Attachment #74758 for bug #115568

Lines 7-13 Link Here

(-)libavcodec/Makefile (-1 / +1 lines)
7	VPATH=$(SRC_PATH)/libavcodec	7	VPATH=$(SRC_PATH)/libavcodec
8		8
9	# NOTE: -I.. is needed to include config.h	9	# NOTE: -I.. is needed to include config.h
10	CFLAGS=$(OPTFLAGS) -DHAVE_AV_CONFIG_H -I.. -I$(SRC_PATH)/libavutil -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE $(AMR_CFLAGS)	10	CFLAGS=$(OPTFLAGS) $(PIC) -DHAVE_AV_CONFIG_H -I.. -I$(SRC_PATH)/libavutil -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE $(AMR_CFLAGS)
11		11
12	OBJS= bitstream.o utils.o mem.o allcodecs.o \	12	OBJS= bitstream.o utils.o mem.o allcodecs.o \
13	mpegvideo.o jrevdct.o jfdctfst.o jfdctint.o\	13	mpegvideo.o jrevdct.o jfdctfst.o jfdctint.o\

Lines 613-648 Link Here

(-)libavcodec/i386/dsputil_mmx.c (-70 / +71 lines)
613	"+m" ((uint64_t)(src - 1*stride)),	613	"+m" ((uint64_t)(src - 1*stride)),
614	"+m" ((uint64_t)(src + 0*stride)),	614	"+m" ((uint64_t)(src + 0*stride)),
615	"+m" ((uint64_t)(src + 1*stride))	615	"+m" ((uint64_t)(src + 1*stride))
616	: "g" (2*strength), "m"(ff_pb_FC)	616	: "g" (2*(long)strength), "m"(ff_pb_FC)
617	);	617	);
618	}	618	}
619		619
620	static inline void transpose4x4(uint8_t dst, uint8_t src, int dst_stride, int src_stride){	620	static inline void transpose4x4(uint8_t dst, uint8_t src, int dst_stride, int src_stride){
		621	long dummy;
621	asm volatile( //FIXME could save 1 instruction if done as 8x4 ...	622	asm volatile( //FIXME could save 1 instruction if done as 8x4 ...
622	"movd %4, %%mm0 \n\t"	623	"movd (%3), %%mm0 \n\t"
623	"movd %5, %%mm1 \n\t"	624	"movd (%3, %4), %%mm1 \n\t"
624	"movd %6, %%mm2 \n\t"	625	"movd (%3, %4, 2), %%mm2 \n\t"
625	"movd %7, %%mm3 \n\t"	626	"lea (%4, %4, 2), %0 \n\t"
		627	"movd (%3, %0), %%mm3 \n\t"
626	"punpcklbw %%mm1, %%mm0 \n\t"	628	"punpcklbw %%mm1, %%mm0 \n\t"
627	"punpcklbw %%mm3, %%mm2 \n\t"	629	"punpcklbw %%mm3, %%mm2 \n\t"
628	"movq %%mm0, %%mm1 \n\t"	630	"movq %%mm0, %%mm1 \n\t"
629	"punpcklwd %%mm2, %%mm0 \n\t"	631	"punpcklwd %%mm2, %%mm0 \n\t"
630	"punpckhwd %%mm2, %%mm1 \n\t"	632	"punpckhwd %%mm2, %%mm1 \n\t"
631	"movd %%mm0, %0 \n\t"	633	"movd %%mm0, (%1) \n\t"
632	"punpckhdq %%mm0, %%mm0 \n\t"	634	"punpckhdq %%mm0, %%mm0 \n\t"
633	"movd %%mm0, %1 \n\t"	635	"movd %%mm0, (%1, %2) \n\t"
634	"movd %%mm1, %2 \n\t"	636	"movd %%mm1, (%1, %2, 2) \n\t"
635	"punpckhdq %%mm1, %%mm1 \n\t"	637	"punpckhdq %%mm1, %%mm1 \n\t"
636	"movd %%mm1, %3 \n\t"	638	"lea (%2, %2, 2), %0 \n\t"
		639	"movd %%mm1, (%1, %0) \n\t"
637		640
638	: "=m" ((uint32_t)(dst + 0*dst_stride)),	641	: "=&r" (dummy)
639	"=m" ((uint32_t)(dst + 1*dst_stride)),	642	: "r" (dst),
640	"=m" ((uint32_t)(dst + 2*dst_stride)),	643	"r" ((long)dst_stride),
641	"=m" ((uint32_t)(dst + 3*dst_stride))	644	"r" (src),
642	: "m" ((uint32_t)(src + 0*src_stride)),	645	"r" ((long)src_stride)
643	"m" ((uint32_t)(src + 1*src_stride)),	646	: "memory"
644	"m" ((uint32_t)(src + 2*src_stride)),
645	"m" ((uint32_t)(src + 3*src_stride))
646	);	647	);
647	}	648	}
648		649
Lines 662-668 Link Here
662	"+m" (temp[1]),	663	"+m" (temp[1]),
663	"+m" (temp[2]),	664	"+m" (temp[2]),
664	"+m" (temp[3])	665	"+m" (temp[3])
665	: "g" (2*strength), "m"(ff_pb_FC)	666	: "g" (2*(long)strength), "m"(ff_pb_FC)
666	);	667	);
667		668
668	asm volatile(	669	asm volatile(
Lines 1727-1733 Link Here
1727		1728
1728	#define QPEL_V_LOW(m3,m4,m5,m6, pw_20, pw_3, rnd, in0, in1, in2, in7, out, OP)\	1729	#define QPEL_V_LOW(m3,m4,m5,m6, pw_20, pw_3, rnd, in0, in1, in2, in7, out, OP)\
1729	"paddw " #m4 ", " #m3 " \n\t" /* x1 */\	1730	"paddw " #m4 ", " #m3 " \n\t" /* x1 */\
1730	"movq "MANGLE(ff_pw_20)", %%mm4 \n\t" /* 20 */\	1731	"movq %5, %%mm4 \n\t" /* 20 */\
1731	"pmullw " #m3 ", %%mm4 \n\t" /* 20x1 */\	1732	"pmullw " #m3 ", %%mm4 \n\t" /* 20x1 */\
1732	"movq "#in7", " #m3 " \n\t" /* d */\	1733	"movq "#in7", " #m3 " \n\t" /* d */\
1733	"movq "#in0", %%mm5 \n\t" /* D */\	1734	"movq "#in0", %%mm5 \n\t" /* D */\
Lines 1739-1745 Link Here
1739	"paddw " #m5 ", %%mm6 \n\t" /* x2 */\	1740	"paddw " #m5 ", %%mm6 \n\t" /* x2 */\
1740	"paddw %%mm6, %%mm6 \n\t" /* 2x2 */\	1741	"paddw %%mm6, %%mm6 \n\t" /* 2x2 */\
1741	"psubw %%mm6, %%mm5 \n\t" /* -2x2 + x3 */\	1742	"psubw %%mm6, %%mm5 \n\t" /* -2x2 + x3 */\
1742	"pmullw "MANGLE(ff_pw_3)", %%mm5 \n\t" /* -6x2 + 3x3 */\	1743	"pmullw %6, %%mm5 \n\t" /* -6x2 + 3x3 */\
1743	"paddw " #rnd ", %%mm4 \n\t" /* x2 */\	1744	"paddw " #rnd ", %%mm4 \n\t" /* x2 */\
1744	"paddw %%mm4, %%mm5 \n\t" /* 20x1 - 6x2 + 3x3 - x4 */\	1745	"paddw %%mm4, %%mm5 \n\t" /* 20x1 - 6x2 + 3x3 - x4 */\
1745	"psraw $5, %%mm5 \n\t"\	1746	"psraw $5, %%mm5 \n\t"\
Lines 1773-1787 Link Here
1773	"paddw %%mm5, %%mm5 \n\t" /* 2b */\	1774	"paddw %%mm5, %%mm5 \n\t" /* 2b */\
1774	"psubw %%mm5, %%mm6 \n\t" /* c - 2b */\	1775	"psubw %%mm5, %%mm6 \n\t" /* c - 2b */\
1775	"pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\	1776	"pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\
1776	"pmullw "MANGLE(ff_pw_3)", %%mm6 \n\t" /* 3c - 6b */\	1777	"pmullw %6, %%mm6 \n\t" /* 3c - 6b */\
1777	"paddw %%mm4, %%mm0 \n\t" /* a */\	1778	"paddw %%mm4, %%mm0 \n\t" /* a */\
1778	"paddw %%mm1, %%mm5 \n\t" /* d */\	1779	"paddw %%mm1, %%mm5 \n\t" /* d */\
1779	"pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\	1780	"pmullw %5, %%mm0 \n\t" /* 20a */\
1780	"psubw %%mm5, %%mm0 \n\t" /* 20a - d */\	1781	"psubw %%mm5, %%mm0 \n\t" /* 20a - d */\
1781	"paddw %6, %%mm6 \n\t"\	1782	"paddw %8, %%mm6 \n\t"\
1782	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\	1783	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\
1783	"psraw $5, %%mm0 \n\t"\	1784	"psraw $5, %%mm0 \n\t"\
1784	"movq %%mm0, %5 \n\t"\	1785	"movq %%mm0, %7 \n\t"\
1785	/* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\	1786	/* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\
1786	\	1787	\
1787	"movq 5(%0), %%mm0 \n\t" /* FGHIJKLM */\	1788	"movq 5(%0), %%mm0 \n\t" /* FGHIJKLM */\
Lines 1799-1813 Link Here
1799	"psrlq $24, %%mm6 \n\t" /* IJKLM000 */\	1800	"psrlq $24, %%mm6 \n\t" /* IJKLM000 */\
1800	"punpcklbw %%mm7, %%mm2 \n\t" /* 0F0G0H0I */\	1801	"punpcklbw %%mm7, %%mm2 \n\t" /* 0F0G0H0I */\
1801	"punpcklbw %%mm7, %%mm6 \n\t" /* 0I0J0K0L */\	1802	"punpcklbw %%mm7, %%mm6 \n\t" /* 0I0J0K0L */\
1802	"pmullw "MANGLE(ff_pw_3)", %%mm3 \n\t" /* 3c - 6b */\	1803	"pmullw %6, %%mm3 \n\t" /* 3c - 6b */\
1803	"paddw %%mm2, %%mm1 \n\t" /* a */\	1804	"paddw %%mm2, %%mm1 \n\t" /* a */\
1804	"paddw %%mm6, %%mm4 \n\t" /* d */\	1805	"paddw %%mm6, %%mm4 \n\t" /* d */\
1805	"pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\	1806	"pmullw %5, %%mm1 \n\t" /* 20a */\
1806	"psubw %%mm4, %%mm3 \n\t" /* - 6b +3c - d */\	1807	"psubw %%mm4, %%mm3 \n\t" /* - 6b +3c - d */\
1807	"paddw %6, %%mm1 \n\t"\	1808	"paddw %8, %%mm1 \n\t"\
1808	"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b +3c - d */\	1809	"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b +3c - d */\
1809	"psraw $5, %%mm3 \n\t"\	1810	"psraw $5, %%mm3 \n\t"\
1810	"movq %5, %%mm1 \n\t"\	1811	"movq %7, %%mm1 \n\t"\
1811	"packuswb %%mm3, %%mm1 \n\t"\	1812	"packuswb %%mm3, %%mm1 \n\t"\
1812	OP_MMX2(%%mm1, (%1),%%mm4, q)\	1813	OP_MMX2(%%mm1, (%1),%%mm4, q)\
1813	/* mm0= GHIJ, mm2=FGHI, mm5=HIJK, mm6=IJKL, mm7=0 */\	1814	/* mm0= GHIJ, mm2=FGHI, mm5=HIJK, mm6=IJKL, mm7=0 */\
Lines 1825-1831 Link Here
1825	"psubw %%mm5, %%mm0 \n\t" /* c - 2b */\	1826	"psubw %%mm5, %%mm0 \n\t" /* c - 2b */\
1826	"movq %%mm3, %%mm5 \n\t" /* JKLMNOPQ */\	1827	"movq %%mm3, %%mm5 \n\t" /* JKLMNOPQ */\
1827	"psrlq $24, %%mm3 \n\t" /* MNOPQ000 */\	1828	"psrlq $24, %%mm3 \n\t" /* MNOPQ000 */\
1828	"pmullw "MANGLE(ff_pw_3)", %%mm0 \n\t" /* 3c - 6b */\	1829	"pmullw %6, %%mm0 \n\t" /* 3c - 6b */\
1829	"punpcklbw %%mm7, %%mm3 \n\t" /* 0M0N0O0P */\	1830	"punpcklbw %%mm7, %%mm3 \n\t" /* 0M0N0O0P */\
1830	"paddw %%mm3, %%mm2 \n\t" /* d */\	1831	"paddw %%mm3, %%mm2 \n\t" /* d */\
1831	"psubw %%mm2, %%mm0 \n\t" /* -6b + 3c - d */\	1832	"psubw %%mm2, %%mm0 \n\t" /* -6b + 3c - d */\
Lines 1833-1840 Link Here
1833	"punpcklbw %%mm7, %%mm2 \n\t" /* 0J0K0L0M */\	1834	"punpcklbw %%mm7, %%mm2 \n\t" /* 0J0K0L0M */\
1834	"punpckhbw %%mm7, %%mm5 \n\t" /* 0N0O0P0Q */\	1835	"punpckhbw %%mm7, %%mm5 \n\t" /* 0N0O0P0Q */\
1835	"paddw %%mm2, %%mm6 \n\t" /* a */\	1836	"paddw %%mm2, %%mm6 \n\t" /* a */\
1836	"pmullw "MANGLE(ff_pw_20)", %%mm6 \n\t" /* 20a */\	1837	"pmullw %5, %%mm6 \n\t" /* 20a */\
1837	"paddw %6, %%mm0 \n\t"\	1838	"paddw %8, %%mm0 \n\t"\
1838	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\	1839	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\
1839	"psraw $5, %%mm0 \n\t"\	1840	"psraw $5, %%mm0 \n\t"\
1840	/* mm1=KLMN, mm2=JKLM, mm3=MNOP, mm4=LMNO, mm5=NOPQ mm7=0 */\	1841	/* mm1=KLMN, mm2=JKLM, mm3=MNOP, mm4=LMNO, mm5=NOPQ mm7=0 */\
Lines 1848-1857 Link Here
1848	"paddw %%mm2, %%mm5 \n\t" /* d */\	1849	"paddw %%mm2, %%mm5 \n\t" /* d */\
1849	"paddw %%mm6, %%mm6 \n\t" /* 2b */\	1850	"paddw %%mm6, %%mm6 \n\t" /* 2b */\
1850	"psubw %%mm6, %%mm4 \n\t" /* c - 2b */\	1851	"psubw %%mm6, %%mm4 \n\t" /* c - 2b */\
