ "paddd %%xmm1,%%xmm7\n" "movdqa %%xmm7,%%xmm1\n" - "psrldq $4, %%xmm7\n" /* shift hi dword to lo */ + "psrldq $4, %%xmm7\n" /* shift hi dword to lo */ "paddd %%xmm1,%%xmm7\n" "movd %%xmm7,%3\n" - : "+r" (pix1), "+r" (pix2), "+r"(h), "=r"(tmp) + : "+r" (pix1), "+r" (pix2), "+r"(h), "=r"(tmp) : "r" ((long)line_size)); return tmp; } @@ -930,7 +930,7 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) { "movl %3,%%ecx\n" "pxor %%mm7,%%mm7\n" "pxor %%mm6,%%mm6\n" - + "movq (%0),%%mm0\n" "movq %%mm0, %%mm1\n" "psllq $8, %%mm0\n" @@ -944,9 +944,9 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) { "punpckhbw %%mm7,%%mm3\n" "psubw %%mm1, %%mm0\n" "psubw %%mm3, %%mm2\n" - + "add %2,%0\n" - + "movq (%0),%%mm4\n" "movq %%mm4, %%mm1\n" "psllq $8, %%mm4\n" @@ -968,14 +968,14 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) { "pcmpgtw %%mm2, %%mm1\n\t" "pxor %%mm3, %%mm0\n" "pxor %%mm1, %%mm2\n" - "psubw %%mm3, %%mm0\n" + "psubw %%mm3, %%mm0\n" "psubw %%mm1, %%mm2\n" "paddw %%mm0, %%mm2\n" "paddw %%mm2, %%mm6\n" "add %2,%0\n" "1:\n" - + "movq (%0),%%mm0\n" "movq %%mm0, %%mm1\n" "psllq $8, %%mm0\n" @@ -997,13 +997,13 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) { "pcmpgtw %%mm5, %%mm1\n\t" "pxor %%mm3, %%mm4\n" "pxor %%mm1, %%mm5\n" - "psubw %%mm3, %%mm4\n" + "psubw %%mm3, %%mm4\n" "psubw %%mm1, %%mm5\n" "paddw %%mm4, %%mm5\n" "paddw %%mm5, %%mm6\n" - + "add %2,%0\n" - + "movq (%0),%%mm4\n" "movq %%mm4, %%mm1\n" "psllq $8, %%mm4\n" @@ -1025,7 +1025,7 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) { "pcmpgtw %%mm2, %%mm1\n\t" "pxor %%mm3, %%mm0\n" "pxor %%mm1, %%mm2\n" - "psubw %%mm3, %%mm0\n" + "psubw %%mm3, %%mm0\n" "psubw %%mm1, %%mm2\n" "paddw %%mm0, %%mm2\n" "paddw %%mm2, %%mm6\n" @@ -1038,12 +1038,12 @@ static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) { "punpcklwd %%mm7,%%mm0\n" "punpckhwd %%mm7,%%mm6\n" "paddd %%mm0, %%mm6\n" - + "movq %%mm6,%%mm0\n" "psrlq $32, %%mm6\n" "paddd %%mm6,%%mm0\n" "movd %%mm0,%1\n" - : "+r" (pix1), "=r"(tmp) + : "+r" (pix1), "=r"(tmp) : "r" ((long)line_size) , "g" (h-2) : "%ecx"); return tmp; @@ -1056,7 +1056,7 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) { "movl %3,%%ecx\n" "pxor %%mm7,%%mm7\n" "pxor %%mm6,%%mm6\n" - + "movq (%0),%%mm0\n" "movq 1(%0),%%mm1\n" "movq %%mm0, %%mm2\n" @@ -1067,9 +1067,9 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) { "punpckhbw %%mm7,%%mm3\n" "psubw %%mm1, %%mm0\n" "psubw %%mm3, %%mm2\n" - + "add %2,%0\n" - + "movq (%0),%%mm4\n" "movq 1(%0),%%mm1\n" "movq %%mm4, %%mm5\n" @@ -1088,14 +1088,14 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) { "pcmpgtw %%mm2, %%mm1\n\t" "pxor %%mm3, %%mm0\n" "pxor %%mm1, %%mm2\n" - "psubw %%mm3, %%mm0\n" + "psubw %%mm3, %%mm0\n" "psubw %%mm1, %%mm2\n" "paddw %%mm0, %%mm2\n" "paddw %%mm2, %%mm6\n" "add %2,%0\n" "1:\n" - + "movq (%0),%%mm0\n" "movq 1(%0),%%mm1\n" "movq %%mm0, %%mm2\n" @@ -1118,9 +1118,9 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) { "psubw %%mm1, %%mm5\n" "paddw %%mm4, %%mm5\n" "paddw %%mm5, %%mm6\n" - + "add %2,%0\n" - + "movq (%0),%%mm4\n" "movq 1(%0),%%mm1\n" "movq %%mm4, %%mm5\n" @@ -1139,7 +1139,7 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) { "pcmpgtw %%mm2, %%mm1\n\t" "pxor %%mm3, %%mm0\n" "pxor %%mm1, %%mm2\n" - "psubw %%mm3, %%mm0\n" + "psubw %%mm3, %%mm0\n" "psubw %%mm1, %%mm2\n" "paddw %%mm0, %%mm2\n" "paddw %%mm2, %%mm6\n" @@ -1152,12 +1152,12 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) { "punpcklwd %%mm7,%%mm0\n" "punpckhwd %%mm7,%%mm6\n" "paddd %%mm0, %%mm6\n" - + "movq %%mm6,%%mm0\n" "psrlq $32, %%mm6\n" "paddd %%mm6,%%mm0\n" "movd %%mm0,%1\n" - : "+r" (pix1), "=r"(tmp) + : "+r" (pix1), "=r"(tmp) : "r" ((long)line_size) , "g" (h-2) : "%ecx"); return tmp + hf_noise8_mmx(pix+8, line_size, h); @@ -1165,8 +1165,11 @@ static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) { static int nsse16_mmx(void *p, uint8_t * pix1, uint8_t * pix2, int line_size, int h) { MpegEncContext *c = p; - int score1= sse16_mmx(c, pix1, pix2, line_size, h); - int score2= hf_noise16_mmx(pix1, line_size, h) - hf_noise16_mmx(pix2, line_size, h); + int score1, score2; + + if(c) score1 = c->dsp.sse[0](c, pix1, pix2, line_size, h); + else score1 = sse16_mmx(c, pix1, pix2, line_size, h); + score2= hf_noise16_mmx(pix1, line_size, h) - hf_noise16_mmx(pix2, line_size, h); if(c) return score1 + ABS(score2)*c->avctx->nsse_weight; else return score1 + ABS(score2)*8; @@ -1183,10 +1186,10 @@ static int nsse8_mmx(void *p, uint8_t * pix1, uint8_t * pix2, int line_size, int static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_size, int h) { int tmp; - + assert( (((int)pix) & 7) == 0); assert((line_size &7) ==0); - + #define SUM(in0, in1, out0, out1) \ "movq (%0), %%mm2\n"\ "movq 8(%0), %%mm3\n"\ @@ -1210,7 +1213,7 @@ static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_si "paddw %%mm2, " #in0 "\n"\ "paddw " #in0 ", %%mm6\n" - + asm volatile ( "movl %3,%%ecx\n" "pxor %%mm6,%%mm6\n" @@ -1221,11 +1224,11 @@ static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_si "subl $2, %%ecx\n" SUM(%%mm0, %%mm1, %%mm4, %%mm5) "1:\n" - + SUM(%%mm4, %%mm5, %%mm0, %%mm1) - + SUM(%%mm0, %%mm1, %%mm4, %%mm5) - + "subl $2, %%ecx\n" "jnz 1b\n" @@ -1236,7 +1239,7 @@ static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_si "psrlq $16, %%mm0\n" "paddw %%mm6,%%mm0\n" "movd %%mm0,%1\n" - : "+r" (pix), "=r"(tmp) + : "+r" (pix), "=r"(tmp) : "r" ((long)line_size) , "m" (h) : "%ecx"); return tmp & 0xFFFF; @@ -1245,10 +1248,10 @@ static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_si static int vsad_intra16_mmx2(void *v, uint8_t * pix, uint8_t * dummy, int line_size, int h) { int tmp; - + assert( (((int)pix) & 7) == 0); assert((line_size &7) ==0); - + #define SUM(in0, in1, out0, out1) \ "movq (%0), " #out0 "\n"\ "movq 8(%0), " #out1 "\n"\ @@ -1268,16 +1271,16 @@ static int vsad_intra16_mmx2(void *v, uint8_t * pix, uint8_t * dummy, int line_s "subl $2, %%ecx\n" SUM(%%mm0, %%mm1, %%mm4, %%mm5) "1:\n" - + SUM(%%mm4, %%mm5, %%mm0, %%mm1) - + SUM(%%mm0, %%mm1, %%mm4, %%mm5) - + "subl $2, %%ecx\n" "jnz 1b\n" "movd %%mm6,%1\n" - : "+r" (pix), "=r"(tmp) + : "+r" (pix), "=r"(tmp) : "r" ((long)line_size) , "m" (h) : "%ecx"); return tmp; @@ -1286,11 +1289,11 @@ static int vsad_intra16_mmx2(void *v, uint8_t * pix, uint8_t * dummy, int line_s static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h) { int tmp; - + assert( (((int)pix1) & 7) == 0); assert( (((int)pix2) & 7) == 0); assert((line_size &7) ==0); - + #define SUM(in0, in1, out0, out1) \ "movq (%0),%%mm2\n"\ "movq (%1)," #out0 "\n"\ @@ -1321,7 +1324,7 @@ static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in "paddw %%mm2, " #in0 "\n"\ "paddw " #in0 ", %%mm6\n" - + asm volatile ( "movl %4,%%ecx\n" "pxor %%mm6,%%mm6\n" @@ -1341,11 +1344,11 @@ static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in "pxor %%mm7, %%mm1\n" SUM(%%mm0, %%mm1, %%mm4, %%mm5) "1:\n" - + SUM(%%mm4, %%mm5, %%mm0, %%mm1) - + SUM(%%mm0, %%mm1, %%mm4, %%mm5) - + "subl $2, %%ecx\n" "jnz 1b\n" @@ -1356,7 +1359,7 @@ static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in "psrlq $16, %%mm0\n" "paddw %%mm6,%%mm0\n" "movd %%mm0,%2\n" - : "+r" (pix1), "+r" (pix2), "=r"(tmp) + : "+r" (pix1), "+r" (pix2), "=r"(tmp) : "r" ((long)line_size) , "m" (h) : "%ecx"); return tmp & 0x7FFF; @@ -1365,11 +1368,11 @@ static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in static int vsad16_mmx2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h) { int tmp; - + assert( (((int)pix1) & 7) == 0); assert( (((int)pix2) & 7) == 0); assert((line_size &7) ==0); - + #define SUM(in0, in1, out0, out1) \ "movq (%0)," #out0 "\n"\ "movq (%1),%%mm2\n"\ @@ -1405,16 +1408,16 @@ static int vsad16_mmx2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, i "pxor %%mm7, %%mm1\n" SUM(%%mm0, %%mm1, %%mm4, %%mm5) "1:\n" - + SUM(%%mm4, %%mm5, %%mm0, %%mm1) - + SUM(%%mm0, %%mm1, %%mm4, %%mm5) - + "subl $2, %%ecx\n" "jnz 1b\n" "movd %%mm6,%2\n" - : "+r" (pix1), "+r" (pix2), "=r"(tmp) + : "+r" (pix1), "+r" (pix2), "=r"(tmp) : "r" ((long)line_size) , "m" (h) : "%ecx"); return tmp; @@ -1424,18 +1427,18 @@ static int vsad16_mmx2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, i static void diff_bytes_mmx(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){ long i=0; asm volatile( - "1: \n\t" - "movq (%2, %0), %%mm0 \n\t" - "movq (%1, %0), %%mm1 \n\t" - "psubb %%mm0, %%mm1 \n\t" - "movq %%mm1, (%3, %0) \n\t" - "movq 8(%2, %0), %%mm0 \n\t" - "movq 8(%1, %0), %%mm1 \n\t" - "psubb %%mm0, %%mm1 \n\t" - "movq %%mm1, 8(%3, %0) \n\t" - "add $16, %0 \n\t" - "cmp %4, %0 \n\t" - " jb 1b \n\t" + "1: \n\t" + "movq (%2, %0), %%mm0 \n\t" + "movq (%1, %0), %%mm1 \n\t" + "psubb %%mm0, %%mm1 \n\t" + "movq %%mm1, (%3, %0) \n\t" + "movq 8(%2, %0), %%mm0 \n\t" + "movq 8(%1, %0), %%mm1 \n\t" + "psubb %%mm0, %%mm1 \n\t" + "movq %%mm1, 8(%3, %0) \n\t" + "add $16, %0 \n\t" + "cmp %4, %0 \n\t" + " jb 1b \n\t" : "+r" (i) : "r"(src1), "r"(src2), "r"(dst), "r"((long)w-15) ); @@ -1446,46 +1449,46 @@ static void diff_bytes_mmx(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){ static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top){ long i=0; uint8_t l, lt; - + asm volatile( - "1: \n\t" - "movq -1(%1, %0), %%mm0 \n\t" // LT - "movq (%1, %0), %%mm1 \n\t" // T - "movq -1(%2, %0), %%mm2 \n\t" // L - "movq (%2, %0), %%mm3 \n\t" // X - "movq %%mm2, %%mm4 \n\t" // L - "psubb %%mm0, %%mm2 \n\t" - "paddb %%mm1, %%mm2 \n\t" // L + T - LT - "movq %%mm4, %%mm5 \n\t" // L - "pmaxub %%mm1, %%mm4 \n\t" // max(T, L) - "pminub %%mm5, %%mm1 \n\t" // min(T, L) - "pminub %%mm2, %%mm4 \n\t" - "pmaxub %%mm1, %%mm4 \n\t" - "psubb %%mm4, %%mm3 \n\t" // dst - pred - "movq %%mm3, (%3, %0) \n\t" - "add $8, %0 \n\t" - "cmp %4, %0 \n\t" - " jb 1b \n\t" + "1: \n\t" + "movq -1(%1, %0), %%mm0 \n\t" // LT + "movq (%1, %0), %%mm1 \n\t" // T + "movq -1(%2, %0), %%mm2 \n\t" // L + "movq (%2, %0), %%mm3 \n\t" // X + "movq %%mm2, %%mm4 \n\t" // L + "psubb %%mm0, %%mm2 \n\t" + "paddb %%mm1, %%mm2 \n\t" // L + T - LT + "movq %%mm4, %%mm5 \n\t" // L + "pmaxub %%mm1, %%mm4 \n\t" // max(T, L) + "pminub %%mm5, %%mm1 \n\t" // min(T, L) + "pminub %%mm2, %%mm4 \n\t" + "pmaxub %%mm1, %%mm4 \n\t" + "psubb %%mm4, %%mm3 \n\t" // dst - pred + "movq %%mm3, (%3, %0) \n\t" + "add $8, %0 \n\t" + "cmp %4, %0 \n\t" + " jb 1b \n\t" : "+r" (i) : "r"(src1), "r"(src2), "r"(dst), "r"((long)w) ); l= *left; lt= *left_top; - + dst[0]= src2[0] - mid_pred(l, src1[0], (l + src1[0] - lt)&0xFF); - + *left_top= src1[w-1]; *left = src2[w-1]; } #define LBUTTERFLY2(a1,b1,a2,b2)\ - "paddw " #b1 ", " #a1 " \n\t"\ - "paddw " #b2 ", " #a2 " \n\t"\ - "paddw " #b1 ", " #b1 " \n\t"\ - "paddw " #b2 ", " #b2 " \n\t"\ - "psubw " #a1 ", " #b1 " \n\t"\ - "psubw " #a2 ", " #b2 " \n\t" + "paddw " #b1 ", " #a1 " \n\t"\ + "paddw " #b2 ", " #a2 " \n\t"\ + "paddw " #b1 ", " #b1 " \n\t"\ + "paddw " #b2 ", " #b2 " \n\t"\ + "psubw " #a1 ", " #b1 " \n\t"\ + "psubw " #a2 ", " #b2 " \n\t" #define HADAMARD48\ LBUTTERFLY2(%%mm0, %%mm1, %%mm2, %%mm3)\ @@ -1496,33 +1499,33 @@ static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *src1, uint8_t LBUTTERFLY2(%%mm2, %%mm6, %%mm3, %%mm7)\ #define MMABS(a,z)\ - "pxor " #z ", " #z " \n\t"\ - "pcmpgtw " #a ", " #z " \n\t"\ - "pxor " #z ", " #a " \n\t"\ - "psubw " #z ", " #a " \n\t" + "pxor " #z ", " #z " \n\t"\ + "pcmpgtw " #a ", " #z " \n\t"\ + "pxor " #z ", " #a " \n\t"\ + "psubw " #z ", " #a " \n\t" #define MMABS_SUM(a,z, sum)\ - "pxor " #z ", " #z " \n\t"\ - "pcmpgtw " #a ", " #z " \n\t"\ - "pxor " #z ", " #a " \n\t"\ - "psubw " #z ", " #a " \n\t"\ - "paddusw " #a ", " #sum " \n\t" + "pxor " #z ", " #z " \n\t"\ + "pcmpgtw " #a ", " #z " \n\t"\ + "pxor " #z ", " #a " \n\t"\ + "psubw " #z ", " #a " \n\t"\ + "paddusw " #a ", " #sum " \n\t" #define MMABS_MMX2(a,z)\ - "pxor " #z ", " #z " \n\t"\ - "psubw " #a ", " #z " \n\t"\ - "pmaxsw " #z ", " #a " \n\t" + "pxor " #z ", " #z " \n\t"\ + "psubw " #a ", " #z " \n\t"\ + "pmaxsw " #z ", " #a " \n\t" #define MMABS_SUM_MMX2(a,z, sum)\ - "pxor " #z ", " #z " \n\t"\ - "psubw " #a ", " #z " \n\t"\ - "pmaxsw " #z ", " #a " \n\t"\ - "paddusw " #a ", " #sum " \n\t" - + "pxor " #z ", " #z " \n\t"\ + "psubw " #a ", " #z " \n\t"\ + "pmaxsw " #z ", " #a " \n\t"\ + "paddusw " #a ", " #sum " \n\t" + #define SBUTTERFLY(a,b,t,n)\ - "movq " #a ", " #t " \n\t" /* abcd */\ - "punpckl" #n " " #b ", " #a " \n\t" /* aebf */\ - "punpckh" #n " " #b ", " #t " \n\t" /* cgdh */\ + "movq " #a ", " #t " \n\t" /* abcd */\ + "punpckl" #n " " #b ", " #a " \n\t" /* aebf */\ + "punpckh" #n " " #b ", " #t " \n\t" /* cgdh */\ #define TRANSPOSE4(a,b,c,d,t)\ SBUTTERFLY(a,b,t,wd) /* a=aebf t=cgdh */\ @@ -1531,21 +1534,21 @@ static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, uint8_t *src1, uint8_t SBUTTERFLY(t,b,c,dq) /* t=cgko c=dhlp */ #define LOAD4(o, a, b, c, d)\ - "movq "#o"(%1), " #a " \n\t"\ - "movq "#o"+16(%1), " #b " \n\t"\ - "movq "#o"+32(%1), " #c " \n\t"\ - "movq "#o"+48(%1), " #d " \n\t" + "movq "#o"(%1), " #a " \n\t"\ + "movq "#o"+16(%1), " #b " \n\t"\ + "movq "#o"+32(%1), " #c " \n\t"\ + "movq "#o"+48(%1), " #d " \n\t" #define STORE4(o, a, b, c, d)\ - "movq "#a", "#o"(%1) \n\t"\ - "movq "#b", "#o"+16(%1) \n\t"\ - "movq "#c", "#o"+32(%1) \n\t"\ - "movq "#d", "#o"+48(%1) \n\t"\ + "movq "#a", "#o"(%1) \n\t"\ + "movq "#b", "#o"+16(%1) \n\t"\ + "movq "#c", "#o"+32(%1) \n\t"\ + "movq "#d", "#o"+48(%1) \n\t"\ static int hadamard8_diff_mmx(void *s, uint8_t *src1, uint8_t *src2, int stride, int h){ - uint64_t temp[16] __align8; + DECLARE_ALIGNED_8(uint64_t, temp[16]); int sum=0; - + assert(h==8); diff_pixels_mmx((DCTELEM*)temp, src1, src2, stride); @@ -1553,40 +1556,40 @@ static int hadamard8_diff_mmx(void *s, uint8_t *src1, uint8_t *src2, int stride, asm volatile( LOAD4(0 , %%mm0, %%mm1, %%mm2, %%mm3) LOAD4(64, %%mm4, %%mm5, %%mm6, %%mm7) - + HADAMARD48 - - "movq %%mm7, 112(%1) \n\t" - + + "movq %%mm7, 112(%1) \n\t" + TRANSPOSE4(%%mm0, %%mm1, %%mm2, %%mm3, %%mm7) STORE4(0 , %%mm0, %%mm3, %%mm7, %%mm2) - - "movq 112(%1), %%mm7 \n\t" + + "movq 112(%1), %%mm7 \n\t" TRANSPOSE4(%%mm4, %%mm5, %%mm6, %%mm7, %%mm0) STORE4(64, %%mm4, %%mm7, %%mm0, %%mm6) LOAD4(8 , %%mm0, %%mm1, %%mm2, %%mm3) LOAD4(72, %%mm4, %%mm5, %%mm6, %%mm7) - + HADAMARD48 - - "movq %%mm7, 120(%1) \n\t" - + + "movq %%mm7, 120(%1) \n\t" + TRANSPOSE4(%%mm0, %%mm1, %%mm2, %%mm3, %%mm7) STORE4(8 , %%mm0, %%mm3, %%mm7, %%mm2) - - "movq 120(%1), %%mm7 \n\t" + + "movq 120(%1), %%mm7 \n\t" TRANSPOSE4(%%mm4, %%mm5, %%mm6, %%mm7, %%mm0) - "movq %%mm7, %%mm5 \n\t"//FIXME remove - "movq %%mm6, %%mm7 \n\t" - "movq %%mm0, %%mm6 \n\t" + "movq %%mm7, %%mm5 \n\t"//FIXME remove + "movq %%mm6, %%mm7 \n\t" + "movq %%mm0, %%mm6 \n\t" // STORE4(72, %%mm4, %%mm7, %%mm0, %%mm6) //FIXME remove - + LOAD4(64, %%mm0, %%mm1, %%mm2, %%mm3) // LOAD4(72, %%mm4, %%mm5, %%mm6, %%mm7) - + HADAMARD48 - "movq %%mm7, 64(%1) \n\t" + "movq %%mm7, 64(%1) \n\t" MMABS(%%mm0, %%mm7) MMABS_SUM(%%mm1, %%mm7, %%mm0) MMABS_SUM(%%mm2, %%mm7, %%mm0) @@ -1594,15 +1597,15 @@ static int hadamard8_diff_mmx(void *s, uint8_t *src1, uint8_t *src2, int stride, MMABS_SUM(%%mm4, %%mm7, %%mm0) MMABS_SUM(%%mm5, %%mm7, %%mm0) MMABS_SUM(%%mm6, %%mm7, %%mm0) - "movq 64(%1), %%mm1 \n\t" + "movq 64(%1), %%mm1 \n\t" MMABS_SUM(%%mm1, %%mm7, %%mm0) - "movq %%mm0, 64(%1) \n\t" - + "movq %%mm0, 64(%1) \n\t" + LOAD4(0 , %%mm0, %%mm1, %%mm2, %%mm3) LOAD4(8 , %%mm4, %%mm5, %%mm6, %%mm7) - + HADAMARD48 - "movq %%mm7, (%1) \n\t" + "movq %%mm7, (%1) \n\t" MMABS(%%mm0, %%mm7) MMABS_SUM(%%mm1, %%mm7, %%mm0) MMABS_SUM(%%mm2, %%mm7, %%mm0) @@ -1610,19 +1613,19 @@ static int hadamard8_diff_mmx(void *s, uint8_t *src1, uint8_t *src2, int stride, MMABS_SUM(%%mm4, %%mm7, %%mm0) MMABS_SUM(%%mm5, %%mm7, %%mm0) MMABS_SUM(%%mm6, %%mm7, %%mm0) - "movq (%1), %%mm1 \n\t" + "movq (%1), %%mm1 \n\t" MMABS_SUM(%%mm1, %%mm7, %%mm0) - "movq 64(%1), %%mm1 \n\t" + "movq 64(%1), %%mm1 \n\t" MMABS_SUM(%%mm1, %%mm7, %%mm0) - - "movq %%mm0, %%mm1 \n\t" - "psrlq $32, %%mm0 \n\t" - "paddusw %%mm1, %%mm0 \n\t" - "movq %%mm0, %%mm1 \n\t" - "psrlq $16, %%mm0 \n\t" - "paddusw %%mm1, %%mm0 \n\t" - "movd %%mm0, %0 \n\t" - + + "movq %%mm0, %%mm1 \n\t" + "psrlq $32, %%mm0 \n\t" + "paddusw %%mm1, %%mm0 \n\t" + "movq %%mm0, %%mm1 \n\t" + "psrlq $16, %%mm0 \n\t" + "paddusw %%mm1, %%mm0 \n\t" + "movd %%mm0, %0 \n\t" + : "=r" (sum) : "r"(temp) ); @@ -1630,9 +1633,9 @@ static int hadamard8_diff_mmx(void *s, uint8_t *src1, uint8_t *src2, int stride, } static int hadamard8_diff_mmx2(void *s, uint8_t *src1, uint8_t *src2, int stride, int h){ - uint64_t temp[16] __align8; + DECLARE_ALIGNED_8(uint64_t, temp[16]); int sum=0; - + assert(h==8); diff_pixels_mmx((DCTELEM*)temp, src1, src2, stride); @@ -1640,40 +1643,40 @@ static int hadamard8_diff_mmx2(void *s, uint8_t *src1, uint8_t *src2, int stride asm volatile( LOAD4(0 , %%mm0, %%mm1, %%mm2, %%mm3) LOAD4(64, %%mm4, %%mm5, %%mm6, %%mm7) - + HADAMARD48 - - "movq %%mm7, 112(%1) \n\t" - + + "movq %%mm7, 112(%1) \n\t" + TRANSPOSE4(%%mm0, %%mm1, %%mm2, %%mm3, %%mm7) STORE4(0 , %%mm0, %%mm3, %%mm7, %%mm2) - - "movq 112(%1), %%mm7 \n\t" + + "movq 112(%1), %%mm7 \n\t" TRANSPOSE4(%%mm4, %%mm5, %%mm6, %%mm7, %%mm0) STORE4(64, %%mm4, %%mm7, %%mm0, %%mm6) LOAD4(8 , %%mm0, %%mm1, %%mm2, %%mm3) LOAD4(72, %%mm4, %%mm5, %%mm6, %%mm7) - + HADAMARD48 - - "movq %%mm7, 120(%1) \n\t" - + + "movq %%mm7, 120(%1) \n\t" + TRANSPOSE4(%%mm0, %%mm1, %%mm2, %%mm3, %%mm7) STORE4(8 , %%mm0, %%mm3, %%mm7, %%mm2) - - "movq 120(%1), %%mm7 \n\t" + + "movq 120(%1), %%mm7 \n\t" TRANSPOSE4(%%mm4, %%mm5, %%mm6, %%mm7, %%mm0) - "movq %%mm7, %%mm5 \n\t"//FIXME remove - "movq %%mm6, %%mm7 \n\t" - "movq %%mm0, %%mm6 \n\t" + "movq %%mm7, %%mm5 \n\t"//FIXME remove + "movq %%mm6, %%mm7 \n\t" + "movq %%mm0, %%mm6 \n\t" // STORE4(72, %%mm4, %%mm7, %%mm0, %%mm6) //FIXME remove - + LOAD4(64, %%mm0, %%mm1, %%mm2, %%mm3) // LOAD4(72, %%mm4, %%mm5, %%mm6, %%mm7) - + HADAMARD48 - "movq %%mm7, 64(%1) \n\t" + "movq %%mm7, 64(%1) \n\t" MMABS_MMX2(%%mm0, %%mm7) MMABS_SUM_MMX2(%%mm1, %%mm7, %%mm0) MMABS_SUM_MMX2(%%mm2, %%mm7, %%mm0) @@ -1681,15 +1684,15 @@ static int hadamard8_diff_mmx2(void *s, uint8_t *src1, uint8_t *src2, int stride MMABS_SUM_MMX2(%%mm4, %%mm7, %%mm0) MMABS_SUM_MMX2(%%mm5, %%mm7, %%mm0) MMABS_SUM_MMX2(%%mm6, %%mm7, %%mm0) - "movq 64(%1), %%mm1 \n\t" + "movq 64(%1), %%mm1 \n\t" MMABS_SUM_MMX2(%%mm1, %%mm7, %%mm0) - "movq %%mm0, 64(%1) \n\t" - + "movq %%mm0, 64(%1) \n\t" + LOAD4(0 , %%mm0, %%mm1, %%mm2, %%mm3) LOAD4(8 , %%mm4, %%mm5, %%mm6, %%mm7) - + HADAMARD48 - "movq %%mm7, (%1) \n\t" + "movq %%mm7, (%1) \n\t" MMABS_MMX2(%%mm0, %%mm7) MMABS_SUM_MMX2(%%mm1, %%mm7, %%mm0) MMABS_SUM_MMX2(%%mm2, %%mm7, %%mm0) @@ -1697,17 +1700,17 @@ static int hadamard8_diff_mmx2(void *s, uint8_t *src1, uint8_t *src2, int stride MMABS_SUM_MMX2(%%mm4, %%mm7, %%mm0) MMABS_SUM_MMX2(%%mm5, %%mm7, %%mm0) MMABS_SUM_MMX2(%%mm6, %%mm7, %%mm0) - "movq (%1), %%mm1 \n\t" + "movq (%1), %%mm1 \n\t" MMABS_SUM_MMX2(%%mm1, %%mm7, %%mm0) - "movq 64(%1), %%mm1 \n\t" + "movq 64(%1), %%mm1 \n\t" MMABS_SUM_MMX2(%%mm1, %%mm7, %%mm0) - + "pshufw $0x0E, %%mm0, %%mm1 \n\t" - "paddusw %%mm1, %%mm0 \n\t" + "paddusw %%mm1, %%mm0 \n\t" "pshufw $0x01, %%mm0, %%mm1 \n\t" - "paddusw %%mm1, %%mm0 \n\t" - "movd %%mm0, %0 \n\t" - + "paddusw %%mm1, %%mm0 \n\t" + "movd %%mm0, %0 \n\t" + : "=r" (sum) : "r"(temp) ); @@ -1723,24 +1726,24 @@ WARPER8_16_SQ(hadamard8_diff_mmx2, hadamard8_diff16_mmx2) #define put_no_rnd_pixels16_mmx(a,b,c,d) put_pixels16_mmx(a,b,c,d) #define QPEL_V_LOW(m3,m4,m5,m6, pw_20, pw_3, rnd, in0, in1, in2, in7, out, OP)\ - "paddw " #m4 ", " #m3 " \n\t" /* x1 */\ - "movq "MANGLE(ff_pw_20)", %%mm4 \n\t" /* 20 */\ - "pmullw " #m3 ", %%mm4 \n\t" /* 20x1 */\ - "movq "#in7", " #m3 " \n\t" /* d */\ - "movq "#in0", %%mm5 \n\t" /* D */\ - "paddw " #m3 ", %%mm5 \n\t" /* x4 */\ - "psubw %%mm5, %%mm4 \n\t" /* 20x1 - x4 */\ - "movq "#in1", %%mm5 \n\t" /* C */\ - "movq "#in2", %%mm6 \n\t" /* B */\ - "paddw " #m6 ", %%mm5 \n\t" /* x3 */\ - "paddw " #m5 ", %%mm6 \n\t" /* x2 */\ - "paddw %%mm6, %%mm6 \n\t" /* 2x2 */\ - "psubw %%mm6, %%mm5 \n\t" /* -2x2 + x3 */\ - "pmullw "MANGLE(ff_pw_3)", %%mm5 \n\t" /* -6x2 + 3x3 */\ - "paddw " #rnd ", %%mm4 \n\t" /* x2 */\ - "paddw %%mm4, %%mm5 \n\t" /* 20x1 - 6x2 + 3x3 - x4 */\ - "psraw $5, %%mm5 \n\t"\ - "packuswb %%mm5, %%mm5 \n\t"\ + "paddw " #m4 ", " #m3 " \n\t" /* x1 */\ + "movq "MANGLE(ff_pw_20)", %%mm4 \n\t" /* 20 */\ + "pmullw " #m3 ", %%mm4 \n\t" /* 20x1 */\ + "movq "#in7", " #m3 " \n\t" /* d */\ + "movq "#in0", %%mm5 \n\t" /* D */\ + "paddw " #m3 ", %%mm5 \n\t" /* x4 */\ + "psubw %%mm5, %%mm4 \n\t" /* 20x1 - x4 */\ + "movq "#in1", %%mm5 \n\t" /* C */\ + "movq "#in2", %%mm6 \n\t" /* B */\ + "paddw " #m6 ", %%mm5 \n\t" /* x3 */\ + "paddw " #m5 ", %%mm6 \n\t" /* x2 */\ + "paddw %%mm6, %%mm6 \n\t" /* 2x2 */\ + "psubw %%mm6, %%mm5 \n\t" /* -2x2 + x3 */\ + "pmullw "MANGLE(ff_pw_3)", %%mm5 \n\t" /* -6x2 + 3x3 */\ + "paddw " #rnd ", %%mm4 \n\t" /* x2 */\ + "paddw %%mm4, %%mm5 \n\t" /* 20x1 - 6x2 + 3x3 - x4 */\ + "psraw $5, %%mm5 \n\t"\ + "packuswb %%mm5, %%mm5 \n\t"\ OP(%%mm5, out, %%mm7, d) #define QPEL_BASE(OPNAME, ROUNDER, RND, OP_MMX2, OP_3DNOW)\ @@ -1748,116 +1751,116 @@ static void OPNAME ## mpeg4_qpel16_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, in uint64_t temp;\ \ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t" /* ABCDEFGH */\ - "movq %%mm0, %%mm1 \n\t" /* ABCDEFGH */\ - "movq %%mm0, %%mm2 \n\t" /* ABCDEFGH */\ - "punpcklbw %%mm7, %%mm0 \n\t" /* 0A0B0C0D */\ - "punpckhbw %%mm7, %%mm1 \n\t" /* 0E0F0G0H */\ - "pshufw $0x90, %%mm0, %%mm5 \n\t" /* 0A0A0B0C */\ - "pshufw $0x41, %%mm0, %%mm6 \n\t" /* 0B0A0A0B */\ - "movq %%mm2, %%mm3 \n\t" /* ABCDEFGH */\ - "movq %%mm2, %%mm4 \n\t" /* ABCDEFGH */\ - "psllq $8, %%mm2 \n\t" /* 0ABCDEFG */\ - "psllq $16, %%mm3 \n\t" /* 00ABCDEF */\ - "psllq $24, %%mm4 \n\t" /* 000ABCDE */\ - "punpckhbw %%mm7, %%mm2 \n\t" /* 0D0E0F0G */\ - "punpckhbw %%mm7, %%mm3 \n\t" /* 0C0D0E0F */\ - "punpckhbw %%mm7, %%mm4 \n\t" /* 0B0C0D0E */\ - "paddw %%mm3, %%mm5 \n\t" /* b */\ - "paddw %%mm2, %%mm6 \n\t" /* c */\ - "paddw %%mm5, %%mm5 \n\t" /* 2b */\ - "psubw %%mm5, %%mm6 \n\t" /* c - 2b */\ - "pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\ - "pmullw "MANGLE(ff_pw_3)", %%mm6 \n\t" /* 3c - 6b */\ - "paddw %%mm4, %%mm0 \n\t" /* a */\ - "paddw %%mm1, %%mm5 \n\t" /* d */\ - "pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\ - "psubw %%mm5, %%mm0 \n\t" /* 20a - d */\ - "paddw %6, %%mm6 \n\t"\ - "paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\ - "psraw $5, %%mm0 \n\t"\ - "movq %%mm0, %5 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t" /* ABCDEFGH */\ + "movq %%mm0, %%mm1 \n\t" /* ABCDEFGH */\ + "movq %%mm0, %%mm2 \n\t" /* ABCDEFGH */\ + "punpcklbw %%mm7, %%mm0 \n\t" /* 0A0B0C0D */\ + "punpckhbw %%mm7, %%mm1 \n\t" /* 0E0F0G0H */\ + "pshufw $0x90, %%mm0, %%mm5 \n\t" /* 0A0A0B0C */\ + "pshufw $0x41, %%mm0, %%mm6 \n\t" /* 0B0A0A0B */\ + "movq %%mm2, %%mm3 \n\t" /* ABCDEFGH */\ + "movq %%mm2, %%mm4 \n\t" /* ABCDEFGH */\ + "psllq $8, %%mm2 \n\t" /* 0ABCDEFG */\ + "psllq $16, %%mm3 \n\t" /* 00ABCDEF */\ + "psllq $24, %%mm4 \n\t" /* 000ABCDE */\ + "punpckhbw %%mm7, %%mm2 \n\t" /* 0D0E0F0G */\ + "punpckhbw %%mm7, %%mm3 \n\t" /* 0C0D0E0F */\ + "punpckhbw %%mm7, %%mm4 \n\t" /* 0B0C0D0E */\ + "paddw %%mm3, %%mm5 \n\t" /* b */\ + "paddw %%mm2, %%mm6 \n\t" /* c */\ + "paddw %%mm5, %%mm5 \n\t" /* 2b */\ + "psubw %%mm5, %%mm6 \n\t" /* c - 2b */\ + "pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\ + "pmullw "MANGLE(ff_pw_3)", %%mm6 \n\t" /* 3c - 6b */\ + "paddw %%mm4, %%mm0 \n\t" /* a */\ + "paddw %%mm1, %%mm5 \n\t" /* d */\ + "pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\ + "psubw %%mm5, %%mm0 \n\t" /* 20a - d */\ + "paddw %6, %%mm6 \n\t"\ + "paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\ + "psraw $5, %%mm0 \n\t"\ + "movq %%mm0, %5 \n\t"\ /* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\ \ - "movq 5(%0), %%mm0 \n\t" /* FGHIJKLM */\ - "movq %%mm0, %%mm5 \n\t" /* FGHIJKLM */\ - "movq %%mm0, %%mm6 \n\t" /* FGHIJKLM */\ - "psrlq $8, %%mm0 \n\t" /* GHIJKLM0 */\ - "psrlq $16, %%mm5 \n\t" /* HIJKLM00 */\ - "punpcklbw %%mm7, %%mm0 \n\t" /* 0G0H0I0J */\ - "punpcklbw %%mm7, %%mm5 \n\t" /* 0H0I0J0K */\ - "paddw %%mm0, %%mm2 \n\t" /* b */\ - "paddw %%mm5, %%mm3 \n\t" /* c */\ - "paddw %%mm2, %%mm2 \n\t" /* 2b */\ - "psubw %%mm2, %%mm3 \n\t" /* c - 2b */\ - "movq %%mm6, %%mm2 \n\t" /* FGHIJKLM */\ - "psrlq $24, %%mm6 \n\t" /* IJKLM000 */\ - "punpcklbw %%mm7, %%mm2 \n\t" /* 0F0G0H0I */\ - "punpcklbw %%mm7, %%mm6 \n\t" /* 0I0J0K0L */\ - "pmullw "MANGLE(ff_pw_3)", %%mm3 \n\t" /* 3c - 6b */\ - "paddw %%mm2, %%mm1 \n\t" /* a */\ - "paddw %%mm6, %%mm4 \n\t" /* d */\ - "pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\ - "psubw %%mm4, %%mm3 \n\t" /* - 6b +3c - d */\ - "paddw %6, %%mm1 \n\t"\ - "paddw %%mm1, %%mm3 \n\t" /* 20a - 6b +3c - d */\ - "psraw $5, %%mm3 \n\t"\ - "movq %5, %%mm1 \n\t"\ - "packuswb %%mm3, %%mm1 \n\t"\ + "movq 5(%0), %%mm0 \n\t" /* FGHIJKLM */\ + "movq %%mm0, %%mm5 \n\t" /* FGHIJKLM */\ + "movq %%mm0, %%mm6 \n\t" /* FGHIJKLM */\ + "psrlq $8, %%mm0 \n\t" /* GHIJKLM0 */\ + "psrlq $16, %%mm5 \n\t" /* HIJKLM00 */\ + "punpcklbw %%mm7, %%mm0 \n\t" /* 0G0H0I0J */\ + "punpcklbw %%mm7, %%mm5 \n\t" /* 0H0I0J0K */\ + "paddw %%mm0, %%mm2 \n\t" /* b */\ + "paddw %%mm5, %%mm3 \n\t" /* c */\ + "paddw %%mm2, %%mm2 \n\t" /* 2b */\ + "psubw %%mm2, %%mm3 \n\t" /* c - 2b */\ + "movq %%mm6, %%mm2 \n\t" /* FGHIJKLM */\ + "psrlq $24, %%mm6 \n\t" /* IJKLM000 */\ + "punpcklbw %%mm7, %%mm2 \n\t" /* 0F0G0H0I */\ + "punpcklbw %%mm7, %%mm6 \n\t" /* 0I0J0K0L */\ + "pmullw "MANGLE(ff_pw_3)", %%mm3 \n\t" /* 3c - 6b */\ + "paddw %%mm2, %%mm1 \n\t" /* a */\ + "paddw %%mm6, %%mm4 \n\t" /* d */\ + "pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\ + "psubw %%mm4, %%mm3 \n\t" /* - 6b +3c - d */\ + "paddw %6, %%mm1 \n\t"\ + "paddw %%mm1, %%mm3 \n\t" /* 20a - 6b +3c - d */\ + "psraw $5, %%mm3 \n\t"\ + "movq %5, %%mm1 \n\t"\ + "packuswb %%mm3, %%mm1 \n\t"\ OP_MMX2(%%mm1, (%1),%%mm4, q)\ /* mm0= GHIJ, mm2=FGHI, mm5=HIJK, mm6=IJKL, mm7=0 */\ \ - "movq 9(%0), %%mm1 \n\t" /* JKLMNOPQ */\ - "movq %%mm1, %%mm4 \n\t" /* JKLMNOPQ */\ - "movq %%mm1, %%mm3 \n\t" /* JKLMNOPQ */\ - "psrlq $8, %%mm1 \n\t" /* KLMNOPQ0 */\ - "psrlq $16, %%mm4 \n\t" /* LMNOPQ00 */\ - "punpcklbw %%mm7, %%mm1 \n\t" /* 0K0L0M0N */\ - "punpcklbw %%mm7, %%mm4 \n\t" /* 0L0M0N0O */\ - "paddw %%mm1, %%mm5 \n\t" /* b */\ - "paddw %%mm4, %%mm0 \n\t" /* c */\ - "paddw %%mm5, %%mm5 \n\t" /* 2b */\ - "psubw %%mm5, %%mm0 \n\t" /* c - 2b */\ - "movq %%mm3, %%mm5 \n\t" /* JKLMNOPQ */\ - "psrlq $24, %%mm3 \n\t" /* MNOPQ000 */\ - "pmullw "MANGLE(ff_pw_3)", %%mm0 \n\t" /* 3c - 6b */\ - "punpcklbw %%mm7, %%mm3 \n\t" /* 0M0N0O0P */\ - "paddw %%mm3, %%mm2 \n\t" /* d */\ - "psubw %%mm2, %%mm0 \n\t" /* -6b + 3c - d */\ - "movq %%mm5, %%mm2 \n\t" /* JKLMNOPQ */\ - "punpcklbw %%mm7, %%mm2 \n\t" /* 0J0K0L0M */\ - "punpckhbw %%mm7, %%mm5 \n\t" /* 0N0O0P0Q */\ - "paddw %%mm2, %%mm6 \n\t" /* a */\ - "pmullw "MANGLE(ff_pw_20)", %%mm6 \n\t" /* 20a */\ - "paddw %6, %%mm0 \n\t"\ - "paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\ - "psraw $5, %%mm0 \n\t"\ + "movq 9(%0), %%mm1 \n\t" /* JKLMNOPQ */\ + "movq %%mm1, %%mm4 \n\t" /* JKLMNOPQ */\ + "movq %%mm1, %%mm3 \n\t" /* JKLMNOPQ */\ + "psrlq $8, %%mm1 \n\t" /* KLMNOPQ0 */\ + "psrlq $16, %%mm4 \n\t" /* LMNOPQ00 */\ + "punpcklbw %%mm7, %%mm1 \n\t" /* 0K0L0M0N */\ + "punpcklbw %%mm7, %%mm4 \n\t" /* 0L0M0N0O */\ + "paddw %%mm1, %%mm5 \n\t" /* b */\ + "paddw %%mm4, %%mm0 \n\t" /* c */\ + "paddw %%mm5, %%mm5 \n\t" /* 2b */\ + "psubw %%mm5, %%mm0 \n\t" /* c - 2b */\ + "movq %%mm3, %%mm5 \n\t" /* JKLMNOPQ */\ + "psrlq $24, %%mm3 \n\t" /* MNOPQ000 */\ + "pmullw "MANGLE(ff_pw_3)", %%mm0 \n\t" /* 3c - 6b */\ + "punpcklbw %%mm7, %%mm3 \n\t" /* 0M0N0O0P */\ + "paddw %%mm3, %%mm2 \n\t" /* d */\ + "psubw %%mm2, %%mm0 \n\t" /* -6b + 3c - d */\ + "movq %%mm5, %%mm2 \n\t" /* JKLMNOPQ */\ + "punpcklbw %%mm7, %%mm2 \n\t" /* 0J0K0L0M */\ + "punpckhbw %%mm7, %%mm5 \n\t" /* 0N0O0P0Q */\ + "paddw %%mm2, %%mm6 \n\t" /* a */\ + "pmullw "MANGLE(ff_pw_20)", %%mm6 \n\t" /* 20a */\ + "paddw %6, %%mm0 \n\t"\ + "paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\ + "psraw $5, %%mm0 \n\t"\ /* mm1=KLMN, mm2=JKLM, mm3=MNOP, mm4=LMNO, mm5=NOPQ mm7=0 */\ \ - "paddw %%mm5, %%mm3 \n\t" /* a */\ - "pshufw $0xF9, %%mm5, %%mm6 \n\t" /* 0O0P0Q0Q */\ - "paddw %%mm4, %%mm6 \n\t" /* b */\ - "pshufw $0xBE, %%mm5, %%mm4 \n\t" /* 0P0Q0Q0P */\ - "pshufw $0x6F, %%mm5, %%mm5 \n\t" /* 0Q0Q0P0O */\ - "paddw %%mm1, %%mm4 \n\t" /* c */\ - "paddw %%mm2, %%mm5 \n\t" /* d */\ - "paddw %%mm6, %%mm6 \n\t" /* 2b */\ - "psubw %%mm6, %%mm4 \n\t" /* c - 2b */\ - "pmullw "MANGLE(ff_pw_20)", %%mm3 \n\t" /* 20a */\ - "pmullw "MANGLE(ff_pw_3)", %%mm4 \n\t" /* 3c - 6b */\ - "psubw %%mm5, %%mm3 \n\t" /* -6b + 3c - d */\ - "paddw %6, %%mm4 \n\t"\ - "paddw %%mm3, %%mm4 \n\t" /* 20a - 6b + 3c - d */\ - "psraw $5, %%mm4 \n\t"\ - "packuswb %%mm4, %%mm0 \n\t"\ + "paddw %%mm5, %%mm3 \n\t" /* a */\ + "pshufw $0xF9, %%mm5, %%mm6 \n\t" /* 0O0P0Q0Q */\ + "paddw %%mm4, %%mm6 \n\t" /* b */\ + "pshufw $0xBE, %%mm5, %%mm4 \n\t" /* 0P0Q0Q0P */\ + "pshufw $0x6F, %%mm5, %%mm5 \n\t" /* 0Q0Q0P0O */\ + "paddw %%mm1, %%mm4 \n\t" /* c */\ + "paddw %%mm2, %%mm5 \n\t" /* d */\ + "paddw %%mm6, %%mm6 \n\t" /* 2b */\ + "psubw %%mm6, %%mm4 \n\t" /* c - 2b */\ + "pmullw "MANGLE(ff_pw_20)", %%mm3 \n\t" /* 20a */\ + "pmullw "MANGLE(ff_pw_3)", %%mm4 \n\t" /* 3c - 6b */\ + "psubw %%mm5, %%mm3 \n\t" /* -6b + 3c - d */\ + "paddw %6, %%mm4 \n\t"\ + "paddw %%mm3, %%mm4 \n\t" /* 20a - 6b + 3c - d */\ + "psraw $5, %%mm4 \n\t"\ + "packuswb %%mm4, %%mm0 \n\t"\ OP_MMX2(%%mm0, 8(%1), %%mm4, q)\ \ - "add %3, %0 \n\t"\ - "add %4, %1 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "add %3, %0 \n\t"\ + "add %4, %1 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ : "+a"(src), "+c"(dst), "+m"(h)\ : "d"((long)srcStride), "S"((long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\ : "memory"\ @@ -1887,21 +1890,21 @@ static void OPNAME ## mpeg4_qpel16_h_lowpass_3dnow(uint8_t *dst, uint8_t *src, i temp[14]= (src[14]+src[15])*20 - (src[13]+src[16])*6 + (src[12]+src[16])*3 - (src[11]+src[15]);\ temp[15]= (src[15]+src[16])*20 - (src[14]+src[16])*6 + (src[13]+src[15])*3 - (src[12]+src[14]);\ asm volatile(\ - "movq (%0), %%mm0 \n\t"\ - "movq 8(%0), %%mm1 \n\t"\ - "paddw %2, %%mm0 \n\t"\ - "paddw %2, %%mm1 \n\t"\ - "psraw $5, %%mm0 \n\t"\ - "psraw $5, %%mm1 \n\t"\ - "packuswb %%mm1, %%mm0 \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "movq 8(%0), %%mm1 \n\t"\ + "paddw %2, %%mm0 \n\t"\ + "paddw %2, %%mm1 \n\t"\ + "psraw $5, %%mm0 \n\t"\ + "psraw $5, %%mm1 \n\t"\ + "packuswb %%mm1, %%mm0 \n\t"\ OP_3DNOW(%%mm0, (%1), %%mm1, q)\ - "movq 16(%0), %%mm0 \n\t"\ - "movq 24(%0), %%mm1 \n\t"\ - "paddw %2, %%mm0 \n\t"\ - "paddw %2, %%mm1 \n\t"\ - "psraw $5, %%mm0 \n\t"\ - "psraw $5, %%mm1 \n\t"\ - "packuswb %%mm1, %%mm0 \n\t"\ + "movq 16(%0), %%mm0 \n\t"\ + "movq 24(%0), %%mm1 \n\t"\ + "paddw %2, %%mm0 \n\t"\ + "paddw %2, %%mm1 \n\t"\ + "psraw $5, %%mm0 \n\t"\ + "psraw $5, %%mm1 \n\t"\ + "packuswb %%mm1, %%mm0 \n\t"\ OP_3DNOW(%%mm0, 8(%1), %%mm1, q)\ :: "r"(temp), "r"(dst), "m"(ROUNDER)\ : "memory"\ @@ -1915,62 +1918,62 @@ static void OPNAME ## mpeg4_qpel8_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, int uint64_t temp;\ \ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t" /* ABCDEFGH */\ - "movq %%mm0, %%mm1 \n\t" /* ABCDEFGH */\ - "movq %%mm0, %%mm2 \n\t" /* ABCDEFGH */\ - "punpcklbw %%mm7, %%mm0 \n\t" /* 0A0B0C0D */\ - "punpckhbw %%mm7, %%mm1 \n\t" /* 0E0F0G0H */\ - "pshufw $0x90, %%mm0, %%mm5 \n\t" /* 0A0A0B0C */\ - "pshufw $0x41, %%mm0, %%mm6 \n\t" /* 0B0A0A0B */\ - "movq %%mm2, %%mm3 \n\t" /* ABCDEFGH */\ - "movq %%mm2, %%mm4 \n\t" /* ABCDEFGH */\ - "psllq $8, %%mm2 \n\t" /* 0ABCDEFG */\ - "psllq $16, %%mm3 \n\t" /* 00ABCDEF */\ - "psllq $24, %%mm4 \n\t" /* 000ABCDE */\ - "punpckhbw %%mm7, %%mm2 \n\t" /* 0D0E0F0G */\ - "punpckhbw %%mm7, %%mm3 \n\t" /* 0C0D0E0F */\ - "punpckhbw %%mm7, %%mm4 \n\t" /* 0B0C0D0E */\ - "paddw %%mm3, %%mm5 \n\t" /* b */\ - "paddw %%mm2, %%mm6 \n\t" /* c */\ - "paddw %%mm5, %%mm5 \n\t" /* 2b */\ - "psubw %%mm5, %%mm6 \n\t" /* c - 2b */\ - "pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\ - "pmullw "MANGLE(ff_pw_3)", %%mm6 \n\t" /* 3c - 6b */\ - "paddw %%mm4, %%mm0 \n\t" /* a */\ - "paddw %%mm1, %%mm5 \n\t" /* d */\ - "pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\ - "psubw %%mm5, %%mm0 \n\t" /* 20a - d */\ - "paddw %6, %%mm6 \n\t"\ - "paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\ - "psraw $5, %%mm0 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t" /* ABCDEFGH */\ + "movq %%mm0, %%mm1 \n\t" /* ABCDEFGH */\ + "movq %%mm0, %%mm2 \n\t" /* ABCDEFGH */\ + "punpcklbw %%mm7, %%mm0 \n\t" /* 0A0B0C0D */\ + "punpckhbw %%mm7, %%mm1 \n\t" /* 0E0F0G0H */\ + "pshufw $0x90, %%mm0, %%mm5 \n\t" /* 0A0A0B0C */\ + "pshufw $0x41, %%mm0, %%mm6 \n\t" /* 0B0A0A0B */\ + "movq %%mm2, %%mm3 \n\t" /* ABCDEFGH */\ + "movq %%mm2, %%mm4 \n\t" /* ABCDEFGH */\ + "psllq $8, %%mm2 \n\t" /* 0ABCDEFG */\ + "psllq $16, %%mm3 \n\t" /* 00ABCDEF */\ + "psllq $24, %%mm4 \n\t" /* 000ABCDE */\ + "punpckhbw %%mm7, %%mm2 \n\t" /* 0D0E0F0G */\ + "punpckhbw %%mm7, %%mm3 \n\t" /* 0C0D0E0F */\ + "punpckhbw %%mm7, %%mm4 \n\t" /* 0B0C0D0E */\ + "paddw %%mm3, %%mm5 \n\t" /* b */\ + "paddw %%mm2, %%mm6 \n\t" /* c */\ + "paddw %%mm5, %%mm5 \n\t" /* 2b */\ + "psubw %%mm5, %%mm6 \n\t" /* c - 2b */\ + "pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\ + "pmullw "MANGLE(ff_pw_3)", %%mm6 \n\t" /* 3c - 6b */\ + "paddw %%mm4, %%mm0 \n\t" /* a */\ + "paddw %%mm1, %%mm5 \n\t" /* d */\ + "pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\ + "psubw %%mm5, %%mm0 \n\t" /* 20a - d */\ + "paddw %6, %%mm6 \n\t"\ + "paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\ + "psraw $5, %%mm0 \n\t"\ /* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\ \ - "movd 5(%0), %%mm5 \n\t" /* FGHI */\ - "punpcklbw %%mm7, %%mm5 \n\t" /* 0F0G0H0I */\ - "pshufw $0xF9, %%mm5, %%mm6 \n\t" /* 0G0H0I0I */\ - "paddw %%mm5, %%mm1 \n\t" /* a */\ - "paddw %%mm6, %%mm2 \n\t" /* b */\ - "pshufw $0xBE, %%mm5, %%mm6 \n\t" /* 0H0I0I0H */\ - "pshufw $0x6F, %%mm5, %%mm5 \n\t" /* 0I0I0H0G */\ - "paddw %%mm6, %%mm3 \n\t" /* c */\ - "paddw %%mm5, %%mm4 \n\t" /* d */\ - "paddw %%mm2, %%mm2 \n\t" /* 2b */\ - "psubw %%mm2, %%mm3 \n\t" /* c - 2b */\ - "pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\ - "pmullw "MANGLE(ff_pw_3)", %%mm3 \n\t" /* 3c - 6b */\ - "psubw %%mm4, %%mm3 \n\t" /* -6b + 3c - d */\ - "paddw %6, %%mm1 \n\t"\ - "paddw %%mm1, %%mm3 \n\t" /* 20a - 6b + 3c - d */\ - "psraw $5, %%mm3 \n\t"\ - "packuswb %%mm3, %%mm0 \n\t"\ + "movd 5(%0), %%mm5 \n\t" /* FGHI */\ + "punpcklbw %%mm7, %%mm5 \n\t" /* 0F0G0H0I */\ + "pshufw $0xF9, %%mm5, %%mm6 \n\t" /* 0G0H0I0I */\ + "paddw %%mm5, %%mm1 \n\t" /* a */\ + "paddw %%mm6, %%mm2 \n\t" /* b */\ + "pshufw $0xBE, %%mm5, %%mm6 \n\t" /* 0H0I0I0H */\ + "pshufw $0x6F, %%mm5, %%mm5 \n\t" /* 0I0I0H0G */\ + "paddw %%mm6, %%mm3 \n\t" /* c */\ + "paddw %%mm5, %%mm4 \n\t" /* d */\ + "paddw %%mm2, %%mm2 \n\t" /* 2b */\ + "psubw %%mm2, %%mm3 \n\t" /* c - 2b */\ + "pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\ + "pmullw "MANGLE(ff_pw_3)", %%mm3 \n\t" /* 3c - 6b */\ + "psubw %%mm4, %%mm3 \n\t" /* -6b + 3c - d */\ + "paddw %6, %%mm1 \n\t"\ + "paddw %%mm1, %%mm3 \n\t" /* 20a - 6b + 3c - d */\ + "psraw $5, %%mm3 \n\t"\ + "packuswb %%mm3, %%mm0 \n\t"\ OP_MMX2(%%mm0, (%1), %%mm4, q)\ \ - "add %3, %0 \n\t"\ - "add %4, %1 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "add %3, %0 \n\t"\ + "add %4, %1 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ : "+a"(src), "+c"(dst), "+m"(h)\ : "S"((long)srcStride), "D"((long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\ : "memory"\ @@ -1992,13 +1995,13 @@ static void OPNAME ## mpeg4_qpel8_h_lowpass_3dnow(uint8_t *dst, uint8_t *src, in temp[ 6]= (src[ 6]+src[ 7])*20 - (src[ 5]+src[ 8])*6 + (src[ 4]+src[ 8])*3 - (src[ 3]+src[ 7]);\ temp[ 7]= (src[ 7]+src[ 8])*20 - (src[ 6]+src[ 8])*6 + (src[ 5]+src[ 7])*3 - (src[ 4]+src[ 6]);\ asm volatile(\ - "movq (%0), %%mm0 \n\t"\ - "movq 8(%0), %%mm1 \n\t"\ - "paddw %2, %%mm0 \n\t"\ - "paddw %2, %%mm1 \n\t"\ - "psraw $5, %%mm0 \n\t"\ - "psraw $5, %%mm1 \n\t"\ - "packuswb %%mm1, %%mm0 \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "movq 8(%0), %%mm1 \n\t"\ + "paddw %2, %%mm0 \n\t"\ + "paddw %2, %%mm1 \n\t"\ + "psraw $5, %%mm0 \n\t"\ + "psraw $5, %%mm1 \n\t"\ + "packuswb %%mm1, %%mm0 \n\t"\ OP_3DNOW(%%mm0, (%1), %%mm1, q)\ :: "r"(temp), "r"(dst), "m"(ROUNDER)\ :"memory"\ @@ -2017,24 +2020,24 @@ static void OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, \ /*FIXME unroll */\ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t"\ - "movq (%0), %%mm1 \n\t"\ - "movq 8(%0), %%mm2 \n\t"\ - "movq 8(%0), %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpckhbw %%mm7, %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpckhbw %%mm7, %%mm3 \n\t"\ - "movq %%mm0, (%1) \n\t"\ - "movq %%mm1, 17*8(%1) \n\t"\ - "movq %%mm2, 2*17*8(%1) \n\t"\ - "movq %%mm3, 3*17*8(%1) \n\t"\ - "add $8, %1 \n\t"\ - "add %3, %0 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "movq (%0), %%mm1 \n\t"\ + "movq 8(%0), %%mm2 \n\t"\ + "movq 8(%0), %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpckhbw %%mm7, %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpckhbw %%mm7, %%mm3 \n\t"\ + "movq %%mm0, (%1) \n\t"\ + "movq %%mm1, 17*8(%1) \n\t"\ + "movq %%mm2, 2*17*8(%1) \n\t"\ + "movq %%mm3, 3*17*8(%1) \n\t"\ + "add $8, %1 \n\t"\ + "add %3, %0 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ : "+r" (src), "+r" (temp_ptr), "+r"(count)\ : "r" ((long)srcStride)\ : "memory"\ @@ -2045,42 +2048,42 @@ static void OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, \ /*FIXME reorder for speed */\ asm volatile(\ - /*"pxor %%mm7, %%mm7 \n\t"*/\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t"\ - "movq 8(%0), %%mm1 \n\t"\ - "movq 16(%0), %%mm2 \n\t"\ - "movq 24(%0), %%mm3 \n\t"\ + /*"pxor %%mm7, %%mm7 \n\t"*/\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "movq 