1851	"pmullw "MANGLE(ff_pw_20)", %%mm3 \n\t" /* 20a */\	1852	"pmullw %5, %%mm3 \n\t" /* 20a */\
1852	"pmullw "MANGLE(ff_pw_3)", %%mm4 \n\t" /* 3c - 6b */\	1853	"pmullw %6, %%mm4 \n\t" /* 3c - 6b */\
1853	"psubw %%mm5, %%mm3 \n\t" /* -6b + 3c - d */\	1854	"psubw %%mm5, %%mm3 \n\t" /* -6b + 3c - d */\
1854	"paddw %6, %%mm4 \n\t"\	1855	"paddw %8, %%mm4 \n\t"\
1855	"paddw %%mm3, %%mm4 \n\t" /* 20a - 6b + 3c - d */\	1856	"paddw %%mm3, %%mm4 \n\t" /* 20a - 6b + 3c - d */\
1856	"psraw $5, %%mm4 \n\t"\	1857	"psraw $5, %%mm4 \n\t"\
1857	"packuswb %%mm4, %%mm0 \n\t"\	1858	"packuswb %%mm4, %%mm0 \n\t"\
Lines 1862-1868 Link Here
1862	"decl %2 \n\t"\	1863	"decl %2 \n\t"\
1863	" jnz 1b \n\t"\	1864	" jnz 1b \n\t"\
1864	: "+a"(src), "+c"(dst), "+m"(h)\	1865	: "+a"(src), "+c"(dst), "+m"(h)\
1865	: "d"((long)srcStride), "S"((long)dstStride), /"m"(ff_pw_20), "m"(ff_pw_3),/ "m"(temp), "m"(ROUNDER)\	1866	: "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(temp), "m"(ROUNDER)\
1866	: "memory"\	1867	: "memory"\
1867	);\	1868	);\
1868	}\	1869	}\
Lines 1940-1951 Link Here
1940	"paddw %%mm5, %%mm5 \n\t" /* 2b */\	1941	"paddw %%mm5, %%mm5 \n\t" /* 2b */\
1941	"psubw %%mm5, %%mm6 \n\t" /* c - 2b */\	1942	"psubw %%mm5, %%mm6 \n\t" /* c - 2b */\
1942	"pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\	1943	"pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\
1943	"pmullw "MANGLE(ff_pw_3)", %%mm6 \n\t" /* 3c - 6b */\	1944	"pmullw %6, %%mm6 \n\t" /* 3c - 6b */\
1944	"paddw %%mm4, %%mm0 \n\t" /* a */\	1945	"paddw %%mm4, %%mm0 \n\t" /* a */\
1945	"paddw %%mm1, %%mm5 \n\t" /* d */\	1946	"paddw %%mm1, %%mm5 \n\t" /* d */\
1946	"pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\	1947	"pmullw %5, %%mm0 \n\t" /* 20a */\
1947	"psubw %%mm5, %%mm0 \n\t" /* 20a - d */\	1948	"psubw %%mm5, %%mm0 \n\t" /* 20a - d */\
1948	"paddw %6, %%mm6 \n\t"\	1949	"paddw %8, %%mm6 \n\t"\
1949	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\	1950	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\
1950	"psraw $5, %%mm0 \n\t"\	1951	"psraw $5, %%mm0 \n\t"\
1951	/* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\	1952	/* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\
Lines 1961-1970 Link Here
1961	"paddw %%mm5, %%mm4 \n\t" /* d */\	1962	"paddw %%mm5, %%mm4 \n\t" /* d */\
1962	"paddw %%mm2, %%mm2 \n\t" /* 2b */\	1963	"paddw %%mm2, %%mm2 \n\t" /* 2b */\
1963	"psubw %%mm2, %%mm3 \n\t" /* c - 2b */\	1964	"psubw %%mm2, %%mm3 \n\t" /* c - 2b */\
1964	"pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\	1965	"pmullw %5, %%mm1 \n\t" /* 20a */\
1965	"pmullw "MANGLE(ff_pw_3)", %%mm3 \n\t" /* 3c - 6b */\	1966	"pmullw %6, %%mm3 \n\t" /* 3c - 6b */\
1966	"psubw %%mm4, %%mm3 \n\t" /* -6b + 3c - d */\	1967	"psubw %%mm4, %%mm3 \n\t" /* -6b + 3c - d */\
1967	"paddw %6, %%mm1 \n\t"\	1968	"paddw %8, %%mm1 \n\t"\
1968	"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b + 3c - d */\	1969	"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b + 3c - d */\
1969	"psraw $5, %%mm3 \n\t"\	1970	"psraw $5, %%mm3 \n\t"\
1970	"packuswb %%mm3, %%mm0 \n\t"\	1971	"packuswb %%mm3, %%mm0 \n\t"\
Lines 1975-1981 Link Here
1975	"decl %2 \n\t"\	1976	"decl %2 \n\t"\
1976	" jnz 1b \n\t"\	1977	" jnz 1b \n\t"\
1977	: "+a"(src), "+c"(dst), "+m"(h)\	1978	: "+a"(src), "+c"(dst), "+m"(h)\
1978	: "S"((long)srcStride), "D"((long)dstStride), /"m"(ff_pw_20), "m"(ff_pw_3),/ "m"(temp), "m"(ROUNDER)\	1979	: "S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(temp), "m"(ROUNDER)\
1979	: "memory"\	1980	: "memory"\
1980	);\	1981	);\
1981	}\	1982	}\
Lines 2054-2092 Link Here
2054	"movq 8(%0), %%mm1 \n\t"\	2055	"movq 8(%0), %%mm1 \n\t"\
2055	"movq 16(%0), %%mm2 \n\t"\	2056	"movq 16(%0), %%mm2 \n\t"\
2056	"movq 24(%0), %%mm3 \n\t"\	2057	"movq 24(%0), %%mm3 \n\t"\
2057	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\	2058	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\
2058	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\	2059	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\
2059	"add %4, %1 \n\t"\	2060	"add %4, %1 \n\t"\
2060	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, (%0), (%0), 8(%0), 48(%0), (%1), OP)\	2061	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, (%0), (%0), 8(%0), 48(%0), (%1), OP)\
2061	\	2062	\
2062	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\	2063	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\
2063	"add %4, %1 \n\t"\	2064	"add %4, %1 \n\t"\
2064	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\	2065	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
2065	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\	2066	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\
2066	"add %4, %1 \n\t"\	2067	"add %4, %1 \n\t"\
2067	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\	2068	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\
2068	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\	2069	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\
2069	"add %4, %1 \n\t"\	2070	"add %4, %1 \n\t"\
2070	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\	2071	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\
2071	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\	2072	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\
2072	"add %4, %1 \n\t"\	2073	"add %4, %1 \n\t"\
2073	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\	2074	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\
2074	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\	2075	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\
2075	"add %4, %1 \n\t"\	2076	"add %4, %1 \n\t"\
2076	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\	2077	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\
2077	\	2078	\
2078	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\	2079	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\
2079	"add %4, %1 \n\t" \	2080	"add %4, %1 \n\t" \
2080	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\	2081	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\
2081	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\	2082	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\
2082	\	2083	\
2083	"add $136, %0 \n\t"\	2084	"add $136, %0 \n\t"\
2084	"add %6, %1 \n\t"\	2085	"add %8, %1 \n\t"\
2085	"decl %2 \n\t"\	2086	"decl %2 \n\t"\
2086	" jnz 1b \n\t"\	2087	" jnz 1b \n\t"\
2087	\	2088	\
2088	: "+r"(temp_ptr), "+r"(dst), "+g"(count)\	2089	: "+r"(temp_ptr), "+r"(dst), "+g"(count)\
2089	: "r"((long)dstStride), "r"(2(long)dstStride), /"m"(ff_pw_20), "m"(ff_pw_3),/ "m"(ROUNDER), "g"(4-14(long)dstStride)\	2090	: "r"((long)dstStride), "r"(2(long)dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(ROUNDER), "g"(4-14(long)dstStride)\
2090	:"memory"\	2091	:"memory"\
2091	);\	2092	);\
2092	}\	2093	}\
Lines 2126-2152 Link Here
2126	"movq 8(%0), %%mm1 \n\t"\	2127	"movq 8(%0), %%mm1 \n\t"\
2127	"movq 16(%0), %%mm2 \n\t"\	2128	"movq 16(%0), %%mm2 \n\t"\
2128	"movq 24(%0), %%mm3 \n\t"\	2129	"movq 24(%0), %%mm3 \n\t"\
2129	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\	2130	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\
2130	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\	2131	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\
2131	"add %4, %1 \n\t"\	2132	"add %4, %1 \n\t"\
2132	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, (%0), (%0), 8(%0), 48(%0), (%1), OP)\	2133	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, (%0), (%0), 8(%0), 48(%0), (%1), OP)\
2133	\	2134	\
2134	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\	2135	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\
2135	"add %4, %1 \n\t"\	2136	"add %4, %1 \n\t"\
2136	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\	2137	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
2137	\	2138	\
2138	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 64(%0), (%1, %3), OP)\	2139	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 16(%0), 24(%0), 32(%0), 64(%0), (%1, %3), OP)\
2139	"add %4, %1 \n\t"\	2140	"add %4, %1 \n\t"\
2140	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 56(%0), (%1), OP)\	2141	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, 24(%0), 32(%0), 40(%0), 56(%0), (%1), OP)\
2141	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 48(%0), (%1, %3), OP)\	2142	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, 32(%0), 40(%0), 48(%0), 48(%0), (%1, %3), OP)\
2142	\	2143	\
2143	"add $72, %0 \n\t"\	2144	"add $72, %0 \n\t"\
2144	"add %6, %1 \n\t"\	2145	"add %8, %1 \n\t"\
2145	"decl %2 \n\t"\	2146	"decl %2 \n\t"\
2146	" jnz 1b \n\t"\	2147	" jnz 1b \n\t"\
2147	\	2148	\
2148	: "+r"(temp_ptr), "+r"(dst), "+g"(count)\	2149	: "+r"(temp_ptr), "+r"(dst), "+g"(count)\
2149	: "r"((long)dstStride), "r"(2(long)dstStride), /"m"(ff_pw_20), "m"(ff_pw_3),/ "m"(ROUNDER), "g"(4-6(long)dstStride)\	2150	: "r"((long)dstStride), "r"(2(long)dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(ROUNDER), "g"(4-6(long)dstStride)\
2150	: "memory"\	2151	: "memory"\
2151	);\	2152	);\
2152	}\	2153	}\

Lines 119-125 Link Here

(-)libavcodec/i386/motion_est_mmx.c (-2 / +2 lines)
119	long len= -(stride*h);	119	long len= -(stride*h);
120	asm volatile(	120	asm volatile(
121	".balign 16 \n\t"	121	".balign 16 \n\t"
122	"movq "MANGLE(bone)", %%mm5 \n\t"	122	"movq %5, %%mm5 \n\t"
123	"1: \n\t"	123	"1: \n\t"
124	"movq (%1, %%"REG_a"), %%mm0 \n\t"	124	"movq (%1, %%"REG_a"), %%mm0 \n\t"
125	"movq (%2, %%"REG_a"), %%mm2 \n\t"	125	"movq (%2, %%"REG_a"), %%mm2 \n\t"
Lines 147-153 Link Here
147	"add %4, %%"REG_a" \n\t"	147	"add %4, %%"REG_a" \n\t"
148	" js 1b \n\t"	148	" js 1b \n\t"
149	: "+a" (len)	149	: "+a" (len)
150	: "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), "r" ((long)stride)	150	: "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), "r" ((long)stride), "m" (bone)
151	);	151	);
152	}	152	}
153		153




    if((s->out_format == FMT_H263 || s->out_format == FMT_H261) && s->mpeg_quant==0){
    
        asm volatile(
            "movd %6, %%mm3			\n\t" // last_non_zero_p1
            SPREADW(%%mm3)
            "pxor %%mm7, %%mm7			\n\t" // 0
            "pxor %%mm4, %%mm4			\n\t" // 0

            "add $8, %%"REG_a"			\n\t"
            " js 1b				\n\t"
	    PMAX(%%mm3, %%mm0)
            "movd %%mm3, %0			\n\t"
            "movzb %b0, %0			\n\t" // last_non_zero_p1
	    : "=q" (last_non_zero_p1)
            : "r" (block+64), "r" (qmat), "r" (bias),
              "r" (inv_zigzag_direct16+64), "r" (temp_block+64), "m" (last_non_zero_p1)
        );
        // note the asm is split cuz gcc doesnt like that many operands ...
        asm volatile(

        );
    }else{ // FMT_H263
        asm volatile(
            "movd %6, %%mm3			\n\t" // last_non_zero_p1
            SPREADW(%%mm3)
            "pxor %%mm7, %%mm7			\n\t" // 0
            "pxor %%mm4, %%mm4			\n\t" // 0

            "add $8, %%"REG_a"			\n\t"
            " js 1b				\n\t"
	    PMAX(%%mm3, %%mm0)
            "movd %%mm3, %0			\n\t"
            "movzb %b0, %0			\n\t" // last_non_zero_p1
	    : "=q" (last_non_zero_p1)
            : "r" (block+64), "r" (qmat+64), "r" (bias+64),
              "r" (inv_zigzag_direct16+64), "r" (temp_block+64), "m" (last_non_zero_p1)
        );
        // note the asm is split cuz gcc doesnt like that many operands ...
        asm volatile(

Lines 363-369 Link Here

(-)libavcodec/i386/simple_idct_mmx.c (-4 / +4 lines)
363	"movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\	363	"movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\
364	"movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\	364	"movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\
365	"movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\	365	"movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\
366	"movq "MANGLE(wm1010)", %%mm4 \n\t"\	366	"movq %3, %%mm4 \n\t"\
367	"pand %%mm0, %%mm4 \n\t"\	367	"pand %%mm0, %%mm4 \n\t"\
368	"por %%mm1, %%mm4 \n\t"\	368	"por %%mm1, %%mm4 \n\t"\
369	"por %%mm2, %%mm4 \n\t"\	369	"por %%mm2, %%mm4 \n\t"\
Lines 471-477 Link Here
471	"movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\	471	"movq " #src4 ", %%mm1 \n\t" /* R6 R2 r6 r2 */\
472	"movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\	472	"movq " #src1 ", %%mm2 \n\t" /* R3 R1 r3 r1 */\
473	"movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\	473	"movq " #src5 ", %%mm3 \n\t" /* R7 R5 r7 r5 */\
474	"movq "MANGLE(wm1010)", %%mm4 \n\t"\	474	"movq %3, %%mm4 \n\t"\
475	"pand %%mm0, %%mm4 \n\t"\	475	"pand %%mm0, %%mm4 \n\t"\
476	"por %%mm1, %%mm4 \n\t"\	476	"por %%mm1, %%mm4 \n\t"\
477	"por %%mm2, %%mm4 \n\t"\	477	"por %%mm2, %%mm4 \n\t"\
Lines 545-551 Link Here
545	"jmp 2f \n\t"\	545	"jmp 2f \n\t"\
546	"1: \n\t"\	546	"1: \n\t"\
547	"pslld $16, %%mm0 \n\t"\	547	"pslld $16, %%mm0 \n\t"\
548	"paddd "MANGLE(d40000)", %%mm0 \n\t"\	548	"paddd %4, %%mm0 \n\t"\
549	"psrad $13, %%mm0 \n\t"\	549	"psrad $13, %%mm0 \n\t"\
550	"packssdw %%mm0, %%mm0 \n\t"\	550	"packssdw %%mm0, %%mm0 \n\t"\
551	"movq %%mm0, " #dst " \n\t"\	551	"movq %%mm0, " #dst " \n\t"\
Lines 1290-1296 Link Here
1290	*/	1290	*/
1291		1291
1292	"9: \n\t"	1292	"9: \n\t"
1293	:: "r" (block), "r" (temp), "r" (coeffs)	1293	:: "r" (block), "r" (temp), "r" (coeffs), "m" (wm1010), "m" (d40000)
1294	: "%eax"	1294	: "%eax"
1295	);	1295	);
1296	}	1296	}