8(%0), %%mm1 \n\t"\ + "movq 16(%0), %%mm2 \n\t"\ + "movq 24(%0), %%mm3 \n\t"\ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, (%0), (%0), 8(%0), 48(%0), (%1), OP)\ \ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\ \ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\ - "add %4, %1 \n\t" \ + "add %4, %1 \n\t" \ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\ \ - "add $136, %0 \n\t"\ - "add %6, %1 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "add $136, %0 \n\t"\ + "add %6, %1 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ \ : "+r"(temp_ptr), "+r"(dst), "+g"(count)\ : "r"((long)dstStride), "r"(2*(long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER), "g"(4-14*(long)dstStride)\ @@ -2095,18 +2098,18 @@ static void OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, \ /*FIXME unroll */\ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t"\ - "movq (%0), %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpckhbw %%mm7, %%mm1 \n\t"\ - "movq %%mm0, (%1) \n\t"\ - "movq %%mm1, 9*8(%1) \n\t"\ - "add $8, %1 \n\t"\ - "add %3, %0 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "movq (%0), %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpckhbw %%mm7, %%mm1 \n\t"\ + "movq %%mm0, (%1) \n\t"\ + "movq %%mm1, 9*8(%1) \n\t"\ + "add $8, %1 \n\t"\ + "add %3, %0 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ : "+r" (src), "+r" (temp_ptr), "+r"(count)\ : "r" ((long)srcStride)\ : "memory"\ @@ -2117,30 +2120,30 @@ static void OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, \ /*FIXME reorder for speed */\ asm volatile(\ - /*"pxor %%mm7, %%mm7 \n\t"*/\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t"\ - "movq 8(%0), %%mm1 \n\t"\ - "movq 16(%0), %%mm2 \n\t"\ - "movq 24(%0), %%mm3 \n\t"\ + /*"pxor %%mm7, %%mm7 \n\t"*/\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "movq 8(%0), %%mm1 \n\t"\ + "movq 16(%0), %%mm2 \n\t"\ + "movq 24(%0), %%mm3 \n\t"\ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, (%0), (%0), 8(%0), 48(%0), (%1), OP)\ \ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\ \ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 64(%0), (%1, %3), OP)\ - "add %4, %1 \n\t"\ + "add %4, %1 \n\t"\ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 56(%0), (%1), OP)\ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 48(%0), (%1, %3), OP)\ \ - "add $72, %0 \n\t"\ - "add %6, %1 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "add $72, %0 \n\t"\ + "add %6, %1 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ \ : "+r"(temp_ptr), "+r"(dst), "+g"(count)\ : "r"((long)dstStride), "r"(2*(long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(ROUNDER), "g"(4-6*(long)dstStride)\ @@ -2371,15 +2374,15 @@ static void OPNAME ## qpel16_mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, stride, 16);\ } -#define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b " \n\t" +#define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b " \n\t" #define AVG_3DNOW_OP(a,b,temp, size) \ -"mov" #size " " #b ", " #temp " \n\t"\ -"pavgusb " #temp ", " #a " \n\t"\ -"mov" #size " " #a ", " #b " \n\t" +"mov" #size " " #b ", " #temp " \n\t"\ +"pavgusb " #temp ", " #a " \n\t"\ +"mov" #size " " #a ", " #b " \n\t" #define AVG_MMX2_OP(a,b,temp, size) \ -"mov" #size " " #b ", " #temp " \n\t"\ -"pavgb " #temp ", " #a " \n\t"\ -"mov" #size " " #a ", " #b " \n\t" +"mov" #size " " #b ", " #temp " \n\t"\ +"pavgb " #temp ", " #a " \n\t"\ +"mov" #size " " #a ", " #b " \n\t" QPEL_BASE(put_ , ff_pw_16, _ , PUT_OP, PUT_OP) QPEL_BASE(avg_ , ff_pw_16, _ , AVG_MMX2_OP, AVG_3DNOW_OP) @@ -2402,46 +2405,46 @@ static void just_return() { return; } static int try_8x8basis_mmx(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale){ long i=0; - + assert(ABS(scale) < 256); scale<<= 16 + 1 - BASIS_SHIFT + RECON_SHIFT; asm volatile( - "pcmpeqw %%mm6, %%mm6 \n\t" // -1w - "psrlw $15, %%mm6 \n\t" // 1w - "pxor %%mm7, %%mm7 \n\t" - "movd %4, %%mm5 \n\t" - "punpcklwd %%mm5, %%mm5 \n\t" - "punpcklwd %%mm5, %%mm5 \n\t" - "1: \n\t" - "movq (%1, %0), %%mm0 \n\t" - "movq 8(%1, %0), %%mm1 \n\t" - "pmulhw %%mm5, %%mm0 \n\t" - "pmulhw %%mm5, %%mm1 \n\t" - "paddw %%mm6, %%mm0 \n\t" - "paddw %%mm6, %%mm1 \n\t" - "psraw $1, %%mm0 \n\t" - "psraw $1, %%mm1 \n\t" - "paddw (%2, %0), %%mm0 \n\t" - "paddw 8(%2, %0), %%mm1 \n\t" - "psraw $6, %%mm0 \n\t" - "psraw $6, %%mm1 \n\t" - "pmullw (%3, %0), %%mm0 \n\t" - "pmullw 8(%3, %0), %%mm1 \n\t" - "pmaddwd %%mm0, %%mm0 \n\t" - "pmaddwd %%mm1, %%mm1 \n\t" - "paddd %%mm1, %%mm0 \n\t" - "psrld $4, %%mm0 \n\t" - "paddd %%mm0, %%mm7 \n\t" - "add $16, %0 \n\t" - "cmp $128, %0 \n\t" //FIXME optimize & bench - " jb 1b \n\t" - "movq %%mm7, %%mm6 \n\t" - "psrlq $32, %%mm7 \n\t" - "paddd %%mm6, %%mm7 \n\t" - "psrld $2, %%mm7 \n\t" - "movd %%mm7, %0 \n\t" - + "pcmpeqw %%mm6, %%mm6 \n\t" // -1w + "psrlw $15, %%mm6 \n\t" // 1w + "pxor %%mm7, %%mm7 \n\t" + "movd %4, %%mm5 \n\t" + "punpcklwd %%mm5, %%mm5 \n\t" + "punpcklwd %%mm5, %%mm5 \n\t" + "1: \n\t" + "movq (%1, %0), %%mm0 \n\t" + "movq 8(%1, %0), %%mm1 \n\t" + "pmulhw %%mm5, %%mm0 \n\t" + "pmulhw %%mm5, %%mm1 \n\t" + "paddw %%mm6, %%mm0 \n\t" + "paddw %%mm6, %%mm1 \n\t" + "psraw $1, %%mm0 \n\t" + "psraw $1, %%mm1 \n\t" + "paddw (%2, %0), %%mm0 \n\t" + "paddw 8(%2, %0), %%mm1 \n\t" + "psraw $6, %%mm0 \n\t" + "psraw $6, %%mm1 \n\t" + "pmullw (%3, %0), %%mm0 \n\t" + "pmullw 8(%3, %0), %%mm1 \n\t" + "pmaddwd %%mm0, %%mm0 \n\t" + "pmaddwd %%mm1, %%mm1 \n\t" + "paddd %%mm1, %%mm0 \n\t" + "psrld $4, %%mm0 \n\t" + "paddd %%mm0, %%mm7 \n\t" + "add $16, %0 \n\t" + "cmp $128, %0 \n\t" //FIXME optimize & bench + " jb 1b \n\t" + "movq %%mm7, %%mm6 \n\t" + "psrlq $32, %%mm7 \n\t" + "paddd %%mm6, %%mm7 \n\t" + "psrld $2, %%mm7 \n\t" + "movd %%mm7, %0 \n\t" + : "+r" (i) : "r"(basis), "r"(rem), "r"(weight), "g"(scale) ); @@ -2450,44 +2453,44 @@ static int try_8x8basis_mmx(int16_t rem[64], int16_t weight[64], int16_t basis[6 static void add_8x8basis_mmx(int16_t rem[64], int16_t basis[64], int scale){ long i=0; - + if(ABS(scale) < 256){ scale<<= 16 + 1 - BASIS_SHIFT + RECON_SHIFT; asm volatile( - "pcmpeqw %%mm6, %%mm6 \n\t" // -1w - "psrlw $15, %%mm6 \n\t" // 1w - "movd %3, %%mm5 \n\t" - "punpcklwd %%mm5, %%mm5 \n\t" - "punpcklwd %%mm5, %%mm5 \n\t" - "1: \n\t" - "movq (%1, %0), %%mm0 \n\t" - "movq 8(%1, %0), %%mm1 \n\t" - "pmulhw %%mm5, %%mm0 \n\t" - "pmulhw %%mm5, %%mm1 \n\t" - "paddw %%mm6, %%mm0 \n\t" - "paddw %%mm6, %%mm1 \n\t" - "psraw $1, %%mm0 \n\t" - "psraw $1, %%mm1 \n\t" - "paddw (%2, %0), %%mm0 \n\t" - "paddw 8(%2, %0), %%mm1 \n\t" - "movq %%mm0, (%2, %0) \n\t" - "movq %%mm1, 8(%2, %0) \n\t" - "add $16, %0 \n\t" - "cmp $128, %0 \n\t" //FIXME optimize & bench - " jb 1b \n\t" - + "pcmpeqw %%mm6, %%mm6 \n\t" // -1w + "psrlw $15, %%mm6 \n\t" // 1w + "movd %3, %%mm5 \n\t" + "punpcklwd %%mm5, %%mm5 \n\t" + "punpcklwd %%mm5, %%mm5 \n\t" + "1: \n\t" + "movq (%1, %0), %%mm0 \n\t" + "movq 8(%1, %0), %%mm1 \n\t" + "pmulhw %%mm5, %%mm0 \n\t" + "pmulhw %%mm5, %%mm1 \n\t" + "paddw %%mm6, %%mm0 \n\t" + "paddw %%mm6, %%mm1 \n\t" + "psraw $1, %%mm0 \n\t" + "psraw $1, %%mm1 \n\t" + "paddw (%2, %0), %%mm0 \n\t" + "paddw 8(%2, %0), %%mm1 \n\t" + "movq %%mm0, (%2, %0) \n\t" + "movq %%mm1, 8(%2, %0) \n\t" + "add $16, %0 \n\t" + "cmp $128, %0 \n\t" //FIXME optimize & bench + " jb 1b \n\t" + : "+r" (i) : "r"(basis), "r"(rem), "g"(scale) ); }else{ for(i=0; i<8*8; i++){ rem[i] += (basis[i]*scale + (1<<(BASIS_SHIFT - RECON_SHIFT-1)))>>(BASIS_SHIFT - RECON_SHIFT); - } + } } } #include "h264dsp_mmx.c" - + /* external functions, from idct_mmx.c */ void ff_mmx_idct(DCTELEM *block); void ff_mmxext_idct(DCTELEM *block); @@ -2560,16 +2563,16 @@ static void ff_idct_xvid_mmx2_add(uint8_t *dest, int line_size, DCTELEM *block) add_pixels_clamped_mmx(block, dest, line_size); } #endif - + void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) { mm_flags = mm_support(); if (avctx->dsp_mask) { - if (avctx->dsp_mask & FF_MM_FORCE) - mm_flags |= (avctx->dsp_mask & 0xffff); - else - mm_flags &= ~(avctx->dsp_mask & 0xffff); + if (avctx->dsp_mask & FF_MM_FORCE) + mm_flags |= (avctx->dsp_mask & 0xffff); + else + mm_flags &= ~(avctx->dsp_mask & 0xffff); } #if 0 @@ -2595,7 +2598,7 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) if(dct_algo==FF_DCT_AUTO || dct_algo==FF_DCT_MMX){ if(mm_flags & MM_SSE2){ c->fdct = ff_fdct_sse2; - }else if(mm_flags & MM_MMXEXT){ + }else if(mm_flags & MM_MMXEXT){ c->fdct = ff_fdct_mmx2; }else{ c->fdct = ff_fdct_mmx; @@ -2698,36 +2701,37 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x2_mmx; c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y2_mmx; c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy2_mmx; - + c->add_bytes= add_bytes_mmx; #ifdef CONFIG_ENCODERS c->diff_bytes= diff_bytes_mmx; - + c->hadamard8_diff[0]= hadamard8_diff16_mmx; c->hadamard8_diff[1]= hadamard8_diff_mmx; - - c->pix_norm1 = pix_norm1_mmx; - c->sse[0] = (mm_flags & MM_SSE2) ? sse16_sse2 : sse16_mmx; - c->sse[1] = sse8_mmx; + + c->pix_norm1 = pix_norm1_mmx; + c->sse[0] = (mm_flags & MM_SSE2) ? sse16_sse2 : sse16_mmx; + c->sse[1] = sse8_mmx; c->vsad[4]= vsad_intra16_mmx; - c->nsse[0] = nsse16_mmx; - c->nsse[1] = nsse8_mmx; + c->nsse[0] = nsse16_mmx; + c->nsse[1] = nsse8_mmx; if(!(avctx->flags & CODEC_FLAG_BITEXACT)){ c->vsad[0] = vsad16_mmx; } - + if(!(avctx->flags & CODEC_FLAG_BITEXACT)){ c->try_8x8basis= try_8x8basis_mmx; } c->add_8x8basis= add_8x8basis_mmx; - + #endif //CONFIG_ENCODERS c->h263_v_loop_filter= h263_v_loop_filter_mmx; - c->h263_h_loop_filter= h263_h_loop_filter_mmx; - c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_mmx; - + c->h263_h_loop_filter= h263_h_loop_filter_mmx; + c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_mmx; + c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_mmx; + if (mm_flags & MM_MMXEXT) { c->put_pixels_tab[0][1] = put_pixels16_x2_mmx2; c->put_pixels_tab[0][2] = put_pixels16_y2_mmx2; @@ -2825,7 +2829,8 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) dspfunc(avg_h264_qpel, 2, 4); #undef dspfunc - c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_mmx2; + c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_mmx2; + c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_mmx2; c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_mmx2; c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_mmx2; c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_mmx2; @@ -2936,10 +2941,11 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx) dspfunc(avg_h264_qpel, 1, 8); dspfunc(avg_h264_qpel, 2, 4); - c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_3dnow; + c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_3dnow; + c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_3dnow; } } - + #ifdef CONFIG_ENCODERS dsputil_init_pix_mmx(c, avctx); #endif //CONFIG_ENCODERS diff --git a/src/libffmpeg/libavcodec/i386/dsputil_mmx_avg.h b/src/libffmpeg/libavcodec/i386/dsputil_mmx_avg.h index c70891304..440c5bb9c 100644 --- a/src/libffmpeg/libavcodec/i386/dsputil_mmx_avg.h +++ b/src/libffmpeg/libavcodec/i386/dsputil_mmx_avg.h @@ -15,603 +15,603 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA * * MMX optimization by Nick Kurshev <nickols_k@mail.ru> * mostly rewritten by Michael Niedermayer <michaelni@gmx.at> * and improved by Zdenek Kabelac <kabi@users.sf.net> */ - + /* XXX: we use explicit registers to avoid a gcc 2.95.2 register asm clobber bug - now it will work with 2.95.2 and also with -fPIC */ static void DEF(put_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "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" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \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" - "add %%"REG_a", %1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "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" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \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" + "add %%"REG_a", %1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } static void DEF(put_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { __asm __volatile( - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movd (%1), %%mm0 \n\t" - "movd (%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "add $4, %2 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - "movd %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - "1: \n\t" - "movd (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movd (%1), %%mm1 \n\t" - "movd (%2), %%mm2 \n\t" - "movd 4(%2), %%mm3 \n\t" - "add %4, %1 \n\t" - PAVGB" %%mm2, %%mm0 \n\t" - PAVGB" %%mm3, %%mm1 \n\t" - "movd %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "movd %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "movd (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movd (%1), %%mm1 \n\t" - "movd 8(%2), %%mm2 \n\t" - "movd 12(%2), %%mm3 \n\t" - "add %4, %1 \n\t" - PAVGB" %%mm2, %%mm0 \n\t" - PAVGB" %%mm3, %%mm1 \n\t" - "movd %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "movd %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "add $16, %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movd (%1), %%mm0 \n\t" + "movd (%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "add $4, %2 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + "movd %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + "1: \n\t" + "movd (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movd (%1), %%mm1 \n\t" + "movd (%2), %%mm2 \n\t" + "movd 4(%2), %%mm3 \n\t" + "add %4, %1 \n\t" + PAVGB" %%mm2, %%mm0 \n\t" + PAVGB" %%mm3, %%mm1 \n\t" + "movd %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "movd %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "movd (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movd (%1), %%mm1 \n\t" + "movd 8(%2), %%mm2 \n\t" + "movd 12(%2), %%mm3 \n\t" + "add %4, %1 \n\t" + PAVGB" %%mm2, %%mm0 \n\t" + PAVGB" %%mm3, %%mm1 \n\t" + "movd %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "movd %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "add $16, %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); } static void DEF(put_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { __asm __volatile( - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "add $8, %2 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movq (%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" 8(%2), %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "movq %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "movq (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movq (%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" 16(%2), %%mm0 \n\t" - PAVGB" 24(%2), %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "movq %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "add $32, %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "add $8, %2 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movq (%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" 8(%2), %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "movq %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "movq (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movq (%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" 16(%2), %%mm0 \n\t" + PAVGB" 24(%2), %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "movq %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "add $32, %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); //the following should be used, though better not with gcc ... -/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) - :"r"(src1Stride), "r"(dstStride) - :"memory");*/ +/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) + :"r"(src1Stride), "r"(dstStride) + :"memory");*/ } static void DEF(put_no_rnd_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { __asm __volatile( - "pcmpeqb %%mm6, %%mm6 \n\t" - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "add $8, %2 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movq (%1), %%mm1 \n\t" - "add %4, %1 \n\t" - "movq (%2), %%mm2 \n\t" - "movq 8(%2), %%mm3 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "pxor %%mm6, %%mm2 \n\t" - "pxor %%mm6, %%mm3 \n\t" - PAVGB" %%mm2, %%mm0 \n\t" - PAVGB" %%mm3, %%mm1 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "movq %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "movq (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movq (%1), %%mm1 \n\t" - "add %4, %1 \n\t" - "movq 16(%2), %%mm2 \n\t" - "movq 24(%2), %%mm3 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "pxor %%mm6, %%mm2 \n\t" - "pxor %%mm6, %%mm3 \n\t" - PAVGB" %%mm2, %%mm0 \n\t" - PAVGB" %%mm3, %%mm1 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "movq %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "add $32, %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" + "pcmpeqb %%mm6, %%mm6 \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "add $8, %2 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movq (%1), %%mm1 \n\t" + "add %4, %1 \n\t" + "movq (%2), %%mm2 \n\t" + "movq 8(%2), %%mm3 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "pxor %%mm6, %%mm2 \n\t" + "pxor %%mm6, %%mm3 \n\t" + PAVGB" %%mm2, %%mm0 \n\t" + PAVGB" %%mm3, %%mm1 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "movq %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "movq (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movq (%1), %%mm1 \n\t" + "add %4, %1 \n\t" + "movq 16(%2), %%mm2 \n\t" + "movq 24(%2), %%mm3 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "pxor %%mm6, %%mm2 \n\t" + "pxor %%mm6, %%mm3 \n\t" + PAVGB" %%mm2, %%mm0 \n\t" + PAVGB" %%mm3, %%mm1 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "movq %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "add $32, %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); //the following should be used, though better not with gcc ... -/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) - :"r"(src1Stride), "r"(dstStride) - :"memory");*/ +/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) + :"r"(src1Stride), "r"(dstStride) + :"memory");*/ } static void DEF(avg_pixels4_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { __asm __volatile( - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movd (%1), %%mm0 \n\t" - "movd (%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "add $4, %2 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" (%3), %%mm0 \n\t" - "movd %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - "1: \n\t" - "movd (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movd (%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" 4(%2), %%mm1 \n\t" - PAVGB" (%3), %%mm0 \n\t" - "movd %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - PAVGB" (%3), %%mm1 \n\t" - "movd %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "movd (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movd (%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" 8(%2), %%mm0 \n\t" - PAVGB" 12(%2), %%mm1 \n\t" - PAVGB" (%3), %%mm0 \n\t" - "movd %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - PAVGB" (%3), %%mm1 \n\t" - "movd %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "add $16, %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movd (%1), %%mm0 \n\t" + "movd (%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "add $4, %2 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" (%3), %%mm0 \n\t" + "movd %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + "1: \n\t" + "movd (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movd (%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" 4(%2), %%mm1 \n\t" + PAVGB" (%3), %%mm0 \n\t" + "movd %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + PAVGB" (%3), %%mm1 \n\t" + "movd %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "movd (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movd (%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" 8(%2), %%mm0 \n\t" + PAVGB" 12(%2), %%mm1 \n\t" + PAVGB" (%3), %%mm0 \n\t" + "movd %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + PAVGB" (%3), %%mm1 \n\t" + "movd %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "add $16, %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); } static void DEF(avg_pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { __asm __volatile( - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "add $8, %2 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" (%3), %%mm0 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movq (%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" 8(%2), %%mm1 \n\t" - PAVGB" (%3), %%mm0 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - PAVGB" (%3), %%mm1 \n\t" - "movq %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "movq (%1), %%mm0 \n\t" - "add %4, %1 \n\t" - "movq (%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" 16(%2), %%mm0 \n\t" - PAVGB" 24(%2), %%mm1 \n\t" - PAVGB" (%3), %%mm0 \n\t" - "movq %%mm0, (%3) \n\t" - "add %5, %3 \n\t" - PAVGB" (%3), %%mm1 \n\t" - "movq %%mm1, (%3) \n\t" - "add %5, %3 \n\t" - "add $32, %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "add $8, %2 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" (%3), %%mm0 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movq (%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" 8(%2), %%mm1 \n\t" + PAVGB" (%3), %%mm0 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + PAVGB" (%3), %%mm1 \n\t" + "movq %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "movq (%1), %%mm0 \n\t" + "add %4, %1 \n\t" + "movq (%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" 16(%2), %%mm0 \n\t" + PAVGB" 24(%2), %%mm1 \n\t" + PAVGB" (%3), %%mm0 \n\t" + "movq %%mm0, (%3) \n\t" + "add %5, %3 \n\t" + PAVGB" (%3), %%mm1 \n\t" + "movq %%mm1, (%3) \n\t" + "add %5, %3 \n\t" + "add $32, %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); //the following should be used, though better not with gcc ... -/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) - :"r"(src1Stride), "r"(dstStride) - :"memory");*/ +/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) + :"r"(src1Stride), "r"(dstStride) + :"memory");*/ } static void DEF(put_pixels16_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq 8(%1), %%mm2 \n\t" - "movq 8(%1, %3), %%mm3 \n\t" - PAVGB" 1(%1), %%mm0 \n\t" - PAVGB" 1(%1, %3), %%mm1 \n\t" - PAVGB" 9(%1), %%mm2 \n\t" - PAVGB" 9(%1, %3), %%mm3 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "movq %%mm2, 8(%2) \n\t" - "movq %%mm3, 8(%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq 8(%1), %%mm2 \n\t" - "movq 8(%1, %3), %%mm3 \n\t" - PAVGB" 1(%1), %%mm0 \n\t" - PAVGB" 1(%1, %3), %%mm1 \n\t" - PAVGB" 9(%1), %%mm2 \n\t" - PAVGB" 9(%1, %3), %%mm3 \n\t" - "add %%"REG_a", %1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "movq %%mm2, 8(%2) \n\t" - "movq %%mm3, 8(%2, %3) \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq 8(%1), %%mm2 \n\t" + "movq 8(%1, %3), %%mm3 \n\t" + PAVGB" 1(%1), %%mm0 \n\t" + PAVGB" 1(%1, %3), %%mm1 \n\t" + PAVGB" 9(%1), %%mm2 \n\t" + PAVGB" 9(%1, %3), %%mm3 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "movq %%mm2, 8(%2) \n\t" + "movq %%mm3, 8(%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq 8(%1), %%mm2 \n\t" + "movq 8(%1, %3), %%mm3 \n\t" + PAVGB" 1(%1), %%mm0 \n\t" + PAVGB" 1(%1, %3), %%mm1 \n\t" + PAVGB" 9(%1), %%mm2 \n\t" + PAVGB" 9(%1, %3), %%mm3 \n\t" + "add %%"REG_a", %1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "movq %%mm2, 8(%2) \n\t" + "movq %%mm3, 8(%2, %3) \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } static void DEF(put_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { __asm __volatile( - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" 8(%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "add $16, %2 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" 8(%2), %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" 16(%2), %%mm0 \n\t" - PAVGB" 24(%2), %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "add $32, %2 \n\t" - "subl $2, %0 \n\t" - "jnz 1b \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" 8(%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "add $16, %2 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" 8(%2), %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" 16(%2), %%mm0 \n\t" + PAVGB" 24(%2), %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "add $32, %2 \n\t" + "subl $2, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); //the following should be used, though better not with gcc ... -/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) - :"r"(src1Stride), "r"(dstStride) - :"memory");*/ +/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) + :"r"(src1Stride), "r"(dstStride) + :"memory");*/ } static void DEF(avg_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { __asm __volatile( - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" 8(%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "add $16, %2 \n\t" - PAVGB" (%3), %%mm0 \n\t" - PAVGB" 8(%3), %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" 8(%2), %%mm1 \n\t" - PAVGB" (%3), %%mm0 \n\t" - PAVGB" 8(%3), %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - "add %4, %1 \n\t" - PAVGB" 16(%2), %%mm0 \n\t" - PAVGB" 24(%2), %%mm1 \n\t" - PAVGB" (%3), %%mm0 \n\t" - PAVGB" 8(%3), %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "add $32, %2 \n\t" - "subl $2, %0 \n\t" - "jnz 1b \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" 8(%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "add $16, %2 \n\t" + PAVGB" (%3), %%mm0 \n\t" + PAVGB" 8(%3), %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" 8(%2), %%mm1 \n\t" + PAVGB" (%3), %%mm0 \n\t" + PAVGB" 8(%3), %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + "add %4, %1 \n\t" + PAVGB" 16(%2), %%mm0 \n\t" + PAVGB" 24(%2), %%mm1 \n\t" + PAVGB" (%3), %%mm0 \n\t" + PAVGB" 8(%3), %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "add $32, %2 \n\t" + "subl $2, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); //the following should be used, though better not with gcc ... -/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) - :"r"(src1Stride), "r"(dstStride) - :"memory");*/ +/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) + :"r"(src1Stride), "r"(dstStride) + :"memory");*/ } static void DEF(put_no_rnd_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { __asm __volatile( - "pcmpeqb %%mm6, %%mm6\n\t" - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - "movq (%2), %%mm2 \n\t" - "movq 8(%2), %%mm3 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "pxor %%mm6, %%mm2 \n\t" - "pxor %%mm6, %%mm3 \n\t" - PAVGB" %%mm2, %%mm0 \n\t" - PAVGB" %%mm3, %%mm1 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "add %4, %1 \n\t" - "add $16, %2 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - "add %4, %1 \n\t" - "movq (%2), %%mm2 \n\t" - "movq 8(%2), %%mm3 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "pxor %%mm6, %%mm2 \n\t" - "pxor %%mm6, %%mm3 \n\t" - PAVGB" %%mm2, %%mm0 \n\t" - PAVGB" %%mm3, %%mm1 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 8(%1), %%mm1 \n\t" - "add %4, %1 \n\t" - "movq 16(%2), %%mm2 \n\t" - "movq 24(%2), %%mm3 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "pxor %%mm6, %%mm2 \n\t" - "pxor %%mm6, %%mm3 \n\t" - PAVGB" %%mm2, %%mm0 \n\t" - PAVGB" %%mm3, %%mm1 \n\t" - "pxor %%mm6, %%mm0 \n\t" - "pxor %%mm6, %%mm1 \n\t" - "movq %%mm0, (%3) \n\t" - "movq %%mm1, 8(%3) \n\t" - "add %5, %3 \n\t" - "add $32, %2 \n\t" - "subl $2, %0 \n\t" - "jnz 1b \n\t" + "pcmpeqb %%mm6, %%mm6 \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + "movq (%2), %%mm2 \n\t" + "movq 8(%2), %%mm3 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "pxor %%mm6, %%mm2 \n\t" + "pxor %%mm6, %%mm3 \n\t" + PAVGB" %%mm2, %%mm0 \n\t" + PAVGB" %%mm3, %%mm1 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "add %4, %1 \n\t" + "add $16, %2 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + "add %4, %1 \n\t" + "movq (%2), %%mm2 \n\t" + "movq 8(%2), %%mm3 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "pxor %%mm6, %%mm2 \n\t" + "pxor %%mm6, %%mm3 \n\t" + PAVGB" %%mm2, %%mm0 \n\t" + PAVGB" %%mm3, %%mm1 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "movq (%1), %%mm0 \n\t" + "movq 8(%1), %%mm1 \n\t" + "add %4, %1 \n\t" + "movq 16(%2), %%mm2 \n\t" + "movq 24(%2), %%mm3 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "pxor %%mm6, %%mm2 \n\t" + "pxor %%mm6, %%mm3 \n\t" + PAVGB" %%mm2, %%mm0 \n\t" + PAVGB" %%mm3, %%mm1 \n\t" + "pxor %%mm6, %%mm0 \n\t" + "pxor %%mm6, %%mm1 \n\t" + "movq %%mm0, (%3) \n\t" + "movq %%mm1, 8(%3) \n\t" + "add %5, %3 \n\t" + "add $32, %2 \n\t" + "subl $2, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); //the following should be used, though better not with gcc ... -/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) - :"r"(src1Stride), "r"(dstStride) - :"memory");*/ +/* :"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst) + :"r"(src1Stride), "r"(dstStride) + :"memory");*/ } - + /* GL: this function does incorrect rounding if overflow */ static void DEF(put_no_rnd_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { MOVQ_BONE(mm6); __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm2 \n\t" - "movq 1(%1), %%mm1 \n\t" - "movq 1(%1, %3), %%mm3 \n\t" - "add %%"REG_a", %1 \n\t" - "psubusb %%mm6, %%mm0 \n\t" - "psubusb %%mm6, %%mm2 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" %%mm3, %%mm2 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm2, (%2, %3) \n\t" - "movq (%1), %%mm0 \n\t" - "movq 1(%1), %%mm1 \n\t" - "movq (%1, %3), %%mm2 \n\t" - "movq 1(%1, %3), %%mm3 \n\t" - "add %%"REG_a", %2 \n\t" - "add %%"REG_a", %1 \n\t" - "psubusb %%mm6, %%mm0 \n\t" - "psubusb %%mm6, %%mm2 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" %%mm3, %%mm2 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm2, (%2, %3) \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%1, %3), %%mm2 \n\t" + "movq 1(%1), %%mm1 \n\t" + "movq 1(%1, %3), %%mm3 \n\t" + "add %%"REG_a", %1 \n\t" + "psubusb %%mm6, %%mm0 \n\t" + "psubusb %%mm6, %%mm2 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" %%mm3, %%mm2 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm2, (%2, %3) \n\t" + "movq (%1), %%mm0 \n\t" + "movq 1(%1), %%mm1 \n\t" + "movq (%1, %3), %%mm2 \n\t" + "movq 1(%1, %3), %%mm3 \n\t" + "add %%"REG_a", %2 \n\t" + "add %%"REG_a", %1 \n\t" + "psubusb %%mm6, %%mm0 \n\t" + "psubusb %%mm6, %%mm2 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" %%mm3, %%mm2 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm2, (%2, %3) \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } static void DEF(put_pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "movq (%1), %%mm0 \n\t" - "sub %3, %2 \n\t" - "1: \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm2 \n\t" - "add %%"REG_a", %1 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" %%mm2, %%mm1 \n\t" - "movq %%mm0, (%2, %3) \n\t" - "movq %%mm1, (%2, %%"REG_a") \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - "add %%"REG_a", %2 \n\t" - "add %%"REG_a", %1 \n\t" - PAVGB" %%mm1, %%mm2 \n\t" - PAVGB" %%mm0, %%mm1 \n\t" - "movq %%mm2, (%2, %3) \n\t" - "movq %%mm1, (%2, %%"REG_a") \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D" (block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "movq (%1), %%mm0 \n\t" + "sub %3, %2 \n\t" + "1: \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm2 \n\t" + "add %%"REG_a", %1 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" %%mm2, %%mm1 \n\t" + "movq %%mm0, (%2, %3) \n\t" + "movq %%mm1, (%2, %%"REG_a") \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + "add %%"REG_a", %2 \n\t" + "add %%"REG_a", %1 \n\t" + PAVGB" %%mm1, %%mm2 \n\t" + PAVGB" %%mm0, %%mm1 \n\t" + "movq %%mm2, (%2, %3) \n\t" + "movq %%mm1, (%2, %%"REG_a") \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D" (block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } /* GL: this function does incorrect rounding if overflow */ @@ -619,173 +619,173 @@ static void DEF(put_no_rnd_pixels8_y2)(uint8_t *block, const uint8_t *pixels, in { MOVQ_BONE(mm6); __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "movq (%1), %%mm0 \n\t" - "sub %3, %2 \n\t" - "1: \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm2 \n\t" - "add %%"REG_a", %1 \n\t" - "psubusb %%mm6, %%mm1 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" %%mm2, %%mm1 \n\t" - "movq %%mm0, (%2, %3) \n\t" - "movq %%mm1, (%2, %%"REG_a") \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - "add %%"REG_a", %2 \n\t" - "add %%"REG_a", %1 \n\t" - "psubusb %%mm6, %%mm1 \n\t" - PAVGB" %%mm1, %%mm2 \n\t" - PAVGB" %%mm0, %%mm1 \n\t" - "movq %%mm2, (%2, %3) \n\t" - "movq %%mm1, (%2, %%"REG_a") \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D" (block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "movq (%1), %%mm0 \n\t" + "sub %3, %2 \n\t" + "1: \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm2 \n\t" + "add %%"REG_a", %1 \n\t" + "psubusb %%mm6, %%mm1 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" %%mm2, %%mm1 \n\t" + "movq %%mm0, (%2, %3) \n\t" + "movq %%mm1, (%2, %%"REG_a") \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + "add %%"REG_a", %2 \n\t" + "add %%"REG_a", %1 \n\t" + "psubusb %%mm6, %%mm1 \n\t" + PAVGB" %%mm1, %%mm2 \n\t" + PAVGB" %%mm0, %%mm1 \n\t" + "movq %%mm2, (%2, %3) \n\t" + "movq %%mm1, (%2, %%"REG_a") \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D" (block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } static void DEF(avg_pixels8)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "1: \n\t" - "movq (%2), %%mm0 \n\t" - "movq (%2, %3), %%mm1 \n\t" - PAVGB" (%1), %%mm0 \n\t" - PAVGB" (%1, %3), %%mm1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%2), %%mm0 \n\t" - "movq (%2, %3), %%mm1 \n\t" - PAVGB" (%1), %%mm0 \n\t" - PAVGB" (%1, %3), %%mm1 \n\t" - "add %%"REG_a", %1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "1: \n\t" + "movq (%2), %%mm0 \n\t" + "movq (%2, %3), %%mm1 \n\t" + PAVGB" (%1), %%mm0 \n\t" + PAVGB" (%1, %3), %%mm1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%2), %%mm0 \n\t" + "movq (%2, %3), %%mm1 \n\t" + PAVGB" (%1), %%mm0 \n\t" + PAVGB" (%1, %3), %%mm1 \n\t" + "add %%"REG_a", %1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } static void DEF(avg_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm2 \n\t" - PAVGB" 1(%1), %%mm0 \n\t" - PAVGB" 1(%1, %3), %%mm2 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" (%2, %3), %%mm2 \n\t" - "add %%"REG_a", %1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm2, (%2, %3) \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%1, %3), %%mm2 \n\t" - PAVGB" 1(%1), %%mm0 \n\t" - PAVGB" 1(%1, %3), %%mm2 \n\t" - "add %%"REG_a", %2 \n\t" - "add %%"REG_a", %1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" (%2, %3), %%mm2 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm2, (%2, %3) \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%1, %3), %%mm2 \n\t" + PAVGB" 1(%1), %%mm0 \n\t" + PAVGB" 1(%1, %3), %%mm2 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" (%2, %3), %%mm2 \n\t" + "add %%"REG_a", %1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm2, (%2, %3) \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%1, %3), %%mm2 \n\t" + PAVGB" 1(%1), %%mm0 \n\t" + PAVGB" 1(%1, %3), %%mm2 \n\t" + "add %%"REG_a", %2 \n\t" + "add %%"REG_a", %1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" (%2, %3), %%mm2 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm2, (%2, %3) \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } static void DEF(avg_pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "movq (%1), %%mm0 \n\t" - "sub %3, %2 \n\t" - "1: \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm2 \n\t" - "add %%"REG_a", %1 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" %%mm2, %%mm1 \n\t" - "movq (%2, %3), %%mm3 \n\t" - "movq (%2, %%"REG_a"), %%mm4 \n\t" - PAVGB" %%mm3, %%mm0 \n\t" - PAVGB" %%mm4, %%mm1 \n\t" - "movq %%mm0, (%2, %3) \n\t" - "movq %%mm1, (%2, %%"REG_a") \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - PAVGB" %%mm1, %%mm2 \n\t" - PAVGB" %%mm0, %%mm1 \n\t" - "add %%"REG_a", %2 \n\t" - "add %%"REG_a", %1 \n\t" - "movq (%2, %3), %%mm3 \n\t" - "movq (%2, %%"REG_a"), %%mm4 \n\t" - PAVGB" %%mm3, %%mm2 \n\t" - PAVGB" %%mm4, %%mm1 \n\t" - "movq %%mm2, (%2, %3) \n\t" - "movq %%mm1, (%2, %%"REG_a") \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "movq (%1), %%mm0 \n\t" + "sub %3, %2 \n\t" + "1: \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm2 \n\t" + "add %%"REG_a", %1 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" %%mm2, %%mm1 \n\t" + "movq (%2, %3), %%mm3 \n\t" + "movq (%2, %%"REG_a"), %%mm4 \n\t" + PAVGB" %%mm3, %%mm0 \n\t" + PAVGB" %%mm4, %%mm1 \n\t" + "movq %%mm0, (%2, %3) \n\t" + "movq %%mm1, (%2, %%"REG_a") \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + PAVGB" %%mm1, %%mm2 \n\t" + PAVGB" %%mm0, %%mm1 \n\t" + "add %%"REG_a", %2 \n\t" + "add %%"REG_a", %1 \n\t" + "movq (%2, %3), %%mm3 \n\t" + "movq (%2, %%"REG_a"), %%mm4 \n\t" + PAVGB" %%mm3, %%mm2 \n\t" + PAVGB" %%mm4, %%mm1 \n\t" + "movq %%mm2, (%2, %3) \n\t" + "movq %%mm1, (%2, %%"REG_a") \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } -// Note this is not correctly rounded, but this function is only used for b frames so it doesnt matter +// Note this is not correctly rounded, but this function is only used for b frames so it doesnt matter static void DEF(avg_pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { MOVQ_BONE(mm6); __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "movq (%1), %%mm0 \n\t" - PAVGB" 1(%1), %%mm0 \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1, %%"REG_a"), %%mm2 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "psubusb %%mm6, %%mm2 \n\t" - PAVGB" 1(%1, %3), %%mm1 \n\t" - PAVGB" 1(%1, %%"REG_a"), %%mm2 \n\t" - "add %%"REG_a", %1 \n\t" - PAVGB" %%mm1, %%mm0 \n\t" - PAVGB" %%mm2, %%mm1 \n\t" - PAVGB" (%2), %%mm0 \n\t" - PAVGB" (%2, %3), %%mm1 \n\t" - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - PAVGB" 1(%1, %3), %%mm1 \n\t" - PAVGB" 1(%1, %%"REG_a"), %%mm0 \n\t" - "add %%"REG_a", %2 \n\t" - "add %%"REG_a", %1 \n\t" - PAVGB" %%mm1, %%mm2 \n\t" - PAVGB" %%mm0, %%mm1 \n\t" - PAVGB" (%2), %%mm2 \n\t" - PAVGB" (%2, %3), %%mm1 \n\t" - "movq %%mm2, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r" ((long)line_size) - :"%"REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "movq (%1), %%mm0 \n\t" + PAVGB" 1(%1), %%mm0 \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1, %%"REG_a"), %%mm2 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "psubusb %%mm6, %%mm2 \n\t" + PAVGB" 1(%1, %3), %%mm1 \n\t" + PAVGB" 1(%1, %%"REG_a"), %%mm2 \n\t" + "add %%"REG_a", %1 \n\t" + PAVGB" %%mm1, %%mm0 \n\t" + PAVGB" %%mm2, %%mm1 \n\t" + PAVGB" (%2), %%mm0 \n\t" + PAVGB" (%2, %3), %%mm1 \n\t" + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + PAVGB" 1(%1, %3), %%mm1 \n\t" + PAVGB" 1(%1, %%"REG_a"), %%mm0 \n\t" + "add %%"REG_a", %2 \n\t" + "add %%"REG_a", %1 \n\t" + PAVGB" %%mm1, %%mm2 \n\t" + PAVGB" %%mm0, %%mm1 \n\t" + PAVGB" (%2), %%mm2 \n\t" + PAVGB" (%2, %3), %%mm1 \n\t" + "movq %%mm2, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r" ((long)line_size) + :"%"REG_a, "memory"); } //FIXME the following could be optimized too ... diff --git a/src/libffmpeg/libavcodec/i386/dsputil_mmx_rnd.h b/src/libffmpeg/libavcodec/i386/dsputil_mmx_rnd.h index a56374b63..3ecd776b8 100644 --- a/src/libffmpeg/libavcodec/i386/dsputil_mmx_rnd.h +++ b/src/libffmpeg/libavcodec/i386/dsputil_mmx_rnd.h @@ -15,7 +15,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA * * MMX optimization by Nick Kurshev <nickols_k@mail.ru> * mostly rewritten by Michael Niedermayer <michaelni@gmx.at> @@ -27,206 +27,206 @@ static void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line { MOVQ_BFE(mm6); __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq 1(%1), %%mm1 \n\t" - "movq (%1, %3), %%mm2 \n\t" - "movq 1(%1, %3), %%mm3 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%2) \n\t" - "movq %%mm5, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 1(%1), %%mm1 \n\t" - "movq (%1, %3), %%mm2 \n\t" - "movq 1(%1, %3), %%mm3 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%2) \n\t" - "movq %%mm5, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r"((long)line_size) - :REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq 1(%1), %%mm1 \n\t" + "movq (%1, %3), %%mm2 \n\t" + "movq 1(%1, %3), %%mm3 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%2) \n\t" + "movq %%mm5, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%1), %%mm0 \n\t" + "movq 1(%1), %%mm1 \n\t" + "movq (%1, %3), %%mm2 \n\t" + "movq 1(%1, %3), %%mm3 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%2) \n\t" + "movq %%mm5, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r"((long)line_size) + :REG_a, "memory"); } static void attribute_unused DEF(put, pixels8_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { MOVQ_BFE(mm6); __asm __volatile( - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "add $8, %2 \n\t" - PAVGB(%%mm0, %%mm1, %%mm4, %%mm6) - "movq %%mm4, (%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "movq (%1), %%mm2 \n\t" - "movq 8(%2), %%mm3 \n\t" - "add %4, %1 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%3) \n\t" - "add %5, %3 \n\t" - "movq %%mm5, (%3) \n\t" - "add %5, %3 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 16(%2), %%mm1 \n\t" - "add %4, %1 \n\t" - "movq (%1), %%mm2 \n\t" - "movq 24(%2), %%mm3 \n\t" - "add %4, %1 \n\t" - "add $32, %2 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%3) \n\t" - "add %5, %3 \n\t" - "movq %%mm5, (%3) \n\t" - "add %5, %3 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "add $8, %2 \n\t" + PAVGB(%%mm0, %%mm1, %%mm4, %%mm6) + "movq %%mm4, (%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "movq (%1), %%mm2 \n\t" + "movq 8(%2), %%mm3 \n\t" + "add %4, %1 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%3) \n\t" + "add %5, %3 \n\t" + "movq %%mm5, (%3) \n\t" + "add %5, %3 \n\t" + "movq (%1), %%mm0 \n\t" + "movq 16(%2), %%mm1 \n\t" + "add %4, %1 \n\t" + "movq (%1), %%mm2 \n\t" + "movq 24(%2), %%mm3 \n\t" + "add %4, %1 \n\t" + "add $32, %2 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%3) \n\t" + "add %5, %3 \n\t" + "movq %%mm5, (%3) \n\t" + "add %5, %3 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); } static void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { MOVQ_BFE(mm6); __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq 1(%1), %%mm1 \n\t" - "movq (%1, %3), %%mm2 \n\t" - "movq 1(%1, %3), %%mm3 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%2) \n\t" - "movq %%mm5, (%2, %3) \n\t" - "movq 8(%1), %%mm0 \n\t" - "movq 9(%1), %%mm1 \n\t" - "movq 8(%1, %3), %%mm2 \n\t" - "movq 9(%1, %3), %%mm3 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, 8(%2) \n\t" - "movq %%mm5, 8(%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 1(%1), %%mm1 \n\t" - "movq (%1, %3), %%mm2 \n\t" - "movq 1(%1, %3), %%mm3 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%2) \n\t" - "movq %%mm5, (%2, %3) \n\t" - "movq 8(%1), %%mm0 \n\t" - "movq 9(%1), %%mm1 \n\t" - "movq 8(%1, %3), %%mm2 \n\t" - "movq 9(%1, %3), %%mm3 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, 8(%2) \n\t" - "movq %%mm5, 8(%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r"((long)line_size) - :REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq 1(%1), %%mm1 \n\t" + "movq (%1, %3), %%mm2 \n\t" + "movq 1(%1, %3), %%mm3 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%2) \n\t" + "movq %%mm5, (%2, %3) \n\t" + "movq 8(%1), %%mm0 \n\t" + "movq 9(%1), %%mm1 \n\t" + "movq 8(%1, %3), %%mm2 \n\t" + "movq 9(%1, %3), %%mm3 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, 8(%2) \n\t" + "movq %%mm5, 8(%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%1), %%mm0 \n\t" + "movq 1(%1), %%mm1 \n\t" + "movq (%1, %3), %%mm2 \n\t" + "movq 1(%1, %3), %%mm3 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%2) \n\t" + "movq %%mm5, (%2, %3) \n\t" + "movq 8(%1), %%mm0 \n\t" + "movq 9(%1), %%mm1 \n\t" + "movq 8(%1, %3), %%mm2 \n\t" + "movq 9(%1, %3), %%mm3 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, 8(%2) \n\t" + "movq %%mm5, 8(%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r"((long)line_size) + :REG_a, "memory"); } static void attribute_unused DEF(put, pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h) { MOVQ_BFE(mm6); __asm __volatile( - "testl $1, %0 \n\t" - " jz 1f \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%2), %%mm1 \n\t" - "movq 8(%1), %%mm2 \n\t" - "movq 8(%2), %%mm3 \n\t" - "add %4, %1 \n\t" - "add $16, %2 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%3) \n\t" - "movq %%mm5, 8(%3) \n\t" - "add %5, %3 \n\t" - "decl %0 \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1), %%mm0 \n\t" - "movq (%2), %%mm1 \n\t" - "movq 8(%1), %%mm2 \n\t" - "movq 8(%2), %%mm3 \n\t" - "add %4, %1 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%3) \n\t" - "movq %%mm5, 8(%3) \n\t" - "add %5, %3 \n\t" - "movq (%1), %%mm0 \n\t" - "movq 16(%2), %%mm1 \n\t" - "movq 8(%1), %%mm2 \n\t" - "movq 24(%2), %%mm3 \n\t" - "add %4, %1 \n\t" - PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) - "movq %%mm4, (%3) \n\t" - "movq %%mm5, 8(%3) \n\t" - "add %5, %3 \n\t" - "add $32, %2 \n\t" - "subl $2, %0 \n\t" - "jnz 1b \n\t" + "testl $1, %0 \n\t" + " jz 1f \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%2), %%mm1 \n\t" + "movq 8(%1), %%mm2 \n\t" + "movq 8(%2), %%mm3 \n\t" + "add %4, %1 \n\t" + "add $16, %2 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%3) \n\t" + "movq %%mm5, 8(%3) \n\t" + "add %5, %3 \n\t" + "decl %0 \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%2), %%mm1 \n\t" + "movq 8(%1), %%mm2 \n\t" + "movq 8(%2), %%mm3 \n\t" + "add %4, %1 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%3) \n\t" + "movq %%mm5, 8(%3) \n\t" + "add %5, %3 \n\t" + "movq (%1), %%mm0 \n\t" + "movq 16(%2), %%mm1 \n\t" + "movq 8(%1), %%mm2 \n\t" + "movq 24(%2), %%mm3 \n\t" + "add %4, %1 \n\t" + PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%3) \n\t" + "movq %%mm5, 8(%3) \n\t" + "add %5, %3 \n\t" + "add $32, %2 \n\t" + "subl $2, %0 \n\t" + "jnz 1b \n\t" #ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used - :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst) #else - :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) + :"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst) #endif - :"S"((long)src1Stride), "D"((long)dstStride) - :"memory"); + :"S"((long)src1Stride), "D"((long)dstStride) + :"memory"); } static void DEF(put, pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) { MOVQ_BFE(mm6); __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "movq (%1), %%mm0 \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"),%%mm2 \n\t" - PAVGBP(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5) - "movq %%mm4, (%2) \n\t" - "movq %%mm5, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"),%%mm0 \n\t" - PAVGBP(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5) - "movq %%mm4, (%2) \n\t" - "movq %%mm5, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r"((long)line_size) - :REG_a, "memory"); + "lea (%3, %3), %%"REG_a" \n\t" + "movq (%1), %%mm0 \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"),%%mm2 \n\t" + PAVGBP(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5) + "movq %%mm4, (%2) \n\t" + "movq %%mm5, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"),%%mm0 \n\t" + PAVGBP(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5) + "movq %%mm4, (%2) \n\t" + "movq %%mm5, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r"((long)line_size) + :REG_a, "memory"); } static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line_size, int h) @@ -234,65 +234,65 @@ static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int lin MOVQ_ZERO(mm7); SET_RND(mm6); // =2 for rnd and =1 for no_rnd version __asm __volatile( - "movq (%1), %%mm0 \n\t" - "movq 1(%1), %%mm4 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm4, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "punpckhbw %%mm7, %%mm5 \n\t" - "paddusw %%mm0, %%mm4 \n\t" - "paddusw %%mm1, %%mm5 \n\t" - "xor %%"REG_a", %%"REG_a" \n\t" - "add %3, %1 \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - "movq 1(%1, %%"REG_a"), %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "punpckhbw %%mm7, %%mm3 \n\t" - "paddusw %%mm2, %%mm0 \n\t" - "paddusw %%mm3, %%mm1 \n\t" - "paddusw %%mm6, %%mm4 \n\t" - "paddusw %%mm6, %%mm5 \n\t" - "paddusw %%mm0, %%mm4 \n\t" - "paddusw %%mm1, %%mm5 \n\t" - "psrlw $2, %%mm4 \n\t" - "psrlw $2, %%mm5 \n\t" - "packuswb %%mm5, %%mm4 \n\t" - "movq %%mm4, (%2, %%"REG_a") \n\t" - "add %3, %%"REG_a" \n\t" + "movq (%1), %%mm0 \n\t" + "movq 1(%1), %%mm4 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm4, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpckhbw %%mm7, %%mm5 \n\t" + "paddusw %%mm0, %%mm4 \n\t" + "paddusw %%mm1, %%mm5 \n\t" + "xor %%"REG_a", %%"REG_a" \n\t" + "add %3, %1 \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + "movq 1(%1, %%"REG_a"), %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "paddusw %%mm2, %%mm0 \n\t" + "paddusw %%mm3, %%mm1 \n\t" + "paddusw %%mm6, %%mm4 \n\t" + "paddusw %%mm6, %%mm5 \n\t" + "paddusw %%mm0, %%mm4 \n\t" + "paddusw %%mm1, %%mm5 \n\t" + "psrlw $2, %%mm4 \n\t" + "psrlw $2, %%mm5 \n\t" + "packuswb %%mm5, %%mm4 \n\t" + "movq %%mm4, (%2, %%"REG_a") \n\t" + "add %3, %%"REG_a" \n\t" - "movq (%1, %%"REG_a"), %%mm2 \n\t" // 0 <-> 2 1 <-> 3 - "movq 1(%1, %%"REG_a"), %%mm4 \n\t" - "movq %%mm2, %%mm3 \n\t" - "movq %%mm4, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "punpckhbw %%mm7, %%mm3 \n\t" - "punpckhbw %%mm7, %%mm5 \n\t" - "paddusw %%mm2, %%mm4 \n\t" - "paddusw %%mm3, %%mm5 \n\t" - "paddusw %%mm6, %%mm0 \n\t" - "paddusw %%mm6, %%mm1 \n\t" - "paddusw %%mm4, %%mm0 \n\t" - "paddusw %%mm5, %%mm1 \n\t" - "psrlw $2, %%mm0 \n\t" - "psrlw $2, %%mm1 \n\t" - "packuswb %%mm1, %%mm0 \n\t" - "movq %%mm0, (%2, %%"REG_a") \n\t" - "add %3, %%"REG_a" \n\t" + "movq (%1, %%"REG_a"), %%mm2 \n\t" // 0 <-> 2 1 <-> 3 + "movq 1(%1, %%"REG_a"), %%mm4 \n\t" + "movq %%mm2, %%mm3 \n\t" + "movq %%mm4, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "punpckhbw %%mm7, %%mm5 \n\t" + "paddusw %%mm2, %%mm4 \n\t" + "paddusw %%mm3, %%mm5 \n\t" + "paddusw %%mm6, %%mm0 \n\t" + "paddusw %%mm6, %%mm1 \n\t" + "paddusw %%mm4, %%mm0 \n\t" + "paddusw %%mm5, %%mm1 \n\t" + "psrlw $2, %%mm0 \n\t" + "psrlw $2, %%mm1 \n\t" + "packuswb %%mm1, %%mm0 \n\t" + "movq %%mm0, (%2, %%"REG_a") \n\t" + "add %3, %%"REG_a" \n\t" - "subl $2, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels) - :"D"(block), "r"((long)line_size) - :REG_a, "memory"); + "subl $2, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels) + :"D"(block), "r"((long)line_size) + :REG_a, "memory"); } // avg_pixels @@ -301,16 +301,16 @@ static void attribute_unused DEF(avg, pixels4)(uint8_t *block, const uint8_t *pi MOVQ_BFE(mm6); JUMPALIGN(); do { - __asm __volatile( - "movd %0, %%mm0 \n\t" - "movd %1, %%mm1 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - "movd %%mm2, %0 \n\t" - :"+m"(*block) - :"m"(*pixels) - :"memory"); - pixels += line_size; - block += line_size; + __asm __volatile( + "movd %0, %%mm0 \n\t" + "movd %1, %%mm1 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + "movd %%mm2, %0 \n\t" + :"+m"(*block) + :"m"(*pixels) + :"memory"); + pixels += line_size; + block += line_size; } while (--h); } @@ -321,16 +321,16 @@ static void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, int line_si MOVQ_BFE(mm6); JUMPALIGN(); do { - __asm __volatile( - "movq %0, %%mm0 \n\t" - "movq %1, %%mm1 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - "movq %%mm2, %0 \n\t" - :"+m"(*block) - :"m"(*pixels) - :"memory"); - pixels += line_size; - block += line_size; + __asm __volatile( + "movq %0, %%mm0 \n\t" + "movq %1, %%mm1 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + "movq %%mm2, %0 \n\t" + :"+m"(*block) + :"m"(*pixels) + :"memory"); + pixels += line_size; + block += line_size; } while (--h); } @@ -340,20 +340,20 @@ static void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, int line_s MOVQ_BFE(mm6); JUMPALIGN(); do { - __asm __volatile( - "movq %0, %%mm0 \n\t" - "movq %1, %%mm1 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - "movq %%mm2, %0 \n\t" - "movq 8%0, %%mm0 \n\t" - "movq 8%1, %%mm1 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - "movq %%mm2, 8%0 \n\t" - :"+m"(*block) - :"m"(*pixels) - :"memory"); - pixels += line_size; - block += line_size; + __asm __volatile( + "movq %0, %%mm0 \n\t" + "movq %1, %%mm1 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + "movq %%mm2, %0 \n\t" + "movq 8%0, %%mm0 \n\t" + "movq 8%1, %%mm1 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + "movq %%mm2, 8%0 \n\t" + :"+m"(*block) + :"m"(*pixels) + :"memory"); + pixels += line_size; + block += line_size; } while (--h); } @@ -363,18 +363,18 @@ static void DEF(avg, pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line MOVQ_BFE(mm6); JUMPALIGN(); do { - __asm __volatile( - "movq %1, %%mm0 \n\t" - "movq 1%1, %%mm1 \n\t" - "movq %0, %%mm3 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) - "movq %%mm0, %0 \n\t" - :"+m"(*block) - :"m"(*pixels) - :"memory"); - pixels += line_size; - block += line_size; + __asm __volatile( + "movq %1, %%mm0 \n\t" + "movq 1%1, %%mm1 \n\t" + "movq %0, %%mm3 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) + "movq %%mm0, %0 \n\t" + :"+m"(*block) + :"m"(*pixels) + :"memory"); + pixels += line_size; + block += line_size; } while (--h); } @@ -383,17 +383,17 @@ static __attribute__((unused)) void DEF(avg, pixels8_l2)(uint8_t *dst, uint8_t * MOVQ_BFE(mm6); JUMPALIGN(); do { - __asm __volatile( - "movq %1, %%mm0 \n\t" - "movq %2, %%mm1 \n\t" - "movq %0, %%mm3 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) - "movq %%mm0, %0 \n\t" - :"+m"(*dst) - :"m"(*src1), "m"(*src2) - :"memory"); - dst += dstStride; + __asm __volatile( + "movq %1, %%mm0 \n\t" + "movq %2, %%mm1 \n\t" + "movq %0, %%mm3 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) + "movq %%mm0, %0 \n\t" + :"+m"(*dst) + :"m"(*src1), "m"(*src2) + :"memory"); + dst += dstStride; src1 += src1Stride; src2 += 8; } while (--h); @@ -404,24 +404,24 @@ static void DEF(avg, pixels16_x2)(uint8_t *block, const uint8_t *pixels, int lin MOVQ_BFE(mm6); JUMPALIGN(); do { - __asm __volatile( - "movq %1, %%mm0 \n\t" - "movq 1%1, %%mm1 \n\t" - "movq %0, %%mm3 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) - "movq %%mm0, %0 \n\t" - "movq 8%1, %%mm0 \n\t" - "movq 9%1, %%mm1 \n\t" - "movq 8%0, %%mm3 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) - "movq %%mm0, 8%0 \n\t" - :"+m"(*block) - :"m"(*pixels) - :"memory"); - pixels += line_size; - block += line_size; + __asm __volatile( + "movq %1, %%mm0 \n\t" + "movq 1%1, %%mm1 \n\t" + "movq %0, %%mm3 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) + "movq %%mm0, %0 \n\t" + "movq 8%1, %%mm0 \n\t" + "movq 9%1, %%mm1 \n\t" + "movq 8%0, %%mm3 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) + "movq %%mm0, 8%0 \n\t" + :"+m"(*block) + :"m"(*pixels) + :"memory"); + pixels += line_size; + block += line_size; } while (--h); } @@ -430,23 +430,23 @@ static __attribute__((unused)) void DEF(avg, pixels16_l2)(uint8_t *dst, uint8_t MOVQ_BFE(mm6); JUMPALIGN(); do { - __asm __volatile( - "movq %1, %%mm0 \n\t" - "movq %2, %%mm1 \n\t" - "movq %0, %%mm3 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) - "movq %%mm0, %0 \n\t" - "movq 8%1, %%mm0 \n\t" - "movq 8%2, %%mm1 \n\t" - "movq 8%0, %%mm3 \n\t" - PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) - PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) - "movq %%mm0, 8%0 \n\t" - :"+m"(*dst) - :"m"(*src1), "m"(*src2) - :"memory"); - dst += dstStride; + __asm __volatile( + "movq %1, %%mm0 \n\t" + "movq %2, %%mm1 \n\t" + "movq %0, %%mm3 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) + "movq %%mm0, %0 \n\t" + "movq 8%1, %%mm0 \n\t" + "movq 8%2, %%mm1 \n\t" + "movq 8%0, %%mm3 \n\t" + PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) + PAVGB(%%mm3, %%mm2, %%mm0, %%mm6) + "movq %%mm0, 8%0 \n\t" + :"+m"(*dst) + :"m"(*src1), "m"(*src2) + :"memory"); + dst += dstStride; src1 += src1Stride; src2 += 16; } while (--h); @@ -456,39 +456,39 @@ static void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line { MOVQ_BFE(mm6); __asm __volatile( - "lea (%3, %3), %%"REG_a" \n\t" - "movq (%1), %%mm0 \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm2 \n\t" - PAVGBP(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5) - "movq (%2), %%mm3 \n\t" - PAVGB(%%mm3, %%mm4, %%mm0, %%mm6) - "movq (%2, %3), %%mm3 \n\t" - PAVGB(%%mm3, %%mm5, %%mm1, %%mm6) - "movq %%mm0, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" + "lea (%3, %3), %%"REG_a" \n\t" + "movq (%1), %%mm0 \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm2 \n\t" + PAVGBP(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5) + "movq (%2), %%mm3 \n\t" + PAVGB(%%mm3, %%mm4, %%mm0, %%mm6) + "movq (%2, %3), %%mm3 \n\t" + PAVGB(%%mm3, %%mm5, %%mm1, %%mm6) + "movq %%mm0, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" - "movq (%1, %3), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - PAVGBP(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5) - "movq (%2), %%mm3 \n\t" - PAVGB(%%mm3, %%mm4, %%mm2, %%mm6) - "movq (%2, %3), %%mm3 \n\t" - PAVGB(%%mm3, %%mm5, %%mm1, %%mm6) - "movq %%mm2, (%2) \n\t" - "movq %%mm1, (%2, %3) \n\t" - "add %%"REG_a", %1 \n\t" - "add %%"REG_a", %2 \n\t" + "movq (%1, %3), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + PAVGBP(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5) + "movq (%2), %%mm3 \n\t" + PAVGB(%%mm3, %%mm4, %%mm2, %%mm6) + "movq (%2, %3), %%mm3 \n\t" + PAVGB(%%mm3, %%mm5, %%mm1, %%mm6) + "movq %%mm2, (%2) \n\t" + "movq %%mm1, (%2, %3) \n\t" + "add %%"REG_a", %1 \n\t" + "add %%"REG_a", %2 \n\t" - "subl $4, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels), "+D"(block) - :"r"((long)line_size) - :REG_a, "memory"); + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels), "+D"(block) + :"r"((long)line_size) + :REG_a, "memory"); } // this routine is 'slightly' suboptimal but mostly unused @@ -497,73 +497,73 @@ static void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int lin MOVQ_ZERO(mm7); SET_RND(mm6); // =2 for rnd and =1 for no_rnd version __asm __volatile( - "movq (%1), %%mm0 \n\t" - "movq 1(%1), %%mm4 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm4, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "punpckhbw %%mm7, %%mm5 \n\t" - "paddusw %%mm0, %%mm4 \n\t" - "paddusw %%mm1, %%mm5 \n\t" - "xor %%"REG_a", %%"REG_a" \n\t" - "add %3, %1 \n\t" - ".balign 8 \n\t" - "1: \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - "movq 1(%1, %%"REG_a"), %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "punpckhbw %%mm7, %%mm3 \n\t" - "paddusw %%mm2, %%mm0 \n\t" - "paddusw %%mm3, %%mm1 \n\t" - "paddusw %%mm6, %%mm4 \n\t" - "paddusw %%mm6, %%mm5 \n\t" - "paddusw %%mm0, %%mm4 \n\t" - "paddusw %%mm1, %%mm5 \n\t" - "psrlw $2, %%mm4 \n\t" - "psrlw $2, %%mm5 \n\t" - "movq (%2, %%"REG_a"), %%mm3 \n\t" - "packuswb %%mm5, %%mm4 \n\t" - "pcmpeqd %%mm2, %%mm2 \n\t" - "paddb %%mm2, %%mm2 \n\t" - PAVGB(%%mm3, %%mm4, %%mm5, %%mm2) - "movq %%mm5, (%2, %%"REG_a") \n\t" - "add %3, %%"REG_a" \n\t" + "movq (%1), %%mm0 \n\t" + "movq 1(%1), %%mm4 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm4, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpckhbw %%mm7, %%mm5 \n\t" + "paddusw %%mm0, %%mm4 \n\t" + "paddusw %%mm1, %%mm5 \n\t" + "xor %%"REG_a", %%"REG_a" \n\t" + "add %3, %1 \n\t" + ".balign 8 \n\t" + "1: \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + "movq 1(%1, %%"REG_a"), %%mm2 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm2, %%mm3 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "paddusw %%mm2, %%mm0 \n\t" + "paddusw %%mm3, %%mm1 \n\t" + "paddusw %%mm6, %%mm4 \n\t" + "paddusw %%mm6, %%mm5 \n\t" + "paddusw %%mm0, %%mm4 \n\t" + "paddusw %%mm1, %%mm5 \n\t" + "psrlw $2, %%mm4 \n\t" + "psrlw $2, %%mm5 \n\t" + "movq (%2, %%"REG_a"), %%mm3 \n\t" + "packuswb %%mm5, %%mm4 \n\t" + "pcmpeqd %%mm2, %%mm2 \n\t" + "paddb %%mm2, %%mm2 \n\t" + PAVGB(%%mm3, %%mm4, %%mm5, %%mm2) + "movq %%mm5, (%2, %%"REG_a") \n\t" + "add %3, %%"REG_a" \n\t" - "movq (%1, %%"REG_a"), %%mm2 \n\t" // 0 <-> 2 1 <-> 3 - "movq 1(%1, %%"REG_a"), %%mm4 \n\t" - "movq %%mm2, %%mm3 \n\t" - "movq %%mm4, %%mm5 \n\t" - "punpcklbw %%mm7, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm4 \n\t" - "punpckhbw %%mm7, %%mm3 \n\t" - "punpckhbw %%mm7, %%mm5 \n\t" - "paddusw %%mm2, %%mm4 \n\t" - "paddusw %%mm3, %%mm5 \n\t" - "paddusw %%mm6, %%mm0 \n\t" - "paddusw %%mm6, %%mm1 \n\t" - "paddusw %%mm4, %%mm0 \n\t" - "paddusw %%mm5, %%mm1 \n\t" - "psrlw $2, %%mm0 \n\t" - "psrlw $2, %%mm1 \n\t" - "movq (%2, %%"REG_a"), %%mm3 \n\t" - "packuswb %%mm1, %%mm0 \n\t" - "pcmpeqd %%mm2, %%mm2 \n\t" - "paddb %%mm2, %%mm2 \n\t" - PAVGB(%%mm3, %%mm0, %%mm1, %%mm2) - "movq %%mm1, (%2, %%"REG_a") \n\t" - "add %3, %%"REG_a" \n\t" + "movq (%1, %%"REG_a"), %%mm2 \n\t" // 0 <-> 2 1 <-> 3 + "movq 1(%1, %%"REG_a"), %%mm4 \n\t" + "movq %%mm2, %%mm3 \n\t" + "movq %%mm4, %%mm5 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm4 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "punpckhbw %%mm7, %%mm5 \n\t" + "paddusw %%mm2, %%mm4 \n\t" + "paddusw %%mm3, %%mm5 \n\t" + "paddusw %%mm6, %%mm0 \n\t" + "paddusw %%mm6, %%mm1 \n\t" + "paddusw %%mm4, %%mm0 \n\t" + "paddusw %%mm5, %%mm1 \n\t" + "psrlw $2, %%mm0 \n\t" + "psrlw $2, %%mm1 \n\t" + "movq (%2, %%"REG_a"), %%mm3 \n\t" + "packuswb %%mm1, %%mm0 \n\t" + "pcmpeqd %%mm2, %%mm2 \n\t" + "paddb %%mm2, %%mm2 \n\t" + PAVGB(%%mm3, %%mm0, %%mm1, %%mm2) + "movq %%mm1, (%2, %%"REG_a") \n\t" + "add %3, %%"REG_a" \n\t" - "subl $2, %0 \n\t" - "jnz 1b \n\t" - :"+g"(h), "+S"(pixels) - :"D"(block), "r"((long)line_size) - :REG_a, "memory"); + "subl $2, %0 \n\t" + "jnz 1b \n\t" + :"+g"(h), "+S"(pixels) + :"D"(block), "r"((long)line_size) + :REG_a, "memory"); } //FIXME optimize diff --git a/src/libffmpeg/libavcodec/i386/fdct_mmx.c b/src/libffmpeg/libavcodec/i386/fdct_mmx.c index 6a13090a1..f6150c83c 100644 --- a/src/libffmpeg/libavcodec/i386/fdct_mmx.c +++ b/src/libffmpeg/libavcodec/i386/fdct_mmx.c @@ -5,7 +5,7 @@ * SSE2 optimization is Copyright (c) 2004 Denes Balatoni. * * from fdctam32.c - AP922 MMX(3D-Now) forward-DCT - * + * * Intel Application Note AP-922 - fast, precise implementation of DCT * http://developer.intel.com/vtune/cbts/appnotes.htm * @@ -30,28 +30,28 @@ // ////////////////////////////////////////////////////////////////////// -#define BITS_FRW_ACC 3 //; 2 or 3 for accuracy -#define SHIFT_FRW_COL BITS_FRW_ACC -#define SHIFT_FRW_ROW (BITS_FRW_ACC + 17 - 3) -#define RND_FRW_ROW (1 << (SHIFT_FRW_ROW-1)) -//#define RND_FRW_COL (1 << (SHIFT_FRW_COL-1)) +#define BITS_FRW_ACC 3 //; 2 or 3 for accuracy +#define SHIFT_FRW_COL BITS_FRW_ACC +#define SHIFT_FRW_ROW (BITS_FRW_ACC + 17 - 3) +#define RND_FRW_ROW (1 << (SHIFT_FRW_ROW-1)) +//#define RND_FRW_COL (1 << (SHIFT_FRW_COL-1)) //concatenated table, for forward DCT transformation static const int16_t fdct_tg_all_16[] ATTR_ALIGN(8) = { - 13036, 13036, 13036, 13036, // tg * (2<<16) + 0.5 - 27146, 27146, 27146, 27146, // tg * (2<<16) + 0.5 - -21746, -21746, -21746, -21746, // tg * (2<<16) + 0.5 + 13036, 13036, 13036, 13036, // tg * (2<<16) + 0.5 + 27146, 27146, 27146, 27146, // tg * (2<<16) + 0.5 + -21746, -21746, -21746, -21746, // tg * (2<<16) + 0.5 }; static const int16_t ocos_4_16[4] ATTR_ALIGN(8) = { - 23170, 23170, 23170, 23170, //cos * (2<<15) + 0.5 + 23170, 23170, 23170, 23170, //cos * (2<<15) + 0.5 }; static const int64_t fdct_one_corr ATTR_ALIGN(8) = 0x0001000100010001LL; static const int32_t fdct_r_row[2] ATTR_ALIGN(8) = {RND_FRW_ROW, RND_FRW_ROW }; -struct +struct { const int32_t fdct_r_row_sse2[4] ATTR_ALIGN(16); } fdct_r_row_sse2 ATTR_ALIGN(16)= @@ -61,90 +61,90 @@ struct //static const long fdct_r_row_sse2[4] ATTR_ALIGN(16) = {RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW}; static const int16_t tab_frw_01234567[] ATTR_ALIGN(8) = { // forward_dct coeff table - 16384, 16384, 22725, 19266, - 16384, 16384, 12873, 4520, - 21407, 8867, 19266, -4520, - -8867, -21407, -22725, -12873, - 16384, -16384, 12873, -22725, - -16384, 16384, 4520, 19266, - 8867, -21407, 4520, -12873, - 21407, -8867, 19266, -22725, - - 22725, 22725, 31521, 26722, - 22725, 22725, 17855, 6270, - 29692, 12299, 26722, -6270, - -12299, -29692, -31521, -17855, - 22725, -22725, 17855, -31521, - -22725, 22725, 6270, 26722, - 12299, -29692, 6270, -17855, - 29692, -12299, 26722, -31521, - - 21407, 21407, 29692, 25172, - 21407, 21407, 16819, 5906, - 27969, 11585, 25172, -5906, - -11585, -27969, -29692, -16819, - 21407, -21407, 16819, -29692, - -21407, 21407, 5906, 25172, - 11585, -27969, 5906, -16819, - 27969, -11585, 25172, -29692, - - 19266, 19266, 26722, 22654, - 19266, 19266, 15137, 5315, - 25172, 10426, 22654, -5315, - -10426, -25172, -26722, -15137, - 19266, -19266, 15137, -26722, - -19266, 19266, 5315, 22654, - 10426, -25172, 5315, -15137, - 25172, -10426, 22654, -26722, - - 16384, 16384, 22725, 19266, - 16384, 16384, 12873, 4520, - 21407, 8867, 19266, -4520, - -8867, -21407, -22725, -12873, - 16384, -16384, 12873, -22725, - -16384, 16384, 4520, 19266, - 8867, -21407, 4520, -12873, - 21407, -8867, 19266, -22725, - - 19266, 19266, 26722, 22654, - 19266, 19266, 15137, 5315, - 25172, 10426, 22654, -5315, - -10426, -25172, -26722, -15137, - 19266, -19266, 15137, -26722, - -19266, 19266, 5315, 22654, - 10426, -25172, 5315, -15137, - 25172, -10426, 22654, -26722, - - 21407, 21407, 29692, 25172, - 21407, 21407, 16819, 5906, - 27969, 11585, 25172, -5906, - -11585, -27969, -29692, -16819, - 21407, -21407, 16819, -29692, - -21407, 21407, 5906, 25172, - 11585, -27969, 5906, -16819, - 27969, -11585, 25172, -29692, - - 22725, 22725, 31521, 26722, - 22725, 22725, 17855, 6270, - 29692, 12299, 26722, -6270, - -12299, -29692, -31521, -17855, - 22725, -22725, 17855, -31521, - -22725, 22725, 6270, 26722, - 12299, -29692, 6270, -17855, - 29692, -12299, 26722, -31521, + 16384, 16384, 22725, 19266, + 16384, 16384, 12873, 4520, + 21407, 8867, 19266, -4520, + -8867, -21407, -22725, -12873, + 16384, -16384, 12873, -22725, + -16384, 16384, 4520, 19266, + 8867, -21407, 4520, -12873, + 21407, -8867, 19266, -22725, + + 22725, 22725, 31521, 26722, + 22725, 22725, 17855, 6270, + 29692, 12299, 26722, -6270, + -12299, -29692, -31521, -17855, + 22725, -22725, 17855, -31521, + -22725, 22725, 6270, 26722, + 12299, -29692, 6270, -17855, + 29692, -12299, 26722, -31521, + + 21407, 21407, 29692, 25172, + 21407, 21407, 16819, 5906, + 27969, 11585, 25172, -5906, + -11585, -27969, -29692, -16819, + 21407, -21407, 16819, -29692, + -21407, 21407, 5906, 25172, + 11585, -27969, 5906, -16819, + 27969, -11585, 25172, -29692, + + 19266, 19266, 26722, 22654, + 19266, 19266, 15137, 5315, + 25172, 10426, 22654, -5315, + -10426, -25172, -26722, -15137, + 19266, -19266, 15137, -26722, + -19266, 19266, 5315, 22654, + 10426, -25172, 5315, -15137, + 25172, -10426, 22654, -26722, + + 16384, 16384, 22725, 19266, + 16384, 16384, 12873, 4520, + 21407, 8867, 19266, -4520, + -8867, -21407, -22725, -12873, + 16384, -16384, 12873, -22725, + -16384, 16384, 4520, 19266, + 8867, -21407, 4520, -12873, + 21407, -8867, 19266, -22725, + + 19266, 19266, 26722, 22654, + 19266, 19266, 15137, 5315, + 25172, 10426, 22654, -5315, + -10426, -25172, -26722, -15137, + 19266, -19266, 15137, -26722, + -19266, 19266, 5315, 22654, + 10426, -25172, 5315, -15137, + 25172, -10426, 22654, -26722, + + 21407, 21407, 29692, 25172, + 21407, 21407, 16819, 5906, + 27969, 11585, 25172, -5906, + -11585, -27969, -29692, -16819, + 21407, -21407, 16819, -29692, + -21407, 21407, 5906, 25172, + 11585, -27969, 5906, -16819, + 27969, -11585, 25172, -29692, + + 22725, 22725, 31521, 26722, + 22725, 22725, 17855, 6270, + 29692, 12299, 26722, -6270, + -12299, -29692, -31521, -17855, + 22725, -22725, 17855, -31521, + -22725, 22725, 6270, 26722, + 12299, -29692, 6270, -17855, + 29692, -12299, 26722, -31521, }; -struct +struct { const int16_t tab_frw_01234567_sse2[256] ATTR_ALIGN(16); } tab_frw_01234567_sse2 ATTR_ALIGN(16) = {{ -//static const int16_t tab_frw_01234567_sse2[] ATTR_ALIGN(16) = { // forward_dct coeff table +//static const int16_t tab_frw_01234567_sse2[] ATTR_ALIGN(16) = { // forward_dct coeff table #define TABLE_SSE2 C4, C4, C1, C3, -C6, -C2, -C1, -C5, \ C4, C4, C5, C7, C2, C6, C3, -C7, \ -C4, C4, C7, C3, C6, -C2, C7, -C5, \ - C4, -C4, C5, -C1, C2, -C6, C3, -C1, -// c1..c7 * cos(pi/4) * 2^15 + C4, -C4, C5, -C1, C2, -C6, C3, -C1, +// c1..c7 * cos(pi/4) * 2^15 #define C1 22725 #define C2 21407 #define C3 19266 @@ -351,67 +351,67 @@ static always_inline void fdct_col(const int16_t *in, int16_t *out, int offset) static always_inline void fdct_row_sse2(const int16_t *in, int16_t *out) { asm volatile( - ".macro FDCT_ROW_SSE2_H1 i t \n\t" - "movq \\i(%0), %%xmm2 \n\t" - "movq \\i+8(%0), %%xmm0 \n\t" - "movdqa \\t+32(%1), %%xmm3 \n\t" - "movdqa \\t+48(%1), %%xmm7 \n\t" - "movdqa \\t(%1), %%xmm4 \n\t" - "movdqa \\t+16(%1), %%xmm5 \n\t" - ".endm \n\t" - ".macro FDCT_ROW_SSE2_H2 i t \n\t" - "movq \\i(%0), %%xmm2 \n\t" - "movq \\i+8(%0), %%xmm0 \n\t" - "movdqa \\t+32(%1), %%xmm3 \n\t" - "movdqa \\t+48(%1), %%xmm7 \n\t" - ".endm \n\t" - ".macro FDCT_ROW_SSE2 i \n\t" - "movq %%xmm2, %%xmm1 \n\t" - "pshuflw $27, %%xmm0, %%xmm0 \n\t" - "paddsw %%xmm0, %%xmm1 \n\t" - "psubsw %%xmm0, %%xmm2 \n\t" - "punpckldq %%xmm2, %%xmm1 \n\t" - "pshufd $78, %%xmm1, %%xmm2 \n\t" - "pmaddwd %%xmm2, %%xmm3 \n\t" - "pmaddwd %%xmm1, %%xmm7 \n\t" - "pmaddwd %%xmm5, %%xmm2 \n\t" - "pmaddwd %%xmm4, %%xmm1 \n\t" - "paddd %%xmm7, %%xmm3 \n\t" - "paddd %%xmm2, %%xmm1 \n\t" - "paddd %%xmm6, %%xmm3 \n\t" - "paddd %%xmm6, %%xmm1 \n\t" - "psrad %3, %%xmm3 \n\t" - "psrad %3, %%xmm1 \n\t" - "packssdw %%xmm3, %%xmm1 \n\t" - "movdqa %%xmm1, \\i(%4) \n\t" - ".endm \n\t" - "movdqa (%2), %%xmm6 \n\t" - "FDCT_ROW_SSE2_H1 0 0 \n\t" - "FDCT_ROW_SSE2 0 \n\t" - "FDCT_ROW_SSE2_H2 64 0 \n\t" - "FDCT_ROW_SSE2 64 \n\t" - - "FDCT_ROW_SSE2_H1 16 64 \n\t" - "FDCT_ROW_SSE2 16 \n\t" - "FDCT_ROW_SSE2_H2 112 64 \n\t" - "FDCT_ROW_SSE2 112 \n\t" - - "FDCT_ROW_SSE2_H1 32 128 \n\t" - "FDCT_ROW_SSE2 32 \n\t" - "FDCT_ROW_SSE2_H2 96 128 \n\t" - "FDCT_ROW_SSE2 96 \n\t" - - "FDCT_ROW_SSE2_H1 48 192 \n\t" - "FDCT_ROW_SSE2 48 \n\t" - "FDCT_ROW_SSE2_H2 80 192 \n\t" - "FDCT_ROW_SSE2 80 \n\t" - : - : "r" (in), "r" (tab_frw_01234567_sse2.tab_frw_01234567_sse2), "r" (fdct_r_row_sse2.fdct_r_row_sse2), "i" (SHIFT_FRW_ROW), "r" (out) + ".macro FDCT_ROW_SSE2_H1 i t \n\t" + "movq \\i(%0), %%xmm2 \n\t" + "movq \\i+8(%0), %%xmm0 \n\t" + "movdqa \\t+32(%1), %%xmm3 \n\t" + "movdqa \\t+48(%1), %%xmm7 \n\t" + "movdqa \\t(%1), %%xmm4 \n\t" + "movdqa \\t+16(%1), %%xmm5 \n\t" + ".endm \n\t" + ".macro FDCT_ROW_SSE2_H2 i t \n\t" + "movq \\i(%0), %%xmm2 \n\t" + "movq \\i+8(%0), %%xmm0 \n\t" + "movdqa \\t+32(%1), %%xmm3 \n\t" + "movdqa \\t+48(%1), %%xmm7 \n\t" + ".endm \n\t" + ".macro FDCT_ROW_SSE2 i \n\t" + "movq %%xmm2, %%xmm1 \n\t" + "pshuflw $27, %%xmm0, %%xmm0 \n\t" + "paddsw %%xmm0, %%xmm1 \n\t" + "psubsw %%xmm0, %%xmm2 \n\t" + "punpckldq %%xmm2, %%xmm1 \n\t" + "pshufd $78, %%xmm1, %%xmm2 \n\t" + "pmaddwd %%xmm2, %%xmm3 \n\t" + "pmaddwd %%xmm1, %%xmm7 \n\t" + "pmaddwd %%xmm5, %%xmm2 \n\t" + "pmaddwd %%xmm4, %%xmm1 \n\t" + "paddd %%xmm7, %%xmm3 \n\t" + "paddd %%xmm2, %%xmm1 \n\t" + "paddd %%xmm6, %%xmm3 \n\t" + "paddd %%xmm6, %%xmm1 \n\t" + "psrad %3, %%xmm3 \n\t" + "psrad %3, %%xmm1 \n\t" + "packssdw %%xmm3, %%xmm1 \n\t" + "movdqa %%xmm1, \\i(%4) \n\t" + ".endm \n\t" + "movdqa (%2), %%xmm6 \n\t" + "FDCT_ROW_SSE2_H1 0 0 \n\t" + "FDCT_ROW_SSE2 0 \n\t" + "FDCT_ROW_SSE2_H2 64 0 \n\t" + "FDCT_ROW_SSE2 64 \n\t" + + "FDCT_ROW_SSE2_H1 16 64 \n\t" + "FDCT_ROW_SSE2 16 \n\t" + "FDCT_ROW_SSE2_H2 112 64 \n\t" + "FDCT_ROW_SSE2 112 \n\t" + + "FDCT_ROW_SSE2_H1 32 128 \n\t" + "FDCT_ROW_SSE2 32 \n\t" + "FDCT_ROW_SSE2_H2 96 128 \n\t" + "FDCT_ROW_SSE2 96 \n\t" + + "FDCT_ROW_SSE2_H1 48 192 \n\t" + "FDCT_ROW_SSE2 48 \n\t" + "FDCT_ROW_SSE2_H2 80 192 \n\t" + "FDCT_ROW_SSE2 80 \n\t" + : + : "r" (in), "r" (tab_frw_01234567_sse2.tab_frw_01234567_sse2), "r" (fdct_r_row_sse2.fdct_r_row_sse2), "i" (SHIFT_FRW_ROW), "r" (out) ); } static always_inline void fdct_row_mmx2(const int16_t *in, int16_t *out, const int16_t *table) -{ +{ pshufw_m2r(*(in + 4), mm5, 0x1B); movq_m2r(*(in + 0), mm0); movq_r2r(mm0, mm1); @@ -454,7 +454,7 @@ static always_inline void fdct_row_mmx2(const int16_t *in, int16_t *out, const i } static always_inline void fdct_row_mmx(const int16_t *in, int16_t *out, const int16_t *table) -{ +{ //FIXME reorder (i dont have a old mmx only cpu here to benchmark ...) movd_m2r(*(in + 6), mm1); punpcklwd_m2r(*(in + 4), mm1); @@ -547,7 +547,7 @@ void ff_fdct_mmx2(int16_t *block) } } -void ff_fdct_sse2(int16_t *block) +void ff_fdct_sse2(int16_t *block) { int64_t align_tmp[16] ATTR_ALIGN(8); int16_t * const block_tmp= (int16_t*)align_tmp; diff --git a/src/libffmpeg/libavcodec/i386/fft_sse.c b/src/libffmpeg/libavcodec/i386/fft_sse.c index d07c943e9..54851fb94 100644 --- a/src/libffmpeg/libavcodec/i386/fft_sse.c +++ b/src/libffmpeg/libavcodec/i386/fft_sse.c @@ -14,7 +14,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include "../dsputil.h" #include <math.h> @@ -23,13 +23,13 @@ #include <xmmintrin.h> -static const float p1p1p1m1[4] __attribute__((aligned(16))) = +static const float p1p1p1m1[4] __attribute__((aligned(16))) = { 1.0, 1.0, 1.0, -1.0 }; -static const float p1p1m1p1[4] __attribute__((aligned(16))) = +static const float p1p1m1p1[4] __attribute__((aligned(16))) = { 1.0, 1.0, -1.0, 1.0 }; -static const float p1p1m1m1[4] __attribute__((aligned(16))) = +static const float p1p1m1m1[4] __attribute__((aligned(16))) = { 1.0, 1.0, -1.0, -1.0 }; #if 0 @@ -45,8 +45,8 @@ static void print_v4sf(const char *str, __m128 a) void ff_fft_calc_sse(FFTContext *s, FFTComplex *z) { int ln = s->nbits; - int j, np, np2; - int nblocks, nloops; + int j, np, np2; + int nblocks, nloops; register FFTComplex *p, *q; FFTComplex *cptr, *cptr1; int k; @@ -107,27 +107,27 @@ void ff_fft_calc_sse(FFTContext *s, FFTComplex *z) a = *(__m128 *)p; b = *(__m128 *)q; - + /* complex mul */ c = *(__m128 *)cptr; /* cre*re cim*re */ - t1 = _mm_mul_ps(c, - _mm_shuffle_ps(b, b, _MM_SHUFFLE(2, 2, 0, 0))); + t1 = _mm_mul_ps(c, + _mm_shuffle_ps(b, b, _MM_SHUFFLE(2, 2, 0, 0))); c = *(__m128 *)(cptr + 2); /* -cim*im cre*im */ t2 = _mm_mul_ps(c, - _mm_shuffle_ps(b, b, _MM_SHUFFLE(3, 3, 1, 1))); + _mm_shuffle_ps(b, b, _MM_SHUFFLE(3, 3, 1, 1))); b = _mm_add_ps(t1, t2); - + /* butterfly */ *(__m128 *)p = _mm_add_ps(a, b); *(__m128 *)q = _mm_sub_ps(a, b); - + p += 2; q += 2; cptr += 4; } while (--k); - + p += nloops; q += nloops; } while (--j); diff --git a/src/libffmpeg/libavcodec/i386/h264dsp_mmx.c b/src/libffmpeg/libavcodec/i386/h264dsp_mmx.c index c278affc8..8ab58f389 100644 --- a/src/libffmpeg/libavcodec/i386/h264dsp_mmx.c +++ b/src/libffmpeg/libavcodec/i386/h264dsp_mmx.c @@ -13,7 +13,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -47,9 +47,9 @@ SUMSUB_BADC( d13, s02, s13, d02 ) #define SBUTTERFLY(a,b,t,n)\ - "movq " #a ", " #t " \n\t" /* abcd */\ - "punpckl" #n " " #b ", " #a " \n\t" /* aebf */\ - "punpckh" #n " " #b ", " #t " \n\t" /* cgdh */\ + "movq " #a ", " #t " \n\t" /* abcd */\ + "punpckl" #n " " #b ", " #a " \n\t" /* aebf */\ + "punpckh" #n " " #b ", " #t " \n\t" /* cgdh */\ #define TRANSPOSE4(a,b,c,d,t)\ SBUTTERFLY(a,b,t,wd) /* a=aebf t=cgdh */\ @@ -369,73 +369,73 @@ static void h264_h_loop_filter_chroma_intra_mmx2(uint8_t *pix, int stride, int a /* motion compensation */ #define QPEL_H264V(A,B,C,D,E,F,OP)\ - "movd (%0), "#F" \n\t"\ - "movq "#C", %%mm6 \n\t"\ - "paddw "#D", %%mm6 \n\t"\ - "psllw $2, %%mm6 \n\t"\ - "psubw "#B", %%mm6 \n\t"\ - "psubw "#E", %%mm6 \n\t"\ - "pmullw %4, %%mm6 \n\t"\ - "add %2, %0 \n\t"\ - "punpcklbw %%mm7, "#F" \n\t"\ - "paddw %5, "#A" \n\t"\ - "paddw "#F", "#A" \n\t"\ - "paddw "#A", %%mm6 \n\t"\ - "psraw $5, %%mm6 \n\t"\ - "packuswb %%mm6, %%mm6 \n\t"\ + "movd (%0), "#F" \n\t"\ + "movq "#C", %%mm6 \n\t"\ + "paddw "#D", %%mm6 \n\t"\ + "psllw $2, %%mm6 \n\t"\ + "psubw "#B", %%mm6 \n\t"\ + "psubw "#E", %%mm6 \n\t"\ + "pmullw %4, %%mm6 \n\t"\ + "add %2, %0 \n\t"\ + "punpcklbw %%mm7, "#F" \n\t"\ + "paddw %5, "#A" \n\t"\ + "paddw "#F", "#A" \n\t"\ + "paddw "#A", %%mm6 \n\t"\ + "psraw $5, %%mm6 \n\t"\ + "packuswb %%mm6, %%mm6 \n\t"\ OP(%%mm6, (%1), A, d)\ - "add %3, %1 \n\t" + "add %3, %1 \n\t" #define QPEL_H264HV(A,B,C,D,E,F,OF)\ - "movd (%0), "#F" \n\t"\ - "movq "#C", %%mm6 \n\t"\ - "paddw "#D", %%mm6 \n\t"\ - "psllw $2, %%mm6 \n\t"\ - "psubw "#B", %%mm6 \n\t"\ - "psubw "#E", %%mm6 \n\t"\ - "pmullw %3, %%mm6 \n\t"\ - "add %2, %0 \n\t"\ - "punpcklbw %%mm7, "#F" \n\t"\ - "paddw "#F", "#A" \n\t"\ - "paddw "#A", %%mm6 \n\t"\ - "movq %%mm6, "#OF"(%1) \n\t" - + "movd (%0), "#F" \n\t"\ + "movq "#C", %%mm6 \n\t"\ + "paddw "#D", %%mm6 \n\t"\ + "psllw $2, %%mm6 \n\t"\ + "psubw "#B", %%mm6 \n\t"\ + "psubw "#E", %%mm6 \n\t"\ + "pmullw %3, %%mm6 \n\t"\ + "add %2, %0 \n\t"\ + "punpcklbw %%mm7, "#F" \n\t"\ + "paddw "#F", "#A" \n\t"\ + "paddw "#A", %%mm6 \n\t"\ + "movq %%mm6, "#OF"(%1) \n\t" + #define QPEL_H264(OPNAME, OP, MMX)\ static void OPNAME ## h264_qpel4_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ int h=4;\ \ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "movq %5, %%mm4 \n\t"\ - "movq %6, %%mm5 \n\t"\ - "1: \n\t"\ - "movd -1(%0), %%mm1 \n\t"\ - "movd (%0), %%mm2 \n\t"\ - "movd 1(%0), %%mm3 \n\t"\ - "movd 2(%0), %%mm0 \n\t"\ - "punpcklbw %%mm7, %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpcklbw %%mm7, %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "paddw %%mm0, %%mm1 \n\t"\ - "paddw %%mm3, %%mm2 \n\t"\ - "movd -2(%0), %%mm0 \n\t"\ - "movd 3(%0), %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpcklbw %%mm7, %%mm3 \n\t"\ - "paddw %%mm3, %%mm0 \n\t"\ - "psllw $2, %%mm2 \n\t"\ - "psubw %%mm1, %%mm2 \n\t"\ - "pmullw %%mm4, %%mm2 \n\t"\ - "paddw %%mm5, %%mm0 \n\t"\ - "paddw %%mm2, %%mm0 \n\t"\ - "psraw $5, %%mm0 \n\t"\ - "packuswb %%mm0, %%mm0 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "movq %5, %%mm4 \n\t"\ + "movq %6, %%mm5 \n\t"\ + "1: \n\t"\ + "movd -1(%0), %%mm1 \n\t"\ + "movd (%0), %%mm2 \n\t"\ + "movd 1(%0), %%mm3 \n\t"\ + "movd 2(%0), %%mm0 \n\t"\ + "punpcklbw %%mm7, %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpcklbw %%mm7, %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "paddw %%mm0, %%mm1 \n\t"\ + "paddw %%mm3, %%mm2 \n\t"\ + "movd -2(%0), %%mm0 \n\t"\ + "movd 3(%0), %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpcklbw %%mm7, %%mm3 \n\t"\ + "paddw %%mm3, %%mm0 \n\t"\ + "psllw $2, %%mm2 \n\t"\ + "psubw %%mm1, %%mm2 \n\t"\ + "pmullw %%mm4, %%mm2 \n\t"\ + "paddw %%mm5, %%mm0 \n\t"\ + "paddw %%mm2, %%mm0 \n\t"\ + "psraw $5, %%mm0 \n\t"\ + "packuswb %%mm0, %%mm0 \n\t"\ OP(%%mm0, (%1),%%mm6, d)\ - "add %3, %0 \n\t"\ - "add %4, %1 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "add %3, %0 \n\t"\ + "add %4, %1 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ : "+a"(src), "+c"(dst), "+m"(h)\ : "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\ : "memory"\ @@ -444,22 +444,22 @@ static void OPNAME ## h264_qpel4_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i static void OPNAME ## h264_qpel4_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ src -= 2*srcStride;\ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "movd (%0), %%mm0 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm1 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm2 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm3 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm4 \n\t"\ - "add %2, %0 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpcklbw %%mm7, %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpcklbw %%mm7, %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm4 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "movd (%0), %%mm0 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm1 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm2 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm3 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm4 \n\t"\ + "add %2, %0 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpcklbw %%mm7, %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpcklbw %%mm7, %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm4 \n\t"\ QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\ QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\ QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\ @@ -476,22 +476,22 @@ static void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, src -= 2*srcStride+2;\ while(w--){\ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "movd (%0), %%mm0 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm1 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm2 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm3 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm4 \n\t"\ - "add %2, %0 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpcklbw %%mm7, %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpcklbw %%mm7, %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm4 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "movd (%0), %%mm0 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm1 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm2 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm3 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm4 \n\t"\ + "add %2, %0 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpcklbw %%mm7, %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpcklbw %%mm7, %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm4 \n\t"\ QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 0*8*3)\ QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 1*8*3)\ QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 2*8*3)\ @@ -506,28 +506,28 @@ static void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, }\ tmp -= 3*4;\ asm volatile(\ - "movq %4, %%mm6 \n\t"\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t"\ - "paddw 10(%0), %%mm0 \n\t"\ - "movq 2(%0), %%mm1 \n\t"\ - "paddw 8(%0), %%mm1 \n\t"\ - "movq 4(%0), %%mm2 \n\t"\ - "paddw 6(%0), %%mm2 \n\t"\ - "psubw %%mm1, %%mm0 \n\t"/*a-b (abccba)*/\ - "psraw $2, %%mm0 \n\t"/*(a-b)/4 */\ - "psubw %%mm1, %%mm0 \n\t"/*(a-b)/4-b */\ - "paddsw %%mm2, %%mm0 \n\t"\ - "psraw $2, %%mm0 \n\t"/*((a-b)/4-b)/4 */\ - "paddw %%mm6, %%mm2 \n\t"\ - "paddw %%mm2, %%mm0 \n\t"\ - "psraw $6, %%mm0 \n\t"\ - "packuswb %%mm0, %%mm0 \n\t"\ + "movq %4, %%mm6 \n\t"\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "paddw 10(%0), %%mm0 \n\t"\ + "movq 2(%0), %%mm1 \n\t"\ + "paddw 8(%0), %%mm1 \n\t"\ + "movq 4(%0), %%mm2 \n\t"\ + "paddw 6(%0), %%mm2 \n\t"\ + "psubw %%mm1, %%mm0 \n\t"/*a-b (abccba)*/\ + "psraw $2, %%mm0 \n\t"/*(a-b)/4 */\ + "psubw %%mm1, %%mm0 \n\t"/*(a-b)/4-b */\ + "paddsw %%mm2, %%mm0 \n\t"\ + "psraw $2, %%mm0 \n\t"/*((a-b)/4-b+c)/4 */\ + "paddw %%mm6, %%mm2 \n\t"\ + "paddw %%mm2, %%mm0 \n\t"/*(a-5*b+20*c)/16 +32 */\ + "psraw $6, %%mm0 \n\t"\ + "packuswb %%mm0, %%mm0 \n\t"\ OP(%%mm0, (%1),%%mm7, d)\ - "add $24, %0 \n\t"\ - "add %3, %1 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "add $24, %0 \n\t"\ + "add %3, %1 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ : "+a"(tmp), "+c"(dst), "+m"(h)\ : "S"((long)dstStride), "m"(ff_pw_32)\ : "memory"\ @@ -537,54 +537,54 @@ static void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, static void OPNAME ## h264_qpel8_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ int h=8;\ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "movq %5, %%mm6 \n\t"\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t"\ - "movq 1(%0), %%mm2 \n\t"\ - "movq %%mm0, %%mm1 \n\t"\ - "movq %%mm2, %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpckhbw %%mm7, %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpckhbw %%mm7, %%mm3 \n\t"\ - "paddw %%mm2, %%mm0 \n\t"\ - "paddw %%mm3, %%mm1 \n\t"\ - "psllw $2, %%mm0 \n\t"\ - "psllw $2, %%mm1 \n\t"\ - "movq -1(%0), %%mm2 \n\t"\ - "movq 2(%0), %%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 %%mm3, %%mm5 \n\t"\ - "psubw %%mm2, %%mm0 \n\t"\ - "psubw %%mm5, %%mm1 \n\t"\ - "pmullw %%mm6, %%mm0 \n\t"\ - "pmullw %%mm6, %%mm1 \n\t"\ - "movd -2(%0), %%mm2 \n\t"\ - "movd 7(%0), %%mm5 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpcklbw %%mm7, %%mm5 \n\t"\ - "paddw %%mm3, %%mm2 \n\t"\ - "paddw %%mm5, %%mm4 \n\t"\ - "movq %6, %%mm5 \n\t"\ - "paddw %%mm5, %%mm2 \n\t"\ - "paddw %%mm5, %%mm4 \n\t"\ - "paddw %%mm2, %%mm0 \n\t"\ - "paddw %%mm4, %%mm1 \n\t"\ - "psraw $5, %%mm0 \n\t"\ - "psraw $5, %%mm1 \n\t"\ - "packuswb %%mm1, %%mm0 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "movq %5, %%mm6 \n\t"\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "movq 1(%0), %%mm2 \n\t"\ + "movq %%mm0, %%mm1 \n\t"\ + "movq %%mm2, %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpckhbw %%mm7, %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpckhbw %%mm7, %%mm3 \n\t"\ + "paddw %%mm2, %%mm0 \n\t"\ + "paddw %%mm3, %%mm1 \n\t"\ + "psllw $2, %%mm0 \n\t"\ + "psllw $2, %%mm1 \n\t"\ + "movq -1(%0), %%mm2 \n\t"\ + "movq 2(%0), %%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 %%mm3, %%mm5 \n\t"\ + "psubw %%mm2, %%mm0 \n\t"\ + "psubw %%mm5, %%mm1 \n\t"\ + "pmullw %%mm6, %%mm0 \n\t"\ + "pmullw %%mm6, %%mm1 \n\t"\ + "movd -2(%0), %%mm2 \n\t"\ + "movd 7(%0), %%mm5 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpcklbw %%mm7, %%mm5 \n\t"\ + "paddw %%mm3, %%mm2 \n\t"\ + "paddw %%mm5, %%mm4 \n\t"\ + "movq %6, %%mm5 \n\t"\ + "paddw %%mm5, %%mm2 \n\t"\ + "paddw %%mm5, %%mm4 \n\t"\ + "paddw %%mm2, %%mm0 \n\t"\ + "paddw %%mm4, %%mm1 \n\t"\ + "psraw $5, %%mm0 \n\t"\ + "psraw $5, %%mm1 \n\t"\ + "packuswb %%mm1, %%mm0 \n\t"\ OP(%%mm0, (%1),%%mm5, q)\ - "add %3, %0 \n\t"\ - "add %4, %1 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "add %3, %0 \n\t"\ + "add %4, %1 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ : "+a"(src), "+c"(dst), "+m"(h)\ : "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\ : "memory"\ @@ -597,22 +597,22 @@ static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, i \ while(h--){\ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "movd (%0), %%mm0 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm1 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm2 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm3 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm4 \n\t"\ - "add %2, %0 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpcklbw %%mm7, %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpcklbw %%mm7, %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm4 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "movd (%0), %%mm0 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm1 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm2 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm3 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm4 \n\t"\ + "add %2, %0 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpcklbw %%mm7, %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpcklbw %%mm7, %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm4 \n\t"\ QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\ QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\ QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\ @@ -636,22 +636,22 @@ static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, src -= 2*srcStride+2;\ while(w--){\ asm volatile(\ - "pxor %%mm7, %%mm7 \n\t"\ - "movd (%0), %%mm0 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm1 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm2 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm3 \n\t"\ - "add %2, %0 \n\t"\ - "movd (%0), %%mm4 \n\t"\ - "add %2, %0 \n\t"\ - "punpcklbw %%mm7, %%mm0 \n\t"\ - "punpcklbw %%mm7, %%mm1 \n\t"\ - "punpcklbw %%mm7, %%mm2 \n\t"\ - "punpcklbw %%mm7, %%mm3 \n\t"\ - "punpcklbw %%mm7, %%mm4 \n\t"\ + "pxor %%mm7, %%mm7 \n\t"\ + "movd (%0), %%mm0 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm1 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm2 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm3 \n\t"\ + "add %2, %0 \n\t"\ + "movd (%0), %%mm4 \n\t"\ + "add %2, %0 \n\t"\ + "punpcklbw %%mm7, %%mm0 \n\t"\ + "punpcklbw %%mm7, %%mm1 \n\t"\ + "punpcklbw %%mm7, %%mm2 \n\t"\ + "punpcklbw %%mm7, %%mm3 \n\t"\ + "punpcklbw %%mm7, %%mm4 \n\t"\ QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 0*8*4)\ QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 1*8*4)\ QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 2*8*4)\ @@ -670,42 +670,42 @@ static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, }\ tmp -= 4*4;\ asm volatile(\ - "movq %4, %%mm6 \n\t"\ - "1: \n\t"\ - "movq (%0), %%mm0 \n\t"\ - "movq 8(%0), %%mm3 \n\t"\ - "movq 2(%0), %%mm1 \n\t"\ - "movq 10(%0), %%mm4 \n\t"\ - "paddw %%mm4, %%mm0 \n\t"\ - "paddw %%mm3, %%mm1 \n\t"\ - "paddw 18(%0), %%mm3 \n\t"\ - "paddw 16(%0), %%mm4 \n\t"\ - "movq 4(%0), %%mm2 \n\t"\ - "movq 12(%0), %%mm5 \n\t"\ - "paddw 6(%0), %%mm2 \n\t"\ - "paddw 14(%0), %%mm5 \n\t"\ - "psubw %%mm1, %%mm0 \n\t"\ - "psubw %%mm4, %%mm3 \n\t"\ - "psraw $2, %%mm0 \n\t"\ - "psraw $2, %%mm3 \n\t"\ - "psubw %%mm1, %%mm0 \n\t"\ - "psubw %%mm4, %%mm3 \n\t"\ - "paddsw %%mm2, %%mm0 \n\t"\ - "paddsw %%mm5, %%mm3 \n\t"\ - "psraw $2, %%mm0 \n\t"\ - "psraw $2, %%mm3 \n\t"\ - "paddw %%mm6, %%mm2 \n\t"\ - "paddw %%mm6, %%mm5 \n\t"\ - "paddw %%mm2, %%mm0 \n\t"\ - "paddw %%mm5, %%mm3 \n\t"\ - "psraw $6, %%mm0 \n\t"\ - "psraw $6, %%mm3 \n\t"\ - "packuswb %%mm3, %%mm0 \n\t"\ + "movq %4, %%mm6 \n\t"\ + "1: \n\t"\ + "movq (%0), %%mm0 \n\t"\ + "movq 8(%0), %%mm3 \n\t"\ + "movq 2(%0), %%mm1 \n\t"\ + "movq 10(%0), %%mm4 \n\t"\ + "paddw %%mm4, %%mm0 \n\t"\ + "paddw %%mm3, %%mm1 \n\t"\ + "paddw 18(%0), %%mm3 \n\t"\ + "paddw 16(%0), %%mm4 \n\t"\ + "movq 4(%0), %%mm2 \n\t"\ + "movq 12(%0), %%mm5 \n\t"\ + "paddw 6(%0), %%mm2 \n\t"\ + "paddw 14(%0), %%mm5 \n\t"\ + "psubw %%mm1, %%mm0 \n\t"\ + "psubw %%mm4, %%mm3 \n\t"\ + "psraw $2, %%mm0 \n\t"\ + "psraw $2, %%mm3 \n\t"\ + "psubw %%mm1, %%mm0 \n\t"\ + "psubw %%mm4, %%mm3 \n\t"\ + "paddsw %%mm2, %%mm0 \n\t"\ + "paddsw %%mm5, %%mm3 \n\t"\ + "psraw $2, %%mm0 \n\t"\ + "psraw $2, %%mm3 \n\t"\ + "paddw %%mm6, %%mm2 \n\t"\ + "paddw %%mm6, %%mm5 \n\t"\ + "paddw %%mm2, %%mm0 \n\t"\ + "paddw %%mm5, %%mm3 \n\t"\ + "psraw $6, %%mm0 \n\t"\ + "psraw $6, %%mm3 \n\t"\ + "packuswb %%mm3, %%mm0 \n\t"\ OP(%%mm0, (%1),%%mm7, q)\ - "add $32, %0 \n\t"\ - "add %3, %1 \n\t"\ - "decl %2 \n\t"\ - " jnz 1b \n\t"\ + "add $32, %0 \n\t"\ + "add %3, %1 \n\t"\ + "decl %2 \n\t"\ + " jnz 1b \n\t"\ : "+a"(tmp), "+c"(dst), "+m"(h)\ : "S"((long)dstStride), "m"(ff_pw_32)\ : "memory"\ @@ -862,15 +862,15 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t * }\ -#define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b " \n\t" +#define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b " \n\t" #define AVG_3DNOW_OP(a,b,temp, size) \ -"mov" #size " " #b ", " #temp " \n\t"\ -"pavgusb " #temp ", " #a " \n\t"\ -"mov" #size " " #a ", " #b " \n\t" +"mov" #size " " #b ", " #temp " \n\t"\ +"pavgusb " #temp ", " #a " \n\t"\ +"mov" #size " " #a ", " #b " \n\t" #define AVG_MMX2_OP(a,b,temp, size) \ -"mov" #size " " #b ", " #temp " \n\t"\ -"pavgb " #temp ", " #a " \n\t"\ -"mov" #size " " #a ", " #b " \n\t" +"mov" #size " " #b ", " #temp " \n\t"\ +"pavgb " #temp ", " #a " \n\t"\ +"mov" #size " " #a ", " #b " \n\t" QPEL_H264(put_, PUT_OP, 3dnow) QPEL_H264(avg_, AVG_3DNOW_OP, 3dnow) @@ -892,22 +892,42 @@ H264_MC(avg_, 16,mmx2) #define H264_CHROMA_OP(S,D) +#define H264_CHROMA_OP4(S,D,T) #define H264_CHROMA_MC8_TMPL put_h264_chroma_mc8_mmx +#define H264_CHROMA_MC4_TMPL put_h264_chroma_mc4_mmx +#define H264_CHROMA_MC8_MV0 put_pixels8_mmx #include "dsputil_h264_template_mmx.c" #undef H264_CHROMA_OP +#undef H264_CHROMA_OP4 #undef H264_CHROMA_MC8_TMPL +#undef H264_CHROMA_MC4_TMPL +#undef H264_CHROMA_MC8_MV0 #define H264_CHROMA_OP(S,D) "pavgb " #S ", " #D " \n\t" +#define H264_CHROMA_OP4(S,D,T) "movd " #S ", " #T " \n\t"\ + "pavgb " #T ", " #D " \n\t" #define H264_CHROMA_MC8_TMPL avg_h264_chroma_mc8_mmx2 +#define H264_CHROMA_MC4_TMPL avg_h264_chroma_mc4_mmx2 +#define H264_CHROMA_MC8_MV0 avg_pixels8_mmx2 #include "dsputil_h264_template_mmx.c" #undef H264_CHROMA_OP +#undef H264_CHROMA_OP4 #undef H264_CHROMA_MC8_TMPL +#undef H264_CHROMA_MC4_TMPL +#undef H264_CHROMA_MC8_MV0 #define H264_CHROMA_OP(S,D) "pavgusb " #S ", " #D " \n\t" +#define H264_CHROMA_OP4(S,D,T) "movd " #S ", " #T " \n\t"\ + "pavgusb " #T ", " #D " \n\t" #define H264_CHROMA_MC8_TMPL avg_h264_chroma_mc8_3dnow +#define H264_CHROMA_MC4_TMPL avg_h264_chroma_mc4_3dnow +#define H264_CHROMA_MC8_MV0 avg_pixels8_3dnow #include "dsputil_h264_template_mmx.c" #undef H264_CHROMA_OP +#undef H264_CHROMA_OP4 #undef H264_CHROMA_MC8_TMPL +#undef H264_CHROMA_MC4_TMPL +#undef H264_CHROMA_MC8_MV0 /***********************************/ /* weighted prediction */ @@ -935,8 +955,8 @@ static inline void ff_h264_weight_WxH_mmx2(uint8_t *dst, int stride, int log2_de "punpcklbw %%mm7, %%mm1 \n\t" "pmullw %%mm4, %%mm0 \n\t" "pmullw %%mm4, %%mm1 \n\t" - "paddw %%mm5, %%mm0 \n\t" - "paddw %%mm5, %%mm1 \n\t" + "paddsw %%mm5, %%mm0 \n\t" + "paddsw %%mm5, %%mm1 \n\t" "psraw %%mm6, %%mm0 \n\t" "psraw %%mm6, %%mm1 \n\t" "packuswb %%mm7, %%mm0 \n\t" @@ -951,10 +971,10 @@ static inline void ff_h264_weight_WxH_mmx2(uint8_t *dst, int stride, int log2_de } } -static inline void ff_h264_biweight_WxH_mmx2(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offsetd, int offsets, int w, int h) +static inline void ff_h264_biweight_WxH_mmx2(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset, int w, int h) { int x, y; - int offset = ((offsets + offsetd + 1) | 1) << log2_denom; + offset = ((offset + 1) | 1) << log2_denom; asm volatile( "movd %0, %%mm3 \n\t" "movd %1, %%mm4 \n\t" @@ -975,8 +995,8 @@ static inline void ff_h264_biweight_WxH_mmx2(uint8_t *dst, uint8_t *src, int str "punpcklbw %%mm7, %%mm1 \n\t" "pmullw %%mm3, %%mm0 \n\t" "pmullw %%mm4, %%mm1 \n\t" - "paddw %%mm5, %%mm0 \n\t" - "paddw %%mm1, %%mm0 \n\t" + "paddsw %%mm1, %%mm0 \n\t" + "paddsw %%mm5, %%mm0 \n\t" "psraw %%mm6, %%mm0 \n\t" "packuswb %%mm0, %%mm0 \n\t" "movd %%mm0, %0 \n\t" @@ -990,8 +1010,8 @@ static inline void ff_h264_biweight_WxH_mmx2(uint8_t *dst, uint8_t *src, int str } #define H264_WEIGHT(W,H) \ -static void ff_h264_biweight_ ## W ## x ## H ## _mmx2(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offsetd, int offsets){ \ - ff_h264_biweight_WxH_mmx2(dst, src, stride, log2_denom, weightd, weights, offsetd, offsets, W, H); \ +static void ff_h264_biweight_ ## W ## x ## H ## _mmx2(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset){ \ + ff_h264_biweight_WxH_mmx2(dst, src, stride, log2_denom, weightd, weights, offset, W, H); \ } \ static void ff_h264_weight_ ## W ## x ## H ## _mmx2(uint8_t *dst, int stride, int log2_denom, int weight, int offset){ \ ff_h264_weight_WxH_mmx2(dst, stride, log2_denom, weight, offset, W, H); \ diff --git a/src/libffmpeg/libavcodec/i386/idct_mmx.c b/src/libffmpeg/libavcodec/i386/idct_mmx.c index d1a84549d..1c8632fb7 100644 --- a/src/libffmpeg/libavcodec/i386/idct_mmx.c +++ b/src/libffmpeg/libavcodec/i386/idct_mmx.c @@ -19,7 +19,7 @@ * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include "common.h" @@ -38,7 +38,7 @@ #if 0 /* C row IDCT - its just here to document the MMXEXT and MMX versions */ static inline void idct_row (int16_t * row, int offset, - int16_t * table, int32_t * rounder) + int16_t * table, int32_t * rounder) { int C1, C2, C3, C4, C5, C6, C7; int a0, a1, a2, a3, b0, b1, b2, b3; @@ -77,241 +77,241 @@ static inline void idct_row (int16_t * row, int offset, /* MMXEXT row IDCT */ -#define mmxext_table(c1,c2,c3,c4,c5,c6,c7) { c4, c2, -c4, -c2, \ - c4, c6, c4, c6, \ - c1, c3, -c1, -c5, \ - c5, c7, c3, -c7, \ - c4, -c6, c4, -c6, \ - -c4, c2, c4, -c2, \ - c5, -c1, c3, -c1, \ - c7, c3, c7, -c5 } +#define mmxext_table(c1,c2,c3,c4,c5,c6,c7) { c4, c2, -c4, -c2, \ + c4, c6, c4, c6, \ + c1, c3, -c1, -c5, \ + c5, c7, c3, -c7, \ + c4, -c6, c4, -c6, \ + -c4, c2, c4, -c2, \ + c5, -c1, c3, -c1, \ + c7, c3, c7, -c5 } static inline void mmxext_row_head (int16_t * row, int offset, const int16_t * table) { - movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 + movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 - movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 - movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 + movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 + movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 - movq_m2r (*table, mm3); // mm3 = -C2 -C4 C2 C4 - movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 + movq_m2r (*table, mm3); // mm3 = -C2 -C4 C2 C4 + movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 - movq_m2r (*(table+4), mm4); // mm4 = C6 C4 C6 C4 - pmaddwd_r2r (mm0, mm3); // mm3 = -C4*x4-C2*x6 C4*x0+C2*x2 + movq_m2r (*(table+4), mm4); // mm4 = C6 C4 C6 C4 + pmaddwd_r2r (mm0, mm3); // mm3 = -C4*x4-C2*x6 C4*x0+C2*x2 - pshufw_r2r (mm2, mm2, 0x4e); // mm2 = x2 x0 x6 x4 + pshufw_r2r (mm2, mm2, 0x4e); // mm2 = x2 x0 x6 x4 } static inline void mmxext_row (const int16_t * table, const int32_t * rounder) { - movq_m2r (*(table+8), mm1); // mm1 = -C5 -C1 C3 C1 - pmaddwd_r2r (mm2, mm4); // mm4 = C4*x0+C6*x2 C4*x4+C6*x6 + movq_m2r (*(table+8), mm1); // mm1 = -C5 -C1 C3 C1 + pmaddwd_r2r (mm2, mm4); // mm4 = C4*x0+C6*x2 C4*x4+C6*x6 - pmaddwd_m2r (*(table+16), mm0); // mm0 = C4*x4-C6*x6 C4*x0-C6*x2 - pshufw_r2r (mm6, mm6, 0x4e); // mm6 = x3 x1 x7 x5 + pmaddwd_m2r (*(table+16), mm0); // mm0 = C4*x4-C6*x6 C4*x0-C6*x2 + pshufw_r2r (mm6, mm6, 0x4e); // mm6 = x3 x1 x7 x5 - movq_m2r (*(table+12), mm7); // mm7 = -C7 C3 C7 C5 - pmaddwd_r2r (mm5, mm1); // mm1 = -C1*x5-C5*x7 C1*x1+C3*x3 + movq_m2r (*(table+12), mm7); // mm7 = -C7 C3 C7 C5 + pmaddwd_r2r (mm5, mm1); // mm1 = -C1*x5-C5*x7 C1*x1+C3*x3 - paddd_m2r (*rounder, mm3); // mm3 += rounder - pmaddwd_r2r (mm6, mm7); // mm7 = C3*x1-C7*x3 C5*x5+C7*x7 + paddd_m2r (*rounder, mm3); // mm3 += rounder + pmaddwd_r2r (mm6, mm7); // mm7 = C3*x1-C7*x3 C5*x5+C7*x7 - pmaddwd_m2r (*(table+20), mm2); // mm2 = C4*x0-C2*x2 -C4*x4+C2*x6 - paddd_r2r (mm4, mm3); // mm3 = a1 a0 + rounder + pmaddwd_m2r (*(table+20), mm2); // mm2 = C4*x0-C2*x2 -C4*x4+C2*x6 + paddd_r2r (mm4, mm3); // mm3 = a1 a0 + rounder - pmaddwd_m2r (*(table+24), mm5); // mm5 = C3*x5-C1*x7 C5*x1-C1*x3 - movq_r2r (mm3, mm4); // mm4 = a1 a0 + rounder + pmaddwd_m2r (*(table+24), mm5); // mm5 = C3*x5-C1*x7 C5*x1-C1*x3 + movq_r2r (mm3, mm4); // mm4 = a1 a0 + rounder - pmaddwd_m2r (*(table+28), mm6); // mm6 = C7*x1-C5*x3 C7*x5+C3*x7 - paddd_r2r (mm7, mm1); // mm1 = b1 b0 + pmaddwd_m2r (*(table+28), mm6); // mm6 = C7*x1-C5*x3 C7*x5+C3*x7 + paddd_r2r (mm7, mm1); // mm1 = b1 b0 - paddd_m2r (*rounder, mm0); // mm0 += rounder - psubd_r2r (mm1, mm3); // mm3 = a1-b1 a0-b0 + rounder + paddd_m2r (*rounder, mm0); // mm0 += rounder + psubd_r2r (mm1, mm3); // mm3 = a1-b1 a0-b0 + rounder - psrad_i2r (ROW_SHIFT, mm3); // mm3 = y6 y7 - paddd_r2r (mm4, mm1); // mm1 = a1+b1 a0+b0 + rounder + psrad_i2r (ROW_SHIFT, mm3); // mm3 = y6 y7 + paddd_r2r (mm4, mm1); // mm1 = a1+b1 a0+b0 + rounder - paddd_r2r (mm2, mm0); // mm0 = a3 a2 + rounder - psrad_i2r (ROW_SHIFT, mm1); // mm1 = y1 y0 + paddd_r2r (mm2, mm0); // mm0 = a3 a2 + rounder + psrad_i2r (ROW_SHIFT, mm1); // mm1 = y1 y0 - paddd_r2r (mm6, mm5); // mm5 = b3 b2 - movq_r2r (mm0, mm4); // mm4 = a3 a2 + rounder + paddd_r2r (mm6, mm5); // mm5 = b3 b2 + movq_r2r (mm0, mm4); // mm4 = a3 a2 + rounder - paddd_r2r (mm5, mm0); // mm0 = a3+b3 a2+b2 + rounder - psubd_r2r (mm5, mm4); // mm4 = a3-b3 a2-b2 + rounder + paddd_r2r (mm5, mm0); // mm0 = a3+b3 a2+b2 + rounder + psubd_r2r (mm5, mm4); // mm4 = a3-b3 a2-b2 + rounder } static inline void mmxext_row_tail (int16_t * row, int store) { - psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 + psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 - psrad_i2r (ROW_SHIFT, mm4); // mm4 = y4 y5 + psrad_i2r (ROW_SHIFT, mm4); // mm4 = y4 y5 - packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 + packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 - packssdw_r2r (mm3, mm4); // mm4 = y6 y7 y4 y5 + packssdw_r2r (mm3, mm4); // mm4 = y6 y7 y4 y5 - movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 - pshufw_r2r (mm4, mm4, 0xb1); // mm4 = y7 y6 y5 y4 + movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 + pshufw_r2r (mm4, mm4, 0xb1); // mm4 = y7 y6 y5 y4 /* slot */ - movq_r2m (mm4, *(row+store+4)); // save y7 y6 y5 y4 + movq_r2m (mm4, *(row+store+4)); // save y7 y6 y5 y4 } static inline void mmxext_row_mid (int16_t * row, int store, - int offset, const int16_t * table) + int offset, const int16_t * table) { - movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 - psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 + movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 + psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 - movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 - psrad_i2r (ROW_SHIFT, mm4); // mm4 = y4 y5 + movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 + psrad_i2r (ROW_SHIFT, mm4); // mm4 = y4 y5 - packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 - movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 + packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 + movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 - packssdw_r2r (mm3, mm4); // mm4 = y6 y7 y4 y5 - movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 + packssdw_r2r (mm3, mm4); // mm4 = y6 y7 y4 y5 + movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 - movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 - pshufw_r2r (mm4, mm4, 0xb1); // mm4 = y7 y6 y5 y4 + movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 + pshufw_r2r (mm4, mm4, 0xb1); // mm4 = y7 y6 y5 y4 - movq_m2r (*table, mm3); // mm3 = -C2 -C4 C2 C4 - movq_r2m (mm4, *(row+store+4)); // save y7 y6 y5 y4 + movq_m2r (*table, mm3); // mm3 = -C2 -C4 C2 C4 + movq_r2m (mm4, *(row+store+4)); // save y7 y6 y5 y4 - pmaddwd_r2r (mm0, mm3); // mm3 = -C4*x4-C2*x6 C4*x0+C2*x2 + pmaddwd_r2r (mm0, mm3); // mm3 = -C4*x4-C2*x6 C4*x0+C2*x2 - movq_m2r (*(table+4), mm4); // mm4 = C6 C4 C6 C4 - pshufw_r2r (mm2, mm2, 0x4e); // mm2 = x2 x0 x6 x4 + movq_m2r (*(table+4), mm4); // mm4 = C6 C4 C6 C4 + pshufw_r2r (mm2, mm2, 0x4e); // mm2 = x2 x0 x6 x4 } /* MMX row IDCT */ -#define mmx_table(c1,c2,c3,c4,c5,c6,c7) { c4, c2, c4, c6, \ - c4, c6, -c4, -c2, \ - c1, c3, c3, -c7, \ - c5, c7, -c1, -c5, \ - c4, -c6, c4, -c2, \ - -c4, c2, c4, -c6, \ - c5, -c1, c7, -c5, \ - c7, c3, c3, -c1 } +#define mmx_table(c1,c2,c3,c4,c5,c6,c7) { c4, c2, c4, c6, \ + c4, c6, -c4, -c2, \ + c1, c3, c3, -c7, \ + c5, c7, -c1, -c5, \ + c4, -c6, c4, -c2, \ + -c4, c2, c4, -c6, \ + c5, -c1, c7, -c5, \ + c7, c3, c3, -c1 } static inline void mmx_row_head (int16_t * row, int offset, const int16_t * table) { - movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 + movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 - movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 - movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 + movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 + movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 - movq_m2r (*table, mm3); // mm3 = C6 C4 C2 C4 - movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 + movq_m2r (*table, mm3); // mm3 = C6 C4 C2 C4 + movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 - punpckldq_r2r (mm0, mm0); // mm0 = x2 x0 x2 x0 + punpckldq_r2r (mm0, mm0); // mm0 = x2 x0 x2 x0 - movq_m2r (*(table+4), mm4); // mm4 = -C2 -C4 C6 C4 - pmaddwd_r2r (mm0, mm3); // mm3 = C4*x0+C6*x2 C4*x0+C2*x2 + movq_m2r (*(table+4), mm4); // mm4 = -C2 -C4 C6 C4 + pmaddwd_r2r (mm0, mm3); // mm3 = C4*x0+C6*x2 C4*x0+C2*x2 - movq_m2r (*(table+8), mm1); // mm1 = -C7 C3 C3 C1 - punpckhdq_r2r (mm2, mm2); // mm2 = x6 x4 x6 x4 + movq_m2r (*(table+8), mm1); // mm1 = -C7 C3 C3 C1 + punpckhdq_r2r (mm2, mm2); // mm2 = x6 x4 x6 x4 } static inline void mmx_row (const int16_t * table, const int32_t * rounder) { - pmaddwd_r2r (mm2, mm4); // mm4 = -C4*x4-C2*x6 C4*x4+C6*x6 - punpckldq_r2r (mm5, mm5); // mm5 = x3 x1 x3 x1 + pmaddwd_r2r (mm2, mm4); // mm4 = -C4*x4-C2*x6 C4*x4+C6*x6 + punpckldq_r2r (mm5, mm5); // mm5 = x3 x1 x3 x1 - pmaddwd_m2r (*(table+16), mm0); // mm0 = C4*x0-C2*x2 C4*x0-C6*x2 - punpckhdq_r2r (mm6, mm6); // mm6 = x7 x5 x7 x5 + pmaddwd_m2r (*(table+16), mm0); // mm0 = C4*x0-C2*x2 C4*x0-C6*x2 + punpckhdq_r2r (mm6, mm6); // mm6 = x7 x5 x7 x5 - movq_m2r (*(table+12), mm7); // mm7 = -C5 -C1 C7 C5 - pmaddwd_r2r (mm5, mm1); // mm1 = C3*x1-C7*x3 C1*x1+C3*x3 + movq_m2r (*(table+12), mm7); // mm7 = -C5 -C1 C7 C5 + pmaddwd_r2r (mm5, mm1); // mm1 = C3*x1-C7*x3 C1*x1+C3*x3 - paddd_m2r (*rounder, mm3); // mm3 += rounder - pmaddwd_r2r (mm6, mm7); // mm7 = -C1*x5-C5*x7 C5*x5+C7*x7 + paddd_m2r (*rounder, mm3); // mm3 += rounder + pmaddwd_r2r (mm6, mm7); // mm7 = -C1*x5-C5*x7 C5*x5+C7*x7 - pmaddwd_m2r (*(table+20), mm2); // mm2 = C4*x4-C6*x6 -C4*x4+C2*x6 - paddd_r2r (mm4, mm3); // mm3 = a1 a0 + rounder + pmaddwd_m2r (*(table+20), mm2); // mm2 = C4*x4-C6*x6 -C4*x4+C2*x6 + paddd_r2r (mm4, mm3); // mm3 = a1 a0 + rounder - pmaddwd_m2r (*(table+24), mm5); // mm5 = C7*x1-C5*x3 C5*x1-C1*x3 - movq_r2r (mm3, mm4); // mm4 = a1 a0 + rounder + pmaddwd_m2r (*(table+24), mm5); // mm5 = C7*x1-C5*x3 C5*x1-C1*x3 + movq_r2r (mm3, mm4); // mm4 = a1 a0 + rounder - pmaddwd_m2r (*(table+28), mm6); // mm6 = C3*x5-C1*x7 C7*x5+C3*x7 - paddd_r2r (mm7, mm1); // mm1 = b1 b0 + pmaddwd_m2r (*(table+28), mm6); // mm6 = C3*x5-C1*x7 C7*x5+C3*x7 + paddd_r2r (mm7, mm1); // mm1 = b1 b0 - paddd_m2r (*rounder, mm0); // mm0 += rounder - psubd_r2r (mm1, mm3); // mm3 = a1-b1 a0-b0 + rounder + paddd_m2r (*rounder, mm0); // mm0 += rounder + psubd_r2r (mm1, mm3); // mm3 = a1-b1 a0-b0 + rounder - psrad_i2r (ROW_SHIFT, mm3); // mm3 = y6 y7 - paddd_r2r (mm4, mm1); // mm1 = a1+b1 a0+b0 + rounder + psrad_i2r (ROW_SHIFT, mm3); // mm3 = y6 y7 + paddd_r2r (mm4, mm1); // mm1 = a1+b1 a0+b0 + rounder - paddd_r2r (mm2, mm0); // mm0 = a3 a2 + rounder - psrad_i2r (ROW_SHIFT, mm1); // mm1 = y1 y0 + paddd_r2r (mm2, mm0); // mm0 = a3 a2 + rounder + psrad_i2r (ROW_SHIFT, mm1); // mm1 = y1 y0 - paddd_r2r (mm6, mm5); // mm5 = b3 b2 - movq_r2r (mm0, mm7); // mm7 = a3 a2 + rounder + paddd_r2r (mm6, mm5); // mm5 = b3 b2 + movq_r2r (mm0, mm7); // mm7 = a3 a2 + rounder - paddd_r2r (mm5, mm0); // mm0 = a3+b3 a2+b2 + rounder - psubd_r2r (mm5, mm7); // mm7 = a3-b3 a2-b2 + rounder + paddd_r2r (mm5, mm0); // mm0 = a3+b3 a2+b2 + rounder + psubd_r2r (mm5, mm7); // mm7 = a3-b3 a2-b2 + rounder } static inline void mmx_row_tail (int16_t * row, int store) { - psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 + psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 - psrad_i2r (ROW_SHIFT, mm7); // mm7 = y4 y5 + psrad_i2r (ROW_SHIFT, mm7); // mm7 = y4 y5 - packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 + packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 - packssdw_r2r (mm3, mm7); // mm7 = y6 y7 y4 y5 + packssdw_r2r (mm3, mm7); // mm7 = y6 y7 y4 y5 - movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 - movq_r2r (mm7, mm4); // mm4 = y6 y7 y4 y5 + movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 + movq_r2r (mm7, mm4); // mm4 = y6 y7 y4 y5 - pslld_i2r (16, mm7); // mm7 = y7 0 y5 0 + pslld_i2r (16, mm7); // mm7 = y7 0 y5 0 - psrld_i2r (16, mm4); // mm4 = 0 y6 0 y4 + psrld_i2r (16, mm4); // mm4 = 0 y6 0 y4 - por_r2r (mm4, mm7); // mm7 = y7 y6 y5 y4 + por_r2r (mm4, mm7); // mm7 = y7 y6 y5 y4 /* slot */ - movq_r2m (mm7, *(row+store+4)); // save y7 y6 y5 y4 + movq_r2m (mm7, *(row+store+4)); // save y7 y6 y5 y4 } static inline void mmx_row_mid (int16_t * row, int store, - int offset, const int16_t * table) + int offset, const int16_t * table) { - movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 - psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 + movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0 + psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2 - movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 - psrad_i2r (ROW_SHIFT, mm7); // mm7 = y4 y5 + movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1 + psrad_i2r (ROW_SHIFT, mm7); // mm7 = y4 y5 - packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 - movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 + packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0 + movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1 - packssdw_r2r (mm3, mm7); // mm7 = y6 y7 y4 y5 - movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 + packssdw_r2r (mm3, mm7); // mm7 = y6 y7 y4 y5 + movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0 - movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 - movq_r2r (mm7, mm1); // mm1 = y6 y7 y4 y5 + movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0 + movq_r2r (mm7, mm1); // mm1 = y6 y7 y4 y5 - punpckldq_r2r (mm0, mm0); // mm0 = x2 x0 x2 x0 - psrld_i2r (16, mm7); // mm7 = 0 y6 0 y4 + punpckldq_r2r (mm0, mm0); // mm0 = x2 x0 x2 x0 + psrld_i2r (16, mm7); // mm7 = 0 y6 0 y4 - movq_m2r (*table, mm3); // mm3 = C6 C4 C2 C4 - pslld_i2r (16, mm1); // mm1 = y7 0 y5 0 + movq_m2r (*table, mm3); // mm3 = C6 C4 C2 C4 + pslld_i2r (16, mm1); // mm1 = y7 0 y5 0 - movq_m2r (*(table+4), mm4); // mm4 = -C2 -C4 C6 C4 - por_r2r (mm1, mm7); // mm7 = y7 y6 y5 y4 + movq_m2r (*(table+4), mm4); // mm4 = -C2 -C4 C6 C4 + por_r2r (mm1, mm7); // mm7 = y7 y6 y5 y4 - movq_m2r (*(table+8), mm1); // mm1 = -C7 C3 C3 C1 - punpckhdq_r2r (mm2, mm2); // mm2 = x6 x4 x6 x4 + movq_m2r (*(table+8), mm1); // mm1 = -C7 C3 C3 C1 + punpckhdq_r2r (mm2, mm2); // mm2 = x6 x4 x6 x4 - movq_r2m (mm7, *(row+store+4)); // save y7 y6 y5 y4 - pmaddwd_r2r (mm0, mm3); // mm3 = C4*x0+C6*x2 C4*x0+C2*x2 + movq_r2m (mm7, *(row+store+4)); // save y7 y6 y5 y4 + pmaddwd_r2r (mm0, mm3); // mm3 = C4*x0+C6*x2 C4*x0+C2*x2 } @@ -403,132 +403,132 @@ static inline void idct_col (int16_t * col, int offset) /* column code adapted from peter gubanov */ /* http://www.elecard.com/peter/idct.shtml */ - movq_m2r (*_T1, mm0); // mm0 = T1 + movq_m2r (*_T1, mm0); // mm0 = T1 - movq_m2r (*(col+offset+1*8), mm1); // mm1 = x1 - movq_r2r (mm0, mm2); // mm2 = T1 + movq_m2r (*(col+offset+1*8), mm1); // mm1 = x1 + movq_r2r (mm0, mm2); // mm2 = T1 - movq_m2r (*(col+offset+7*8), mm4); // mm4 = x7 - pmulhw_r2r (mm1, mm0); // mm0 = T1*x1 + movq_m2r (*(col+offset+7*8), mm4); // mm4 = x7 + pmulhw_r2r (mm1, mm0); // mm0 = T1*x1 - movq_m2r (*_T3, mm5); // mm5 = T3 - pmulhw_r2r (mm4, mm2); // mm2 = T1*x7 + movq_m2r (*_T3, mm5); // mm5 = T3 + pmulhw_r2r (mm4, mm2); // mm2 = T1*x7 - movq_m2r (*(col+offset+5*8), mm6); // mm6 = x5 - movq_r2r (mm5, mm7); // mm7 = T3-1 + movq_m2r (*(col+offset+5*8), mm6); // mm6 = x5 + movq_r2r (mm5, mm7); // mm7 = T3-1 - movq_m2r (*(col+offset+3*8), mm3); // mm3 = x3 - psubsw_r2r (mm4, mm0); // mm0 = v17 + movq_m2r (*(col+offset+3*8), mm3); // mm3 = x3 + psubsw_r2r (mm4, mm0); // mm0 = v17 - movq_m2r (*_T2, mm4); // mm4 = T2 - pmulhw_r2r (mm3, mm5); // mm5 = (T3-1)*x3 + movq_m2r (*_T2, mm4); // mm4 = T2 + pmulhw_r2r (mm3, mm5); // mm5 = (T3-1)*x3 - paddsw_r2r (mm2, mm1); // mm1 = u17 - pmulhw_r2r (mm6, mm7); // mm7 = (T3-1)*x5 + paddsw_r2r (mm2, mm1); // mm1 = u17 + pmulhw_r2r (mm6, mm7); // mm7 = (T3-1)*x5 /* slot */ - movq_r2r (mm4, mm2); // mm2 = T2 - paddsw_r2r (mm3, mm5); // mm5 = T3*x3 + movq_r2r (mm4, mm2); // mm2 = T2 + paddsw_r2r (mm3, mm5); // mm5 = T3*x3 pmulhw_m2r (*(col+offset+2*8), mm4);// mm4 = T2*x2 - paddsw_r2r (mm6, mm7); // mm7 = T3*x5 + paddsw_r2r (mm6, mm7); // mm7 = T3*x5 - psubsw_r2r (mm6, mm5); // mm5 = v35 - paddsw_r2r (mm3, mm7); // mm7 = u35 + psubsw_r2r (mm6, mm5); // mm5 = v35 + paddsw_r2r (mm3, mm7); // mm7 = u35 - movq_m2r (*(col+offset+6*8), mm3); // mm3 = x6 - movq_r2r (mm0, mm6); // mm6 = v17 + movq_m2r (*(col+offset+6*8), mm3); // mm3 = x6 + movq_r2r (mm0, mm6); // mm6 = v17 - pmulhw_r2r (mm3, mm2); // mm2 = T2*x6 - psubsw_r2r (mm5, mm0); // mm0 = b3 + pmulhw_r2r (mm3, mm2); // mm2 = T2*x6 + psubsw_r2r (mm5, mm0); // mm0 = b3 - psubsw_r2r (mm3, mm4); // mm4 = v26 - paddsw_r2r (mm6, mm5); // mm5 = v12 + psubsw_r2r (mm3, mm4); // mm4 = v26 + paddsw_r2r (mm6, mm5); // mm5 = v12 - movq_r2m (mm0, *(col+offset+3*8)); // save b3 in scratch0 - movq_r2r (mm1, mm6); // mm6 = u17 + movq_r2m (mm0, *(col+offset+3*8)); // save b3 in scratch0 + movq_r2r (mm1, mm6); // mm6 = u17 paddsw_m2r (*(col+offset+2*8), mm2);// mm2 = u26 - paddsw_r2r (mm7, mm6); // mm6 = b0 + paddsw_r2r (mm7, mm6); // mm6 = b0 - psubsw_r2r (mm7, mm1); // mm1 = u12 - movq_r2r (mm1, mm7); // mm7 = u12 + psubsw_r2r (mm7, mm1); // mm1 = u12 + movq_r2r (mm1, mm7); // mm7 = u12 - movq_m2r (*(col+offset+0*8), mm3); // mm3 = x0 - paddsw_r2r (mm5, mm1); // mm1 = u12+v12 + movq_m2r (*(col+offset+0*8), mm3); // mm3 = x0 + paddsw_r2r (mm5, mm1); // mm1 = u12+v12 - movq_m2r (*_C4, mm0); // mm0 = C4/2 - psubsw_r2r (mm5, mm7); // mm7 = u12-v12 + movq_m2r (*_C4, mm0); // mm0 = C4/2 + psubsw_r2r (mm5, mm7); // mm7 = u12-v12 - movq_r2m (mm6, *(col+offset+5*8)); // save b0 in scratch1 - pmulhw_r2r (mm0, mm1); // mm1 = b1/2 + movq_r2m (mm6, *(col+offset+5*8)); // save b0 in scratch1 + pmulhw_r2r (mm0, mm1); // mm1 = b1/2 - movq_r2r (mm4, mm6); // mm6 = v26 - pmulhw_r2r (mm0, mm7); // mm7 = b2/2 + movq_r2r (mm4, mm6); // mm6 = v26 + pmulhw_r2r (mm0, mm7); // mm7 = b2/2 - movq_m2r (*(col+offset+4*8), mm5); // mm5 = x4 - movq_r2r (mm3, mm0); // mm0 = x0 + movq_m2r (*(col+offset+4*8), mm5); // mm5 = x4 + movq_r2r (mm3, mm0); // mm0 = x0 - psubsw_r2r (mm5, mm3); // mm3 = v04 - paddsw_r2r (mm5, mm0); // mm0 = u04 + psubsw_r2r (mm5, mm3); // mm3 = v04 + paddsw_r2r (mm5, mm0); // mm0 = u04 - paddsw_r2r (mm3, mm4); // mm4 = a1 - movq_r2r (mm0, mm5); // mm5 = u04 + paddsw_r2r (mm3, mm4); // mm4 = a1 + movq_r2r (mm0, mm5); // mm5 = u04 - psubsw_r2r (mm6, mm3); // mm3 = a2 - paddsw_r2r (mm2, mm5); // mm5 = a0 + psubsw_r2r (mm6, mm3); // mm3 = a2 + paddsw_r2r (mm2, mm5); // mm5 = a0 - paddsw_r2r (mm1, mm1); // mm1 = b1 - psubsw_r2r (mm2, mm0); // mm0 = a3 + paddsw_r2r (mm1, mm1); // mm1 = b1 + psubsw_r2r (mm2, mm0); // mm0 = a3 - paddsw_r2r (mm7, mm7); // mm7 = b2 - movq_r2r (mm3, mm2); // mm2 = a2 + paddsw_r2r (mm7, mm7); // mm7 = b2 + movq_r2r (mm3, mm2); // mm2 = a2 - movq_r2r (mm4, mm6); // mm6 = a1 - paddsw_r2r (mm7, mm3); // mm3 = a2+b2 + movq_r2r (mm4, mm6); // mm6 = a1 + paddsw_r2r (mm7, mm3); // mm3 = a2+b2 - psraw_i2r (COL_SHIFT, mm3); // mm3 = y2 - paddsw_r2r (mm1, mm4); // mm4 = a1+b1 + psraw_i2r (COL_SHIFT, mm3); // mm3 = y2 + paddsw_r2r (mm1, mm4); // mm4 = a1+b1 - psraw_i2r (COL_SHIFT, mm4); // mm4 = y1 - psubsw_r2r (mm1, mm6); // mm6 = a1-b1 + psraw_i2r (COL_SHIFT, mm4); // mm4 = y1 + psubsw_r2r (mm1, mm6); // mm6 = a1-b1 - movq_m2r (*(col+offset+5*8), mm1); // mm1 = b0 - psubsw_r2r (mm7, mm2); // mm2 = a2-b2 + movq_m2r (*(col+offset+5*8), mm1); // mm1 = b0 + psubsw_r2r (mm7, mm2); // mm2 = a2-b2 - psraw_i2r (COL_SHIFT, mm6); // mm6 = y6 - movq_r2r (mm5, mm7); // mm7 = a0 + psraw_i2r (COL_SHIFT, mm6); // mm6 = y6 + movq_r2r (mm5, mm7); // mm7 = a0 - movq_r2m (mm4, *(col+offset+1*8)); // save y1 - psraw_i2r (COL_SHIFT, mm2); // mm2 = y5 + movq_r2m (mm4, *(col+offset+1*8)); // save y1 + psraw_i2r (COL_SHIFT, mm2); // mm2 = y5 - movq_r2m (mm3, *(col+offset+2*8)); // save y2 - paddsw_r2r (mm1, mm5); // mm5 = a0+b0 + movq_r2m (mm3, *(col+offset+2*8)); // save y2 + paddsw_r2r (mm1, mm5); // mm5 = a0+b0 - movq_m2r (*(col+offset+3*8), mm4); // mm4 = b3 - psubsw_r2r (mm1, mm7); // mm7 = a0-b0 + movq_m2r (*(col+offset+3*8), mm4); // mm4 = b3 + psubsw_r2r (mm1, mm7); // mm7 = a0-b0 - psraw_i2r (COL_SHIFT, mm5); // mm5 = y0 - movq_r2r (mm0, mm3); // mm3 = a3 + psraw_i2r (COL_SHIFT, mm5); // mm5 = y0 + movq_r2r (mm0, mm3); // mm3 = a3 - movq_r2m (mm2, *(col+offset+5*8)); // save y5 - psubsw_r2r (mm4, mm3); // mm3 = a3-b3 + movq_r2m (mm2, *(col+offset+5*8)); // save y5 + psubsw_r2r (mm4, mm3); // mm3 = a3-b3 - psraw_i2r (COL_SHIFT, mm7); // mm7 = y7 - paddsw_r2r (mm0, mm4); // mm4 = a3+b3 + psraw_i2r (COL_SHIFT, mm7); // mm7 = y7 + paddsw_r2r (mm0, mm4); // mm4 = a3+b3 - movq_r2m (mm5, *(col+offset+0*8)); // save y0 - psraw_i2r (COL_SHIFT, mm3); // mm3 = y4 + movq_r2m (mm5, *(col+offset+0*8)); // save y0 + psraw_i2r (COL_SHIFT, mm3); // mm3 = y4 - movq_r2m (mm6, *(col+offset+6*8)); // save y6 - psraw_i2r (COL_SHIFT, mm4); // mm4 = y3 + movq_r2m (mm6, *(col+offset+6*8)); // save y6 + psraw_i2r (COL_SHIFT, mm4); // mm4 = y3 - movq_r2m (mm7, *(col+offset+7*8)); // save y7 + movq_r2m (mm7, *(col+offset+7*8)); // save y7 - movq_r2m (mm3, *(col+offset+4*8)); // save y4 + movq_r2m (mm3, *(col+offset+4*8)); // save y4 - movq_r2m (mm4, *(col+offset+3*8)); // save y3 + movq_r2m (mm4, *(col+offset+3*8)); // save y3 #undef T1 #undef T2 @@ -540,61 +540,61 @@ static const int32_t rounder0[] ATTR_ALIGN(8) = rounder ((1 << (COL_SHIFT - 1)) - 0.5); static const int32_t rounder4[] ATTR_ALIGN(8) = rounder (0); static const int32_t rounder1[] ATTR_ALIGN(8) = - rounder (1.25683487303); /* C1*(C1/C4+C1+C7)/2 */ + rounder (1.25683487303); /* C1*(C1/C4+C1+C7)/2 */ static const int32_t rounder7[] ATTR_ALIGN(8) = - rounder (-0.25); /* C1*(C7/C4+C7-C1)/2 */ + rounder (-0.25); /* C1*(C7/C4+C7-C1)/2 */ static const int32_t rounder2[] ATTR_ALIGN(8) = - rounder (0.60355339059); /* C2 * (C6+C2)/2 */ + rounder (0.60355339059); /* C2 * (C6+C2)/2 */ static const int32_t rounder6[] ATTR_ALIGN(8) = - rounder (-0.25); /* C2 * (C6-C2)/2 */ + rounder (-0.25); /* C2 * (C6-C2)/2 */ static const int32_t rounder3[] ATTR_ALIGN(8) = - rounder (0.087788325588); /* C3*(-C3/C4+C3+C5)/2 */ + rounder (0.087788325588); /* C3*(-C3/C4+C3+C5)/2 */ static const int32_t rounder5[] ATTR_ALIGN(8) = - rounder (-0.441341716183); /* C3*(-C5/C4+C5-C3)/2 */ + rounder (-0.441341716183); /* C3*(-C5/C4+C5-C3)/2 */ #undef COL_SHIFT #undef ROW_SHIFT -#define declare_idct(idct,table,idct_row_head,idct_row,idct_row_tail,idct_row_mid) \ -void idct (int16_t * block) \ -{ \ - static const int16_t table04[] ATTR_ALIGN(16) = \ - table (22725, 21407, 19266, 16384, 12873, 8867, 4520); \ - static const int16_t table17[] ATTR_ALIGN(16) = \ - table (31521, 29692, 26722, 22725, 17855, 12299, 6270); \ - static const int16_t table26[] ATTR_ALIGN(16) = \ - table (29692, 27969, 25172, 21407, 16819, 11585, 5906); \ - static const int16_t table35[] ATTR_ALIGN(16) = \ - table (26722, 25172, 22654, 19266, 15137, 10426, 5315); \ - \ - idct_row_head (block, 0*8, table04); \ - idct_row (table04, rounder0); \ - idct_row_mid (block, 0*8, 4*8, table04); \ - idct_row (table04, rounder4); \ - idct_row_mid (block, 4*8, 1*8, table17); \ - idct_row (table17, rounder1); \ - idct_row_mid (block, 1*8, 7*8, table17); \ - idct_row (table17, rounder7); \ - idct_row_mid (block, 7*8, 2*8, table26); \ - idct_row (table26, rounder2); \ - idct_row_mid (block, 2*8, 6*8, table26); \ - idct_row (table26, rounder6); \ - idct_row_mid (block, 6*8, 3*8, table35); \ - idct_row (table35, rounder3); \ - idct_row_mid (block, 3*8, 5*8, table35); \ - idct_row (table35, rounder5); \ - idct_row_tail (block, 5*8); \ - \ - idct_col (block, 0); \ - idct_col (block, 4); \ +#define declare_idct(idct,table,idct_row_head,idct_row,idct_row_tail,idct_row_mid) \ +void idct (int16_t * block) \ +{ \ + static const int16_t table04[] ATTR_ALIGN(16) = \ + table (22725, 21407, 19266, 16384, 12873, 8867, 4520); \ + static const int16_t table17[] ATTR_ALIGN(16) = \ + table (31521, 29692, 26722, 22725, 17855, 12299, 6270); \ + static const int16_t table26[] ATTR_ALIGN(16) = \ + table (29692, 27969, 25172, 21407, 16819, 11585, 5906); \ + static const int16_t table35[] ATTR_ALIGN(16) = \ + table (26722, 25172, 22654, 19266, 15137, 10426, 5315); \ + \ + idct_row_head (block, 0*8, table04); \ + idct_row (table04, rounder0); \ + idct_row_mid (block, 0*8, 4*8, table04); \ + idct_row (table04, rounder4); \ + idct_row_mid (block, 4*8, 1*8, table17); \ + idct_row (table17, rounder1); \ + idct_row_mid (block, 1*8, 7*8, table17); \ + idct_row (table17, rounder7); \ + idct_row_mid (block, 7*8, 2*8, table26); \ + idct_row (table26, rounder2); \ + idct_row_mid (block, 2*8, 6*8, table26); \ + idct_row (table26, rounder6); \ + idct_row_mid (block, 6*8, 3*8, table35); \ + idct_row (table35, rounder3); \ + idct_row_mid (block, 3*8, 5*8, table35); \ + idct_row (table35, rounder5); \ + idct_row_tail (block, 5*8); \ + \ + idct_col (block, 0); \ + idct_col (block, 4); \ } void ff_mmx_idct(DCTELEM *block); void ff_mmxext_idct(DCTELEM *block); declare_idct (ff_mmxext_idct, mmxext_table, - mmxext_row_head, mmxext_row, mmxext_row_tail, mmxext_row_mid) + mmxext_row_head, mmxext_row, mmxext_row_tail, mmxext_row_mid) declare_idct (ff_mmx_idct, mmx_table, - mmx_row_head, mmx_row, mmx_row_tail, mmx_row_mid) + mmx_row_head, mmx_row, mmx_row_tail, mmx_row_mid) diff --git a/src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c b/src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c index aff57e3fa..7bc6f5f78 100644 --- a/src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c +++ b/src/libffmpeg/libavcodec/i386/idct_mmx_xvid.c @@ -16,10 +16,11 @@ // * GNU General Public License for more details. // * // * You should have received a copy of the GNU General Public License -// * along with this program; if not, write to the Free Software -// * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA +// * along with this program; if not, write to the Free Software Foundation, +// * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA +// // * -// * $Id: idct_mmx_xvid.c,v 1.1 2005/10/23 02:11:44 miguelfreitas Exp $ +// * $Id: idct_mmx_xvid.c,v 1.2 2006/02/05 14:11:36 miguelfreitas Exp $ // * // ***************************************************************************/ @@ -72,13 +73,13 @@ //----------------------------------------------------------------------------- -static const int16_t tg_1_16[4*4] attribute_used __attribute__ ((aligned(8))) = { +static const int16_t tg_1_16[4*4] attribute_used __attribute__ ((aligned(8))) = { 13036,13036,13036,13036, // tg * (2<<16) + 0.5 27146,27146,27146,27146, // tg * (2<<16) + 0.5 -21746,-21746,-21746,-21746, // tg * (2<<16) + 0.5 23170,23170,23170,23170}; // cos * (2<<15) + 0.5 -static const int32_t rounder_0[2*8] attribute_used __attribute__ ((aligned(8))) = { +static const int32_t rounder_0[2*8] attribute_used __attribute__ ((aligned(8))) = { 65536,65536, 3597,3597, 2260,2260, @@ -148,7 +149,7 @@ static const int32_t rounder_0[2*8] attribute_used __attribute__ ((aligned(8))) //----------------------------------------------------------------------------- // Table for rows 0,4 - constants are multiplied by cos_4_16 -static const int16_t tab_i_04_mmx[32*4] attribute_used __attribute__ ((aligned(8))) = { +static const int16_t tab_i_04_mmx[32*4] attribute_used __attribute__ ((aligned(8))) = { 16384,16384,16384,-16384, // movq-> w06 w04 w02 w00 21407,8867,8867,-21407, // w07 w05 w03 w01 16384,-16384,16384,16384, // w14 w12 w10 w08 @@ -190,7 +191,7 @@ static const int16_t tab_i_04_mmx[32*4] attribute_used __attribute__ ((aligned(8 //----------------------------------------------------------------------------- // %3 for rows 0,4 - constants are multiplied by cos_4_16 -static const int16_t tab_i_04_xmm[32*4] attribute_used __attribute__ ((aligned(8))) = { +static const int16_t tab_i_04_xmm[32*4] attribute_used __attribute__ ((aligned(8))) = { 16384,21407,16384,8867, // movq-> w05 w04 w01 w00 16384,8867,-16384,-21407, // w07 w06 w03 w02 16384,-8867,16384,-21407, // w13 w12 w09 w08 @@ -501,7 +502,7 @@ asm volatile( DCT_8_INV_ROW_MMX(5*16(%0), 5*16(%0), 64*3(%2), 8*5(%1)) DCT_8_INV_ROW_MMX(6*16(%0), 6*16(%0), 64*2(%2), 8*6(%1)) DCT_8_INV_ROW_MMX(7*16(%0), 7*16(%0), 64*1(%2), 8*7(%1)) - + //# Process the columns (4 at a time) DCT_8_INV_COL(0(%0), 0(%0)) DCT_8_INV_COL(8(%0), 8(%0)) @@ -524,7 +525,7 @@ asm volatile( DCT_8_INV_ROW_XMM(5*16(%0), 5*16(%0), 64*3(%2), 8*5(%1)) DCT_8_INV_ROW_XMM(6*16(%0), 6*16(%0), 64*2(%2), 8*6(%1)) DCT_8_INV_ROW_XMM(7*16(%0), 7*16(%0), 64*1(%2), 8*7(%1)) - + //# Process the columns (4 at a time) DCT_8_INV_COL(0(%0), 0(%0)) DCT_8_INV_COL(8(%0), 8(%0)) diff --git a/src/libffmpeg/libavcodec/i386/mmx.h b/src/libffmpeg/libavcodec/i386/mmx.h index f0ef1b79e..df4620e0a 100644 --- a/src/libffmpeg/libavcodec/i386/mmx.h +++ b/src/libffmpeg/libavcodec/i386/mmx.h @@ -7,8 +7,18 @@ #ifdef ARCH_X86_64 # define REG_a "rax" +# define REG_b "rbx" +# define REG_c "rcx" +# define REG_d "rdx" +# define REG_D "rdi" +# define REG_S "rsi" #else # define REG_a "eax" +# define REG_b "ebx" +# define REG_c "ecx" +# define REG_d "edx" +# define REG_D "edi" +# define REG_S "esi" #endif /* @@ -17,257 +27,257 @@ * values by ULL, lest they be truncated by the compiler) */ -typedef union { - long long q; /* Quadword (64-bit) value */ - unsigned long long uq; /* Unsigned Quadword */ - int d[2]; /* 2 Doubleword (32-bit) values */ - unsigned int ud[2]; /* 2 Unsigned Doubleword */ - short w[4]; /* 4 Word (16-bit) values */ - unsigned short uw[4]; /* 4 Unsigned Word */ - char b[8]; /* 8 Byte (8-bit) values */ - unsigned char ub[8]; /* 8 Unsigned Byte */ - float s[2]; /* Single-precision (32-bit) value */ -} mmx_t; /* On an 8-byte (64-bit) boundary */ - - -#define mmx_i2r(op,imm,reg) \ - __asm__ __volatile__ (#op " %0, %%" #reg \ - : /* nothing */ \ - : "i" (imm) ) - -#define mmx_m2r(op,mem,reg) \ - __asm__ __volatile__ (#op " %0, %%" #reg \ - : /* nothing */ \ - : "m" (mem)) - -#define mmx_r2m(op,reg,mem) \ - __asm__ __volatile__ (#op " %%" #reg ", %0" \ - : "=m" (mem) \ - : /* nothing */ ) - -#define mmx_r2r(op,regs,regd) \ - __asm__ __volatile__ (#op " %" #regs ", %" #regd) - - -#define emms() __asm__ __volatile__ ("emms") - -#define movd_m2r(var,reg) mmx_m2r (movd, var, reg) -#define movd_r2m(reg,var) mmx_r2m (movd, reg, var) -#define movd_r2r(regs,regd) mmx_r2r (movd, regs, regd) - -#define movq_m2r(var,reg) mmx_m2r (movq, var, reg) -#define movq_r2m(reg,var) mmx_r2m (movq, reg, var) -#define movq_r2r(regs,regd) mmx_r2r (movq, regs, regd) - -#define packssdw_m2r(var,reg) mmx_m2r (packssdw, var, reg) -#define packssdw_r2r(regs,regd) mmx_r2r (packssdw, regs, regd) -#define packsswb_m2r(var,reg) mmx_m2r (packsswb, var, reg) -#define packsswb_r2r(regs,regd) mmx_r2r (packsswb, regs, regd) - -#define packuswb_m2r(var,reg) mmx_m2r (packuswb, var, reg) -#define packuswb_r2r(regs,regd) mmx_r2r (packuswb, regs, regd) - -#define paddb_m2r(var,reg) mmx_m2r (paddb, var, reg) -#define paddb_r2r(regs,regd) mmx_r2r (paddb, regs, regd) -#define paddd_m2r(var,reg) mmx_m2r (paddd, var, reg) -#define paddd_r2r(regs,regd) mmx_r2r (paddd, regs, regd) -#define paddw_m2r(var,reg) mmx_m2r (paddw, var, reg) -#define paddw_r2r(regs,regd) mmx_r2r (paddw, regs, regd) - -#define paddsb_m2r(var,reg) mmx_m2r (paddsb, var, reg) -#define paddsb_r2r(regs,regd) mmx_r2r (paddsb, regs, regd) -#define paddsw_m2r(var,reg) mmx_m2r (paddsw, var, reg) -#define paddsw_r2r(regs,regd) mmx_r2r (paddsw, regs, regd) - -#define paddusb_m2r(var,reg) mmx_m2r (paddusb, var, reg) -#define paddusb_r2r(regs,regd) mmx_r2r (paddusb, regs, regd) -#define paddusw_m2r(var,reg) mmx_m2r (paddusw, var, reg) -#define paddusw_r2r(regs,regd) mmx_r2r (paddusw, regs, regd) - -#define pand_m2r(var,reg) mmx_m2r (pand, var, reg) -#define pand_r2r(regs,regd) mmx_r2r (pand, regs, regd) - -#define pandn_m2r(var,reg) mmx_m2r (pandn, var, reg) -#define pandn_r2r(regs,regd) mmx_r2r (pandn, regs, regd) - -#define pcmpeqb_m2r(var,reg) mmx_m2r (pcmpeqb, var, reg) -#define pcmpeqb_r2r(regs,regd) mmx_r2r (pcmpeqb, regs, regd) -#define pcmpeqd_m2r(var,reg) mmx_m2r (pcmpeqd, var, reg) -#define pcmpeqd_r2r(regs,regd) mmx_r2r (pcmpeqd, regs, regd) -#define pcmpeqw_m2r(var,reg) mmx_m2r (pcmpeqw, var, reg) -#define pcmpeqw_r2r(regs,regd) mmx_r2r (pcmpeqw, regs, regd) - -#define pcmpgtb_m2r(var,reg) mmx_m2r (pcmpgtb, var, reg) -#define pcmpgtb_r2r(regs,regd) mmx_r2r (pcmpgtb, regs, regd) -#define pcmpgtd_m2r(var,reg) mmx_m2r (pcmpgtd, var, reg) -#define pcmpgtd_r2r(regs,regd) mmx_r2r (pcmpgtd, regs, regd) -#define pcmpgtw_m2r(var,reg) mmx_m2r (pcmpgtw, var, reg) -#define pcmpgtw_r2r(regs,regd) mmx_r2r (pcmpgtw, regs, regd) - -#define pmaddwd_m2r(var,reg) mmx_m2r (pmaddwd, var, reg) -#define pmaddwd_r2r(regs,regd) mmx_r2r (pmaddwd, regs, regd) - -#define pmulhw_m2r(var,reg) mmx_m2r (pmulhw, var, reg) -#define pmulhw_r2r(regs,regd) mmx_r2r (pmulhw, regs, regd) - -#define pmullw_m2r(var,reg) mmx_m2r (pmullw, var, reg) -#define pmullw_r2r(regs,regd) mmx_r2r (pmullw, regs, regd) - -#define por_m2r(var,reg) mmx_m2r (por, var, reg) -#define por_r2r(regs,regd) mmx_r2r (por, regs, regd) - -#define pslld_i2r(imm,reg) mmx_i2r (pslld, imm, reg) -#define pslld_m2r(var,reg) mmx_m2r (pslld, var, reg) -#define pslld_r2r(regs,regd) mmx_r2r (pslld, regs, regd) -#define psllq_i2r(imm,reg) mmx_i2r (psllq, imm, reg) -#define psllq_m2r(var,reg) mmx_m2r (psllq, var, reg) -#define psllq_r2r(regs,regd) mmx_r2r (psllq, regs, regd) -#define psllw_i2r(imm,reg) mmx_i2r (psllw, imm, reg) -#define psllw_m2r(var,reg) mmx_m2r (psllw, var, reg) -#define psllw_r2r(regs,regd) mmx_r2r (psllw, regs, regd) - -#define psrad_i2r(imm,reg) mmx_i2r (psrad, imm, reg) -#define psrad_m2r(var,reg) mmx_m2r (psrad, var, reg) -#define psrad_r2r(regs,regd) mmx_r2r (psrad, regs, regd) -#define psraw_i2r(imm,reg) mmx_i2r (psraw, imm, reg) -#define psraw_m2r(var,reg) mmx_m2r (psraw, var, reg) -#define psraw_r2r(regs,regd) mmx_r2r (psraw, regs, regd) - -#define psrld_i2r(imm,reg) mmx_i2r (psrld, imm, reg) -#define psrld_m2r(var,reg) mmx_m2r (psrld, var, reg) -#define psrld_r2r(regs,regd) mmx_r2r (psrld, regs, regd) -#define psrlq_i2r(imm,reg) mmx_i2r (psrlq, imm, reg) -#define psrlq_m2r(var,reg) mmx_m2r (psrlq, var, reg) -#define psrlq_r2r(regs,regd) mmx_r2r (psrlq, regs, regd) -#define psrlw_i2r(imm,reg) mmx_i2r (psrlw, imm, reg) -#define psrlw_m2r(var,reg) mmx_m2r (psrlw, var, reg) -#define psrlw_r2r(regs,regd) mmx_r2r (psrlw, regs, regd) - -#define psubb_m2r(var,reg) mmx_m2r (psubb, var, reg) -#define psubb_r2r(regs,regd) mmx_r2r (psubb, regs, regd) -#define psubd_m2r(var,reg) mmx_m2r (psubd, var, reg) -#define psubd_r2r(regs,regd) mmx_r2r (psubd, regs, regd) -#define psubw_m2r(var,reg) mmx_m2r (psubw, var, reg) -#define psubw_r2r(regs,regd) mmx_r2r (psubw, regs, regd) - -#define psubsb_m2r(var,reg) mmx_m2r (psubsb, var, reg) -#define psubsb_r2r(regs,regd) mmx_r2r (psubsb, regs, regd) -#define psubsw_m2r(var,reg) mmx_m2r (psubsw, var, reg) -#define psubsw_r2r(regs,regd) mmx_r2r (psubsw, regs, regd) - -#define psubusb_m2r(var,reg) mmx_m2r (psubusb, var, reg) -#define psubusb_r2r(regs,regd) mmx_r2r (psubusb, regs, regd) -#define psubusw_m2r(var,reg) mmx_m2r (psubusw, var, reg) -#define psubusw_r2r(regs,regd) mmx_r2r (psubusw, regs, regd) - -#define punpckhbw_m2r(var,reg) mmx_m2r (punpckhbw, var, reg) -#define punpckhbw_r2r(regs,regd) mmx_r2r (punpckhbw, regs, regd) -#define punpckhdq_m2r(var,reg) mmx_m2r (punpckhdq, var, reg) -#define punpckhdq_r2r(regs,regd) mmx_r2r (punpckhdq, regs, regd) -#define punpckhwd_m2r(var,reg) mmx_m2r (punpckhwd, var, reg) -#define punpckhwd_r2r(regs,regd) mmx_r2r (punpckhwd, regs, regd) - -#define punpcklbw_m2r(var,reg) mmx_m2r (punpcklbw, var, reg) -#define punpcklbw_r2r(regs,regd) mmx_r2r (punpcklbw, regs, regd) -#define punpckldq_m2r(var,reg) mmx_m2r (punpckldq, var, reg) -#define punpckldq_r2r(regs,regd) mmx_r2r (punpckldq, regs, regd) -#define punpcklwd_m2r(var,reg) mmx_m2r (punpcklwd, var, reg) -#define punpcklwd_r2r(regs,regd) mmx_r2r (punpcklwd, regs, regd) - -#define pxor_m2r(var,reg) mmx_m2r (pxor, var, reg) -#define pxor_r2r(regs,regd) mmx_r2r (pxor, regs, regd) +typedef union { + long long q; /* Quadword (64-bit) value */ + unsigned long long uq; /* Unsigned Quadword */ + int d[2]; /* 2 Doubleword (32-bit) values */ + unsigned int ud[2]; /* 2 Unsigned Doubleword */ + short w[4]; /* 4 Word (16-bit) values */ + unsigned short uw[4]; /* 4 Unsigned Word */ + char b[8]; /* 8 Byte (8-bit) values */ + unsigned char ub[8]; /* 8 Unsigned Byte */ + float s[2]; /* Single-precision (32-bit) value */ +} mmx_t; /* On an 8-byte (64-bit) boundary */ + + +#define mmx_i2r(op,imm,reg) \ + __asm__ __volatile__ (#op " %0, %%" #reg \ + : /* nothing */ \ + : "i" (imm) ) + +#define mmx_m2r(op,mem,reg) \ + __asm__ __volatile__ (#op " %0, %%" #reg \ + : /* nothing */ \ + : "m" (mem)) + +#define mmx_r2m(op,reg,mem) \ + __asm__ __volatile__ (#op " %%" #reg ", %0" \ + : "=m" (mem) \ + : /* nothing */ ) + +#define mmx_r2r(op,regs,regd) \ + __asm__ __volatile__ (#op " %" #regs ", %" #regd) + + +#define emms() __asm__ __volatile__ ("emms") + +#define movd_m2r(var,reg) mmx_m2r (movd, var, reg) +#define movd_r2m(reg,var) mmx_r2m (movd, reg, var) +#define movd_r2r(regs,regd) mmx_r2r (movd, regs, regd) + +#define movq_m2r(var,reg) mmx_m2r (movq, var, reg) +#define movq_r2m(reg,var) mmx_r2m (movq, reg, var) +#define movq_r2r(regs,regd) mmx_r2r (movq, regs, regd) + +#define packssdw_m2r(var,reg) mmx_m2r (packssdw, var, reg) +#define packssdw_r2r(regs,regd) mmx_r2r (packssdw, regs, regd) +#define packsswb_m2r(var,reg) mmx_m2r (packsswb, var, reg) +#define packsswb_r2r(regs,regd) mmx_r2r (packsswb, regs, regd) + +#define packuswb_m2r(var,reg) mmx_m2r (packuswb, var, reg) +#define packuswb_r2r(regs,regd) mmx_r2r (packuswb, regs, regd) + +#define paddb_m2r(var,reg) mmx_m2r (paddb, var, reg) +#define paddb_r2r(regs,regd) mmx_r2r (paddb, regs, regd) +#define paddd_m2r(var,reg) mmx_m2r (paddd, var, reg) +#define paddd_r2r(regs,regd) mmx_r2r (paddd, regs, regd) +#define paddw_m2r(var,reg) mmx_m2r (paddw, var, reg) +#define paddw_r2r(regs,regd) mmx_r2r (paddw, regs, regd) + +#define paddsb_m2r(var,reg) mmx_m2r (paddsb, var, reg) +#define paddsb_r2r(regs,regd) mmx_r2r (paddsb, regs, regd) +#define paddsw_m2r(var,reg) mmx_m2r (paddsw, var, reg) +#define paddsw_r2r(regs,regd) mmx_r2r (paddsw, regs, regd) + +#define paddusb_m2r(var,reg) mmx_m2r (paddusb, var, reg) +#define paddusb_r2r(regs,regd) mmx_r2r (paddusb, regs, regd) +#define paddusw_m2r(var,reg) mmx_m2r (paddusw, var, reg) +#define paddusw_r2r(regs,regd) mmx_r2r (paddusw, regs, regd) + +#define pand_m2r(var,reg) mmx_m2r (pand, var, reg) +#define pand_r2r(regs,regd) mmx_r2r (pand, regs, regd) + +#define pandn_m2r(var,reg) mmx_m2r (pandn, var, reg) +#define pandn_r2r(regs,regd) mmx_r2r (pandn, regs, regd) + +#define pcmpeqb_m2r(var,reg) mmx_m2r (pcmpeqb, var, reg) +#define pcmpeqb_r2r(regs,regd) mmx_r2r (pcmpeqb, regs, regd) +#define pcmpeqd_m2r(var,reg) mmx_m2r (pcmpeqd, var, reg) +#define pcmpeqd_r2r(regs,regd) mmx_r2r (pcmpeqd, regs, regd) +#define pcmpeqw_m2r(var,reg) mmx_m2r (pcmpeqw, var, reg) +#define pcmpeqw_r2r(regs,regd) mmx_r2r (pcmpeqw, regs, regd) + +#define pcmpgtb_m2r(var,reg) mmx_m2r (pcmpgtb, var, reg) +#define pcmpgtb_r2r(regs,regd) mmx_r2r (pcmpgtb, regs, regd) +#define pcmpgtd_m2r(var,reg) mmx_m2r (pcmpgtd, var, reg) +#define pcmpgtd_r2r(regs,regd) mmx_r2r (pcmpgtd, regs, regd) +#define pcmpgtw_m2r(var,reg) mmx_m2r (pcmpgtw, var, reg) +#define pcmpgtw_r2r(regs,regd) mmx_r2r (pcmpgtw, regs, regd) + +#define pmaddwd_m2r(var,reg) mmx_m2r (pmaddwd, var, reg) +#define pmaddwd_r2r(regs,regd) mmx_r2r (pmaddwd, regs, regd) + +#define pmulhw_m2r(var,reg) mmx_m2r (pmulhw, var, reg) +#define pmulhw_r2r(regs,regd) mmx_r2r (pmulhw, regs, regd) + +#define pmullw_m2r(var,reg) mmx_m2r (pmullw, var, reg) +#define pmullw_r2r(regs,regd) mmx_r2r (pmullw, regs, regd) + +#define por_m2r(var,reg) mmx_m2r (por, var, reg) +#define por_r2r(regs,regd) mmx_r2r (por, regs, regd) + +#define pslld_i2r(imm,reg) mmx_i2r (pslld, imm, reg) +#define pslld_m2r(var,reg) mmx_m2r (pslld, var, reg) +#define pslld_r2r(regs,regd) mmx_r2r (pslld, regs, regd) +#define psllq_i2r(imm,reg) mmx_i2r (psllq, imm, reg) +#define psllq_m2r(var,reg) mmx_m2r (psllq, var, reg) +#define psllq_r2r(regs,regd) mmx_r2r (psllq, regs, regd) +#define psllw_i2r(imm,reg) mmx_i2r (psllw, imm, reg) +#define psllw_m2r(var,reg) mmx_m2r (psllw, var, reg) +#define psllw_r2r(regs,regd) mmx_r2r (psllw, regs, regd) + +#define psrad_i2r(imm,reg) mmx_i2r (psrad, imm, reg) +#define psrad_m2r(var,reg) mmx_m2r (psrad, var, reg) +#define psrad_r2r(regs,regd) mmx_r2r (psrad, regs, regd) +#define psraw_i2r(imm,reg) mmx_i2r (psraw, imm, reg) +#define psraw_m2r(var,reg) mmx_m2r (psraw, var, reg) +#define psraw_r2r(regs,regd) mmx_r2r (psraw, regs, regd) + +#define psrld_i2r(imm,reg) mmx_i2r (psrld, imm, reg) +#define psrld_m2r(var,reg) mmx_m2r (psrld, var, reg) +#define psrld_r2r(regs,regd) mmx_r2r (psrld, regs, regd) +#define psrlq_i2r(imm,reg) mmx_i2r (psrlq, imm, reg) +#define psrlq_m2r(var,reg) mmx_m2r (psrlq, var, reg) +#define psrlq_r2r(regs,regd) mmx_r2r (psrlq, regs, regd) +#define psrlw_i2r(imm,reg) mmx_i2r (psrlw, imm, reg) +#define psrlw_m2r(var,reg) mmx_m2r (psrlw, var, reg) +#define psrlw_r2r(regs,regd) mmx_r2r (psrlw, regs, regd) + +#define psubb_m2r(var,reg) mmx_m2r (psubb, var, reg) +#define psubb_r2r(regs,regd) mmx_r2r (psubb, regs, regd) +#define psubd_m2r(var,reg) mmx_m2r (psubd, var, reg) +#define psubd_r2r(regs,regd) mmx_r2r (psubd, regs, regd) +#define psubw_m2r(var,reg) mmx_m2r (psubw, var, reg) +#define psubw_r2r(regs,regd) mmx_r2r (psubw, regs, regd) + +#define psubsb_m2r(var,reg) mmx_m2r (psubsb, var, reg) +#define psubsb_r2r(regs,regd) mmx_r2r (psubsb, regs, regd) +#define psubsw_m2r(var,reg) mmx_m2r (psubsw, var, reg) +#define psubsw_r2r(regs,regd) mmx_r2r (psubsw, regs, regd) + +#define psubusb_m2r(var,reg) mmx_m2r (psubusb, var, reg) +#define psubusb_r2r(regs,regd) mmx_r2r (psubusb, regs, regd) +#define psubusw_m2r(var,reg) mmx_m2r (psubusw, var, reg) +#define psubusw_r2r(regs,regd) mmx_r2r (psubusw, regs, regd) + +#define punpckhbw_m2r(var,reg) mmx_m2r (punpckhbw, var, reg) +#define punpckhbw_r2r(regs,regd) mmx_r2r (punpckhbw, regs, regd) +#define punpckhdq_m2r(var,reg) mmx_m2r (punpckhdq, var, reg) +#define punpckhdq_r2r(regs,regd) mmx_r2r (punpckhdq, regs, regd) +#define punpckhwd_m2r(var,reg) mmx_m2r (punpckhwd, var, reg) +#define punpckhwd_r2r(regs,regd) mmx_r2r (punpckhwd, regs, regd) + +#define punpcklbw_m2r(var,reg) mmx_m2r (punpcklbw, var, reg) +#define punpcklbw_r2r(regs,regd) mmx_r2r (punpcklbw, regs, regd) +#define punpckldq_m2r(var,reg) mmx_m2r (punpckldq, var, reg) +#define punpckldq_r2r(regs,regd) mmx_r2r (punpckldq, regs, regd) +#define punpcklwd_m2r(var,reg) mmx_m2r (punpcklwd, var, reg) +#define punpcklwd_r2r(regs,regd) mmx_r2r (punpcklwd, regs, regd) + +#define pxor_m2r(var,reg) mmx_m2r (pxor, var, reg) +#define pxor_r2r(regs,regd) mmx_r2r (pxor, regs, regd) /* 3DNOW extensions */ -#define pavgusb_m2r(var,reg) mmx_m2r (pavgusb, var, reg) -#define pavgusb_r2r(regs,regd) mmx_r2r (pavgusb, regs, regd) +#define pavgusb_m2r(var,reg) mmx_m2r (pavgusb, var, reg) +#define pavgusb_r2r(regs,regd) mmx_r2r (pavgusb, regs, regd) /* AMD MMX extensions - also available in intel SSE */ -#define mmx_m2ri(op,mem,reg,imm) \ +#define mmx_m2ri(op,mem,reg,imm) \ __asm__ __volatile__ (#op " %1, %0, %%" #reg \ : /* nothing */ \ : "X" (mem), "X" (imm)) -#define mmx_r2ri(op,regs,regd,imm) \ +#define mmx_r2ri(op,regs,regd,imm) \ __asm__ __volatile__ (#op " %0, %%" #regs ", %%" #regd \ : /* nothing */ \ : "X" (imm) ) -#define mmx_fetch(mem,hint) \ - __asm__ __volatile__ ("prefetch" #hint " %0" \ - : /* nothing */ \ - : "X" (mem)) +#define mmx_fetch(mem,hint) \ + __asm__ __volatile__ ("prefetch" #hint " %0" \ + : /* nothing */ \ + : "X" (mem)) -#define maskmovq(regs,maskreg) mmx_r2ri (maskmovq, regs, maskreg) +#define maskmovq(regs,maskreg) mmx_r2ri (maskmovq, regs, maskreg) -#define movntq_r2m(mmreg,var) mmx_r2m (movntq, mmreg, var) +#define movntq_r2m(mmreg,var) mmx_r2m (movntq, mmreg, var) -#define pavgb_m2r(var,reg) mmx_m2r (pavgb, var, reg) -#define pavgb_r2r(regs,regd) mmx_r2r (pavgb, regs, regd) -#define pavgw_m2r(var,reg) mmx_m2r (pavgw, var, reg) -#define pavgw_r2r(regs,regd) mmx_r2r (pavgw, regs, regd) +#define pavgb_m2r(var,reg) mmx_m2r (pavgb, var, reg) +#define pavgb_r2r(regs,regd) mmx_r2r (pavgb, regs, regd) +#define pavgw_m2r(var,reg) mmx_m2r (pavgw, var, reg) +#define pavgw_r2r(regs,regd) mmx_r2r (pavgw, regs, regd) -#define pextrw_r2r(mmreg,reg,imm) mmx_r2ri (pextrw, mmreg, reg, imm) +#define pextrw_r2r(mmreg,reg,imm) mmx_r2ri (pextrw, mmreg, reg, imm) -#define pinsrw_r2r(reg,mmreg,imm) mmx_r2ri (pinsrw, reg, mmreg, imm) +#define pinsrw_r2r(reg,mmreg,imm) mmx_r2ri (pinsrw, reg, mmreg, imm) -#define pmaxsw_m2r(var,reg) mmx_m2r (pmaxsw, var, reg) -#define pmaxsw_r2r(regs,regd) mmx_r2r (pmaxsw, regs, regd) +#define pmaxsw_m2r(var,reg) mmx_m2r (pmaxsw, var, reg) +#define pmaxsw_r2r(regs,regd) mmx_r2r (pmaxsw, regs, regd) -#define pmaxub_m2r(var,reg) mmx_m2r (pmaxub, var, reg) -#define pmaxub_r2r(regs,regd) mmx_r2r (pmaxub, regs, regd) +#define pmaxub_m2r(var,reg) mmx_m2r (pmaxub, var, reg) +#define pmaxub_r2r(regs,regd) mmx_r2r (pmaxub, regs, regd) -#define pminsw_m2r(var,reg) mmx_m2r (pminsw, var, reg) -#define pminsw_r2r(regs,regd) mmx_r2r (pminsw, regs, regd) +#define pminsw_m2r(var,reg) mmx_m2r (pminsw, var, reg) +#define pminsw_r2r(regs,regd) mmx_r2r (pminsw, regs, regd) -#define pminub_m2r(var,reg) mmx_m2r (pminub, var, reg) -#define pminub_r2r(regs,regd) mmx_r2r (pminub, regs, regd) +#define pminub_m2r(var,reg) mmx_m2r (pminub, var, reg) +#define pminub_r2r(regs,regd) mmx_r2r (pminub, regs, regd) -#define pmovmskb(mmreg,reg) \ - __asm__ __volatile__ ("movmskps %" #mmreg ", %" #reg) +#define pmovmskb(mmreg,reg) \ + __asm__ __volatile__ ("movmskps %" #mmreg ", %" #reg) -#define pmulhuw_m2r(var,reg) mmx_m2r (pmulhuw, var, reg) -#define pmulhuw_r2r(regs,regd) mmx_r2r (pmulhuw, regs, regd) +#define pmulhuw_m2r(var,reg) mmx_m2r (pmulhuw, var, reg) +#define pmulhuw_r2r(regs,regd) mmx_r2r (pmulhuw, regs, regd) -#define prefetcht0(mem) mmx_fetch (mem, t0) -#define prefetcht1(mem) mmx_fetch (mem, t1) -#define prefetcht2(mem) mmx_fetch (mem, t2) -#define prefetchnta(mem) mmx_fetch (mem, nta) +#define prefetcht0(mem) mmx_fetch (mem, t0) +#define prefetcht1(mem) mmx_fetch (mem, t1) +#define prefetcht2(mem) mmx_fetch (mem, t2) +#define prefetchnta(mem) mmx_fetch (mem, nta) -#define psadbw_m2r(var,reg) mmx_m2r (psadbw, var, reg) -#define psadbw_r2r(regs,regd) mmx_r2r (psadbw, regs, regd) +#define psadbw_m2r(var,reg) mmx_m2r (psadbw, var, reg) +#define psadbw_r2r(regs,regd) mmx_r2r (psadbw, regs, regd) -#define pshufw_m2r(var,reg,imm) mmx_m2ri(pshufw, var, reg, imm) -#define pshufw_r2r(regs,regd,imm) mmx_r2ri(pshufw, regs, regd, imm) +#define pshufw_m2r(var,reg,imm) mmx_m2ri(pshufw, var, reg, imm) +#define pshufw_r2r(regs,regd,imm) mmx_r2ri(pshufw, regs, regd, imm) -#define sfence() __asm__ __volatile__ ("sfence\n\t") +#define sfence() __asm__ __volatile__ ("sfence\n\t") /* SSE2 */ -#define pshufhw_m2r(var,reg,imm) mmx_m2ri(pshufhw, var, reg, imm) -#define pshufhw_r2r(regs,regd,imm) mmx_r2ri(pshufhw, regs, regd, imm) -#define pshuflw_m2r(var,reg,imm) mmx_m2ri(pshuflw, var, reg, imm) -#define pshuflw_r2r(regs,regd,imm) mmx_r2ri(pshuflw, regs, regd, imm) +#define pshufhw_m2r(var,reg,imm) mmx_m2ri(pshufhw, var, reg, imm) +#define pshufhw_r2r(regs,regd,imm) mmx_r2ri(pshufhw, regs, regd, imm) +#define pshuflw_m2r(var,reg,imm) mmx_m2ri(pshuflw, var, reg, imm) +#define pshuflw_r2r(regs,regd,imm) mmx_r2ri(pshuflw, regs, regd, imm) -#define pshufd_r2r(regs,regd,imm) mmx_r2ri(pshufd, regs, regd, imm) +#define pshufd_r2r(regs,regd,imm) mmx_r2ri(pshufd, regs, regd, imm) -#define movdqa_m2r(var,reg) mmx_m2r (movdqa, var, reg) -#define movdqa_r2m(reg,var) mmx_r2m (movdqa, reg, var) -#define movdqa_r2r(regs,regd) mmx_r2r (movdqa, regs, regd) -#define movdqu_m2r(var,reg) mmx_m2r (movdqu, var, reg) -#define movdqu_r2m(reg,var) mmx_r2m (movdqu, reg, var) -#define movdqu_r2r(regs,regd) mmx_r2r (movdqu, regs, regd) +#define movdqa_m2r(var,reg) mmx_m2r (movdqa, var, reg) +#define movdqa_r2m(reg,var) mmx_r2m (movdqa, reg, var) +#define movdqa_r2r(regs,regd) mmx_r2r (movdqa, regs, regd) +#define movdqu_m2r(var,reg) mmx_m2r (movdqu, var, reg) +#define movdqu_r2m(reg,var) mmx_r2m (movdqu, reg, var) +#define movdqu_r2r(regs,regd) mmx_r2r (movdqu, regs, regd) -#define pmullw_r2m(reg,var) mmx_r2m (pmullw, reg, var) +#define pmullw_r2m(reg,var) mmx_r2m (pmullw, reg, var) -#define pslldq_i2r(imm,reg) mmx_i2r (pslldq, imm, reg) -#define psrldq_i2r(imm,reg) mmx_i2r (psrldq, imm, reg) +#define pslldq_i2r(imm,reg) mmx_i2r (pslldq, imm, reg) +#define psrldq_i2r(imm,reg) mmx_i2r (psrldq, imm, reg) -#define punpcklqdq_r2r(regs,regd) mmx_r2r (punpcklqdq, regs, regd) -#define punpckhqdq_r2r(regs,regd) mmx_r2r (punpckhqdq, regs, regd) +#define punpcklqdq_r2r(regs,regd) mmx_r2r (punpcklqdq, regs, regd) +#define punpckhqdq_r2r(regs,regd) mmx_r2r (punpckhqdq, regs, regd) #endif /* AVCODEC_I386MMX_H */ diff --git a/src/libffmpeg/libavcodec/i386/motion_est_mmx.c b/src/libffmpeg/libavcodec/i386/motion_est_mmx.c index 1b90f8e40..c14b79384 100644 --- a/src/libffmpeg/libavcodec/i386/motion_est_mmx.c +++ b/src/libffmpeg/libavcodec/i386/motion_est_mmx.c @@ -15,7 +15,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA * * mostly by Michael Niedermayer <michaelni@gmx.at> */ @@ -34,33 +34,33 @@ static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) { long len= -(stride*h); asm volatile( - ".balign 16 \n\t" - "1: \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - "movq (%2, %%"REG_a"), %%mm2 \n\t" - "movq (%2, %%"REG_a"), %%mm4 \n\t" - "add %3, %%"REG_a" \n\t" - "psubusb %%mm0, %%mm2 \n\t" - "psubusb %%mm4, %%mm0 \n\t" - "movq (%1, %%"REG_a"), %%mm1 \n\t" - "movq (%2, %%"REG_a"), %%mm3 \n\t" - "movq (%2, %%"REG_a"), %%mm5 \n\t" - "psubusb %%mm1, %%mm3 \n\t" - "psubusb %%mm5, %%mm1 \n\t" - "por %%mm2, %%mm0 \n\t" - "por %%mm1, %%mm3 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm3, %%mm2 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "punpcklbw %%mm7, %%mm3 \n\t" - "punpckhbw %%mm7, %%mm2 \n\t" - "paddw %%mm1, %%mm0 \n\t" - "paddw %%mm3, %%mm2 \n\t" - "paddw %%mm2, %%mm0 \n\t" - "paddw %%mm0, %%mm6 \n\t" - "add %3, %%"REG_a" \n\t" - " js 1b \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + "movq (%2, %%"REG_a"), %%mm2 \n\t" + "movq (%2, %%"REG_a"), %%mm4 \n\t" + "add %3, %%"REG_a" \n\t" + "psubusb %%mm0, %%mm2 \n\t" + "psubusb %%mm4, %%mm0 \n\t" + "movq (%1, %%"REG_a"), %%mm1 \n\t" + "movq (%2, %%"REG_a"), %%mm3 \n\t" + "movq (%2, %%"REG_a"), %%mm5 \n\t" + "psubusb %%mm1, %%mm3 \n\t" + "psubusb %%mm5, %%mm1 \n\t" + "por %%mm2, %%mm0 \n\t" + "por %%mm1, %%mm3 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm3, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm3 \n\t" + "punpckhbw %%mm7, %%mm2 \n\t" + "paddw %%mm1, %%mm0 \n\t" + "paddw %%mm3, %%mm2 \n\t" + "paddw %%mm2, %%mm0 \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) ); @@ -70,19 +70,19 @@ static inline void sad8_1_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h) { long len= -(stride*h); asm volatile( - ".balign 16 \n\t" - "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" - "paddw %%mm0, %%mm6 \n\t" - "add %3, %%"REG_a" \n\t" - " js 1b \n\t" + ".balign 16 \n\t" + "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" + "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) ); @@ -92,23 +92,23 @@ static inline void sad8_2_mmx2(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, in { long len= -(stride*h); asm volatile( - ".balign 16 \n\t" - "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" - "paddw %%mm0, %%mm6 \n\t" - "add %4, %%"REG_a" \n\t" - " js 1b \n\t" + ".balign 16 \n\t" + "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" + "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) ); @@ -118,34 +118,34 @@ static inline void sad8_4_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h) { //FIXME reuse src long len= -(stride*h); asm volatile( - ".balign 16 \n\t" - "movq "MANGLE(bone)", %%mm5 \n\t" - "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" - "paddw %%mm0, %%mm6 \n\t" - "add %4, %%"REG_a" \n\t" - " js 1b \n\t" + ".balign 16 \n\t" + "movq "MANGLE(bone)", %%mm5 \n\t" + "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" + "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) ); @@ -155,35 +155,35 @@ static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int { long len= -(stride*h); asm volatile( - ".balign 16 \n\t" - "1: \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - "movq (%2, %%"REG_a"), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm2 \n\t" - "movq (%2, %%"REG_a"), %%mm3 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpcklbw %%mm7, %%mm1 \n\t" - "punpckhbw %%mm7, %%mm2 \n\t" - "punpckhbw %%mm7, %%mm3 \n\t" - "paddw %%mm0, %%mm1 \n\t" - "paddw %%mm2, %%mm3 \n\t" - "movq (%3, %%"REG_a"), %%mm4 \n\t" - "movq (%3, %%"REG_a"), %%mm2 \n\t" - "paddw %%mm5, %%mm1 \n\t" - "paddw %%mm5, %%mm3 \n\t" - "psrlw $1, %%mm1 \n\t" - "psrlw $1, %%mm3 \n\t" - "packuswb %%mm3, %%mm1 \n\t" - "psubusb %%mm1, %%mm4 \n\t" - "psubusb %%mm2, %%mm1 \n\t" - "por %%mm4, %%mm1 \n\t" - "movq %%mm1, %%mm0 \n\t" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpckhbw %%mm7, %%mm1 \n\t" - "paddw %%mm1, %%mm0 \n\t" - "paddw %%mm0, %%mm6 \n\t" - "add %4, %%"REG_a" \n\t" - " js 1b \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%1, %%"REG_a"), %%mm0 \n\t" + "movq (%2, %%"REG_a"), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm2 \n\t" + "movq (%2, %%"REG_a"), %%mm3 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm1 \n\t" + "punpckhbw %%mm7, %%mm2 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "paddw %%mm0, %%mm1 \n\t" + "paddw %%mm2, %%mm3 \n\t" + "movq (%3, %%"REG_a"), %%mm4 \n\t" + "movq (%3, %%"REG_a"), %%mm2 \n\t" + "paddw %%mm5, %%mm1 \n\t" + "paddw %%mm5, %%mm3 \n\t" + "psrlw $1, %%mm1 \n\t" + "psrlw $1, %%mm3 \n\t" + "packuswb %%mm3, %%mm1 \n\t" + "psubusb %%mm1, %%mm4 \n\t" + "psubusb %%mm2, %%mm1 \n\t" + "por %%mm4, %%mm1 \n\t" + "movq %%mm1, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "paddw %%mm1, %%mm0 \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) ); @@ -193,47 +193,47 @@ static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h) { long len= -(stride*h); asm volatile( - ".balign 16 \n\t" - "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" - "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" - "paddw %%mm5, %%mm1 \n\t" - "psrlw $2, %%mm2 \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" - "punpcklbw %%mm7, %%mm0 \n\t" - "punpckhbw %%mm7, %%mm2 \n\t" - "paddw %%mm2, %%mm0 \n\t" - "paddw %%mm0, %%mm6 \n\t" - "add %4, %%"REG_a" \n\t" - " js 1b \n\t" + ".balign 16 \n\t" + "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" + "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" + "paddw %%mm5, %%mm1 \n\t" + "psrlw $2, %%mm2 \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" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm2 \n\t" + "paddw %%mm2, %%mm0 \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) ); @@ -243,13 +243,13 @@ static inline int sum_mmx(void) { int ret; asm volatile( - "movq %%mm6, %%mm0 \n\t" - "psrlq $32, %%mm6 \n\t" - "paddw %%mm0, %%mm6 \n\t" - "movq %%mm6, %%mm0 \n\t" - "psrlq $16, %%mm6 \n\t" - "paddw %%mm0, %%mm6 \n\t" - "movd %%mm6, %0 \n\t" + "movq %%mm6, %%mm0 \n\t" + "psrlq $32, %%mm6 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "movq %%mm6, %%mm0 \n\t" + "psrlq $16, %%mm6 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "movd %%mm6, %0 \n\t" : "=r" (ret) ); return ret&0xFFFF; @@ -259,7 +259,7 @@ static inline int sum_mmx2(void) { int ret; asm volatile( - "movd %%mm6, %0 \n\t" + "movd %%mm6, %0 \n\t" : "=r" (ret) ); return ret; @@ -270,8 +270,8 @@ static inline int sum_mmx2(void) static int sad8_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\ {\ assert(h==8);\ - asm volatile("pxor %%mm7, %%mm7 \n\t"\ - "pxor %%mm6, %%mm6 \n\t":);\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t":);\ \ sad8_1_ ## suf(blk1, blk2, stride, 8);\ \ @@ -280,9 +280,9 @@ static int sad8_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h static int sad8_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\ {\ assert(h==8);\ - asm volatile("pxor %%mm7, %%mm7 \n\t"\ - "pxor %%mm6, %%mm6 \n\t"\ - "movq %0, %%mm5 \n\t"\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ :: "m"(round_tab[1]) \ );\ \ @@ -294,9 +294,9 @@ static int sad8_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, in static int sad8_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\ {\ assert(h==8);\ - asm volatile("pxor %%mm7, %%mm7 \n\t"\ - "pxor %%mm6, %%mm6 \n\t"\ - "movq %0, %%mm5 \n\t"\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ :: "m"(round_tab[1]) \ );\ \ @@ -308,9 +308,9 @@ static int sad8_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, in static int sad8_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\ {\ assert(h==8);\ - asm volatile("pxor %%mm7, %%mm7 \n\t"\ - "pxor %%mm6, %%mm6 \n\t"\ - "movq %0, %%mm5 \n\t"\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ :: "m"(round_tab[2]) \ );\ \ @@ -321,8 +321,8 @@ static int sad8_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, i \ static int sad16_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\ {\ - asm volatile("pxor %%mm7, %%mm7 \n\t"\ - "pxor %%mm6, %%mm6 \n\t":);\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t":);\ \ sad8_1_ ## suf(blk1 , blk2 , stride, h);\ sad8_1_ ## suf(blk1+8, blk2+8, stride, h);\ @@ -331,9 +331,9 @@ static int sad16_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int }\ static int sad16_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\ {\ - asm volatile("pxor %%mm7, %%mm7 \n\t"\ - "pxor %%mm6, %%mm6 \n\t"\ - "movq %0, %%mm5 \n\t"\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ :: "m"(round_tab[1]) \ );\ \ @@ -344,9 +344,9 @@ static int sad16_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, i }\ static int sad16_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\ {\ - asm volatile("pxor %%mm7, %%mm7 \n\t"\ - "pxor %%mm6, %%mm6 \n\t"\ - "movq %0, %%mm5 \n\t"\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ :: "m"(round_tab[1]) \ );\ \ @@ -357,9 +357,9 @@ static int sad16_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, i }\ static int sad16_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\ {\ - asm volatile("pxor %%mm7, %%mm7 \n\t"\ - "pxor %%mm6, %%mm6 \n\t"\ - "movq %0, %%mm5 \n\t"\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ :: "m"(round_tab[2]) \ );\ \ @@ -384,16 +384,16 @@ void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx) c->pix_abs[1][2] = sad8_y2_mmx; c->pix_abs[1][3] = sad8_xy2_mmx; - c->sad[0]= sad16_mmx; + c->sad[0]= sad16_mmx; c->sad[1]= sad8_mmx; } if (mm_flags & MM_MMXEXT) { - c->pix_abs[0][0] = sad16_mmx2; - c->pix_abs[1][0] = sad8_mmx2; + c->pix_abs[0][0] = sad16_mmx2; + c->pix_abs[1][0] = sad8_mmx2; + + c->sad[0]= sad16_mmx2; + c->sad[1]= sad8_mmx2; - c->sad[0]= sad16_mmx2; - c->sad[1]= sad8_mmx2; - if(!