Lines 16-25 Link Here

(-)libavcodec/liba52/resample_mmx.c (-23 / +23 lines)
16	int32_t * f = (int32_t *) _f;	16	int32_t * f = (int32_t *) _f;
17	asm volatile(	17	asm volatile(
18	"movl $-512, %%esi \n\t"	18	"movl $-512, %%esi \n\t"
19	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	19	"movq %2, %%mm7 \n\t"
20	"movq "MANGLE(wm1100)", %%mm3 \n\t"	20	"movq %3, %%mm3 \n\t"
21	"movq "MANGLE(wm0101)", %%mm4 \n\t"	21	"movq %4, %%mm4 \n\t"
22	"movq "MANGLE(wm1010)", %%mm5 \n\t"	22	"movq %5, %%mm5 \n\t"
23	"pxor %%mm6, %%mm6 \n\t"	23	"pxor %%mm6, %%mm6 \n\t"
24	"1: \n\t"	24	"1: \n\t"
25	"movq (%1, %%esi, 2), %%mm0 \n\t"	25	"movq (%1, %%esi, 2), %%mm0 \n\t"
Lines 43-49 Link Here
43	"addl $8, %%esi \n\t"	43	"addl $8, %%esi \n\t"
44	" jnz 1b \n\t"	44	" jnz 1b \n\t"
45	"emms \n\t"	45	"emms \n\t"
46	:: "r" (s16+1280), "r" (f+256)	46	:: "r" (s16+1280), "r" (f+256), "m" (magicF2W), "m" (wm1100), "m" (wm0101), "m" (wm1010)
47	:"%esi", "%edi", "memory"	47	:"%esi", "%edi", "memory"
48	);	48	);
49	return 5*256;	49	return 5*256;
Lines 71-77 Link Here
71	);*/	71	);*/
72	asm volatile(	72	asm volatile(
73	"movl $-1024, %%esi \n\t"	73	"movl $-1024, %%esi \n\t"
74	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	74	"movq %2, %%mm7 \n\t"
75	"1: \n\t"	75	"1: \n\t"
76	"movq (%1, %%esi), %%mm0 \n\t"	76	"movq (%1, %%esi), %%mm0 \n\t"
77	"movq 8(%1, %%esi), %%mm1 \n\t"	77	"movq 8(%1, %%esi), %%mm1 \n\t"
Lines 91-97 Link Here
91	"addl $16, %%esi \n\t"	91	"addl $16, %%esi \n\t"
92	" jnz 1b \n\t"	92	" jnz 1b \n\t"
93	"emms \n\t"	93	"emms \n\t"
94	:: "r" (s16+512), "r" (f+256)	94	:: "r" (s16+512), "r" (f+256), "m" (magicF2W)
95	:"%esi", "memory"	95	:"%esi", "memory"
96	);	96	);
97	return 2*256;	97	return 2*256;
Lines 101-107 Link Here
101	int32_t * f = (int32_t *) _f;	101	int32_t * f = (int32_t *) _f;
102	asm volatile(	102	asm volatile(
103	"movl $-1024, %%esi \n\t"	103	"movl $-1024, %%esi \n\t"
104	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	104	"movq %2, %%mm7 \n\t"
105	"pxor %%mm6, %%mm6 \n\t"	105	"pxor %%mm6, %%mm6 \n\t"
106	"movq %%mm7, %%mm5 \n\t"	106	"movq %%mm7, %%mm5 \n\t"
107	"punpckldq %%mm6, %%mm5 \n\t"	107	"punpckldq %%mm6, %%mm5 \n\t"
Lines 146-152 Link Here
146	"addl $16, %%esi \n\t"	146	"addl $16, %%esi \n\t"
147	" jnz 1b \n\t"	147	" jnz 1b \n\t"
148	"emms \n\t"	148	"emms \n\t"
149	:: "r" (s16+1280), "r" (f+256)	149	:: "r" (s16+1280), "r" (f+256), "m" (magicF2W)
150	:"%esi", "%edi", "memory"	150	:"%esi", "%edi", "memory"
151	);	151	);
152	return 5*256;	152	return 5*256;
Lines 156-162 Link Here
156	int32_t * f = (int32_t *) _f;	156	int32_t * f = (int32_t *) _f;
157	asm volatile(	157	asm volatile(
158	"movl $-1024, %%esi \n\t"	158	"movl $-1024, %%esi \n\t"
159	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	159	"movq %2, %%mm7 \n\t"
160	"1: \n\t"	160	"1: \n\t"
161	"movq (%1, %%esi), %%mm0 \n\t"	161	"movq (%1, %%esi), %%mm0 \n\t"
162	"movq 8(%1, %%esi), %%mm1 \n\t"	162	"movq 8(%1, %%esi), %%mm1 \n\t"
Lines 197-203 Link Here
197	"addl $16, %%esi \n\t"	197	"addl $16, %%esi \n\t"
198	" jnz 1b \n\t"	198	" jnz 1b \n\t"
199	"emms \n\t"	199	"emms \n\t"
200	:: "r" (s16+1024), "r" (f+256)	200	:: "r" (s16+1024), "r" (f+256), "m" (magicF2W)
201	:"%esi", "memory"	201	:"%esi", "memory"
202	);	202	);
203	return 4*256;	203	return 4*256;
Lines 207-213 Link Here
207	int32_t * f = (int32_t *) _f;	207	int32_t * f = (int32_t *) _f;
208	asm volatile(	208	asm volatile(
209	"movl $-1024, %%esi \n\t"	209	"movl $-1024, %%esi \n\t"
210	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	210	"movq %2, %%mm7 \n\t"
211	"1: \n\t"	211	"1: \n\t"
212	"movd (%1, %%esi), %%mm0 \n\t"	212	"movd (%1, %%esi), %%mm0 \n\t"
213	"punpckldq 2048(%1, %%esi), %%mm0\n\t"	213	"punpckldq 2048(%1, %%esi), %%mm0\n\t"
Lines 256-262 Link Here
256	"addl $16, %%esi \n\t"	256	"addl $16, %%esi \n\t"
257	" jnz 1b \n\t"	257	" jnz 1b \n\t"
258	"emms \n\t"	258	"emms \n\t"
259	:: "r" (s16+1280), "r" (f+256)	259	:: "r" (s16+1280), "r" (f+256), "m" (magicF2W)
260	:"%esi", "%edi", "memory"	260	:"%esi", "%edi", "memory"
261	);	261	);
262	return 5*256;	262	return 5*256;
Lines 266-272 Link Here
266	int32_t * f = (int32_t *) _f;	266	int32_t * f = (int32_t *) _f;
267	asm volatile(	267	asm volatile(
268	"movl $-1024, %%esi \n\t"	268	"movl $-1024, %%esi \n\t"
269	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	269	"movq %2, %%mm7 \n\t"
270	"pxor %%mm6, %%mm6 \n\t"	270	"pxor %%mm6, %%mm6 \n\t"
271	"1: \n\t"	271	"1: \n\t"
272	"movq 1024(%1, %%esi), %%mm0 \n\t"	272	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 296-302 Link Here
296	"addl $16, %%esi \n\t"	296	"addl $16, %%esi \n\t"
297	" jnz 1b \n\t"	297	" jnz 1b \n\t"
298	"emms \n\t"	298	"emms \n\t"
299	:: "r" (s16+1536), "r" (f+256)	299	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
300	:"%esi", "%edi", "memory"	300	:"%esi", "%edi", "memory"
301	);	301	);
302	return 6*256;	302	return 6*256;
Lines 306-312 Link Here
306	int32_t * f = (int32_t *) _f;	306	int32_t * f = (int32_t *) _f;
307	asm volatile(	307	asm volatile(
308	"movl $-1024, %%esi \n\t"	308	"movl $-1024, %%esi \n\t"
309	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	309	"movq %2, %%mm7 \n\t"
310	"pxor %%mm6, %%mm6 \n\t"	310	"pxor %%mm6, %%mm6 \n\t"
311	"1: \n\t"	311	"1: \n\t"
312	"movq 1024(%1, %%esi), %%mm0 \n\t"	312	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 334-340 Link Here
334	"addl $8, %%esi \n\t"	334	"addl $8, %%esi \n\t"
335	" jnz 1b \n\t"	335	" jnz 1b \n\t"
336	"emms \n\t"	336	"emms \n\t"
337	:: "r" (s16+1536), "r" (f+256)	337	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
338	:"%esi", "%edi", "memory"	338	:"%esi", "%edi", "memory"
339	);	339	);
340	return 6*256;	340	return 6*256;
Lines 344-350 Link Here
344	int32_t * f = (int32_t *) _f;	344	int32_t * f = (int32_t *) _f;
345	asm volatile(	345	asm volatile(
346	"movl $-1024, %%esi \n\t"	346	"movl $-1024, %%esi \n\t"
347	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	347	"movq %2, %%mm7 \n\t"
348	"pxor %%mm6, %%mm6 \n\t"	348	"pxor %%mm6, %%mm6 \n\t"
349	"1: \n\t"	349	"1: \n\t"
350	"movq 1024(%1, %%esi), %%mm0 \n\t"	350	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 374-380 Link Here
374	"addl $8, %%esi \n\t"	374	"addl $8, %%esi \n\t"
375	" jnz 1b \n\t"	375	" jnz 1b \n\t"
376	"emms \n\t"	376	"emms \n\t"
377	:: "r" (s16+1536), "r" (f+256)	377	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
378	:"%esi", "%edi", "memory"	378	:"%esi", "%edi", "memory"
379	);	379	);
380	return 6*256;	380	return 6*256;
Lines 384-390 Link Here
384	int32_t * f = (int32_t *) _f;	384	int32_t * f = (int32_t *) _f;
385	asm volatile(	385	asm volatile(
386	"movl $-1024, %%esi \n\t"	386	"movl $-1024, %%esi \n\t"
387	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	387	"movq %2, %%mm7 \n\t"
388	// "pxor %%mm6, %%mm6 \n\t"	388	// "pxor %%mm6, %%mm6 \n\t"
389	"1: \n\t"	389	"1: \n\t"
390	"movq 1024(%1, %%esi), %%mm0 \n\t"	390	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 420-426 Link Here
420	"addl $8, %%esi \n\t"	420	"addl $8, %%esi \n\t"
421	" jnz 1b \n\t"	421	" jnz 1b \n\t"
422	"emms \n\t"	422	"emms \n\t"
423	:: "r" (s16+1536), "r" (f+256)	423	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
424	:"%esi", "%edi", "memory"	424	:"%esi", "%edi", "memory"
425	);	425	);
426	return 6*256;	426	return 6*256;
Lines 430-436 Link Here
430	int32_t * f = (int32_t *) _f;	430	int32_t * f = (int32_t *) _f;
431	asm volatile(	431	asm volatile(
432	"movl $-1024, %%esi \n\t"	432	"movl $-1024, %%esi \n\t"
433	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	433	"movq %2, %%mm7 \n\t"
434	// "pxor %%mm6, %%mm6 \n\t"	434	// "pxor %%mm6, %%mm6 \n\t"
435	"1: \n\t"	435	"1: \n\t"
436	"movq 1024(%1, %%esi), %%mm0 \n\t"	436	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 468-474 Link Here
468	"addl $8, %%esi \n\t"	468	"addl $8, %%esi \n\t"
469	" jnz 1b \n\t"	469	" jnz 1b \n\t"
470	"emms \n\t"	470	"emms \n\t"
471	:: "r" (s16+1536), "r" (f+256)	471	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
472	:"%esi", "%edi", "memory"	472	:"%esi", "%edi", "memory"
473	);	473	);
474	return 6*256;	474	return 6*256;

Lines 20-26 Link Here

(-)libavcodec/libpostproc/Makefile (-1 / +1 lines)
20	PPOBJS=postprocess.o	20	PPOBJS=postprocess.o
21	SPPOBJS=postprocess_pic.o	21	SPPOBJS=postprocess_pic.o
22		22
23	CFLAGS = $(OPTFLAGS) $(MLIB_INC) -I. -I.. $(EXTRA_INC)	23	CFLAGS = $(OPTFLAGS) $(PIC) $(MLIB_INC) -I. -I.. $(EXTRA_INC)
24	# -I/usr/X11R6/include/	24	# -I/usr/X11R6/include/
25		25
26	.SUFFIXES: .c .o	26	.SUFFIXES: .c .o

Lines 515-521 Link Here

(-)libavcodec/libpostproc/postprocess_template.c (-62 / +57 lines)
515	"paddusb %%mm0, %%mm0 \n\t"	515	"paddusb %%mm0, %%mm0 \n\t"
516	"psubusb %%mm0, %%mm4 \n\t"	516	"psubusb %%mm0, %%mm4 \n\t"
517	"pcmpeqb %%mm7, %%mm4 \n\t" // d <= QP ? -1 : 0	517	"pcmpeqb %%mm7, %%mm4 \n\t" // d <= QP ? -1 : 0
518	"psubusb "MANGLE(b01)", %%mm3 \n\t"	518	"psubusb %3, %%mm3 \n\t"
519	"pand %%mm4, %%mm3 \n\t" // d <= QP ? d : 0	519	"pand %%mm4, %%mm3 \n\t" // d <= QP ? d : 0
520		520
521	PAVGB(%%mm7, %%mm3) // d/2	521	PAVGB(%%mm7, %%mm3) // d/2
Lines 564-570 Link Here
564	"movq %%mm0, (%%"REG_c", %1, 2) \n\t" // line 7	564	"movq %%mm0, (%%"REG_c", %1, 2) \n\t" // line 7
565		565
566	:	566	:
567	: "r" (src), "r" ((long)stride), "m" (co->pQPb)	567	: "r" (src), "r" ((long)stride), "m" (co->pQPb), "m" (b01)
568	: "%"REG_a, "%"REG_c	568	: "%"REG_a, "%"REG_c
569	);	569	);
570	#else	570	#else
Lines 697-713 Link Here
697		697
698	PMINUB(%%mm2, %%mm1, %%mm4) // MIN(\|lenergy\|,\|renergy\|)/8	698	PMINUB(%%mm2, %%mm1, %%mm4) // MIN(\|lenergy\|,\|renergy\|)/8
699	"movq %2, %%mm4 \n\t" // QP //FIXME QP+1 ?	699	"movq %2, %%mm4 \n\t" // QP //FIXME QP+1 ?
700	"paddusb "MANGLE(b01)", %%mm4 \n\t"	700	"paddusb %3, %%mm4 \n\t"
701	"pcmpgtb %%mm3, %%mm4 \n\t" // \|menergy\|/8 < QP	701	"pcmpgtb %%mm3, %%mm4 \n\t" // \|menergy\|/8 < QP
702	"psubusb %%mm1, %%mm3 \n\t" // d=\|menergy\|/8-MIN(\|lenergy\|,\|renergy\|)/8	702	"psubusb %%mm1, %%mm3 \n\t" // d=\|menergy\|/8-MIN(\|lenergy\|,\|renergy\|)/8
703	"pand %%mm4, %%mm3 \n\t"	703	"pand %%mm4, %%mm3 \n\t"
704		704
705	"movq %%mm3, %%mm1 \n\t"	705	"movq %%mm3, %%mm1 \n\t"
706	// "psubusb "MANGLE(b01)", %%mm3 \n\t"	706	// "psubusb %3, %%mm3 \n\t"
707	PAVGB(%%mm7, %%mm3)	707	PAVGB(%%mm7, %%mm3)
708	PAVGB(%%mm7, %%mm3)	708	PAVGB(%%mm7, %%mm3)
709	"paddusb %%mm1, %%mm3 \n\t"	709	"paddusb %%mm1, %%mm3 \n\t"
710	// "paddusb "MANGLE(b01)", %%mm3 \n\t"	710	// "paddusb %3, %%mm3 \n\t"
711		711
712	"movq (%%"REG_a", %1, 2), %%mm6 \n\t" //l3	712	"movq (%%"REG_a", %1, 2), %%mm6 \n\t" //l3
713	"movq (%0, %1, 4), %%mm5 \n\t" //l4	713	"movq (%0, %1, 4), %%mm5 \n\t" //l4
Lines 720-726 Link Here
720	"pand %%mm0, %%mm3 \n\t"	720	"pand %%mm0, %%mm3 \n\t"
721	PMINUB(%%mm5, %%mm3, %%mm0)	721	PMINUB(%%mm5, %%mm3, %%mm0)
722		722
723	"psubusb "MANGLE(b01)", %%mm3 \n\t"	723	"psubusb %3, %%mm3 \n\t"
724	PAVGB(%%mm7, %%mm3)	724	PAVGB(%%mm7, %%mm3)
725		725
726	"movq (%%"REG_a", %1, 2), %%mm0 \n\t"	726	"movq (%%"REG_a", %1, 2), %%mm0 \n\t"
Lines 752-758 Link Here
752	"movq (%%"REG_a", %1), %%mm3 \n\t" // l2	752	"movq (%%"REG_a", %1), %%mm3 \n\t" // l2
753	"pxor %%mm6, %%mm2 \n\t" // -l5-1	753	"pxor %%mm6, %%mm2 \n\t" // -l5-1
754	"movq %%mm2, %%mm5 \n\t" // -l5-1	754	"movq %%mm2, %%mm5 \n\t" // -l5-1
755	"movq "MANGLE(b80)", %%mm4 \n\t" // 128	755	"movq %4, %%mm4 \n\t" // 128
756	"lea (%%"REG_a", %1, 4), %%"REG_c" \n\t"	756	"lea (%%"REG_a", %1, 4), %%"REG_c" \n\t"
757	PAVGB(%%mm3, %%mm2) // (l2-l5+256)/2	757	PAVGB(%%mm3, %%mm2) // (l2-l5+256)/2
758	PAVGB(%%mm0, %%mm4) // ~(l4-l3)/4 + 128	758	PAVGB(%%mm0, %%mm4) // ~(l4-l3)/4 + 128
Lines 764-770 Link Here
764	"pxor %%mm6, %%mm2 \n\t" // -l1-1	764	"pxor %%mm6, %%mm2 \n\t" // -l1-1
765	PAVGB(%%mm3, %%mm2) // (l2-l1+256)/2	765	PAVGB(%%mm3, %%mm2) // (l2-l1+256)/2
766	PAVGB((%0), %%mm1) // (l0-l3+256)/2	766	PAVGB((%0), %%mm1) // (l0-l3+256)/2
767	"movq "MANGLE(b80)", %%mm3 \n\t" // 128	767	"movq %4, %%mm3 \n\t" // 128
768	PAVGB(%%mm2, %%mm3) // ~(l2-l1)/4 + 128	768	PAVGB(%%mm2, %%mm3) // ~(l2-l1)/4 + 128
769	PAVGB(%%mm1, %%mm3) // ~(l0-l3)/4 +(l2-l1)/8 + 128	769	PAVGB(%%mm1, %%mm3) // ~(l0-l3)/4 +(l2-l1)/8 + 128
770	PAVGB(%%mm2, %%mm3) // ~(l0-l3)/8 +5(l2-l1)/16 + 128	770	PAVGB(%%mm2, %%mm3) // ~(l0-l3)/8 +5(l2-l1)/16 + 128
Lines 774-787 Link Here
774	"movq (%%"REG_c", %1, 2), %%mm1 \n\t" // l7	774	"movq (%%"REG_c", %1, 2), %%mm1 \n\t" // l7
775	"pxor %%mm6, %%mm1 \n\t" // -l7-1	775	"pxor %%mm6, %%mm1 \n\t" // -l7-1
776	PAVGB((%0, %1, 4), %%mm1) // (l4-l7+256)/2	776	PAVGB((%0, %1, 4), %%mm1) // (l4-l7+256)/2
777	"movq "MANGLE(b80)", %%mm2 \n\t" // 128	777	"movq %4, %%mm2 \n\t" // 128
778	PAVGB(%%mm5, %%mm2) // ~(l6-l5)/4 + 128	778	PAVGB(%%mm5, %%mm2) // ~(l6-l5)/4 + 128
779	PAVGB(%%mm1, %%mm2) // ~(l4-l7)/4 +(l6-l5)/8 + 128	779	PAVGB(%%mm1, %%mm2) // ~(l4-l7)/4 +(l6-l5)/8 + 128
780	PAVGB(%%mm5, %%mm2) // ~(l4-l7)/8 +5(l6-l5)/16 + 128	780	PAVGB(%%mm5, %%mm2) // ~(l4-l7)/8 +5(l6-l5)/16 + 128
781	// mm0=128-q, mm2=renergy/16 + 128, mm3=lenergy/16 + 128, mm4= menergy/16 + 128	781	// mm0=128-q, mm2=renergy/16 + 128, mm3=lenergy/16 + 128, mm4= menergy/16 + 128
782		782
783	"movq "MANGLE(b00)", %%mm1 \n\t" // 0	783	"movq %5, %%mm1 \n\t" // 0
784	"movq "MANGLE(b00)", %%mm5 \n\t" // 0	784	"movq %5, %%mm5 \n\t" // 0
785	"psubb %%mm2, %%mm1 \n\t" // 128 - renergy/16	785	"psubb %%mm2, %%mm1 \n\t" // 128 - renergy/16
786	"psubb %%mm3, %%mm5 \n\t" // 128 - lenergy/16	786	"psubb %%mm3, %%mm5 \n\t" // 128 - lenergy/16
787	PMAXUB(%%mm1, %%mm2) // 128 + \|renergy/16\|	787	PMAXUB(%%mm1, %%mm2) // 128 + \|renergy/16\|
Lines 790-796 Link Here
790		790
791	// mm0=128-q, mm3=128 + MIN(\|lenergy\|,\|renergy\|)/16, mm4= menergy/16 + 128	791	// mm0=128-q, mm3=128 + MIN(\|lenergy\|,\|renergy\|)/16, mm4= menergy/16 + 128
792		792
793	"movq "MANGLE(b00)", %%mm7 \n\t" // 0	793	"movq %5, %%mm7 \n\t" // 0
794	"movq %2, %%mm2 \n\t" // QP	794	"movq %2, %%mm2 \n\t" // QP
795	PAVGB(%%mm6, %%mm2) // 128 + QP/2	795	PAVGB(%%mm6, %%mm2) // 128 + QP/2
796	"psubb %%mm6, %%mm2 \n\t"	796	"psubb %%mm6, %%mm2 \n\t"
Lines 804-816 Link Here
804	// mm0=128-q, mm1= SIGN(menergy), mm2= \|menergy\|/16 < QP/2, mm4= d/16	804	// mm0=128-q, mm1= SIGN(menergy), mm2= \|menergy\|/16 < QP/2, mm4= d/16