(avctx->flags & CODEC_FLAG_BITEXACT)){ c->pix_abs[0][1] = sad16_x2_mmx2; c->pix_abs[0][2] = sad16_y2_mmx2; diff --git a/src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c b/src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c index 70c81f675..f83df3a19 100644 --- a/src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c +++ b/src/libffmpeg/libavcodec/i386/mpegvideo_mmx.c @@ -14,7 +14,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA * * Optimized for ia32 cpus by Nick Kurshev <nickols_k@mail.ru> * h263, mpeg1, mpeg2 dequantizer & draw_edges by Michael Niedermayer <michaelni@gmx.at> @@ -40,7 +40,7 @@ static void dct_unquantize_h263_intra_mmx(MpegEncContext *s, qmul = qscale << 1; assert(s->block_last_index[n]>=0 || s->h263_aic); - + if (!s->h263_aic) { if (n < 4) level = block[0] * s->y_dc_scale; @@ -57,52 +57,52 @@ static void dct_unquantize_h263_intra_mmx(MpegEncContext *s, nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ]; //printf("%d %d ", qmul, qadd); asm volatile( - "movd %1, %%mm6 \n\t" //qmul - "packssdw %%mm6, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "movd %2, %%mm5 \n\t" //qadd - "pxor %%mm7, %%mm7 \n\t" - "packssdw %%mm5, %%mm5 \n\t" - "packssdw %%mm5, %%mm5 \n\t" - "psubw %%mm5, %%mm7 \n\t" - "pxor %%mm4, %%mm4 \n\t" - ".balign 16\n\t" - "1: \n\t" - "movq (%0, %3), %%mm0 \n\t" - "movq 8(%0, %3), %%mm1 \n\t" - - "pmullw %%mm6, %%mm0 \n\t" - "pmullw %%mm6, %%mm1 \n\t" - - "movq (%0, %3), %%mm2 \n\t" - "movq 8(%0, %3), %%mm3 \n\t" - - "pcmpgtw %%mm4, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 - "pcmpgtw %%mm4, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 - - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - - "paddw %%mm7, %%mm0 \n\t" - "paddw %%mm7, %%mm1 \n\t" - - "pxor %%mm0, %%mm2 \n\t" - "pxor %%mm1, %%mm3 \n\t" - - "pcmpeqw %%mm7, %%mm0 \n\t" // block[i] == 0 ? -1 : 0 - "pcmpeqw %%mm7, %%mm1 \n\t" // block[i] == 0 ? -1 : 0 - - "pandn %%mm2, %%mm0 \n\t" - "pandn %%mm3, %%mm1 \n\t" - - "movq %%mm0, (%0, %3) \n\t" - "movq %%mm1, 8(%0, %3) \n\t" - - "add $16, %3 \n\t" - "jng 1b \n\t" - ::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(-nCoeffs)) - : "memory" - ); + "movd %1, %%mm6 \n\t" //qmul + "packssdw %%mm6, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "movd %2, %%mm5 \n\t" //qadd + "pxor %%mm7, %%mm7 \n\t" + "packssdw %%mm5, %%mm5 \n\t" + "packssdw %%mm5, %%mm5 \n\t" + "psubw %%mm5, %%mm7 \n\t" + "pxor %%mm4, %%mm4 \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%0, %3), %%mm0 \n\t" + "movq 8(%0, %3), %%mm1 \n\t" + + "pmullw %%mm6, %%mm0 \n\t" + "pmullw %%mm6, %%mm1 \n\t" + + "movq (%0, %3), %%mm2 \n\t" + "movq 8(%0, %3), %%mm3 \n\t" + + "pcmpgtw %%mm4, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 + "pcmpgtw %%mm4, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 + + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + + "paddw %%mm7, %%mm0 \n\t" + "paddw %%mm7, %%mm1 \n\t" + + "pxor %%mm0, %%mm2 \n\t" + "pxor %%mm1, %%mm3 \n\t" + + "pcmpeqw %%mm7, %%mm0 \n\t" // block[i] == 0 ? -1 : 0 + "pcmpeqw %%mm7, %%mm1 \n\t" // block[i] == 0 ? -1 : 0 + + "pandn %%mm2, %%mm0 \n\t" + "pandn %%mm3, %%mm1 \n\t" + + "movq %%mm0, (%0, %3) \n\t" + "movq %%mm1, 8(%0, %3) \n\t" + + "add $16, %3 \n\t" + "jng 1b \n\t" + ::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(-nCoeffs)) + : "memory" + ); block[0]= level; } @@ -116,56 +116,56 @@ static void dct_unquantize_h263_inter_mmx(MpegEncContext *s, qadd = (qscale - 1) | 1; assert(s->block_last_index[n]>=0 || s->h263_aic); - + nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ]; //printf("%d %d ", qmul, qadd); asm volatile( - "movd %1, %%mm6 \n\t" //qmul - "packssdw %%mm6, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "movd %2, %%mm5 \n\t" //qadd - "pxor %%mm7, %%mm7 \n\t" - "packssdw %%mm5, %%mm5 \n\t" - "packssdw %%mm5, %%mm5 \n\t" - "psubw %%mm5, %%mm7 \n\t" - "pxor %%mm4, %%mm4 \n\t" - ".balign 16\n\t" - "1: \n\t" - "movq (%0, %3), %%mm0 \n\t" - "movq 8(%0, %3), %%mm1 \n\t" - - "pmullw %%mm6, %%mm0 \n\t" - "pmullw %%mm6, %%mm1 \n\t" - - "movq (%0, %3), %%mm2 \n\t" - "movq 8(%0, %3), %%mm3 \n\t" - - "pcmpgtw %%mm4, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 - "pcmpgtw %%mm4, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 - - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - - "paddw %%mm7, %%mm0 \n\t" - "paddw %%mm7, %%mm1 \n\t" - - "pxor %%mm0, %%mm2 \n\t" - "pxor %%mm1, %%mm3 \n\t" - - "pcmpeqw %%mm7, %%mm0 \n\t" // block[i] == 0 ? -1 : 0 - "pcmpeqw %%mm7, %%mm1 \n\t" // block[i] == 0 ? -1 : 0 - - "pandn %%mm2, %%mm0 \n\t" - "pandn %%mm3, %%mm1 \n\t" - - "movq %%mm0, (%0, %3) \n\t" - "movq %%mm1, 8(%0, %3) \n\t" - - "add $16, %3 \n\t" - "jng 1b \n\t" - ::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(-nCoeffs)) - : "memory" - ); + "movd %1, %%mm6 \n\t" //qmul + "packssdw %%mm6, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "movd %2, %%mm5 \n\t" //qadd + "pxor %%mm7, %%mm7 \n\t" + "packssdw %%mm5, %%mm5 \n\t" + "packssdw %%mm5, %%mm5 \n\t" + "psubw %%mm5, %%mm7 \n\t" + "pxor %%mm4, %%mm4 \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%0, %3), %%mm0 \n\t" + "movq 8(%0, %3), %%mm1 \n\t" + + "pmullw %%mm6, %%mm0 \n\t" + "pmullw %%mm6, %%mm1 \n\t" + + "movq (%0, %3), %%mm2 \n\t" + "movq 8(%0, %3), %%mm3 \n\t" + + "pcmpgtw %%mm4, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 + "pcmpgtw %%mm4, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 + + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + + "paddw %%mm7, %%mm0 \n\t" + "paddw %%mm7, %%mm1 \n\t" + + "pxor %%mm0, %%mm2 \n\t" + "pxor %%mm1, %%mm3 \n\t" + + "pcmpeqw %%mm7, %%mm0 \n\t" // block[i] == 0 ? -1 : 0 + "pcmpeqw %%mm7, %%mm1 \n\t" // block[i] == 0 ? -1 : 0 + + "pandn %%mm2, %%mm0 \n\t" + "pandn %%mm3, %%mm1 \n\t" + + "movq %%mm0, (%0, %3) \n\t" + "movq %%mm1, 8(%0, %3) \n\t" + + "add $16, %3 \n\t" + "jng 1b \n\t" + ::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(-nCoeffs)) + : "memory" + ); } @@ -209,61 +209,61 @@ static void dct_unquantize_mpeg1_intra_mmx(MpegEncContext *s, nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ]+1; - if (n < 4) + if (n < 4) block0 = block[0] * s->y_dc_scale; else block0 = block[0] * s->c_dc_scale; /* XXX: only mpeg1 */ quant_matrix = s->intra_matrix; asm volatile( - "pcmpeqw %%mm7, %%mm7 \n\t" - "psrlw $15, %%mm7 \n\t" - "movd %2, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "mov %3, %%"REG_a" \n\t" - ".balign 16\n\t" - "1: \n\t" - "movq (%0, %%"REG_a"), %%mm0 \n\t" - "movq 8(%0, %%"REG_a"), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm4 \n\t" - "movq 8(%1, %%"REG_a"), %%mm5 \n\t" - "pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i] - "pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i] - "pxor %%mm2, %%mm2 \n\t" - "pxor %%mm3, %%mm3 \n\t" - "pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 - "pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - "psubw %%mm2, %%mm0 \n\t" // abs(block[i]) - "psubw %%mm3, %%mm1 \n\t" // abs(block[i]) - "pmullw %%mm4, %%mm0 \n\t" // abs(block[i])*q - "pmullw %%mm5, %%mm1 \n\t" // abs(block[i])*q - "pxor %%mm4, %%mm4 \n\t" - "pxor %%mm5, %%mm5 \n\t" // FIXME slow - "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0 - "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0 - "psraw $3, %%mm0 \n\t" - "psraw $3, %%mm1 \n\t" - "psubw %%mm7, %%mm0 \n\t" - "psubw %%mm7, %%mm1 \n\t" - "por %%mm7, %%mm0 \n\t" - "por %%mm7, %%mm1 \n\t" - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - "psubw %%mm2, %%mm0 \n\t" - "psubw %%mm3, %%mm1 \n\t" - "pandn %%mm0, %%mm4 \n\t" - "pandn %%mm1, %%mm5 \n\t" - "movq %%mm4, (%0, %%"REG_a") \n\t" - "movq %%mm5, 8(%0, %%"REG_a") \n\t" - - "add $16, %%"REG_a" \n\t" - "js 1b \n\t" - ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs) - : "%"REG_a, "memory" - ); + "pcmpeqw %%mm7, %%mm7 \n\t" + "psrlw $15, %%mm7 \n\t" + "movd %2, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "mov %3, %%"REG_a" \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%0, %%"REG_a"), %%mm0 \n\t" + "movq 8(%0, %%"REG_a"), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm4 \n\t" + "movq 8(%1, %%"REG_a"), %%mm5 \n\t" + "pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i] + "pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i] + "pxor %%mm2, %%mm2 \n\t" + "pxor %%mm3, %%mm3 \n\t" + "pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 + "pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + "psubw %%mm2, %%mm0 \n\t" // abs(block[i]) + "psubw %%mm3, %%mm1 \n\t" // abs(block[i]) + "pmullw %%mm4, %%mm0 \n\t" // abs(block[i])*q + "pmullw %%mm5, %%mm1 \n\t" // abs(block[i])*q + "pxor %%mm4, %%mm4 \n\t" + "pxor %%mm5, %%mm5 \n\t" // FIXME slow + "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0 + "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0 + "psraw $3, %%mm0 \n\t" + "psraw $3, %%mm1 \n\t" + "psubw %%mm7, %%mm0 \n\t" + "psubw %%mm7, %%mm1 \n\t" + "por %%mm7, %%mm0 \n\t" + "por %%mm7, %%mm1 \n\t" + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + "psubw %%mm2, %%mm0 \n\t" + "psubw %%mm3, %%mm1 \n\t" + "pandn %%mm0, %%mm4 \n\t" + "pandn %%mm1, %%mm5 \n\t" + "movq %%mm4, (%0, %%"REG_a") \n\t" + "movq %%mm5, 8(%0, %%"REG_a") \n\t" + + "add $16, %%"REG_a" \n\t" + "js 1b \n\t" + ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs) + : "%"REG_a, "memory" + ); block[0]= block0; } @@ -279,58 +279,58 @@ static void dct_unquantize_mpeg1_inter_mmx(MpegEncContext *s, quant_matrix = s->inter_matrix; asm volatile( - "pcmpeqw %%mm7, %%mm7 \n\t" - "psrlw $15, %%mm7 \n\t" - "movd %2, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "mov %3, %%"REG_a" \n\t" - ".balign 16\n\t" - "1: \n\t" - "movq (%0, %%"REG_a"), %%mm0 \n\t" - "movq 8(%0, %%"REG_a"), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm4 \n\t" - "movq 8(%1, %%"REG_a"), %%mm5 \n\t" - "pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i] - "pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i] - "pxor %%mm2, %%mm2 \n\t" - "pxor %%mm3, %%mm3 \n\t" - "pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 - "pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - "psubw %%mm2, %%mm0 \n\t" // abs(block[i]) - "psubw %%mm3, %%mm1 \n\t" // abs(block[i]) - "paddw %%mm0, %%mm0 \n\t" // abs(block[i])*2 - "paddw %%mm1, %%mm1 \n\t" // abs(block[i])*2 - "paddw %%mm7, %%mm0 \n\t" // abs(block[i])*2 + 1 - "paddw %%mm7, %%mm1 \n\t" // abs(block[i])*2 + 1 - "pmullw %%mm4, %%mm0 \n\t" // (abs(block[i])*2 + 1)*q - "pmullw %%mm5, %%mm1 \n\t" // (abs(block[i])*2 + 1)*q - "pxor %%mm4, %%mm4 \n\t" - "pxor %%mm5, %%mm5 \n\t" // FIXME slow - "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0 - "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0 - "psraw $4, %%mm0 \n\t" - "psraw $4, %%mm1 \n\t" - "psubw %%mm7, %%mm0 \n\t" - "psubw %%mm7, %%mm1 \n\t" - "por %%mm7, %%mm0 \n\t" - "por %%mm7, %%mm1 \n\t" - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - "psubw %%mm2, %%mm0 \n\t" - "psubw %%mm3, %%mm1 \n\t" - "pandn %%mm0, %%mm4 \n\t" - "pandn %%mm1, %%mm5 \n\t" - "movq %%mm4, (%0, %%"REG_a") \n\t" - "movq %%mm5, 8(%0, %%"REG_a") \n\t" - - "add $16, %%"REG_a" \n\t" - "js 1b \n\t" - ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs) - : "%"REG_a, "memory" - ); + "pcmpeqw %%mm7, %%mm7 \n\t" + "psrlw $15, %%mm7 \n\t" + "movd %2, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "mov %3, %%"REG_a" \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%0, %%"REG_a"), %%mm0 \n\t" + "movq 8(%0, %%"REG_a"), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm4 \n\t" + "movq 8(%1, %%"REG_a"), %%mm5 \n\t" + "pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i] + "pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i] + "pxor %%mm2, %%mm2 \n\t" + "pxor %%mm3, %%mm3 \n\t" + "pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 + "pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + "psubw %%mm2, %%mm0 \n\t" // abs(block[i]) + "psubw %%mm3, %%mm1 \n\t" // abs(block[i]) + "paddw %%mm0, %%mm0 \n\t" // abs(block[i])*2 + "paddw %%mm1, %%mm1 \n\t" // abs(block[i])*2 + "paddw %%mm7, %%mm0 \n\t" // abs(block[i])*2 + 1 + "paddw %%mm7, %%mm1 \n\t" // abs(block[i])*2 + 1 + "pmullw %%mm4, %%mm0 \n\t" // (abs(block[i])*2 + 1)*q + "pmullw %%mm5, %%mm1 \n\t" // (abs(block[i])*2 + 1)*q + "pxor %%mm4, %%mm4 \n\t" + "pxor %%mm5, %%mm5 \n\t" // FIXME slow + "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0 + "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0 + "psraw $4, %%mm0 \n\t" + "psraw $4, %%mm1 \n\t" + "psubw %%mm7, %%mm0 \n\t" + "psubw %%mm7, %%mm1 \n\t" + "por %%mm7, %%mm0 \n\t" + "por %%mm7, %%mm1 \n\t" + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + "psubw %%mm2, %%mm0 \n\t" + "psubw %%mm3, %%mm1 \n\t" + "pandn %%mm0, %%mm4 \n\t" + "pandn %%mm1, %%mm5 \n\t" + "movq %%mm4, (%0, %%"REG_a") \n\t" + "movq %%mm5, 8(%0, %%"REG_a") \n\t" + + "add $16, %%"REG_a" \n\t" + "js 1b \n\t" + ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs) + : "%"REG_a, "memory" + ); } static void dct_unquantize_mpeg2_intra_mmx(MpegEncContext *s, @@ -339,62 +339,62 @@ static void dct_unquantize_mpeg2_intra_mmx(MpegEncContext *s, long nCoeffs; const uint16_t *quant_matrix; int block0; - + assert(s->block_last_index[n]>=0); if(s->alternate_scan) nCoeffs= 63; //FIXME else nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ]; - if (n < 4) + if (n < 4) block0 = block[0] * s->y_dc_scale; else block0 = block[0] * s->c_dc_scale; quant_matrix = s->intra_matrix; asm volatile( - "pcmpeqw %%mm7, %%mm7 \n\t" - "psrlw $15, %%mm7 \n\t" - "movd %2, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "mov %3, %%"REG_a" \n\t" - ".balign 16\n\t" - "1: \n\t" - "movq (%0, %%"REG_a"), %%mm0 \n\t" - "movq 8(%0, %%"REG_a"), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm4 \n\t" - "movq 8(%1, %%"REG_a"), %%mm5 \n\t" - "pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i] - "pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i] - "pxor %%mm2, %%mm2 \n\t" - "pxor %%mm3, %%mm3 \n\t" - "pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 - "pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - "psubw %%mm2, %%mm0 \n\t" // abs(block[i]) - "psubw %%mm3, %%mm1 \n\t" // abs(block[i]) - "pmullw %%mm4, %%mm0 \n\t" // abs(block[i])*q - "pmullw %%mm5, %%mm1 \n\t" // abs(block[i])*q - "pxor %%mm4, %%mm4 \n\t" - "pxor %%mm5, %%mm5 \n\t" // FIXME slow - "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0 - "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0 - "psraw $3, %%mm0 \n\t" - "psraw $3, %%mm1 \n\t" - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - "psubw %%mm2, %%mm0 \n\t" - "psubw %%mm3, %%mm1 \n\t" - "pandn %%mm0, %%mm4 \n\t" - "pandn %%mm1, %%mm5 \n\t" - "movq %%mm4, (%0, %%"REG_a") \n\t" - "movq %%mm5, 8(%0, %%"REG_a") \n\t" - - "add $16, %%"REG_a" \n\t" - "jng 1b \n\t" - ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs) - : "%"REG_a, "memory" - ); + "pcmpeqw %%mm7, %%mm7 \n\t" + "psrlw $15, %%mm7 \n\t" + "movd %2, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "mov %3, %%"REG_a" \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%0, %%"REG_a"), %%mm0 \n\t" + "movq 8(%0, %%"REG_a"), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm4 \n\t" + "movq 8(%1, %%"REG_a"), %%mm5 \n\t" + "pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i] + "pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i] + "pxor %%mm2, %%mm2 \n\t" + "pxor %%mm3, %%mm3 \n\t" + "pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 + "pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + "psubw %%mm2, %%mm0 \n\t" // abs(block[i]) + "psubw %%mm3, %%mm1 \n\t" // abs(block[i]) + "pmullw %%mm4, %%mm0 \n\t" // abs(block[i])*q + "pmullw %%mm5, %%mm1 \n\t" // abs(block[i])*q + "pxor %%mm4, %%mm4 \n\t" + "pxor %%mm5, %%mm5 \n\t" // FIXME slow + "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0 + "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0 + "psraw $3, %%mm0 \n\t" + "psraw $3, %%mm1 \n\t" + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + "psubw %%mm2, %%mm0 \n\t" + "psubw %%mm3, %%mm1 \n\t" + "pandn %%mm0, %%mm4 \n\t" + "pandn %%mm1, %%mm5 \n\t" + "movq %%mm4, (%0, %%"REG_a") \n\t" + "movq %%mm5, 8(%0, %%"REG_a") \n\t" + + "add $16, %%"REG_a" \n\t" + "jng 1b \n\t" + ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs) + : "%"REG_a, "memory" + ); block[0]= block0; //Note, we dont do mismatch control for intra as errors cannot accumulate } @@ -404,7 +404,7 @@ static void dct_unquantize_mpeg2_inter_mmx(MpegEncContext *s, { long nCoeffs; const uint16_t *quant_matrix; - + assert(s->block_last_index[n]>=0); if(s->alternate_scan) nCoeffs= 63; //FIXME @@ -412,71 +412,71 @@ static void dct_unquantize_mpeg2_inter_mmx(MpegEncContext *s, quant_matrix = s->inter_matrix; asm volatile( - "pcmpeqw %%mm7, %%mm7 \n\t" - "psrlq $48, %%mm7 \n\t" - "movd %2, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "packssdw %%mm6, %%mm6 \n\t" - "mov %3, %%"REG_a" \n\t" - ".balign 16\n\t" - "1: \n\t" - "movq (%0, %%"REG_a"), %%mm0 \n\t" - "movq 8(%0, %%"REG_a"), %%mm1 \n\t" - "movq (%1, %%"REG_a"), %%mm4 \n\t" - "movq 8(%1, %%"REG_a"), %%mm5 \n\t" - "pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i] - "pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i] - "pxor %%mm2, %%mm2 \n\t" - "pxor %%mm3, %%mm3 \n\t" - "pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 - "pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - "psubw %%mm2, %%mm0 \n\t" // abs(block[i]) - "psubw %%mm3, %%mm1 \n\t" // abs(block[i]) - "paddw %%mm0, %%mm0 \n\t" // abs(block[i])*2 - "paddw %%mm1, %%mm1 \n\t" // abs(block[i])*2 - "pmullw %%mm4, %%mm0 \n\t" // abs(block[i])*2*q - "pmullw %%mm5, %%mm1 \n\t" // abs(block[i])*2*q - "paddw %%mm4, %%mm0 \n\t" // (abs(block[i])*2 + 1)*q - "paddw %%mm5, %%mm1 \n\t" // (abs(block[i])*2 + 1)*q - "pxor %%mm4, %%mm4 \n\t" - "pxor %%mm5, %%mm5 \n\t" // FIXME slow - "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0 - "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0 - "psrlw $4, %%mm0 \n\t" - "psrlw $4, %%mm1 \n\t" - "pxor %%mm2, %%mm0 \n\t" - "pxor %%mm3, %%mm1 \n\t" - "psubw %%mm2, %%mm0 \n\t" - "psubw %%mm3, %%mm1 \n\t" - "pandn %%mm0, %%mm4 \n\t" - "pandn %%mm1, %%mm5 \n\t" - "pxor %%mm4, %%mm7 \n\t" - "pxor %%mm5, %%mm7 \n\t" - "movq %%mm4, (%0, %%"REG_a") \n\t" - "movq %%mm5, 8(%0, %%"REG_a") \n\t" - - "add $16, %%"REG_a" \n\t" - "jng 1b \n\t" - "movd 124(%0, %3), %%mm0 \n\t" - "movq %%mm7, %%mm6 \n\t" - "psrlq $32, %%mm7 \n\t" - "pxor %%mm6, %%mm7 \n\t" - "movq %%mm7, %%mm6 \n\t" - "psrlq $16, %%mm7 \n\t" - "pxor %%mm6, %%mm7 \n\t" - "pslld $31, %%mm7 \n\t" - "psrlq $15, %%mm7 \n\t" - "pxor %%mm7, %%mm0 \n\t" - "movd %%mm0, 124(%0, %3) \n\t" - - ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "r" (-2*nCoeffs) - : "%"REG_a, "memory" - ); + "pcmpeqw %%mm7, %%mm7 \n\t" + "psrlq $48, %%mm7 \n\t" + "movd %2, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "packssdw %%mm6, %%mm6 \n\t" + "mov %3, %%"REG_a" \n\t" + ".balign 16 \n\t" + "1: \n\t" + "movq (%0, %%"REG_a"), %%mm0 \n\t" + "movq 8(%0, %%"REG_a"), %%mm1 \n\t" + "movq (%1, %%"REG_a"), %%mm4 \n\t" + "movq 8(%1, %%"REG_a"), %%mm5 \n\t" + "pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i] + "pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i] + "pxor %%mm2, %%mm2 \n\t" + "pxor %%mm3, %%mm3 \n\t" + "pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0 + "pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0 + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + "psubw %%mm2, %%mm0 \n\t" // abs(block[i]) + "psubw %%mm3, %%mm1 \n\t" // abs(block[i]) + "paddw %%mm0, %%mm0 \n\t" // abs(block[i])*2 + "paddw %%mm1, %%mm1 \n\t" // abs(block[i])*2 + "pmullw %%mm4, %%mm0 \n\t" // abs(block[i])*2*q + "pmullw %%mm5, %%mm1 \n\t" // abs(block[i])*2*q + "paddw %%mm4, %%mm0 \n\t" // (abs(block[i])*2 + 1)*q + "paddw %%mm5, %%mm1 \n\t" // (abs(block[i])*2 + 1)*q + "pxor %%mm4, %%mm4 \n\t" + "pxor %%mm5, %%mm5 \n\t" // FIXME slow + "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0 + "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0 + "psrlw $4, %%mm0 \n\t" + "psrlw $4, %%mm1 \n\t" + "pxor %%mm2, %%mm0 \n\t" + "pxor %%mm3, %%mm1 \n\t" + "psubw %%mm2, %%mm0 \n\t" + "psubw %%mm3, %%mm1 \n\t" + "pandn %%mm0, %%mm4 \n\t" + "pandn %%mm1, %%mm5 \n\t" + "pxor %%mm4, %%mm7 \n\t" + "pxor %%mm5, %%mm7 \n\t" + "movq %%mm4, (%0, %%"REG_a") \n\t" + "movq %%mm5, 8(%0, %%"REG_a") \n\t" + + "add $16, %%"REG_a" \n\t" + "jng 1b \n\t" + "movd 124(%0, %3), %%mm0 \n\t" + "movq %%mm7, %%mm6 \n\t" + "psrlq $32, %%mm7 \n\t" + "pxor %%mm6, %%mm7 \n\t" + "movq %%mm7, %%mm6 \n\t" + "psrlq $16, %%mm7 \n\t" + "pxor %%mm6, %%mm7 \n\t" + "pslld $31, %%mm7 \n\t" + "psrlq $15, %%mm7 \n\t" + "pxor %%mm7, %%mm0 \n\t" + "movd %%mm0, 124(%0, %3) \n\t" + + ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "r" (-2*nCoeffs) + : "%"REG_a, "memory" + ); } -/* draw the edges of width 'w' of an image of size width, height +/* draw the edges of width 'w' of an image of size width, height this mmx version can only handle w==8 || w==16 */ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w) { @@ -488,79 +488,79 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w) ptr = buf; if(w==8) { - asm volatile( - "1: \n\t" - "movd (%0), %%mm0 \n\t" - "punpcklbw %%mm0, %%mm0 \n\t" - "punpcklwd %%mm0, %%mm0 \n\t" - "punpckldq %%mm0, %%mm0 \n\t" - "movq %%mm0, -8(%0) \n\t" - "movq -8(%0, %2), %%mm1 \n\t" - "punpckhbw %%mm1, %%mm1 \n\t" - "punpckhwd %%mm1, %%mm1 \n\t" - "punpckhdq %%mm1, %%mm1 \n\t" - "movq %%mm1, (%0, %2) \n\t" - "add %1, %0 \n\t" - "cmp %3, %0 \n\t" - " jb 1b \n\t" - : "+r" (ptr) - : "r" ((long)wrap), "r" ((long)width), "r" (ptr + wrap*height) - ); + asm volatile( + "1: \n\t" + "movd (%0), %%mm0 \n\t" + "punpcklbw %%mm0, %%mm0 \n\t" + "punpcklwd %%mm0, %%mm0 \n\t" + "punpckldq %%mm0, %%mm0 \n\t" + "movq %%mm0, -8(%0) \n\t" + "movq -8(%0, %2), %%mm1 \n\t" + "punpckhbw %%mm1, %%mm1 \n\t" + "punpckhwd %%mm1, %%mm1 \n\t" + "punpckhdq %%mm1, %%mm1 \n\t" + "movq %%mm1, (%0, %2) \n\t" + "add %1, %0 \n\t" + "cmp %3, %0 \n\t" + " jb 1b \n\t" + : "+r" (ptr) + : "r" ((long)wrap), "r" ((long)width), "r" (ptr + wrap*height) + ); } else { - asm volatile( - "1: \n\t" - "movd (%0), %%mm0 \n\t" - "punpcklbw %%mm0, %%mm0 \n\t" - "punpcklwd %%mm0, %%mm0 \n\t" - "punpckldq %%mm0, %%mm0 \n\t" - "movq %%mm0, -8(%0) \n\t" - "movq %%mm0, -16(%0) \n\t" - "movq -8(%0, %2), %%mm1 \n\t" - "punpckhbw %%mm1, %%mm1 \n\t" - "punpckhwd %%mm1, %%mm1 \n\t" - "punpckhdq %%mm1, %%mm1 \n\t" - "movq %%mm1, (%0, %2) \n\t" - "movq %%mm1, 8(%0, %2) \n\t" - "add %1, %0 \n\t" - "cmp %3, %0 \n\t" - " jb 1b \n\t" - : "+r" (ptr) - : "r" ((long)wrap), "r" ((long)width), "r" (ptr + wrap*height) - ); + asm volatile( + "1: \n\t" + "movd (%0), %%mm0 \n\t" + "punpcklbw %%mm0, %%mm0 \n\t" + "punpcklwd %%mm0, %%mm0 \n\t" + "punpckldq %%mm0, %%mm0 \n\t" + "movq %%mm0, -8(%0) \n\t" + "movq %%mm0, -16(%0) \n\t" + "movq -8(%0, %2), %%mm1 \n\t" + "punpckhbw %%mm1, %%mm1 \n\t" + "punpckhwd %%mm1, %%mm1 \n\t" + "punpckhdq %%mm1, %%mm1 \n\t" + "movq %%mm1, (%0, %2) \n\t" + "movq %%mm1, 8(%0, %2) \n\t" + "add %1, %0 \n\t" + "cmp %3, %0 \n\t" + " jb 1b \n\t" + : "+r" (ptr) + : "r" ((long)wrap), "r" ((long)width), "r" (ptr + wrap*height) + ); } - + for(i=0;i<w;i+=4) { /* top and bottom (and hopefully also the corners) */ - ptr= buf - (i + 1) * wrap - w; - asm volatile( - "1: \n\t" - "movq (%1, %0), %%mm0 \n\t" - "movq %%mm0, (%0) \n\t" - "movq %%mm0, (%0, %2) \n\t" - "movq %%mm0, (%0, %2, 2) \n\t" - "movq %%mm0, (%0, %3) \n\t" - "add $8, %0 \n\t" - "cmp %4, %0 \n\t" - " jb 1b \n\t" - : "+r" (ptr) - : "r" ((long)buf - (long)ptr - w), "r" ((long)-wrap), "r" ((long)-wrap*3), "r" (ptr+width+2*w) - ); - ptr= last_line + (i + 1) * wrap - w; - asm volatile( - "1: \n\t" - "movq (%1, %0), %%mm0 \n\t" - "movq %%mm0, (%0) \n\t" - "movq %%mm0, (%0, %2) \n\t" - "movq %%mm0, (%0, %2, 2) \n\t" - "movq %%mm0, (%0, %3) \n\t" - "add $8, %0 \n\t" - "cmp %4, %0 \n\t" - " jb 1b \n\t" - : "+r" (ptr) - : "r" ((long)last_line - (long)ptr - w), "r" ((long)wrap), "r" ((long)wrap*3), "r" (ptr+width+2*w) - ); + ptr= buf - (i + 1) * wrap - w; + asm volatile( + "1: \n\t" + "movq (%1, %0), %%mm0 \n\t" + "movq %%mm0, (%0) \n\t" + "movq %%mm0, (%0, %2) \n\t" + "movq %%mm0, (%0, %2, 2) \n\t" + "movq %%mm0, (%0, %3) \n\t" + "add $8, %0 \n\t" + "cmp %4, %0 \n\t" + " jb 1b \n\t" + : "+r" (ptr) + : "r" ((long)buf - (long)ptr - w), "r" ((long)-wrap), "r" ((long)-wrap*3), "r" (ptr+width+2*w) + ); + ptr= last_line + (i + 1) * wrap - w; + asm volatile( + "1: \n\t" + "movq (%1, %0), %%mm0 \n\t" + "movq %%mm0, (%0) \n\t" + "movq %%mm0, (%0, %2) \n\t" + "movq %%mm0, (%0, %2, 2) \n\t" + "movq %%mm0, (%0, %3) \n\t" + "add $8, %0 \n\t" + "cmp %4, %0 \n\t" + " jb 1b \n\t" + : "+r" (ptr) + : "r" ((long)last_line - (long)ptr - w), "r" ((long)wrap), "r" ((long)wrap*3), "r" (ptr+width+2*w) + ); } } @@ -572,47 +572,47 @@ static void denoise_dct_mmx(MpegEncContext *s, DCTELEM *block){ s->dct_count[intra]++; asm volatile( - "pxor %%mm7, %%mm7 \n\t" - "1: \n\t" - "pxor %%mm0, %%mm0 \n\t" - "pxor %%mm1, %%mm1 \n\t" - "movq (%0), %%mm2 \n\t" - "movq 8(%0), %%mm3 \n\t" - "pcmpgtw %%mm2, %%mm0 \n\t" - "pcmpgtw %%mm3, %%mm1 \n\t" - "pxor %%mm0, %%mm2 \n\t" - "pxor %%mm1, %%mm3 \n\t" - "psubw %%mm0, %%mm2 \n\t" - "psubw %%mm1, %%mm3 \n\t" - "movq %%mm2, %%mm4 \n\t" - "movq %%mm3, %%mm5 \n\t" - "psubusw (%2), %%mm2 \n\t" - "psubusw 8(%2), %%mm3 \n\t" - "pxor %%mm0, %%mm2 \n\t" - "pxor %%mm1, %%mm3 \n\t" - "psubw %%mm0, %%mm2 \n\t" - "psubw %%mm1, %%mm3 \n\t" - "movq %%mm2, (%0) \n\t" - "movq %%mm3, 8(%0) \n\t" - "movq %%mm4, %%mm2 \n\t" - "movq %%mm5, %%mm3 \n\t" - "punpcklwd %%mm7, %%mm4 \n\t" - "punpckhwd %%mm7, %%mm2 \n\t" - "punpcklwd %%mm7, %%mm5 \n\t" - "punpckhwd %%mm7, %%mm3 \n\t" - "paddd (%1), %%mm4 \n\t" - "paddd 8(%1), %%mm2 \n\t" - "paddd 16(%1), %%mm5 \n\t" - "paddd 24(%1), %%mm3 \n\t" - "movq %%mm4, (%1) \n\t" - "movq %%mm2, 8(%1) \n\t" - "movq %%mm5, 16(%1) \n\t" - "movq %%mm3, 24(%1) \n\t" - "add $16, %0 \n\t" - "add $32, %1 \n\t" - "add $16, %2 \n\t" - "cmp %3, %0 \n\t" - " jb 1b \n\t" + "pxor %%mm7, %%mm7 \n\t" + "1: \n\t" + "pxor %%mm0, %%mm0 \n\t" + "pxor %%mm1, %%mm1 \n\t" + "movq (%0), %%mm2 \n\t" + "movq 8(%0), %%mm3 \n\t" + "pcmpgtw %%mm2, %%mm0 \n\t" + "pcmpgtw %%mm3, %%mm1 \n\t" + "pxor %%mm0, %%mm2 \n\t" + "pxor %%mm1, %%mm3 \n\t" + "psubw %%mm0, %%mm2 \n\t" + "psubw %%mm1, %%mm3 \n\t" + "movq %%mm2, %%mm4 \n\t" + "movq %%mm3, %%mm5 \n\t" + "psubusw (%2), %%mm2 \n\t" + "psubusw 8(%2), %%mm3 \n\t" + "pxor %%mm0, %%mm2 \n\t" + "pxor %%mm1, %%mm3 \n\t" + "psubw %%mm0, %%mm2 \n\t" + "psubw %%mm1, %%mm3 \n\t" + "movq %%mm2, (%0) \n\t" + "movq %%mm3, 8(%0) \n\t" + "movq %%mm4, %%mm2 \n\t" + "movq %%mm5, %%mm3 \n\t" + "punpcklwd %%mm7, %%mm4 \n\t" + "punpckhwd %%mm7, %%mm2 \n\t" + "punpcklwd %%mm7, %%mm5 \n\t" + "punpckhwd %%mm7, %%mm3 \n\t" + "paddd (%1), %%mm4 \n\t" + "paddd 8(%1), %%mm2 \n\t" + "paddd 16(%1), %%mm5 \n\t" + "paddd 24(%1), %%mm3 \n\t" + "movq %%mm4, (%1) \n\t" + "movq %%mm2, 8(%1) \n\t" + "movq %%mm5, 16(%1) \n\t" + "movq %%mm3, 24(%1) \n\t" + "add $16, %0 \n\t" + "add $32, %1 \n\t" + "add $16, %2 \n\t" + "cmp %3, %0 \n\t" + " jb 1b \n\t" : "+r" (block), "+r" (sum), "+r" (offset) : "r"(block+64) ); @@ -626,47 +626,47 @@ static void denoise_dct_sse2(MpegEncContext *s, DCTELEM *block){ s->dct_count[intra]++; asm volatile( - "pxor %%xmm7, %%xmm7 \n\t" - "1: \n\t" - "pxor %%xmm0, %%xmm0 \n\t" - "pxor %%xmm1, %%xmm1 \n\t" - "movdqa (%0), %%xmm2 \n\t" - "movdqa 16(%0), %%xmm3 \n\t" - "pcmpgtw %%xmm2, %%xmm0 \n\t" - "pcmpgtw %%xmm3, %%xmm1 \n\t" - "pxor %%xmm0, %%xmm2 \n\t" - "pxor %%xmm1, %%xmm3 \n\t" - "psubw %%xmm0, %%xmm2 \n\t" - "psubw %%xmm1, %%xmm3 \n\t" - "movdqa %%xmm2, %%xmm4 \n\t" - "movdqa %%xmm3, %%xmm5 \n\t" - "psubusw (%2), %%xmm2 \n\t" - "psubusw 16(%2), %%xmm3 \n\t" - "pxor %%xmm0, %%xmm2 \n\t" - "pxor %%xmm1, %%xmm3 \n\t" - "psubw %%xmm0, %%xmm2 \n\t" - "psubw %%xmm1, %%xmm3 \n\t" - "movdqa %%xmm2, (%0) \n\t" - "movdqa %%xmm3, 16(%0) \n\t" - "movdqa %%xmm4, %%xmm6 \n\t" - "movdqa %%xmm5, %%xmm0 \n\t" - "punpcklwd %%xmm7, %%xmm4 \n\t" - "punpckhwd %%xmm7, %%xmm6 \n\t" - "punpcklwd %%xmm7, %%xmm5 \n\t" - "punpckhwd %%xmm7, %%xmm0 \n\t" - "paddd (%1), %%xmm4 \n\t" - "paddd 16(%1), %%xmm6 \n\t" - "paddd 32(%1), %%xmm5 \n\t" - "paddd 48(%1), %%xmm0 \n\t" - "movdqa %%xmm4, (%1) \n\t" - "movdqa %%xmm6, 16(%1) \n\t" - "movdqa %%xmm5, 32(%1) \n\t" - "movdqa %%xmm0, 48(%1) \n\t" - "add $32, %0 \n\t" - "add $64, %1 \n\t" - "add $32, %2 \n\t" - "cmp %3, %0 \n\t" - " jb 1b \n\t" + "pxor %%xmm7, %%xmm7 \n\t" + "1: \n\t" + "pxor %%xmm0, %%xmm0 \n\t" + "pxor %%xmm1, %%xmm1 \n\t" + "movdqa (%0), %%xmm2 \n\t" + "movdqa 16(%0), %%xmm3 \n\t" + "pcmpgtw %%xmm2, %%xmm0 \n\t" + "pcmpgtw %%xmm3, %%xmm1 \n\t" + "pxor %%xmm0, %%xmm2 \n\t" + "pxor %%xmm1, %%xmm3 \n\t" + "psubw %%xmm0, %%xmm2 \n\t" + "psubw %%xmm1, %%xmm3 \n\t" + "movdqa %%xmm2, %%xmm4 \n\t" + "movdqa %%xmm3, %%xmm5 \n\t" + "psubusw (%2), %%xmm2 \n\t" + "psubusw 16(%2), %%xmm3 \n\t" + "pxor %%xmm0, %%xmm2 \n\t" + "pxor %%xmm1, %%xmm3 \n\t" + "psubw %%xmm0, %%xmm2 \n\t" + "psubw %%xmm1, %%xmm3 \n\t" + "movdqa %%xmm2, (%0) \n\t" + "movdqa %%xmm3, 16(%0) \n\t" + "movdqa %%xmm4, %%xmm6 \n\t" + "movdqa %%xmm5, %%xmm0 \n\t" + "punpcklwd %%xmm7, %%xmm4 \n\t" + "punpckhwd %%xmm7, %%xmm6 \n\t" + "punpcklwd %%xmm7, %%xmm5 \n\t" + "punpckhwd %%xmm7, %%xmm0 \n\t" + "paddd (%1), %%xmm4 \n\t" + "paddd 16(%1), %%xmm6 \n\t" + "paddd 32(%1), %%xmm5 \n\t" + "paddd 48(%1), %%xmm0 \n\t" + "movdqa %%xmm4, (%1) \n\t" + "movdqa %%xmm6, 16(%1) \n\t" + "movdqa %%xmm5, 32(%1) \n\t" + "movdqa %%xmm0, 48(%1) \n\t" + "add $32, %0 \n\t" + "add $64, %1 \n\t" + "add $32, %2 \n\t" + "cmp %3, %0 \n\t" + " jb 1b \n\t" : "+r" (block), "+r" (sum), "+r" (offset) : "r"(block+64) ); @@ -694,7 +694,7 @@ void MPV_common_init_mmx(MpegEncContext *s) { if (mm_flags & MM_MMX) { const int dct_algo = s->avctx->dct_algo; - + s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_mmx; s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_mmx; s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_mmx; @@ -703,12 +703,12 @@ void MPV_common_init_mmx(MpegEncContext *s) s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_mmx; draw_edges = draw_edges_mmx; - + if (mm_flags & MM_SSE2) { - s->denoise_dct= denoise_dct_sse2; - } else { - s->denoise_dct= denoise_dct_mmx; - } + s->denoise_dct= denoise_dct_sse2; + } else { + s->denoise_dct= denoise_dct_mmx; + } if(dct_algo==FF_DCT_AUTO || dct_algo==FF_DCT_MMX){ if(mm_flags & MM_SSE2){ diff --git a/src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c b/src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c index 93f156ee5..2c50df232 100644 --- a/src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c +++ b/src/libffmpeg/libavcodec/i386/mpegvideo_mmx_template.c @@ -15,32 +15,32 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #undef SPREADW #undef PMAXW #ifdef HAVE_MMX2 #define SPREADW(a) "pshufw $0, " #a ", " #a " \n\t" -#define PMAXW(a,b) "pmaxsw " #a ", " #b " \n\t" +#define PMAXW(a,b) "pmaxsw " #a ", " #b " \n\t" #define PMAX(a,b) \ - "pshufw $0x0E," #a ", " #b " \n\t"\ - PMAXW(b, a)\ - "pshufw $0x01," #a ", " #b " \n\t"\ - PMAXW(b, a) + "pshufw $0x0E," #a ", " #b " \n\t"\ + PMAXW(b, a)\ + "pshufw $0x01," #a ", " #b " \n\t"\ + PMAXW(b, a) #else #define SPREADW(a) \ - "punpcklwd " #a ", " #a " \n\t"\ - "punpcklwd " #a ", " #a " \n\t" + "punpcklwd " #a ", " #a " \n\t"\ + "punpcklwd " #a ", " #a " \n\t" #define PMAXW(a,b) \ - "psubusw " #a ", " #b " \n\t"\ - "paddw " #a ", " #b " \n\t" + "psubusw " #a ", " #b " \n\t"\ + "paddw " #a ", " #b " \n\t" #define PMAX(a,b) \ - "movq " #a ", " #b " \n\t"\ - "psrlq $32, " #a " \n\t"\ - PMAXW(b, a)\ - "movq " #a ", " #b " \n\t"\ - "psrlq $16, " #a " \n\t"\ - PMAXW(b, a) + "movq " #a ", " #b " \n\t"\ + "psrlq $32, " #a " \n\t"\ + PMAXW(b, a)\ + "movq " #a ", " #b " \n\t"\ + "psrlq $16, " #a " \n\t"\ + PMAXW(b, a) #endif @@ -51,8 +51,8 @@ static int RENAME(dct_quantize)(MpegEncContext *s, long last_non_zero_p1; int level=0, q; //=0 is cuz gcc says uninitalized ... const uint16_t *qmat, *bias; - __align8 int16_t temp_block[64]; - + DECLARE_ALIGNED_8(int16_t, temp_block[64]); + assert((7&(int)(&temp_block[0])) == 0); //did gcc align it correctly? //s->fdct (block); @@ -71,24 +71,24 @@ static int RENAME(dct_quantize)(MpegEncContext *s, if (!s->h263_aic) { #if 1 asm volatile ( - "mul %%ecx \n\t" - : "=d" (level), "=a"(dummy) - : "a" ((block[0]>>2) + q), "c" (inverse[q<<1]) + "mul %%ecx \n\t" + : "=d" (level), "=a"(dummy) + : "a" ((block[0]>>2) + q), "c" (inverse[q<<1]) ); #else asm volatile ( - "xorl %%edx, %%edx \n\t" - "divw %%cx \n\t" - "movzwl %%ax, %%eax \n\t" - : "=a" (level) - : "a" ((block[0]>>2) + q), "c" (q<<1) - : "%edx" + "xorl %%edx, %%edx \n\t" + "divw %%cx \n\t" + "movzwl %%ax, %%eax \n\t" + : "=a" (level) + : "a" ((block[0]>>2) + q), "c" (q<<1) + : "%edx" ); #endif } else /* For AIC we skip quant/dequant of INTRADC */ level = (block[0] + 4)>>3; - + block[0]=0; //avoid fake overflow // temp_block[0] = (block[0] + (q >> 1)) / q; last_non_zero_p1 = 1; @@ -101,96 +101,96 @@ static int RENAME(dct_quantize)(MpegEncContext *s, } if((s->out_format == FMT_H263 || s->out_format == FMT_H261) && s->mpeg_quant==0){ - + asm volatile( - "movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1 + "movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1 SPREADW(%%mm3) - "pxor %%mm7, %%mm7 \n\t" // 0 - "pxor %%mm4, %%mm4 \n\t" // 0 - "movq (%2), %%mm5 \n\t" // qmat[0] - "pxor %%mm6, %%mm6 \n\t" - "psubw (%3), %%mm6 \n\t" // -bias[0] - "mov $-128, %%"REG_a" \n\t" - ".balign 16 \n\t" - "1: \n\t" - "pxor %%mm1, %%mm1 \n\t" // 0 - "movq (%1, %%"REG_a"), %%mm0 \n\t" // block[i] - "pcmpgtw %%mm0, %%mm1 \n\t" // block[i] <= 0 ? 0xFF : 0x00 - "pxor %%mm1, %%mm0 \n\t" - "psubw %%mm1, %%mm0 \n\t" // ABS(block[i]) - "psubusw %%mm6, %%mm0 \n\t" // ABS(block[i]) + bias[0] - "pmulhw %%mm5, %%mm0 \n\t" // (ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16 - "por %%mm0, %%mm4 \n\t" - "pxor %%mm1, %%mm0 \n\t" - "psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i]) - "movq %%mm0, (%5, %%"REG_a") \n\t" - "pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00 - "movq (%4, %%"REG_a"), %%mm1 \n\t" - "movq %%mm7, (%1, %%"REG_a") \n\t" // 0 - "pandn %%mm1, %%mm0 \n\t" - PMAXW(%%mm0, %%mm3) - "add $8, %%"REG_a" \n\t" - " js 1b \n\t" - PMAX(%%mm3, %%mm0) - "movd %%mm3, %%"REG_a" \n\t" - "movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1 - : "+a" (last_non_zero_p1) + "pxor %%mm7, %%mm7 \n\t" // 0 + "pxor %%mm4, %%mm4 \n\t" // 0 + "movq (%2), %%mm5 \n\t" // qmat[0] + "pxor %%mm6, %%mm6 \n\t" + "psubw (%3), %%mm6 \n\t" // -bias[0] + "mov $-128, %%"REG_a" \n\t" + ".balign 16 \n\t" + "1: \n\t" + "pxor %%mm1, %%mm1 \n\t" // 0 + "movq (%1, %%"REG_a"), %%mm0 \n\t" // block[i] + "pcmpgtw %%mm0, %%mm1 \n\t" // block[i] <= 0 ? 0xFF : 0x00 + "pxor %%mm1, %%mm0 \n\t" + "psubw %%mm1, %%mm0 \n\t" // ABS(block[i]) + "psubusw %%mm6, %%mm0 \n\t" // ABS(block[i]) + bias[0] + "pmulhw %%mm5, %%mm0 \n\t" // (ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16 + "por %%mm0, %%mm4 \n\t" + "pxor %%mm1, %%mm0 \n\t" + "psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i]) + "movq %%mm0, (%5, %%"REG_a") \n\t" + "pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00 + "movq (%4, %%"REG_a"), %%mm1 \n\t" + "movq %%mm7, (%1, %%"REG_a") \n\t" // 0 + "pandn %%mm1, %%mm0 \n\t" + PMAXW(%%mm0, %%mm3) + "add $8, %%"REG_a" \n\t" + " js 1b \n\t" + PMAX(%%mm3, %%mm0) + "movd %%mm3, %%"REG_a" \n\t" + "movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1 + : "+a" (last_non_zero_p1) : "r" (block+64), "r" (qmat), "r" (bias), "r" (inv_zigzag_direct16+64), "r" (temp_block+64) ); // note the asm is split cuz gcc doesnt like that many operands ... asm volatile( - "movd %1, %%mm1 \n\t" // max_qcoeff - SPREADW(%%mm1) - "psubusw %%mm1, %%mm4 \n\t" - "packuswb %%mm4, %%mm4 \n\t" - "movd %%mm4, %0 \n\t" // *overflow + "movd %1, %%mm1 \n\t" // max_qcoeff + SPREADW(%%mm1) + "psubusw %%mm1, %%mm4 \n\t" + "packuswb %%mm4, %%mm4 \n\t" + "movd %%mm4, %0 \n\t" // *overflow : "=g" (*overflow) : "g" (s->max_qcoeff) ); }else{ // FMT_H263 asm volatile( - "movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1 + "movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1 SPREADW(%%mm3) - "pxor %%mm7, %%mm7 \n\t" // 0 - "pxor %%mm4, %%mm4 \n\t" // 0 - "mov $-128, %%"REG_a" \n\t" - ".balign 16 \n\t" - "1: \n\t" - "pxor %%mm1, %%mm1 \n\t" // 0 - "movq (%1, %%"REG_a"), %%mm0 \n\t" // block[i] - "pcmpgtw %%mm0, %%mm1 \n\t" // block[i] <= 0 ? 0xFF : 0x00 - "pxor %%mm1, %%mm0 \n\t" - "psubw %%mm1, %%mm0 \n\t" // ABS(block[i]) - "movq (%3, %%"REG_a"), %%mm6 \n\t" // bias[0] - "paddusw %%mm6, %%mm0 \n\t" // ABS(block[i]) + bias[0] - "movq (%2, %%"REG_a"), %%mm5 \n\t" // qmat[i] - "pmulhw %%mm5, %%mm0 \n\t" // (ABS(block[i])*qmat[0] + bias[0]*qmat[0])>>16 - "por %%mm0, %%mm4 \n\t" - "pxor %%mm1, %%mm0 \n\t" - "psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i]) - "movq %%mm0, (%5, %%"REG_a") \n\t" - "pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00 - "movq (%4, %%"REG_a"), %%mm1 \n\t" - "movq %%mm7, (%1, %%"REG_a") \n\t" // 0 - "pandn %%mm1, %%mm0 \n\t" - PMAXW(%%mm0, %%mm3) - "add $8, %%"REG_a" \n\t" - " js 1b \n\t" - PMAX(%%mm3, %%mm0) - "movd %%mm3, %%"REG_a" \n\t" - "movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1 - : "+a" (last_non_zero_p1) + "pxor %%mm7, %%mm7 \n\t" // 0 + "pxor %%mm4, %%mm4 \n\t" // 0 + "mov $-128, %%"REG_a" \n\t" + ".balign 16 \n\t" + "1: \n\t" + "pxor %%mm1, %%mm1 \n\t" // 0 + "movq (%1, %%"REG_a"), %%mm0 \n\t" // block[i] + "pcmpgtw %%mm0, %%mm1 \n\t" // block[i] <= 0 ? 0xFF : 0x00 + "pxor %%mm1, %%mm0 \n\t" + "psubw %%mm1, %%mm0 \n\t" // ABS(block[i]) + "movq (%3, %%"REG_a"), %%mm6 \n\t" // bias[0] + "paddusw %%mm6, %%mm0 \n\t" // ABS(block[i]) + bias[0] + "movq (%2, %%"REG_a"), %%mm5 \n\t" // qmat[i] + "pmulhw %%mm5, %%mm0 \n\t" // (ABS(block[i])*qmat[0] + bias[0]*qmat[0])>>16 + "por %%mm0, %%mm4 \n\t" + "pxor %%mm1, %%mm0 \n\t" + "psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i]) + "movq %%mm0, (%5, %%"REG_a") \n\t" + "pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00 + "movq (%4, %%"REG_a"), %%mm1 \n\t" + "movq %%mm7, (%1, %%"REG_a") \n\t" // 0 + "pandn %%mm1, %%mm0 \n\t" + PMAXW(%%mm0, %%mm3) + "add $8, %%"REG_a" \n\t" + " js 1b \n\t" + PMAX(%%mm3, %%mm0) + "movd %%mm3, %%"REG_a" \n\t" + "movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1 + : "+a" (last_non_zero_p1) : "r" (block+64), "r" (qmat+64), "r" (bias+64), "r" (inv_zigzag_direct16+64), "r" (temp_block+64) ); // note the asm is split cuz gcc doesnt like that many operands ... asm volatile( - "movd %1, %%mm1 \n\t" // max_qcoeff - SPREADW(%%mm1) - "psubusw %%mm1, %%mm4 \n\t" - "packuswb %%mm4, %%mm4 \n\t" - "movd %%mm4, %0 \n\t" // *overflow + "movd %1, %%mm1 \n\t" // max_qcoeff + SPREADW(%%mm1) + "psubusw %%mm1, %%mm4 \n\t" + "packuswb %%mm4, %%mm4 \n\t" + "movd %%mm4, %0 \n\t" // *overflow : "=g" (*overflow) : "g" (s->max_qcoeff) ); @@ -201,135 +201,135 @@ static int RENAME(dct_quantize)(MpegEncContext *s, if(s->dsp.idct_permutation_type == FF_SIMPLE_IDCT_PERM){ if(last_non_zero_p1 <= 1) goto end; - block[0x08] = temp_block[0x01]; block[0x10] = temp_block[0x08]; - block[0x20] = temp_block[0x10]; + block[0x08] = temp_block[0x01]; block[0x10] = temp_block[0x08]; + block[0x20] = temp_block[0x10]; if(last_non_zero_p1 <= 4) goto end; - block[0x18] = temp_block[0x09]; block[0x04] = temp_block[0x02]; - block[0x09] = temp_block[0x03]; + block[0x18] = temp_block[0x09]; block[0x04] = temp_block[0x02]; + block[0x09] = temp_block[0x03]; if(last_non_zero_p1 <= 7) goto end; - block[0x14] = temp_block[0x0A]; block[0x28] = temp_block[0x11]; - block[0x12] = temp_block[0x18]; block[0x02] = temp_block[0x20]; + block[0x14] = temp_block[0x0A]; block[0x28] = temp_block[0x11]; + block[0x12] = temp_block[0x18]; block[0x02] = temp_block[0x20]; if(last_non_zero_p1 <= 11) goto end; - block[0x1A] = temp_block[0x19]; block[0x24] = temp_block[0x12]; - block[0x19] = temp_block[0x0B]; block[0x01] = temp_block[0x04]; - block[0x0C] = temp_block[0x05]; + block[0x1A] = temp_block[0x19]; block[0x24] = temp_block[0x12]; + block[0x19] = temp_block[0x0B]; block[0x01] = temp_block[0x04]; + block[0x0C] = temp_block[0x05]; if(last_non_zero_p1 <= 16) goto end; - block[0x11] = temp_block[0x0C]; block[0x29] = temp_block[0x13]; - block[0x16] = temp_block[0x1A]; block[0x0A] = temp_block[0x21]; - block[0x30] = temp_block[0x28]; block[0x22] = temp_block[0x30]; - block[0x38] = temp_block[0x29]; block[0x06] = temp_block[0x22]; + block[0x11] = temp_block[0x0C]; block[0x29] = temp_block[0x13]; + block[0x16] = temp_block[0x1A]; block[0x0A] = temp_block[0x21]; + block[0x30] = temp_block[0x28]; block[0x22] = temp_block[0x30]; + block[0x38] = temp_block[0x29]; block[0x06] = temp_block[0x22]; if(last_non_zero_p1 <= 24) goto end; - block[0x1B] = temp_block[0x1B]; block[0x21] = temp_block[0x14]; - block[0x1C] = temp_block[0x0D]; block[0x05] = temp_block[0x06]; - block[0x0D] = temp_block[0x07]; block[0x15] = temp_block[0x0E]; - block[0x2C] = temp_block[0x15]; block[0x13] = temp_block[0x1C]; + block[0x1B] = temp_block[0x1B]; block[0x21] = temp_block[0x14]; + block[0x1C] = temp_block[0x0D]; block[0x05] = temp_block[0x06]; + block[0x0D] = temp_block[0x07]; block[0x15] = temp_block[0x0E]; + block[0x2C] = temp_block[0x15]; block[0x13] = temp_block[0x1C]; if(last_non_zero_p1 <= 32) goto end; - block[0x0B] = temp_block[0x23]; block[0x34] = temp_block[0x2A]; - block[0x2A] = temp_block[0x31]; block[0x32] = temp_block[0x38]; - block[0x3A] = temp_block[0x39]; block[0x26] = temp_block[0x32]; - block[0x39] = temp_block[0x2B]; block[0x03] = temp_block[0x24]; + block[0x0B] = temp_block[0x23]; block[0x34] = temp_block[0x2A]; + block[0x2A] = temp_block[0x31]; block[0x32] = temp_block[0x38]; + block[0x3A] = temp_block[0x39]; block[0x26] = temp_block[0x32]; + block[0x39] = temp_block[0x2B]; block[0x03] = temp_block[0x24]; if(last_non_zero_p1 <= 40) goto end; - block[0x1E] = temp_block[0x1D]; block[0x25] = temp_block[0x16]; - block[0x1D] = temp_block[0x0F]; block[0x2D] = temp_block[0x17]; - block[0x17] = temp_block[0x1E]; block[0x0E] = temp_block[0x25]; - block[0x31] = temp_block[0x2C]; block[0x2B] = temp_block[0x33]; + block[0x1E] = temp_block[0x1D]; block[0x25] = temp_block[0x16]; + block[0x1D] = temp_block[0x0F]; block[0x2D] = temp_block[0x17]; + block[0x17] = temp_block[0x1E]; block[0x0E] = temp_block[0x25]; + block[0x31] = temp_block[0x2C]; block[0x2B] = temp_block[0x33]; if(last_non_zero_p1 <= 48) goto end; - block[0x36] = temp_block[0x3A]; block[0x3B] = temp_block[0x3B]; - block[0x23] = temp_block[0x34]; block[0x3C] = temp_block[0x2D]; - block[0x07] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; - block[0x0F] = temp_block[0x27]; block[0x35] = temp_block[0x2E]; + block[0x36] = temp_block[0x3A]; block[0x3B] = temp_block[0x3B]; + block[0x23] = temp_block[0x34]; block[0x3C] = temp_block[0x2D]; + block[0x07] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; + block[0x0F] = temp_block[0x27]; block[0x35] = temp_block[0x2E]; if(last_non_zero_p1 <= 56) goto end; - block[0x2E] = temp_block[0x35]; block[0x33] = temp_block[0x3C]; - block[0x3E] = temp_block[0x3D]; block[0x27] = temp_block[0x36]; - block[0x3D] = temp_block[0x2F]; block[0x2F] = temp_block[0x37]; + block[0x2E] = temp_block[0x35]; block[0x33] = temp_block[0x3C]; + block[0x3E] = temp_block[0x3D]; block[0x27] = temp_block[0x36]; + block[0x3D] = temp_block[0x2F]; block[0x2F] = temp_block[0x37]; block[0x37] = temp_block[0x3E]; block[0x3F] = temp_block[0x3F]; }else if(s->dsp.idct_permutation_type == FF_LIBMPEG2_IDCT_PERM){ if(last_non_zero_p1 <= 1) goto end; - block[0x04] = temp_block[0x01]; - block[0x08] = temp_block[0x08]; block[0x10] = temp_block[0x10]; + block[0x04] = temp_block[0x01]; + block[0x08] = temp_block[0x08]; block[0x10] = temp_block[0x10]; if(last_non_zero_p1 <= 4) goto end; - block[0x0C] = temp_block[0x09]; block[0x01] = temp_block[0x02]; - block[0x05] = temp_block[0x03]; + block[0x0C] = temp_block[0x09]; block[0x01] = temp_block[0x02]; + block[0x05] = temp_block[0x03]; if(last_non_zero_p1 <= 7) goto end; - block[0x09] = temp_block[0x0A]; block[0x14] = temp_block[0x11]; - block[0x18] = temp_block[0x18]; block[0x20] = temp_block[0x20]; + block[0x09] = temp_block[0x0A]; block[0x14] = temp_block[0x11]; + block[0x18] = temp_block[0x18]; block[0x20] = temp_block[0x20]; if(last_non_zero_p1 <= 11) goto end; - block[0x1C] = temp_block[0x19]; - block[0x11] = temp_block[0x12]; block[0x0D] = temp_block[0x0B]; - block[0x02] = temp_block[0x04]; block[0x06] = temp_block[0x05]; + block[0x1C] = temp_block[0x19]; + block[0x11] = temp_block[0x12]; block[0x0D] = temp_block[0x0B]; + block[0x02] = temp_block[0x04]; block[0x06] = temp_block[0x05]; if(last_non_zero_p1 <= 16) goto end; - block[0x0A] = temp_block[0x0C]; block[0x15] = temp_block[0x13]; - block[0x19] = temp_block[0x1A]; block[0x24] = temp_block[0x21]; - block[0x28] = temp_block[0x28]; block[0x30] = temp_block[0x30]; - block[0x2C] = temp_block[0x29]; block[0x21] = temp_block[0x22]; + block[0x0A] = temp_block[0x0C]; block[0x15] = temp_block[0x13]; + block[0x19] = temp_block[0x1A]; block[0x24] = temp_block[0x21]; + block[0x28] = temp_block[0x28]; block[0x30] = temp_block[0x30]; + block[0x2C] = temp_block[0x29]; block[0x21] = temp_block[0x22]; if(last_non_zero_p1 <= 24) goto end; - block[0x1D] = temp_block[0x1B]; block[0x12] = temp_block[0x14]; - block[0x0E] = temp_block[0x0D]; block[0x03] = temp_block[0x06]; - block[0x07] = temp_block[0x07]; block[0x0B] = temp_block[0x0E]; - block[0x16] = temp_block[0x15]; block[0x1A] = temp_block[0x1C]; + block[0x1D] = temp_block[0x1B]; block[0x12] = temp_block[0x14]; + block[0x0E] = temp_block[0x0D]; block[0x03] = temp_block[0x06]; + block[0x07] = temp_block[0x07]; block[0x0B] = temp_block[0x0E]; + block[0x16] = temp_block[0x15]; block[0x1A] = temp_block[0x1C]; if(last_non_zero_p1 <= 32) goto end; - block[0x25] = temp_block[0x23]; block[0x29] = temp_block[0x2A]; - block[0x34] = temp_block[0x31]; block[0x38] = temp_block[0x38]; - block[0x3C] = temp_block[0x39]; block[0x31] = temp_block[0x32]; - block[0x2D] = temp_block[0x2B]; block[0x22] = temp_block[0x24]; + block[0x25] = temp_block[0x23]; block[0x29] = temp_block[0x2A]; + block[0x34] = temp_block[0x31]; block[0x38] = temp_block[0x38]; + block[0x3C] = temp_block[0x39]; block[0x31] = temp_block[0x32]; + block[0x2D] = temp_block[0x2B]; block[0x22] = temp_block[0x24]; if(last_non_zero_p1 <= 40) goto end; - block[0x1E] = temp_block[0x1D]; block[0x13] = temp_block[0x16]; - block[0x0F] = temp_block[0x0F]; block[0x17] = temp_block[0x17]; - block[0x1B] = temp_block[0x1E]; block[0x26] = temp_block[0x25]; - block[0x2A] = temp_block[0x2C]; block[0x35] = temp_block[0x33]; + block[0x1E] = temp_block[0x1D]; block[0x13] = temp_block[0x16]; + block[0x0F] = temp_block[0x0F]; block[0x17] = temp_block[0x17]; + block[0x1B] = temp_block[0x1E]; block[0x26] = temp_block[0x25]; + block[0x2A] = temp_block[0x2C]; block[0x35] = temp_block[0x33]; if(last_non_zero_p1 <= 48) goto end; - block[0x39] = temp_block[0x3A]; block[0x3D] = temp_block[0x3B]; - block[0x32] = temp_block[0x34]; block[0x2E] = temp_block[0x2D]; - block[0x23] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; - block[0x27] = temp_block[0x27]; block[0x2B] = temp_block[0x2E]; + block[0x39] = temp_block[0x3A]; block[0x3D] = temp_block[0x3B]; + block[0x32] = temp_block[0x34]; block[0x2E] = temp_block[0x2D]; + block[0x23] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; + block[0x27] = temp_block[0x27]; block[0x2B] = temp_block[0x2E]; if(last_non_zero_p1 <= 56) goto end; - block[0x36] = temp_block[0x35]; block[0x3A] = temp_block[0x3C]; - block[0x3E] = temp_block[0x3D]; block[0x33] = temp_block[0x36]; - block[0x2F] = temp_block[0x2F]; block[0x37] = temp_block[0x37]; + block[0x36] = temp_block[0x35]; block[0x3A] = temp_block[0x3C]; + block[0x3E] = temp_block[0x3D]; block[0x33] = temp_block[0x36]; + block[0x2F] = temp_block[0x2F]; block[0x37] = temp_block[0x37]; block[0x3B] = temp_block[0x3E]; block[0x3F] = temp_block[0x3F]; }else{ if(last_non_zero_p1 <= 1) goto end; - block[0x01] = temp_block[0x01]; - block[0x08] = temp_block[0x08]; block[0x10] = temp_block[0x10]; + block[0x01] = temp_block[0x01]; + block[0x08] = temp_block[0x08]; block[0x10] = temp_block[0x10]; if(last_non_zero_p1 <= 4) goto end; - block[0x09] = temp_block[0x09]; block[0x02] = temp_block[0x02]; - block[0x03] = temp_block[0x03]; + block[0x09] = temp_block[0x09]; block[0x02] = temp_block[0x02]; + block[0x03] = temp_block[0x03]; if(last_non_zero_p1 <= 7) goto end; - block[0x0A] = temp_block[0x0A]; block[0x11] = temp_block[0x11]; - block[0x18] = temp_block[0x18]; block[0x20] = temp_block[0x20]; + block[0x0A] = temp_block[0x0A]; block[0x11] = temp_block[0x11]; + block[0x18] = temp_block[0x18]; block[0x20] = temp_block[0x20]; if(last_non_zero_p1 <= 11) goto end; - block[0x19] = temp_block[0x19]; - block[0x12] = temp_block[0x12]; block[0x0B] = temp_block[0x0B]; - block[0x04] = temp_block[0x04]; block[0x05] = temp_block[0x05]; + block[0x19] = temp_block[0x19]; + block[0x12] = temp_block[0x12]; block[0x0B] = temp_block[0x0B]; + block[0x04] = temp_block[0x04]; block[0x05] = temp_block[0x05]; if(last_non_zero_p1 <= 16) goto end; - block[0x0C] = temp_block[0x0C]; block[0x13] = temp_block[0x13]; - block[0x1A] = temp_block[0x1A]; block[0x21] = temp_block[0x21]; - block[0x28] = temp_block[0x28]; block[0x30] = temp_block[0x30]; - block[0x29] = temp_block[0x29]; block[0x22] = temp_block[0x22]; + block[0x0C] = temp_block[0x0C]; block[0x13] = temp_block[0x13]; + block[0x1A] = temp_block[0x1A]; block[0x21] = temp_block[0x21]; + block[0x28] = temp_block[0x28]; block[0x30] = temp_block[0x30]; + block[0x29] = temp_block[0x29]; block[0x22] = temp_block[0x22]; if(last_non_zero_p1 <= 24) goto end; - block[0x1B] = temp_block[0x1B]; block[0x14] = temp_block[0x14]; - block[0x0D] = temp_block[0x0D]; block[0x06] = temp_block[0x06]; - block[0x07] = temp_block[0x07]; block[0x0E] = temp_block[0x0E]; - block[0x15] = temp_block[0x15]; block[0x1C] = temp_block[0x1C]; + block[0x1B] = temp_block[0x1B]; block[0x14] = temp_block[0x14]; + block[0x0D] = temp_block[0x0D]; block[0x06] = temp_block[0x06]; + block[0x07] = temp_block[0x07]; block[0x0E] = temp_block[0x0E]; + block[0x15] = temp_block[0x15]; block[0x1C] = temp_block[0x1C]; if(last_non_zero_p1 <= 32) goto end; - block[0x23] = temp_block[0x23]; block[0x2A] = temp_block[0x2A]; - block[0x31] = temp_block[0x31]; block[0x38] = temp_block[0x38]; - block[0x39] = temp_block[0x39]; block[0x32] = temp_block[0x32]; - block[0x2B] = temp_block[0x2B]; block[0x24] = temp_block[0x24]; + block[0x23] = temp_block[0x23]; block[0x2A] = temp_block[0x2A]; + block[0x31] = temp_block[0x31]; block[0x38] = temp_block[0x38]; + block[0x39] = temp_block[0x39]; block[0x32] = temp_block[0x32]; + block[0x2B] = temp_block[0x2B]; block[0x24] = temp_block[0x24]; if(last_non_zero_p1 <= 40) goto end; - block[0x1D] = temp_block[0x1D]; block[0x16] = temp_block[0x16]; - block[0x0F] = temp_block[0x0F]; block[0x17] = temp_block[0x17]; - block[0x1E] = temp_block[0x1E]; block[0x25] = temp_block[0x25]; - block[0x2C] = temp_block[0x2C]; block[0x33] = temp_block[0x33]; + block[0x1D] = temp_block[0x1D]; block[0x16] = temp_block[0x16]; + block[0x0F] = temp_block[0x0F]; block[0x17] = temp_block[0x17]; + block[0x1E] = temp_block[0x1E]; block[0x25] = temp_block[0x25]; + block[0x2C] = temp_block[0x2C]; block[0x33] = temp_block[0x33]; if(last_non_zero_p1 <= 48) goto end; - block[0x3A] = temp_block[0x3A]; block[0x3B] = temp_block[0x3B]; - block[0x34] = temp_block[0x34]; block[0x2D] = temp_block[0x2D]; - block[0x26] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; - block[0x27] = temp_block[0x27]; block[0x2E] = temp_block[0x2E]; + block[0x3A] = temp_block[0x3A]; block[0x3B] = temp_block[0x3B]; + block[0x34] = temp_block[0x34]; block[0x2D] = temp_block[0x2D]; + block[0x26] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; + block[0x27] = temp_block[0x27]; block[0x2E] = temp_block[0x2E]; if(last_non_zero_p1 <= 56) goto end; - block[0x35] = temp_block[0x35]; block[0x3C] = temp_block[0x3C]; - block[0x3D] = temp_block[0x3D]; block[0x36] = temp_block[0x36]; - block[0x2F] = temp_block[0x2F]; block[0x37] = temp_block[0x37]; + block[0x35] = temp_block[0x35]; block[0x3C] = temp_block[0x3C]; + block[0x3D] = temp_block[0x3D]; block[0x36] = temp_block[0x36]; + block[0x2F] = temp_block[0x2F]; block[0x37] = temp_block[0x37]; block[0x3E] = temp_block[0x3E]; block[0x3F] = temp_block[0x3F]; } end: diff --git a/src/libffmpeg/libavcodec/i386/simple_idct_mmx.c b/src/libffmpeg/libavcodec/i386/simple_idct_mmx.c index 92a366f21..b033a12b8 100644 --- a/src/libffmpeg/libavcodec/i386/simple_idct_mmx.c +++ b/src/libffmpeg/libavcodec/i386/simple_idct_mmx.c @@ -15,7 +15,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include "../dsputil.h" #include "../simple_idct.h" @@ -40,8 +40,8 @@ #define C4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) - 0.5 #endif #define C5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 -#define C6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 -#define C7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 +#define C6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 +#define C7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 #define ROW_SHIFT 11 #define COL_SHIFT 20 // 6 @@ -50,37 +50,37 @@ static const uint64_t attribute_used __attribute__((aligned(8))) wm1010= 0xFFFF0 static const uint64_t attribute_used __attribute__((aligned(8))) d40000= 0x0000000000040000ULL; static const int16_t __attribute__((aligned(8))) coeffs[]= { - 1<<(ROW_SHIFT-1), 0, 1<<(ROW_SHIFT-1), 0, -// 1<<(COL_SHIFT-1), 0, 1<<(COL_SHIFT-1), 0, -// 0, 1<<(COL_SHIFT-1-16), 0, 1<<(COL_SHIFT-1-16), - 1<<(ROW_SHIFT-1), 1, 1<<(ROW_SHIFT-1), 0, - // the 1 = ((1<<(COL_SHIFT-1))/C4)<<ROW_SHIFT :) -// 0, 0, 0, 0, -// 0, 0, 0, 0, + 1<<(ROW_SHIFT-1), 0, 1<<(ROW_SHIFT-1), 0, +// 1<<(COL_SHIFT-1), 0, 1<<(COL_SHIFT-1), 0, +// 0, 1<<(COL_SHIFT-1-16), 0, 1<<(COL_SHIFT-1-16), + 1<<(ROW_SHIFT-1), 1, 1<<(ROW_SHIFT-1), 0, + // the 1 = ((1<<(COL_SHIFT-1))/C4)<<ROW_SHIFT :) +// 0, 0, 0, 0, +// 0, 0, 0, 0, C4, C4, C4, C4, C4, -C4, C4, -C4, - + C2, C6, C2, C6, C6, -C2, C6, -C2, - + C1, C3, C1, C3, C5, C7, C5, C7, - + C3, -C7, C3, -C7, -C1, -C5, -C1, -C5, - + C5, -C1, C5, -C1, C7, C3, C7, C3, - + C7, -C5, C7, -C5, C3, -C1, C3, -C1 }; #if 0 static void unused_var_killer(){ - int a= wm1010 + d40000; - temp[0]=a; + int a= wm1010 + d40000; + temp[0]=a; } static void inline idctCol (int16_t * col, int16_t *input) @@ -93,21 +93,21 @@ static void inline idctCol (int16_t * col, int16_t *input) #undef C5 #undef C6 #undef C7 - int a0, a1, a2, a3, b0, b1, b2, b3; - const int C0 = 23170; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C1 = 22725; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C2 = 21407; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C3 = 19266; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C4 = 16383; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C5 = 12873; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C6 = 8867; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C7 = 4520; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + int a0, a1, a2, a3, b0, b1, b2, b3; + const int C0 = 23170; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C1 = 22725; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C2 = 21407; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C3 = 19266; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C4 = 16383; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C5 = 12873; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C6 = 8867; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C7 = 4520; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 /* - if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) { - col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] = - col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3; - return; - }*/ + if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) { + col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] = + col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3; + return; + }*/ col[8*0] = input[8*0 + 0]; col[8*1] = input[8*2 + 0]; @@ -118,39 +118,39 @@ col[8*5] = input[8*6 + 0]; col[8*6] = input[8*4 + 1]; col[8*7] = input[8*6 + 1]; - a0 = C4*col[8*0] + C2*col[8*2] + C4*col[8*4] + C6*col[8*6] + (1<<(COL_SHIFT-1)); - a1 = C4*col[8*0] + C6*col[8*2] - C4*col[8*4] - C2*col[8*6] + (1<<(COL_SHIFT-1)); - a2 = C4*col[8*0] - C6*col[8*2] - C4*col[8*4] + C2*col[8*6] + (1<<(COL_SHIFT-1)); - a3 = C4*col[8*0] - C2*col[8*2] + C4*col[8*4] - C6*col[8*6] + (1<<(COL_SHIFT-1)); - - b0 = C1*col[8*1] + C3*col[8*3] + C5*col[8*5] + C7*col[8*7]; - b1 = C3*col[8*1] - C7*col[8*3] - C1*col[8*5] - C5*col[8*7]; - b2 = C5*col[8*1] - C1*col[8*3] + C7*col[8*5] + C3*col[8*7]; - b3 = C7*col[8*1] - C5*col[8*3] + C3*col[8*5] - C1*col[8*7]; - - col[8*0] = (a0 + b0) >> COL_SHIFT; - col[8*1] = (a1 + b1) >> COL_SHIFT; - col[8*2] = (a2 + b2) >> COL_SHIFT; - col[8*3] = (a3 + b3) >> COL_SHIFT; - col[8*4] = (a3 - b3) >> COL_SHIFT; - col[8*5] = (a2 - b2) >> COL_SHIFT; - col[8*6] = (a1 - b1) >> COL_SHIFT; - col[8*7] = (a0 - b0) >> COL_SHIFT; + a0 = C4*col[8*0] + C2*col[8*2] + C4*col[8*4] + C6*col[8*6] + (1<<(COL_SHIFT-1)); + a1 = C4*col[8*0] + C6*col[8*2] - C4*col[8*4] - C2*col[8*6] + (1<<(COL_SHIFT-1)); + a2 = C4*col[8*0] - C6*col[8*2] - C4*col[8*4] + C2*col[8*6] + (1<<(COL_SHIFT-1)); + a3 = C4*col[8*0] - C2*col[8*2] + C4*col[8*4] - C6*col[8*6] + (1<<(COL_SHIFT-1)); + + b0 = C1*col[8*1] + C3*col[8*3] + C5*col[8*5] + C7*col[8*7]; + b1 = C3*col[8*1] - C7*col[8*3] - C1*col[8*5] - C5*col[8*7]; + b2 = C5*col[8*1] - C1*col[8*3] + C7*col[8*5] + C3*col[8*7]; + b3 = C7*col[8*1] - C5*col[8*3] + C3*col[8*5] - C1*col[8*7]; + + col[8*0] = (a0 + b0) >> COL_SHIFT; + col[8*1] = (a1 + b1) >> COL_SHIFT; + col[8*2] = (a2 + b2) >> COL_SHIFT; + col[8*3] = (a3 + b3) >> COL_SHIFT; + col[8*4] = (a3 - b3) >> COL_SHIFT; + col[8*5] = (a2 - b2) >> COL_SHIFT; + col[8*6] = (a1 - b1) >> COL_SHIFT; + col[8*7] = (a0 - b0) >> COL_SHIFT; } static void inline idctRow (int16_t * output, int16_t * input) { - int16_t row[8]; - - int a0, a1, a2, a3, b0, b1, b2, b3; - const int C0 = 23170; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C1 = 22725; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C2 = 21407; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C3 = 19266; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C4 = 16383; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C5 = 12873; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C6 = 8867; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 - const int C7 = 4520; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + int16_t row[8]; + + int a0, a1, a2, a3, b0, b1, b2, b3; + const int C0 = 23170; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C1 = 22725; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C2 = 21407; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C3 = 19266; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C4 = 16383; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C5 = 12873; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C6 = 8867; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 + const int C7 = 4520; //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 row[0] = input[0]; row[2] = input[1]; @@ -161,290 +161,290 @@ row[3] = input[9]; row[5] = input[12]; row[7] = input[13]; - if( !(row[1] | row[2] |row[3] |row[4] |row[5] |row[6] | row[7]) ) { - row[0] = row[1] = row[2] = row[3] = row[4] = - row[5] = row[6] = row[7] = row[0]<<3; - output[0] = row[0]; - output[2] = row[1]; - output[4] = row[2]; - output[6] = row[3]; - output[8] = row[4]; - output[10] = row[5]; - output[12] = row[6]; - output[14] = row[7]; - return; - } - - a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6] + (1<<(ROW_SHIFT-1)); - a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6] + (1<<(ROW_SHIFT-1)); - a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6] + (1<<(ROW_SHIFT-1)); - a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6] + (1<<(ROW_SHIFT-1)); - - b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7]; - b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7]; - b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7]; - b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7]; - - row[0] = (a0 + b0) >> ROW_SHIFT; - row[1] = (a1 + b1) >> ROW_SHIFT; - row[2] = (a2 + b2) >> ROW_SHIFT; - row[3] = (a3 + b3) >> ROW_SHIFT; - row[4] = (a3 - b3) >> ROW_SHIFT; - row[5] = (a2 - b2) >> ROW_SHIFT; - row[6] = (a1 - b1) >> ROW_SHIFT; - row[7] = (a0 - b0) >> ROW_SHIFT; - - output[0] = row[0]; - output[2] = row[1]; - output[4] = row[2]; - output[6] = row[3]; - output[8] = row[4]; - output[10] = row[5]; - output[12] = row[6]; - output[14] = row[7]; + if( !(row[1] | row[2] |row[3] |row[4] |row[5] |row[6] | row[7]) ) { + row[0] = row[1] = row[2] = row[3] = row[4] = + row[5] = row[6] = row[7] = row[0]<<3; + output[0] = row[0]; + output[2] = row[1]; + output[4] = row[2]; + output[6] = row[3]; + output[8] = row[4]; + output[10] = row[5]; + output[12] = row[6]; + output[14] = row[7]; + return; + } + + a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6] + (1<<(ROW_SHIFT-1)); + a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6] + (1<<(ROW_SHIFT-1)); + a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6] + (1<<(ROW_SHIFT-1)); + a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6] + (1<<(ROW_SHIFT-1)); + + b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7]; + b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7]; + b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7]; + b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7]; + + row[0] = (a0 + b0) >> ROW_SHIFT; + row[1] = (a1 + b1) >> ROW_SHIFT; + row[2] = (a2 + b2) >> ROW_SHIFT; + row[3] = (a3 + b3) >> ROW_SHIFT; + row[4] = (a3 - b3) >> ROW_SHIFT; + row[5] = (a2 - b2) >> ROW_SHIFT; + row[6] = (a1 - b1) >> ROW_SHIFT; + row[7] = (a0 - b0) >> ROW_SHIFT; + + output[0] = row[0]; + output[2] = row[1]; + output[4] = row[2]; + output[6] = row[3]; + output[8] = row[4]; + output[10] = row[5]; + output[12] = row[6]; + output[14] = row[7]; } #endif static inline void idct(int16_t *block) { - int64_t __attribute__((aligned(8))) align_tmp[16]; - int16_t * const temp= (int16_t*)align_tmp; + int64_t __attribute__((aligned(8))) align_tmp[16]; + int16_t * const temp= (int16_t*)align_tmp; - asm volatile( + asm volatile( #if 0 //Alternative, simpler variant #define ROW_IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - #rounder ", %%mm0 \n\t"\ - "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ - "paddd %%mm0, %%mm0 \n\t" \ - "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ - "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ - "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm1 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ - "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ - "movq %%mm7, " #dst " \n\t"\ - "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "movq %%mm2, 24+" #dst " \n\t"\ - "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ - "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm0 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ - "movq %%mm2, 8+" #dst " \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ - "movq %%mm4, 16+" #dst " \n\t"\ + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + #rounder ", %%mm0 \n\t"\ + "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ + "paddd %%mm0, %%mm0 \n\t" \ + "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ + "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ + "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm1 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ + "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ + "movq %%mm7, " #dst " \n\t"\ + "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "movq %%mm2, 24+" #dst " \n\t"\ + "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ + "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm0 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ + "movq %%mm2, 8+" #dst " \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ + "movq %%mm4, 16+" #dst " \n\t"\ #define COL_IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - #rounder ", %%mm0 \n\t"\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ - "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ - "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm1, %%mm7 \n\t" /* B0 b0 */\ - "movq 72(%2), %%mm1 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "paddd %%mm2, %%mm1 \n\t" /* B1 b1 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm1, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm0 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "movd %%mm7, " #dst " \n\t"\ - "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "movd %%mm0, 16+" #dst " \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "movd %%mm2, 96+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "movd %%mm4, 112+" #dst " \n\t"\ - "movq " #src1 ", %%mm0 \n\t" /* R3 R1 r3 r1 */\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd 96(%2), %%mm0 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ - "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm5 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm0, %%mm3 \n\t" /* B3 b3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "movd %%mm2, 32+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ - "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ - "movd %%mm6, 48+" #dst " \n\t"\ - "movd %%mm4, 64+" #dst " \n\t"\ - "movd %%mm5, 80+" #dst " \n\t"\ - - + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + #rounder ", %%mm0 \n\t"\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ + "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ + "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm1, %%mm7 \n\t" /* B0 b0 */\ + "movq 72(%2), %%mm1 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "paddd %%mm2, %%mm1 \n\t" /* B1 b1 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm1, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm0 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "movd %%mm7, " #dst " \n\t"\ + "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "movd %%mm0, 16+" #dst " \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "movd %%mm2, 96+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "movd %%mm4, 112+" #dst " \n\t"\ + "movq " #src1 ", %%mm0 \n\t" /* R3 R1 r3 r1 */\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd 96(%2), %%mm0 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ + "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm5 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm0, %%mm3 \n\t" /* B3 b3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "movd %%mm2, 32+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ + "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ + "movd %%mm6, 48+" #dst " \n\t"\ + "movd %%mm4, 64+" #dst " \n\t"\ + "movd %%mm5, 80+" #dst " \n\t"\ + + #define DC_COND_ROW_IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq "MANGLE(wm1010)", %%mm4 \n\t"\ - "pand %%mm0, %%mm4 \n\t"\ - "por %%mm1, %%mm4 \n\t"\ - "por %%mm2, %%mm4 \n\t"\ - "por %%mm3, %%mm4 \n\t"\ - "packssdw %%mm4,%%mm4 \n\t"\ - "movd %%mm4, %%eax \n\t"\ - "orl %%eax, %%eax \n\t"\ - "jz 1f \n\t"\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - #rounder ", %%mm0 \n\t"\ - "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ - "paddd %%mm0, %%mm0 \n\t" \ - "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ - "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ - "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm1 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ - "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ - "movq %%mm7, " #dst " \n\t"\ - "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "movq %%mm2, 24+" #dst " \n\t"\ - "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ - "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm0 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ - "movq %%mm2, 8+" #dst " \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ - "movq %%mm4, 16+" #dst " \n\t"\ - "jmp 2f \n\t"\ - "1: \n\t"\ - "pslld $16, %%mm0 \n\t"\ - "#paddd "MANGLE(d40000)", %%mm0 \n\t"\ - "psrad $13, %%mm0 \n\t"\ - "packssdw %%mm0, %%mm0 \n\t"\ - "movq %%mm0, " #dst " \n\t"\ - "movq %%mm0, 8+" #dst " \n\t"\ - "movq %%mm0, 16+" #dst " \n\t"\ - "movq %%mm0, 24+" #dst " \n\t"\ - "2: \n\t" + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq "MANGLE(wm1010)", %%mm4 \n\t"\ + "pand %%mm0, %%mm4 \n\t"\ + "por %%mm1, %%mm4 \n\t"\ + "por %%mm2, %%mm4 \n\t"\ + "por %%mm3, %%mm4 \n\t"\ + "packssdw %%mm4,%%mm4 \n\t"\ + "movd %%mm4, %%eax \n\t"\ + "orl %%eax, %%eax \n\t"\ + "jz 1f \n\t"\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + #rounder ", %%mm0 \n\t"\ + "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ + "paddd %%mm0, %%mm0 \n\t" \ + "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ + "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ + "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm1 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ + "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ + "movq %%mm7, " #dst " \n\t"\ + "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "movq %%mm2, 24+" #dst " \n\t"\ + "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ + "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm0 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ + "movq %%mm2, 8+" #dst " \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ + "movq %%mm4, 16+" #dst " \n\t"\ + "jmp 2f \n\t"\ + "1: \n\t"\ + "pslld $16, %%mm0 \n\t"\ + "#paddd "MANGLE(d40000)", %%mm0 \n\t"\ + "psrad $13, %%mm0 \n\t"\ + "packssdw %%mm0, %%mm0 \n\t"\ + "movq %%mm0, " #dst " \n\t"\ + "movq %%mm0, 8+" #dst " \n\t"\ + "movq %%mm0, 16+" #dst " \n\t"\ + "movq %%mm0, 24+" #dst " \n\t"\ + "2: \n\t" //IDCT( src0, src4, src1, src5, dst, rounder, shift) @@ -467,236 +467,236 @@ COL_IDCT( 24(%1), 88(%1), 56(%1), 120(%1), 12(%0),/nop, 20) #else #define DC_COND_IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq "MANGLE(wm1010)", %%mm4 \n\t"\ - "pand %%mm0, %%mm4 \n\t"\ - "por %%mm1, %%mm4 \n\t"\ - "por %%mm2, %%mm4 \n\t"\ - "por %%mm3, %%mm4 \n\t"\ - "packssdw %%mm4,%%mm4 \n\t"\ - "movd %%mm4, %%eax \n\t"\ - "orl %%eax, %%eax \n\t"\ - "jz 1f \n\t"\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - #rounder ", %%mm0 \n\t"\ - "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ - "paddd %%mm0, %%mm0 \n\t" \ - "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ - "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ - "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm1 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ - "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ - "movq %%mm7, " #dst " \n\t"\ - "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "movq %%mm2, 24+" #dst " \n\t"\ - "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ - "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm0 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ - "movq %%mm2, 8+" #dst " \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ - "movq %%mm4, 16+" #dst " \n\t"\ - "jmp 2f \n\t"\ - "1: \n\t"\ - "pslld $16, %%mm0 \n\t"\ - "paddd "MANGLE(d40000)", %%mm0 \n\t"\ - "psrad $13, %%mm0 \n\t"\ - "packssdw %%mm0, %%mm0 \n\t"\ - "movq %%mm0, " #dst " \n\t"\ - "movq %%mm0, 8+" #dst " \n\t"\ - "movq %%mm0, 16+" #dst " \n\t"\ - "movq %%mm0, 24+" #dst " \n\t"\ - "2: \n\t" + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq "MANGLE(wm1010)", %%mm4 \n\t"\ + "pand %%mm0, %%mm4 \n\t"\ + "por %%mm1, %%mm4 \n\t"\ + "por %%mm2, %%mm4 \n\t"\ + "por %%mm3, %%mm4 \n\t"\ + "packssdw %%mm4,%%mm4 \n\t"\ + "movd %%mm4, %%eax \n\t"\ + "orl %%eax, %%eax \n\t"\ + "jz 1f \n\t"\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + #rounder ", %%mm0 \n\t"\ + "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ + "paddd %%mm0, %%mm0 \n\t" \ + "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ + "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ + "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm1 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ + "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ + "movq %%mm7, " #dst " \n\t"\ + "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "movq %%mm2, 24+" #dst " \n\t"\ + "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ + "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm0 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ + "movq %%mm2, 8+" #dst " \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ + "movq %%mm4, 16+" #dst " \n\t"\ + "jmp 2f \n\t"\ + "1: \n\t"\ + "pslld $16, %%mm0 \n\t"\ + "paddd "MANGLE(d40000)", %%mm0 \n\t"\ + "psrad $13, %%mm0 \n\t"\ + "packssdw %%mm0, %%mm0 \n\t"\ + "movq %%mm0, " #dst " \n\t"\ + "movq %%mm0, 8+" #dst " \n\t"\ + "movq %%mm0, 16+" #dst " \n\t"\ + "movq %%mm0, 24+" #dst " \n\t"\ + "2: \n\t" #define Z_COND_IDCT(src0, src4, src1, src5, dst, rounder, shift, bt) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq %%mm0, %%mm4 \n\t"\ - "por %%mm1, %%mm4 \n\t"\ - "por %%mm2, %%mm4 \n\t"\ - "por %%mm3, %%mm4 \n\t"\ - "packssdw %%mm4,%%mm4 \n\t"\ - "movd %%mm4, %%eax \n\t"\ - "orl %%eax, %%eax \n\t"\ - "jz " #bt " \n\t"\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - #rounder ", %%mm0 \n\t"\ - "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ - "paddd %%mm0, %%mm0 \n\t" \ - "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ - "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ - "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm1 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ - "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ - "movq %%mm7, " #dst " \n\t"\ - "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "movq %%mm2, 24+" #dst " \n\t"\ - "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ - "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm0 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ - "movq %%mm2, 8+" #dst " \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ - "movq %%mm4, 16+" #dst " \n\t"\ + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq %%mm0, %%mm4 \n\t"\ + "por %%mm1, %%mm4 \n\t"\ + "por %%mm2, %%mm4 \n\t"\ + "por %%mm3, %%mm4 \n\t"\ + "packssdw %%mm4,%%mm4 \n\t"\ + "movd %%mm4, %%eax \n\t"\ + "orl %%eax, %%eax \n\t"\ + "jz " #bt " \n\t"\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + #rounder ", %%mm0 \n\t"\ + "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ + "paddd %%mm0, %%mm0 \n\t" \ + "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ + "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ + "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm1 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ + "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ + "movq %%mm7, " #dst " \n\t"\ + "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "movq %%mm2, 24+" #dst " \n\t"\ + "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ + "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm0 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ + "movq %%mm2, 8+" #dst " \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ + "movq %%mm4, 16+" #dst " \n\t"\ #define ROW_IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - #rounder ", %%mm0 \n\t"\ - "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ - "paddd %%mm0, %%mm0 \n\t" \ - "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ - "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ - "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm1 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ - "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ - "movq %%mm7, " #dst " \n\t"\ - "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "movq %%mm2, 24+" #dst " \n\t"\ - "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ - "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm0 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ - "movq %%mm2, 8+" #dst " \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ - "movq %%mm4, 16+" #dst " \n\t"\ + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq 56(%2), %%mm5 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + #rounder ", %%mm0 \n\t"\ + "paddd %%mm0, %%mm1 \n\t" /* A1 a1 */\ + "paddd %%mm0, %%mm0 \n\t" \ + "psubd %%mm1, %%mm0 \n\t" /* A2 a2 */\ + "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm5, %%mm7 \n\t" /* B0 b0 */\ + "movq 72(%2), %%mm5 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm5 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "paddd %%mm2, %%mm5 \n\t" /* B1 b1 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm1, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm5, %%mm1 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm5, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm1 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm1, %%mm7 \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0 */\ + "packssdw %%mm4, %%mm2 \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1 */\ + "movq %%mm7, " #dst " \n\t"\ + "movq " #src1 ", %%mm1 \n\t" /* R3 R1 r3 r1 */\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "movq %%mm2, 24+" #dst " \n\t"\ + "pmaddwd %%mm1, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd 96(%2), %%mm1 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm0, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ + "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm0 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm0 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm1, %%mm3 \n\t" /* B3 b3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "packssdw %%mm6, %%mm2 \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2 */\ + "movq %%mm2, 8+" #dst " \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm0, %%mm4 \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3 */\ + "movq %%mm4, 16+" #dst " \n\t"\ //IDCT( src0, src4, src1, src5, dst, rounder, shift) DC_COND_IDCT( 0(%0), 8(%0), 16(%0), 24(%0), 0(%1),paddd 8(%2), 11) @@ -706,80 +706,80 @@ Z_COND_IDCT( 96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11, 1f) #undef IDCT #define IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - #rounder ", %%mm0 \n\t"\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ - "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ - "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm1, %%mm7 \n\t" /* B0 b0 */\ - "movq 72(%2), %%mm1 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "paddd %%mm2, %%mm1 \n\t" /* B1 b1 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm1, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm0 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "movd %%mm7, " #dst " \n\t"\ - "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "movd %%mm0, 16+" #dst " \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "movd %%mm2, 96+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "movd %%mm4, 112+" #dst " \n\t"\ - "movq " #src1 ", %%mm0 \n\t" /* R3 R1 r3 r1 */\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd 96(%2), %%mm0 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ - "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm5 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm0, %%mm3 \n\t" /* B3 b3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "movd %%mm2, 32+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ - "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ - "movd %%mm6, 48+" #dst " \n\t"\ - "movd %%mm4, 64+" #dst " \n\t"\ - "movd %%mm5, 80+" #dst " \n\t" + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + #rounder ", %%mm0 \n\t"\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ + "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ + "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm1, %%mm7 \n\t" /* B0 b0 */\ + "movq 72(%2), %%mm1 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "paddd %%mm2, %%mm1 \n\t" /* B1 b1 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm1, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm0 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "movd %%mm7, " #dst " \n\t"\ + "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "movd %%mm0, 16+" #dst " \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "movd %%mm2, 96+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "movd %%mm4, 112+" #dst " \n\t"\ + "movq " #src1 ", %%mm0 \n\t" /* R3 R1 r3 r1 */\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd 96(%2), %%mm0 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ + "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm5 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm0, %%mm3 \n\t" /* B3 b3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "movd %%mm2, 32+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ + "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ + "movd %%mm6, 48+" #dst " \n\t"\ + "movd %%mm4, 64+" #dst " \n\t"\ + "movd %%mm5, 80+" #dst " \n\t" //IDCT( src0, src4, src1, src5, dst, rounder, shift) @@ -787,144 +787,144 @@ IDCT( (%1), 64(%1), 32(%1), 96(%1), 0(%0),/nop, 20) IDCT( 8(%1), 72(%1), 40(%1), 104(%1), 4(%0),/nop, 20) IDCT( 16(%1), 80(%1), 48(%1), 112(%1), 8(%0),/nop, 20) IDCT( 24(%1), 88(%1), 56(%1), 120(%1), 12(%0),/nop, 20) - "jmp 9f \n\t" + "jmp 9f \n\t" - "#.balign 16 \n\t"\ - "4: \n\t" + "#.balign 16 \n\t"\ + "4: \n\t" Z_COND_IDCT( 64(%0), 72(%0), 80(%0), 88(%0), 64(%1),paddd (%2), 11, 6f) Z_COND_IDCT( 96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11, 5f) #undef IDCT #define IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - #rounder ", %%mm0 \n\t"\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ - "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ - "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - "movq 72(%2), %%mm7 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm1 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm1, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "psrad $" #shift ", %%mm1 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm7, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm7, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm0 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm1, %%mm1 \n\t" /* A0+B0 a0+b0 */\ - "movd %%mm1, " #dst " \n\t"\ - "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "movd %%mm0, 16+" #dst " \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "movd %%mm2, 96+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "movd %%mm4, 112+" #dst " \n\t"\ - "movq 88(%2), %%mm1 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm1, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm1, %%mm5 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm5 \n\t"\ - "movq %%mm6, %%mm1 \n\t" /* A3 a3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm1 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "psrad $" #shift ", %%mm1 \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "movd %%mm2, 32+" #dst " \n\t"\ - "packssdw %%mm1, %%mm1 \n\t" /* A3-B3 a3-b3 */\ - "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ - "movd %%mm6, 48+" #dst " \n\t"\ - "movd %%mm1, 64+" #dst " \n\t"\ - "movd %%mm5, 80+" #dst " \n\t" + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + #rounder ", %%mm0 \n\t"\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ + "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ + "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + "movq 72(%2), %%mm7 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm1 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm1, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "psrad $" #shift ", %%mm1 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm7, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm7, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm0 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm1, %%mm1 \n\t" /* A0+B0 a0+b0 */\ + "movd %%mm1, " #dst " \n\t"\ + "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "movd %%mm0, 16+" #dst " \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "movd %%mm2, 96+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "movd %%mm4, 112+" #dst " \n\t"\ + "movq 88(%2), %%mm1 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm1, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm1, %%mm5 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm5 \n\t"\ + "movq %%mm6, %%mm1 \n\t" /* A3 a3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm1 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "psrad $" #shift ", %%mm1 \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "movd %%mm2, 32+" #dst " \n\t"\ + "packssdw %%mm1, %%mm1 \n\t" /* A3-B3 a3-b3 */\ + "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ + "movd %%mm6, 48+" #dst " \n\t"\ + "movd %%mm1, 64+" #dst " \n\t"\ + "movd %%mm5, 80+" #dst " \n\t" //IDCT( src0, src4, src1, src5, dst, rounder, shift) IDCT( (%1), 64(%1), 32(%1), 96(%1), 0(%0),/nop, 20) IDCT( 8(%1), 72(%1), 40(%1), 104(%1), 4(%0),/nop, 20) IDCT( 16(%1), 80(%1), 48(%1), 112(%1), 8(%0),/nop, 20) IDCT( 24(%1), 88(%1), 56(%1), 120(%1), 12(%0),/nop, 20) - "jmp 9f \n\t" + "jmp 9f \n\t" - "#.balign 16 \n\t"\ - "6: \n\t" + "#.balign 16 \n\t"\ + "6: \n\t" Z_COND_IDCT( 96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11, 7f) #undef IDCT #define IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - #rounder ", %%mm0 \n\t"\ - "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - "movq 72(%2), %%mm7 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm1 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm1, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "psrad $" #shift ", %%mm1 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm7, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm7, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm0 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm1, %%mm1 \n\t" /* A0+B0 a0+b0 */\ - "movd %%mm1, " #dst " \n\t"\ - "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "movd %%mm0, 16+" #dst " \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "movd %%mm2, 96+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "movd %%mm4, 112+" #dst " \n\t"\ - "movq 88(%2), %%mm1 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm1, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm1, %%mm5 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm5 \n\t"\ - "movq %%mm6, %%mm1 \n\t" /* A3 a3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm1 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "psrad $" #shift ", %%mm1 \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "movd %%mm2, 32+" #dst " \n\t"\ - "packssdw %%mm1, %%mm1 \n\t" /* A3-B3 a3-b3 */\ - "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ - "movd %%mm6, 48+" #dst " \n\t"\ - "movd %%mm1, 64+" #dst " \n\t"\ - "movd %%mm5, 80+" #dst " \n\t" + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + #rounder ", %%mm0 \n\t"\ + "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + "movq 72(%2), %%mm7 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm1 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm1, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "psrad $" #shift ", %%mm1 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm7, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm7, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm0 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm1, %%mm1 \n\t" /* A0+B0 a0+b0 */\ + "movd %%mm1, " #dst " \n\t"\ + "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "movd %%mm0, 16+" #dst " \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "movd %%mm2, 96+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "movd %%mm4, 112+" #dst " \n\t"\ + "movq 88(%2), %%mm1 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm1, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm1, %%mm5 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm5 \n\t"\ + "movq %%mm6, %%mm1 \n\t" /* A3 a3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm1 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "psrad $" #shift ", %%mm1 \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "movd %%mm2, 32+" #dst " \n\t"\ + "packssdw %%mm1, %%mm1 \n\t" /* A3-B3 a3-b3 */\ + "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ + "movd %%mm6, 48+" #dst " \n\t"\ + "movd %%mm1, 64+" #dst " \n\t"\ + "movd %%mm5, 80+" #dst " \n\t" //IDCT( src0, src4, src1, src5, dst, rounder, shift) @@ -932,144 +932,144 @@ IDCT( (%1), 64(%1), 32(%1), 96(%1), 0(%0),/nop, 20) IDCT( 8(%1), 72(%1), 40(%1), 104(%1), 4(%0),/nop, 20) IDCT( 16(%1), 80(%1), 48(%1), 112(%1), 8(%0),/nop, 20) IDCT( 24(%1), 88(%1), 56(%1), 120(%1), 12(%0),/nop, 20) - "jmp 9f \n\t" + "jmp 9f \n\t" - "#.balign 16 \n\t"\ - "2: \n\t" + "#.balign 16 \n\t"\ + "2: \n\t" Z_COND_IDCT( 96(%0),104(%0),112(%0),120(%0), 96(%1),paddd (%2), 11, 3f) #undef IDCT #define IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - #rounder ", %%mm0 \n\t"\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ - "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm1, %%mm7 \n\t" /* B0 b0 */\ - "movq 72(%2), %%mm1 \n\t" /* -C5 -C1 -C5 -C1 */\ - "pmaddwd %%mm3, %%mm1 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "paddd %%mm2, %%mm1 \n\t" /* B1 b1 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm1, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm0 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "movd %%mm7, " #dst " \n\t"\ - "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "movd %%mm0, 16+" #dst " \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ - "movd %%mm2, 96+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "movd %%mm4, 112+" #dst " \n\t"\ - "movq " #src1 ", %%mm0 \n\t" /* R3 R1 r3 r1 */\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ - "pmaddwd 96(%2), %%mm0 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ - "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ - "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ - "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ - "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm5 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm0, %%mm3 \n\t" /* B3 b3 */\ - "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ - "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "movd %%mm2, 32+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ - "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ - "movd %%mm6, 48+" #dst " \n\t"\ - "movd %%mm4, 64+" #dst " \n\t"\ - "movd %%mm5, 80+" #dst " \n\t" + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + #rounder ", %%mm0 \n\t"\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 56(%2), %%mm1 \n\t" /* C7 C5 C7 C5 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* C7R7+C5R5 C7r7+C5r5 */\ + "pmaddwd 64(%2), %%mm2 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm1, %%mm7 \n\t" /* B0 b0 */\ + "movq 72(%2), %%mm1 \n\t" /* -C5 -C1 -C5 -C1 */\ + "pmaddwd %%mm3, %%mm1 \n\t" /* -C5R7-C1R5 -C5r7-C1r5 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "paddd %%mm2, %%mm1 \n\t" /* B1 b1 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm0, %%mm2 \n\t" /* A1 a1 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm1, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm0 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "movd %%mm7, " #dst " \n\t"\ + "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "movd %%mm0, 16+" #dst " \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A1-B1 a1-b1 */\ + "movd %%mm2, 96+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "movd %%mm4, 112+" #dst " \n\t"\ + "movq " #src1 ", %%mm0 \n\t" /* R3 R1 r3 r1 */\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "movq 88(%2), %%mm7 \n\t" /* C3 C7 C3 C7 */\ + "pmaddwd 96(%2), %%mm0 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C3R7+C7R5 C3r7+C7r5 */\ + "movq %%mm5, %%mm2 \n\t" /* A2 a2 */\ + "pmaddwd 104(%2), %%mm3 \n\t" /* -C1R7+C3R5 -C1r7+C3r5 */\ + "paddd %%mm7, %%mm4 \n\t" /* B2 b2 */\ + "paddd %%mm4, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm5 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm0, %%mm3 \n\t" /* B3 b3 */\ + "paddd %%mm3, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm3, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm2, %%mm2 \n\t" /* A2+B2 a2+b2 */\ + "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "movd %%mm2, 32+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ + "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ + "movd %%mm6, 48+" #dst " \n\t"\ + "movd %%mm4, 64+" #dst " \n\t"\ + "movd %%mm5, 80+" #dst " \n\t" //IDCT( src0, src4, src1, src5, dst, rounder, shift) IDCT( (%1), 64(%1), 32(%1), 96(%1), 0(%0),/nop, 20) IDCT( 8(%1), 72(%1), 40(%1), 104(%1), 4(%0),/nop, 20) IDCT( 16(%1), 80(%1), 48(%1), 112(%1), 8(%0),/nop, 20) IDCT( 24(%1), 88(%1), 56(%1), 120(%1), 12(%0),/nop, 20) - "jmp 9f \n\t" + "jmp 9f \n\t" - "#.balign 16 \n\t"\ - "3: \n\t" + "#.balign 16 \n\t"\ + "3: \n\t" #undef IDCT #define IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - #rounder ", %%mm0 \n\t"\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 64(%2), %%mm3 \n\t"\ - "pmaddwd %%mm2, %%mm3 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm0, %%mm1 \n\t" /* A1 a1 */\ - "paddd %%mm3, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm3, %%mm1 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm0 \n\t"\ - "psrad $" #shift ", %%mm1 \n\t"\ - "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "movd %%mm7, " #dst " \n\t"\ - "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "movd %%mm0, 16+" #dst " \n\t"\ - "packssdw %%mm1, %%mm1 \n\t" /* A1-B1 a1-b1 */\ - "movd %%mm1, 96+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "movd %%mm4, 112+" #dst " \n\t"\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "pmaddwd %%mm2, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "pmaddwd 96(%2), %%mm2 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "movq %%mm5, %%mm1 \n\t" /* A2 a2 */\ - "paddd %%mm4, %%mm1 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm1 \n\t"\ - "psrad $" #shift ", %%mm5 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm2, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm2, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm1, %%mm1 \n\t" /* A2+B2 a2+b2 */\ - "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "movd %%mm1, 32+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ - "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ - "movd %%mm6, 48+" #dst " \n\t"\ - "movd %%mm4, 64+" #dst " \n\t"\ - "movd %%mm5, 80+" #dst " \n\t" + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + #rounder ", %%mm0 \n\t"\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 64(%2), %%mm3 \n\t"\ + "pmaddwd %%mm2, %%mm3 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm0, %%mm1 \n\t" /* A1 a1 */\ + "paddd %%mm3, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm3, %%mm1 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm0 \n\t"\ + "psrad $" #shift ", %%mm1 \n\t"\ + "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "movd %%mm7, " #dst " \n\t"\ + "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "movd %%mm0, 16+" #dst " \n\t"\ + "packssdw %%mm1, %%mm1 \n\t" /* A1-B1 a1-b1 */\ + "movd %%mm1, 96+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "movd %%mm4, 112+" #dst " \n\t"\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "pmaddwd %%mm2, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "pmaddwd 96(%2), %%mm2 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "movq %%mm5, %%mm1 \n\t" /* A2 a2 */\ + "paddd %%mm4, %%mm1 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm1 \n\t"\ + "psrad $" #shift ", %%mm5 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm2, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm2, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm1, %%mm1 \n\t" /* A2+B2 a2+b2 */\ + "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "movd %%mm1, 32+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ + "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ + "movd %%mm6, 48+" #dst " \n\t"\ + "movd %%mm4, 64+" #dst " \n\t"\ + "movd %%mm5, 80+" #dst " \n\t" //IDCT( src0, src4, src1, src5, dst, rounder, shift) @@ -1077,186 +1077,186 @@ IDCT( (%1), 64(%1), 32(%1), 96(%1), 0(%0),/nop, 20) IDCT( 8(%1), 72(%1), 40(%1), 104(%1), 4(%0),/nop, 20) IDCT( 16(%1), 80(%1), 48(%1), 112(%1), 8(%0),/nop, 20) IDCT( 24(%1), 88(%1), 56(%1), 120(%1), 12(%0),/nop, 20) - "jmp 9f \n\t" + "jmp 9f \n\t" - "#.balign 16 \n\t"\ - "5: \n\t" + "#.balign 16 \n\t"\ + "5: \n\t" #undef IDCT #define IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - #rounder ", %%mm0 \n\t"\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ - "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ - "movq 8+" #src0 ", %%mm2 \n\t" /* R4 R0 r4 r0 */\ - "movq 8+" #src4 ", %%mm3 \n\t" /* R6 R2 r6 r2 */\ - "movq 16(%2), %%mm1 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm2, %%mm1 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm7 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm7, %%mm2 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm7 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm3, %%mm7 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "pmaddwd 40(%2), %%mm3 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - #rounder ", %%mm1 \n\t"\ - "paddd %%mm1, %%mm7 \n\t" /* A0 a0 */\ - "paddd %%mm1, %%mm1 \n\t" /* 2C0 2c0 */\ - #rounder ", %%mm2 \n\t"\ - "psubd %%mm7, %%mm1 \n\t" /* A3 a3 */\ - "paddd %%mm2, %%mm3 \n\t" /* A1 a1 */\ - "paddd %%mm2, %%mm2 \n\t" /* 2C1 2c1 */\ - "psubd %%mm3, %%mm2 \n\t" /* A2 a2 */\ - "psrad $" #shift ", %%mm4 \n\t"\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm3 \n\t"\ - "packssdw %%mm7, %%mm4 \n\t" /* A0 a0 */\ - "movq %%mm4, " #dst " \n\t"\ - "psrad $" #shift ", %%mm0 \n\t"\ - "packssdw %%mm3, %%mm0 \n\t" /* A1 a1 */\ - "movq %%mm0, 16+" #dst " \n\t"\ - "movq %%mm0, 96+" #dst " \n\t"\ - "movq %%mm4, 112+" #dst " \n\t"\ - "psrad $" #shift ", %%mm5 \n\t"\ - "psrad $" #shift ", %%mm6 \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm2, %%mm5 \n\t" /* A2-B2 a2-b2 */\ - "movq %%mm5, 32+" #dst " \n\t"\ - "psrad $" #shift ", %%mm1 \n\t"\ - "packssdw %%mm1, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "movq %%mm6, 48+" #dst " \n\t"\ - "movq %%mm6, 64+" #dst " \n\t"\ - "movq %%mm5, 80+" #dst " \n\t" - + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + #rounder ", %%mm0 \n\t"\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ + "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ + "movq 8+" #src0 ", %%mm2 \n\t" /* R4 R0 r4 r0 */\ + "movq 8+" #src4 ", %%mm3 \n\t" /* R6 R2 r6 r2 */\ + "movq 16(%2), %%mm1 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm2, %%mm1 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm7 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm7, %%mm2 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm7 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm3, %%mm7 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "pmaddwd 40(%2), %%mm3 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + #rounder ", %%mm1 \n\t"\ + "paddd %%mm1, %%mm7 \n\t" /* A0 a0 */\ + "paddd %%mm1, %%mm1 \n\t" /* 2C0 2c0 */\ + #rounder ", %%mm2 \n\t"\ + "psubd %%mm7, %%mm1 \n\t" /* A3 a3 */\ + "paddd %%mm2, %%mm3 \n\t" /* A1 a1 */\ + "paddd %%mm2, %%mm2 \n\t" /* 2C1 2c1 */\ + "psubd %%mm3, %%mm2 \n\t" /* A2 a2 */\ + "psrad $" #shift ", %%mm4 \n\t"\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm3 \n\t"\ + "packssdw %%mm7, %%mm4 \n\t" /* A0 a0 */\ + "movq %%mm4, " #dst " \n\t"\ + "psrad $" #shift ", %%mm0 \n\t"\ + "packssdw %%mm3, %%mm0 \n\t" /* A1 a1 */\ + "movq %%mm0, 16+" #dst " \n\t"\ + "movq %%mm0, 96+" #dst " \n\t"\ + "movq %%mm4, 112+" #dst " \n\t"\ + "psrad $" #shift ", %%mm5 \n\t"\ + "psrad $" #shift ", %%mm6 \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm2, %%mm5 \n\t" /* A2-B2 a2-b2 */\ + "movq %%mm5, 32+" #dst " \n\t"\ + "psrad $" #shift ", %%mm1 \n\t"\ + "packssdw %%mm1, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "movq %%mm6, 48+" #dst " \n\t"\ + "movq %%mm6, 64+" #dst " \n\t"\ + "movq %%mm5, 80+" #dst " \n\t" + //IDCT( src0, src4, src1, src5, dst, rounder, shift) IDCT( 0(%1), 64(%1), 32(%1), 96(%1), 0(%0),/nop, 20) //IDCT( 8(%1), 72(%1), 40(%1), 104(%1), 4(%0),/nop, 20) IDCT( 16(%1), 80(%1), 48(%1), 112(%1), 8(%0),/nop, 20) //IDCT( 24(%1), 88(%1), 56(%1), 120(%1), 12(%0),/nop, 20) - "jmp 9f \n\t" + "jmp 9f \n\t" - "#.balign 16 \n\t"\ - "1: \n\t" + "#.balign 16 \n\t"\ + "1: \n\t" #undef IDCT #define IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ - "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ - "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ - "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ - "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ - #rounder ", %%mm4 \n\t"\ - "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ - #rounder ", %%mm0 \n\t"\ - "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ - "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ - "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ - "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ - "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ - "movq 64(%2), %%mm1 \n\t"\ - "pmaddwd %%mm2, %%mm1 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ - "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ - "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "psrad $" #shift ", %%mm7 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "movq %%mm0, %%mm3 \n\t" /* A1 a1 */\ - "paddd %%mm1, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "psubd %%mm1, %%mm3 \n\t" /* A1-B1 a1-b1 */\ - "psrad $" #shift ", %%mm0 \n\t"\ - "psrad $" #shift ", %%mm3 \n\t"\ - "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ - "movd %%mm7, " #dst " \n\t"\ - "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ - "movd %%mm0, 16+" #dst " \n\t"\ - "packssdw %%mm3, %%mm3 \n\t" /* A1-B1 a1-b1 */\ - "movd %%mm3, 96+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ - "movd %%mm4, 112+" #dst " \n\t"\ - "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ - "pmaddwd %%mm2, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ - "pmaddwd 96(%2), %%mm2 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ - "movq %%mm5, %%mm3 \n\t" /* A2 a2 */\ - "paddd %%mm4, %%mm3 \n\t" /* A2+B2 a2+b2 */\ - "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ - "psrad $" #shift ", %%mm3 \n\t"\ - "psrad $" #shift ", %%mm5 \n\t"\ - "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ - "paddd %%mm2, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "psubd %%mm2, %%mm4 \n\t" /* a3-B3 a3-b3 */\ - "psrad $" #shift ", %%mm6 \n\t"\ - "packssdw %%mm3, %%mm3 \n\t" /* A2+B2 a2+b2 */\ - "movd %%mm3, 32+" #dst " \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ - "movd %%mm6, 48+" #dst " \n\t"\ - "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ - "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ - "movd %%mm4, 64+" #dst " \n\t"\ - "movd %%mm5, 80+" #dst " \n\t" - + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\ + "movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm5 \n\t" /* C6 C2 C6 C2 */\ + "pmaddwd %%mm1, %%mm5 \n\t" /* C6R6+C2R2 C6r6+C2r2 */\ + "movq 40(%2), %%mm6 \n\t" /* -C2 C6 -C2 C6 */\ + "pmaddwd %%mm6, %%mm1 \n\t" /* -C2R6+C6R2 -C2r6+C6r2 */\ + #rounder ", %%mm4 \n\t"\ + "movq %%mm4, %%mm6 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 48(%2), %%mm7 \n\t" /* C3 C1 C3 C1 */\ + #rounder ", %%mm0 \n\t"\ + "pmaddwd %%mm2, %%mm7 \n\t" /* C3R3+C1R1 C3r3+C1r1 */\ + "paddd %%mm5, %%mm4 \n\t" /* A0 a0 */\ + "psubd %%mm5, %%mm6 \n\t" /* A3 a3 */\ + "movq %%mm0, %%mm5 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1 a1 */\ + "psubd %%mm1, %%mm5 \n\t" /* A2 a2 */\ + "movq 64(%2), %%mm1 \n\t"\ + "pmaddwd %%mm2, %%mm1 \n\t" /* -C7R3+C3R1 -C7r3+C3r1 */\ + "paddd %%mm4, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "paddd %%mm4, %%mm4 \n\t" /* 2A0 2a0 */\ + "psubd %%mm7, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "psrad $" #shift ", %%mm7 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "movq %%mm0, %%mm3 \n\t" /* A1 a1 */\ + "paddd %%mm1, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "psubd %%mm1, %%mm3 \n\t" /* A1-B1 a1-b1 */\ + "psrad $" #shift ", %%mm0 \n\t"\ + "psrad $" #shift ", %%mm3 \n\t"\ + "packssdw %%mm7, %%mm7 \n\t" /* A0+B0 a0+b0 */\ + "movd %%mm7, " #dst " \n\t"\ + "packssdw %%mm0, %%mm0 \n\t" /* A1+B1 a1+b1 */\ + "movd %%mm0, 16+" #dst " \n\t"\ + "packssdw %%mm3, %%mm3 \n\t" /* A1-B1 a1-b1 */\ + "movd %%mm3, 96+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A0-B0 a0-b0 */\ + "movd %%mm4, 112+" #dst " \n\t"\ + "movq 80(%2), %%mm4 \n\t" /* -C1 C5 -C1 C5 */\ + "pmaddwd %%mm2, %%mm4 \n\t" /* -C1R3+C5R1 -C1r3+C5r1 */\ + "pmaddwd 96(%2), %%mm2 \n\t" /* -C5R3+C7R1 -C5r3+C7r1 */\ + "movq %%mm5, %%mm3 \n\t" /* A2 a2 */\ + "paddd %%mm4, %%mm3 \n\t" /* A2+B2 a2+b2 */\ + "psubd %%mm4, %%mm5 \n\t" /* a2-B2 a2-b2 */\ + "psrad $" #shift ", %%mm3 \n\t"\ + "psrad $" #shift ", %%mm5 \n\t"\ + "movq %%mm6, %%mm4 \n\t" /* A3 a3 */\ + "paddd %%mm2, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "psubd %%mm2, %%mm4 \n\t" /* a3-B3 a3-b3 */\ + "psrad $" #shift ", %%mm6 \n\t"\ + "packssdw %%mm3, %%mm3 \n\t" /* A2+B2 a2+b2 */\ + "movd %%mm3, 32+" #dst " \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "packssdw %%mm6, %%mm6 \n\t" /* A3+B3 a3+b3 */\ + "movd %%mm6, 48+" #dst " \n\t"\ + "packssdw %%mm4, %%mm4 \n\t" /* A3-B3 a3-b3 */\ + "packssdw %%mm5, %%mm5 \n\t" /* A2-B2 a2-b2 */\ + "movd %%mm4, 64+" #dst " \n\t"\ + "movd %%mm5, 80+" #dst " \n\t" + //IDCT( src0, src4, src1, src5, dst, rounder, shift) IDCT( (%1), 64(%1), 32(%1), 96(%1), 0(%0),/nop, 20) IDCT( 8(%1), 72(%1), 40(%1), 104(%1), 4(%0),/nop, 20) IDCT( 16(%1), 80(%1), 48(%1), 112(%1), 8(%0),/nop, 20) IDCT( 24(%1), 88(%1), 56(%1), 120(%1), 12(%0),/nop, 20) - "jmp 9f \n\t" + "jmp 9f \n\t" - "#.balign 16 \n\t" - "7: \n\t" + "#.balign 16 \n\t" + "7: \n\t" #undef IDCT #define IDCT(src0, src4, src1, src5, dst, rounder, shift) \ - "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ - "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - #rounder ", %%mm4 \n\t"\ - #rounder ", %%mm0 \n\t"\ - "psrad $" #shift ", %%mm4 \n\t"\ - "psrad $" #shift ", %%mm0 \n\t"\ - "movq 8+" #src0 ", %%mm2 \n\t" /* R4 R0 r4 r0 */\ - "movq 16(%2), %%mm1 \n\t" /* C4 C4 C4 C4 */\ - "pmaddwd %%mm2, %%mm1 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ - "movq 24(%2), %%mm7 \n\t" /* -C4 C4 -C4 C4 */\ - "pmaddwd %%mm7, %%mm2 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ - "movq 32(%2), %%mm7 \n\t" /* C6 C2 C6 C2 */\ - #rounder ", %%mm1 \n\t"\ - #rounder ", %%mm2 \n\t"\ - "psrad $" #shift ", %%mm1 \n\t"\ - "packssdw %%mm1, %%mm4 \n\t" /* A0 a0 */\ - "movq %%mm4, " #dst " \n\t"\ - "psrad $" #shift ", %%mm2 \n\t"\ - "packssdw %%mm2, %%mm0 \n\t" /* A1 a1 */\ - "movq %%mm0, 16+" #dst " \n\t"\ - "movq %%mm0, 96+" #dst " \n\t"\ - "movq %%mm4, 112+" #dst " \n\t"\ - "movq %%mm0, 32+" #dst " \n\t"\ - "movq %%mm4, 48+" #dst " \n\t"\ - "movq %%mm4, 64+" #dst " \n\t"\ - "movq %%mm0, 80+" #dst " \n\t" + "movq " #src0 ", %%mm0 \n\t" /* R4 R0 r4 r0 */\ + "movq 16(%2), %%mm4 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm0, %%mm4 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm5 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm5, %%mm0 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + #rounder ", %%mm4 \n\t"\ + #rounder ", %%mm0 \n\t"\ + "psrad $" #shift ", %%mm4 \n\t"\ + "psrad $" #shift ", %%mm0 \n\t"\ + "movq 8+" #src0 ", %%mm2 \n\t" /* R4 R0 r4 r0 */\ + "movq 16(%2), %%mm1 \n\t" /* C4 C4 C4 C4 */\ + "pmaddwd %%mm2, %%mm1 \n\t" /* C4R4+C4R0 C4r4+C4r0 */\ + "movq 24(%2), %%mm7 \n\t" /* -C4 C4 -C4 C4 */\ + "pmaddwd %%mm7, %%mm2 \n\t" /* -C4R4+C4R0 -C4r4+C4r0 */\ + "movq 32(%2), %%mm7 \n\t" /* C6 C2 C6 C2 */\ + #rounder ", %%mm1 \n\t"\ + #rounder ", %%mm2 \n\t"\ + "psrad $" #shift ", %%mm1 \n\t"\ + "packssdw %%mm1, %%mm4 \n\t" /* A0 a0 */\ + "movq %%mm4, " #dst " \n\t"\ + "psrad $" #shift ", %%mm2 \n\t"\ + "packssdw %%mm2, %%mm0 \n\t" /* A1 a1 */\ + "movq %%mm0, 16+" #dst " \n\t"\ + "movq %%mm0, 96+" #dst " \n\t"\ + "movq %%mm4, 112+" #dst " \n\t"\ + "movq %%mm0, 32+" #dst " \n\t"\ + "movq %%mm4, 48+" #dst " \n\t"\ + "movq %%mm4, 64+" #dst " \n\t"\ + "movq %%mm0, 80+" #dst " \n\t" //IDCT( src0, src4, src1, src5, dst, rounder, shift) IDCT( 0(%1), 64(%1), 32(%1), 96(%1), 0(%0),/nop, 20) @@ -1277,7 +1277,7 @@ Input 12 32 16 36 52 72 56 76 05 45 07 47 25 65 27 67 15 35 17 37 55 75 57 77 - + Temp 00 04 10 14 20 24 30 34 40 44 50 54 60 64 70 74 @@ -1290,9 +1290,9 @@ Temp */ "9: \n\t" - :: "r" (block), "r" (temp), "r" (coeffs) - : "%eax" - ); + :: "r" (block), "r" (temp), "r" (coeffs) + : "%eax" + ); } void ff_simple_idct_mmx(int16_t *block) diff --git a/src/libffmpeg/libavcodec/i386/vp3dsp_mmx.c b/src/libffmpeg/libavcodec/i386/vp3dsp_mmx.c index 3d220c1d4..0684531ae 100644 --- a/src/libffmpeg/libavcodec/i386/vp3dsp_mmx.c +++ b/src/libffmpeg/libavcodec/i386/vp3dsp_mmx.c @@ -13,7 +13,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ /** @@ -208,7 +208,7 @@ static const uint16_t idct_cosine_table[7] = { I(1) = d1 c1 b1 a1 I(2) = d2 c2 b2 a2 I(3) = d3 c3 b3 a3 - + J(4) = h0 g0 f0 e0 J(5) = h1 g1 f1 e1 J(6) = h2 g2 f2 e2 diff --git a/src/libffmpeg/libavcodec/i386/vp3dsp_sse2.c b/src/libffmpeg/libavcodec/i386/vp3dsp_sse2.c index ed17891bf..cf822f7d4 100644 --- a/src/libffmpeg/libavcodec/i386/vp3dsp_sse2.c +++ b/src/libffmpeg/libavcodec/i386/vp3dsp_sse2.c @@ -13,7 +13,7 @@ * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ /** @@ -24,7 +24,7 @@ #include "../dsputil.h" #include "mmx.h" -static const unsigned short __align16 SSE2_dequant_const[] = +static DECLARE_ALIGNED_16(const unsigned short, SSE2_dequant_const[]) = { 0,65535,65535,0,0,0,0,0, // 0x0000 0000 0000 0000 0000 FFFF FFFF 0000 0,0,0,0,65535,65535,0,0, // 0x0000 0000 FFFF FFFF 0000 0000 0000 0000 @@ -35,22 +35,22 @@ static const unsigned short __align16 SSE2_dequant_const[] = 0,0,65535,65535, 0,0,0,0 // 0x0000 0000 0000 0000 FFFF FFFF 0000 0000 }; -static const unsigned int __align16 eight_data[] = -{ - 0x00080008, +static DECLARE_ALIGNED_16(const unsigned int, eight_data[]) = +{ + 0x00080008, + 0x00080008, 0x00080008, - 0x00080008, - 0x00080008 -}; + 0x00080008 +}; -static const unsigned short __align16 SSE2_idct_data[7 * 8] = +static DECLARE_ALIGNED_16(const unsigned short, SSE2_idct_data[7 * 8]) = { - 64277,64277,64277,64277,64277,64277,64277,64277, - 60547,60547,60547,60547,60547,60547,60547,60547, - 54491,54491,54491,54491,54491,54491,54491,54491, - 46341,46341,46341,46341,46341,46341,46341,46341, - 36410,36410,36410,36410,36410,36410,36410,36410, - 25080,25080,25080,25080,25080,25080,25080,25080, + 64277,64277,64277,64277,64277,64277,64277,64277, + 60547,60547,60547,60547,60547,60547,60547,60547, + 54491,54491,54491,54491,54491,54491,54491,54491, + 46341,46341,46341,46341,46341,46341,46341,46341, + 36410,36410,36410,36410,36410,36410,36410,36410, + 25080,25080,25080,25080,25080,25080,25080,25080, 12785,12785,12785,12785,12785,12785,12785,12785 }; @@ -820,6 +820,6 @@ void ff_vp3_idct_sse2(int16_t *input_data) SSE2_Row_IDCT(); SSE2_Transpose(); - + SSE2_Column_IDCT(); } |