805		805
806	"movq %%mm4, %%mm3 \n\t" // d	806	"movq %%mm4, %%mm3 \n\t" // d
807	"psubusb "MANGLE(b01)", %%mm4 \n\t"	807	"psubusb %3, %%mm4 \n\t"
808	PAVGB(%%mm7, %%mm4) // d/32	808	PAVGB(%%mm7, %%mm4) // d/32
809	PAVGB(%%mm7, %%mm4) // (d + 32)/64	809	PAVGB(%%mm7, %%mm4) // (d + 32)/64
810	"paddb %%mm3, %%mm4 \n\t" // 5d/64	810	"paddb %%mm3, %%mm4 \n\t" // 5d/64
811	"pand %%mm2, %%mm4 \n\t"	811	"pand %%mm2, %%mm4 \n\t"
812		812
813	"movq "MANGLE(b80)", %%mm5 \n\t" // 128	813	"movq %4, %%mm5 \n\t" // 128
814	"psubb %%mm0, %%mm5 \n\t" // q	814	"psubb %%mm0, %%mm5 \n\t" // q
815	"paddsb %%mm6, %%mm5 \n\t" // fix bad rounding	815	"paddsb %%mm6, %%mm5 \n\t" // fix bad rounding
816	"pcmpgtb %%mm5, %%mm7 \n\t" // SIGN(q)	816	"pcmpgtb %%mm5, %%mm7 \n\t" // SIGN(q)
Lines 832-838 Link Here
832	"movq %%mm2, (%0, %1, 4) \n\t"	832	"movq %%mm2, (%0, %1, 4) \n\t"
833		833
834	:	834	:
835	: "r" (src), "r" ((long)stride), "m" (c->pQPb)	835	: "r" (src), "r" ((long)stride), "m" (c->pQPb), "m" (b01), "m" (b80), "m" (b00)
836	: "%"REG_a, "%"REG_c	836	: "%"REG_a, "%"REG_c
837	);	837	);
838		838
Lines 1076-1085 Link Here
1076	"psubusw %%mm1, %%mm5 \n\t" // ld	1076	"psubusw %%mm1, %%mm5 \n\t" // ld
1077		1077
1078		1078
1079	"movq "MANGLE(w05)", %%mm2 \n\t" // 5	1079	"movq %3, %%mm2 \n\t" // 5
1080	"pmullw %%mm2, %%mm4 \n\t"	1080	"pmullw %%mm2, %%mm4 \n\t"
1081	"pmullw %%mm2, %%mm5 \n\t"	1081	"pmullw %%mm2, %%mm5 \n\t"
1082	"movq "MANGLE(w20)", %%mm2 \n\t" // 32	1082	"movq %4, %%mm2 \n\t" // 32
1083	"paddw %%mm2, %%mm4 \n\t"	1083	"paddw %%mm2, %%mm4 \n\t"
1084	"paddw %%mm2, %%mm5 \n\t"	1084	"paddw %%mm2, %%mm5 \n\t"
1085	"psrlw $6, %%mm4 \n\t"	1085	"psrlw $6, %%mm4 \n\t"
Lines 1129-1135 Link Here
1129	"movq %%mm0, (%0, %1) \n\t"	1129	"movq %%mm0, (%0, %1) \n\t"
1130		1130
1131	: "+r" (src)	1131	: "+r" (src)
1132	: "r" ((long)stride), "m" (c->pQPb)	1132	: "r" ((long)stride), "m" (c->pQPb), "m" (w05), "m" (w20)
1133	: "%"REG_a, "%"REG_c	1133	: "%"REG_a, "%"REG_c
1134	);	1134	);
1135	#else	1135	#else
Lines 1273-1279 Link Here
1273	"movq %%mm6, %%mm0 \n\t" // max	1273	"movq %%mm6, %%mm0 \n\t" // max
1274	"psubb %%mm7, %%mm6 \n\t" // max - min	1274	"psubb %%mm7, %%mm6 \n\t" // max - min
1275	"movd %%mm6, %%ecx \n\t"	1275	"movd %%mm6, %%ecx \n\t"
1276	"cmpb "MANGLE(deringThreshold)", %%cl \n\t"	1276	"cmpb %4, %%cl \n\t"
1277	" jb 1f \n\t"	1277	" jb 1f \n\t"
1278	"lea -24(%%"REG_SP"), %%"REG_c" \n\t"	1278	"lea -24(%%"REG_SP"), %%"REG_c" \n\t"
1279	"and "ALIGN_MASK", %%"REG_c" \n\t"	1279	"and "ALIGN_MASK", %%"REG_c" \n\t"
Lines 1300-1308 Link Here
1300	"psubusb %%mm7, %%mm0 \n\t"	1300	"psubusb %%mm7, %%mm0 \n\t"
1301	"psubusb %%mm7, %%mm2 \n\t"	1301	"psubusb %%mm7, %%mm2 \n\t"
1302	"psubusb %%mm7, %%mm3 \n\t"	1302	"psubusb %%mm7, %%mm3 \n\t"
1303	"pcmpeqb "MANGLE(b00)", %%mm0 \n\t" // L10 > a ? 0 : -1	1303	"pcmpeqb %5, %%mm0 \n\t" // L10 > a ? 0 : -1
1304	"pcmpeqb "MANGLE(b00)", %%mm2 \n\t" // L20 > a ? 0 : -1	1304	"pcmpeqb %5, %%mm2 \n\t" // L20 > a ? 0 : -1
1305	"pcmpeqb "MANGLE(b00)", %%mm3 \n\t" // L00 > a ? 0 : -1	1305	"pcmpeqb %5, %%mm3 \n\t" // L00 > a ? 0 : -1
1306	"paddb %%mm2, %%mm0 \n\t"	1306	"paddb %%mm2, %%mm0 \n\t"
1307	"paddb %%mm3, %%mm0 \n\t"	1307	"paddb %%mm3, %%mm0 \n\t"
1308		1308
Lines 1323-1331 Link Here
1323	"psubusb %%mm7, %%mm2 \n\t"	1323	"psubusb %%mm7, %%mm2 \n\t"
1324	"psubusb %%mm7, %%mm4 \n\t"	1324	"psubusb %%mm7, %%mm4 \n\t"
1325	"psubusb %%mm7, %%mm5 \n\t"	1325	"psubusb %%mm7, %%mm5 \n\t"
1326	"pcmpeqb "MANGLE(b00)", %%mm2 \n\t" // L11 > a ? 0 : -1	1326	"pcmpeqb %5, %%mm2 \n\t" // L11 > a ? 0 : -1
1327	"pcmpeqb "MANGLE(b00)", %%mm4 \n\t" // L21 > a ? 0 : -1	1327	"pcmpeqb %5, %%mm4 \n\t" // L21 > a ? 0 : -1
1328	"pcmpeqb "MANGLE(b00)", %%mm5 \n\t" // L01 > a ? 0 : -1	1328	"pcmpeqb %5, %%mm5 \n\t" // L01 > a ? 0 : -1
1329	"paddb %%mm4, %%mm2 \n\t"	1329	"paddb %%mm4, %%mm2 \n\t"
1330	"paddb %%mm5, %%mm2 \n\t"	1330	"paddb %%mm5, %%mm2 \n\t"
1331	// 0, 2, 3, 1	1331	// 0, 2, 3, 1
Lines 1350-1356 Link Here
1350	"psubusb " #lx ", " #t1 " \n\t"\	1350	"psubusb " #lx ", " #t1 " \n\t"\
1351	"psubusb " #lx ", " #t0 " \n\t"\	1351	"psubusb " #lx ", " #t0 " \n\t"\
1352	"psubusb " #lx ", " #sx " \n\t"\	1352	"psubusb " #lx ", " #sx " \n\t"\
1353	"movq "MANGLE(b00)", " #lx " \n\t"\	1353	"movq %5, " #lx " \n\t"\
1354	"pcmpeqb " #lx ", " #t1 " \n\t" /* src[-1] > a ? 0 : -1*/\	1354	"pcmpeqb " #lx ", " #t1 " \n\t" /* src[-1] > a ? 0 : -1*/\
1355	"pcmpeqb " #lx ", " #t0 " \n\t" /* src[+1] > a ? 0 : -1*/\	1355	"pcmpeqb " #lx ", " #t0 " \n\t" /* src[+1] > a ? 0 : -1*/\
1356	"pcmpeqb " #lx ", " #sx " \n\t" /* src[0] > a ? 0 : -1*/\	1356	"pcmpeqb " #lx ", " #sx " \n\t" /* src[0] > a ? 0 : -1*/\
Lines 1366-1373 Link Here
1366	PMINUB(t1, pplx, t0)\	1366	PMINUB(t1, pplx, t0)\
1367	"paddb " #sx ", " #ppsx " \n\t"\	1367	"paddb " #sx ", " #ppsx " \n\t"\
1368	"paddb " #psx ", " #ppsx " \n\t"\	1368	"paddb " #psx ", " #ppsx " \n\t"\
1369	"#paddb "MANGLE(b02)", " #ppsx " \n\t"\	1369	"#paddb %6, " #ppsx " \n\t"\
1370	"pand "MANGLE(b08)", " #ppsx " \n\t"\	1370	"pand %7, " #ppsx " \n\t"\
1371	"pcmpeqb " #lx ", " #ppsx " \n\t"\	1371	"pcmpeqb " #lx ", " #ppsx " \n\t"\
1372	"pand " #ppsx ", " #pplx " \n\t"\	1372	"pand " #ppsx ", " #pplx " \n\t"\
1373	"pandn " #dst ", " #ppsx " \n\t"\	1373	"pandn " #dst ", " #ppsx " \n\t"\
Lines 1403-1409 Link Here
1403	DERING_CORE((%0, %1, 8),(%%REGd, %1, 4) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)	1403	DERING_CORE((%0, %1, 8),(%%REGd, %1, 4) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
1404		1404
1405	"1: \n\t"	1405	"1: \n\t"
1406	: : "r" (src), "r" ((long)stride), "m" (c->pQPb), "m"(c->pQPb2)	1406	: : "r" (src), "r" ((long)stride), "m" (c->pQPb), "m"(c->pQPb2), "m" (deringThreshold), "m" (b00), "m" (b02), "m" (b08)
1407	: "%"REG_a, "%"REG_d, "%"REG_c	1407	: "%"REG_a, "%"REG_d, "%"REG_c
1408	);	1408	);
1409	#else	1409	#else
Lines 2281-2287 Link Here
2281	#else	2281	#else
2282	#if defined (FAST_L2_DIFF)	2282	#if defined (FAST_L2_DIFF)
2283	"pcmpeqb %%mm7, %%mm7 \n\t"	2283	"pcmpeqb %%mm7, %%mm7 \n\t"
2284	"movq "MANGLE(b80)", %%mm6 \n\t"	2284	"movq %4, %%mm6 \n\t"
2285	"pxor %%mm0, %%mm0 \n\t"	2285	"pxor %%mm0, %%mm0 \n\t"
2286	#define REAL_L2_DIFF_CORE(a, b)\	2286	#define REAL_L2_DIFF_CORE(a, b)\
2287	"movq " #a ", %%mm5 \n\t"\	2287	"movq " #a ", %%mm5 \n\t"\
Lines 2530-2536 Link Here
2530		2530
2531	"4: \n\t"	2531	"4: \n\t"
2532		2532
2533	:: "r" (src), "r" (tempBlured), "r"((long)stride), "m" (tempBluredPast)	2533	:: "r" (src), "r" (tempBlured), "r"((long)stride), "m" (tempBluredPast), "m" (b80)
2534	: "%"REG_a, "%"REG_d, "%"REG_c, "memory"	2534	: "%"REG_a, "%"REG_d, "%"REG_c, "memory"
2535	);	2535	);
2536	//printf("%d\n", test);	2536	//printf("%d\n", test);
Lines 2805-2812 Link Here
2805	"movq %%mm6, %%mm1 \n\t"	2805	"movq %%mm6, %%mm1 \n\t"
2806	"psllw $2, %%mm0 \n\t"	2806	"psllw $2, %%mm0 \n\t"
2807	"psllw $2, %%mm1 \n\t"	2807	"psllw $2, %%mm1 \n\t"
2808	"paddw "MANGLE(w04)", %%mm0 \n\t"	2808	"paddw %5, %%mm0 \n\t"
2809	"paddw "MANGLE(w04)", %%mm1 \n\t"	2809	"paddw %5, %%mm1 \n\t"
2810		2810
2811	#define NEXT\	2811	#define NEXT\
2812	"movq (%0), %%mm2 \n\t"\	2812	"movq (%0), %%mm2 \n\t"\
Lines 2895-2901 Link Here
2895	"mov %4, %0 \n\t" //FIXME	2895	"mov %4, %0 \n\t" //FIXME
2896		2896
2897	: "+&r"(src)	2897	: "+&r"(src)
2898	: "r" ((long)step), "m" (c->pQPb), "r"(sums), "g"(src)	2898	: "r" ((long)step), "m" (c->pQPb), "r"(sums), "g"(src), "m" (w04)
2899	);	2899	);
2900		2900
2901	src+= step; // src points to begin of the 8x8 Block	2901	src+= step; // src points to begin of the 8x8 Block
Lines 3112-3121 Link Here
3112	"psubusw %%mm1, %%mm5 \n\t" // ld	3112	"psubusw %%mm1, %%mm5 \n\t" // ld
3113		3113
3114		3114
3115	"movq "MANGLE(w05)", %%mm2 \n\t" // 5	3115	"movq %4, %%mm2 \n\t" // 5
3116	"pmullw %%mm2, %%mm4 \n\t"	3116	"pmullw %%mm2, %%mm4 \n\t"
3117	"pmullw %%mm2, %%mm5 \n\t"	3117	"pmullw %%mm2, %%mm5 \n\t"
3118	"movq "MANGLE(w20)", %%mm2 \n\t" // 32	3118	"movq %5, %%mm2 \n\t" // 32
3119	"paddw %%mm2, %%mm4 \n\t"	3119	"paddw %%mm2, %%mm4 \n\t"
3120	"paddw %%mm2, %%mm5 \n\t"	3120	"paddw %%mm2, %%mm5 \n\t"
3121	"psrlw $6, %%mm4 \n\t"	3121	"psrlw $6, %%mm4 \n\t"
Lines 3167-3173 Link Here
3167	"movq %%mm0, (%0, %1) \n\t"	3167	"movq %%mm0, (%0, %1) \n\t"
3168		3168
3169	: "+r" (temp_src)	3169	: "+r" (temp_src)
3170	: "r" ((long)step), "m" (c->pQPb), "m"(eq_mask)	3170	: "r" ((long)step), "m" (c->pQPb), "m"(eq_mask), "m" (w05), "m" (w20)
3171	: "%"REG_a, "%"REG_c	3171	: "%"REG_a, "%"REG_c
3172	);	3172	);
3173	}	3173	}
Lines 3198-3207 Link Here
3198	{	3198	{
3199	#ifdef HAVE_MMX	3199	#ifdef HAVE_MMX
3200	asm volatile(	3200	asm volatile(
3201	"movq (%%"REG_a"), %%mm2 \n\t" // packedYOffset	3201	"movq (%0), %%mm2 \n\t" // packedYOffset
3202	"movq 8(%%"REG_a"), %%mm3 \n\t" // packedYScale	3202	"movq 8(%0), %%mm3 \n\t" // packedYScale
3203	"lea (%2,%4), %%"REG_a" \n\t"
3204	"lea (%3,%5), %%"REG_d" \n\t"
3205	"pxor %%mm4, %%mm4 \n\t"	3203	"pxor %%mm4, %%mm4 \n\t"
3206	#ifdef HAVE_MMX2	3204	#ifdef HAVE_MMX2
3207	#define REAL_SCALED_CPY(src1, src2, dst1, dst2) \	3205	#define REAL_SCALED_CPY(src1, src2, dst1, dst2) \
Lines 3257-3277 Link Here
3257	#define SCALED_CPY(src1, src2, dst1, dst2)\	3255	#define SCALED_CPY(src1, src2, dst1, dst2)\
3258	REAL_SCALED_CPY(src1, src2, dst1, dst2)	3256	REAL_SCALED_CPY(src1, src2, dst1, dst2)
3259		3257
3260	SCALED_CPY((%2) , (%2, %4) , (%3) , (%3, %5))	3258	SCALED_CPY((%1) , (%1, %3) , (%2) , (%2, %4))
3261	SCALED_CPY((%2, %4, 2), (%%REGa, %4, 2), (%3, %5, 2), (%%REGd, %5, 2))	3259	SCALED_CPY((%1, %3, 2), (%1, %3, 4) , (%2, %4, 2), (%2, %4, 4))
3262	SCALED_CPY((%2, %4, 4), (%%REGa, %4, 4), (%3, %5, 4), (%%REGd, %5, 4))	3260	"lea (%1,%3,2), %1 \n\t"
3263	"lea (%%"REG_a",%4,4), %%"REG_a" \n\t"	3261	"lea (%2,%4,2), %2 \n\t"
3264	"lea (%%"REG_d",%5,4), %%"REG_d" \n\t"	3262	SCALED_CPY((%1, %3, 2), (%1, %3, 4) , (%2, %4, 2), (%2, %4, 4))
3265	SCALED_CPY((%%REGa, %4), (%%REGa, %4, 2), (%%REGd, %5), (%%REGd, %5, 2))	3263	"lea (%1,%3), %1 \n\t"
3266		3264	"lea (%2,%4), %2 \n\t"
3267		3265	SCALED_CPY((%1, %3, 2), (%1, %3, 4) , (%2, %4, 2), (%2, %4, 4))
3268	: "=&a" (packedOffsetAndScale)	3266	: : "r" (packedOffsetAndScale),
3269	: "0" (packedOffsetAndScale),
3270	"r"(src),	3267	"r"(src),
3271	"r"(dst),	3268	"r"(dst),
3272	"r" ((long)srcStride),	3269	"r" ((long)srcStride),
3273	"r" ((long)dstStride)	3270	"r" ((long)dstStride)
3274	: "%"REG_d	3271	: "memory"
3275	);	3272	);
3276	#else	3273	#else
3277	for(i=0; i<8; i++)	3274	for(i=0; i<8; i++)
Lines 3283-3291 Link Here
3283	{	3280	{
3284	#ifdef HAVE_MMX	3281	#ifdef HAVE_MMX
3285	asm volatile(	3282	asm volatile(
3286	"lea (%0,%2), %%"REG_a" \n\t"
3287	"lea (%1,%3), %%"REG_d" \n\t"
3288
3289	#define REAL_SIMPLE_CPY(src1, src2, dst1, dst2) \	3283	#define REAL_SIMPLE_CPY(src1, src2, dst1, dst2) \
3290	"movq " #src1 ", %%mm0 \n\t"\	3284	"movq " #src1 ", %%mm0 \n\t"\
3291	"movq " #src2 ", %%mm1 \n\t"\	3285	"movq " #src2 ", %%mm1 \n\t"\
Lines 3296-3312 Link Here
3296	REAL_SIMPLE_CPY(src1, src2, dst1, dst2)	3290	REAL_SIMPLE_CPY(src1, src2, dst1, dst2)
3297		3291
3298	SIMPLE_CPY((%0) , (%0, %2) , (%1) , (%1, %3))	3292	SIMPLE_CPY((%0) , (%0, %2) , (%1) , (%1, %3))
3299	SIMPLE_CPY((%0, %2, 2), (%%REGa, %2, 2), (%1, %3, 2), (%%REGd, %3, 2))	3293	SIMPLE_CPY((%0, %2, 2), (%0, %2, 4) , (%1, %3, 2), (%1, %3, 4))
3300	SIMPLE_CPY((%0, %2, 4), (%%REGa, %2, 4), (%1, %3, 4), (%%REGd, %3, 4))	3294	"lea (%0,%2,2), %0 \n\t"
3301	"lea (%%"REG_a",%2,4), %%"REG_a" \n\t"	3295	"lea (%1,%3,2), %1 \n\t"
3302	"lea (%%"REG_d",%3,4), %%"REG_d" \n\t"	3296	SIMPLE_CPY((%0, %2) , (%0, %2, 4) , (%1, %3) , (%1, %3, 4))
3303	SIMPLE_CPY((%%REGa, %2), (%%REGa, %2, 2), (%%REGd, %3), (%%REGd, %3, 2))	3297	"lea (%0,%2), %0 \n\t"
3304		3298	"lea (%1,%3), %1 \n\t"
		3299	SIMPLE_CPY((%0, %2, 2), (%0, %2, 4) , (%1, %3, 2), (%1, %3, 4))
3305	: : "r" (src),	3300	: : "r" (src),
3306	"r" (dst),	3301	"r" (dst),
3307	"r" ((long)srcStride),	3302	"r" ((long)srcStride),
3308	"r" ((long)dstStride)	3303	"r" ((long)dstStride)
3309	: "%"REG_a, "%"REG_d	3304	: "memory"
3310	);	3305	);
3311	#else	3306	#else
3312	for(i=0; i<8; i++)	3307	for(i=0; i<8; i++)

Lines 6-12 Link Here

(-)libavformat/Makefile (-1 / +1 lines)
6		6
7	VPATH=$(SRC_PATH)/libavformat	7	VPATH=$(SRC_PATH)/libavformat
8		8
9	CFLAGS=$(OPTFLAGS) -I.. -I$(SRC_PATH) -I$(SRC_PATH)/libavutil -I$(SRC_PATH)/libavcodec -DHAVE_AV_CONFIG_H -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE	9	CFLAGS=$(OPTFLAGS) $(PIC) -I.. -I$(SRC_PATH) -I$(SRC_PATH)/libavutil -I$(SRC_PATH)/libavcodec -DHAVE_AV_CONFIG_H -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE
10		10
11	OBJS= utils.o cutils.o os_support.o allformats.o	11	OBJS= utils.o cutils.o os_support.o allformats.o
12	PPOBJS=	12	PPOBJS=

Lines 6-12 Link Here

(-)libavutil/Makefile (-1 / +1 lines)
6	VPATH=$(SRC_PATH)/libavutil	6	VPATH=$(SRC_PATH)/libavutil
7		7
8	# NOTE: -I.. is needed to include config.h	8	# NOTE: -I.. is needed to include config.h
9	CFLAGS=$(OPTFLAGS) -DHAVE_AV_CONFIG_H -I.. -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE	9	CFLAGS=$(OPTFLAGS) $(PIC) -DHAVE_AV_CONFIG_H -I.. -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE
10		10
11	OBJS= mathematics.o \	11	OBJS= mathematics.o \
12	integer.o \	12	integer.o \

Return to bug 115568

Lines 103-109 Link Here

(-)libavcodec/i386/mpegvideo_mmx_template.c (-10 / +10 lines)
103	if((s->out_format == FMT_H263 \|\| s->out_format == FMT_H261) && s->mpeg_quant==0){	103	if((s->out_format == FMT_H263 \|\| s->out_format == FMT_H261) && s->mpeg_quant==0){
104		104
105	asm volatile(	105	asm volatile(
106	"movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1	106	"movd %6, %%mm3 \n\t" // last_non_zero_p1
107	SPREADW(%%mm3)	107	SPREADW(%%mm3)
108	"pxor %%mm7, %%mm7 \n\t" // 0	108	"pxor %%mm7, %%mm7 \n\t" // 0
109	"pxor %%mm4, %%mm4 \n\t" // 0	109	"pxor %%mm4, %%mm4 \n\t" // 0
Lines 132-142 Link Here
132	"add $8, %%"REG_a" \n\t"	132	"add $8, %%"REG_a" \n\t"
133	" js 1b \n\t"	133	" js 1b \n\t"
134	PMAX(%%mm3, %%mm0)	134	PMAX(%%mm3, %%mm0)
135	"movd %%mm3, %%"REG_a" \n\t"	135	"movd %%mm3, %0 \n\t"
136	"movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1	136	"movzb %b0, %0 \n\t" // last_non_zero_p1
137	: "+a" (last_non_zero_p1)	137	: "=q" (last_non_zero_p1)
138	: "r" (block+64), "r" (qmat), "r" (bias),	138	: "r" (block+64), "r" (qmat), "r" (bias),
139	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)	139	"r" (inv_zigzag_direct16+64), "r" (temp_block+64), "m" (last_non_zero_p1)
140	);	140	);
141	// note the asm is split cuz gcc doesnt like that many operands ...	141	// note the asm is split cuz gcc doesnt like that many operands ...
142	asm volatile(	142	asm volatile(
Lines 150-156 Link Here
150	);	150	);
151	}else{ // FMT_H263	151	}else{ // FMT_H263
152	asm volatile(	152	asm volatile(
153	"movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1	153	"movd %6, %%mm3 \n\t" // last_non_zero_p1
154	SPREADW(%%mm3)	154	SPREADW(%%mm3)
155	"pxor %%mm7, %%mm7 \n\t" // 0	155	"pxor %%mm7, %%mm7 \n\t" // 0
156	"pxor %%mm4, %%mm4 \n\t" // 0	156	"pxor %%mm4, %%mm4 \n\t" // 0
Lines 178-188 Link Here
178	"add $8, %%"REG_a" \n\t"	178	"add $8, %%"REG_a" \n\t"
179	" js 1b \n\t"	179	" js 1b \n\t"
180	PMAX(%%mm3, %%mm0)	180	PMAX(%%mm3, %%mm0)
181	"movd %%mm3, %%"REG_a" \n\t"	181	"movd %%mm3, %0 \n\t"
182	"movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1	182	"movzb %b0, %0 \n\t" // last_non_zero_p1
183	: "+a" (last_non_zero_p1)	183	: "=q" (last_non_zero_p1)
184	: "r" (block+64), "r" (qmat+64), "r" (bias+64),	184	: "r" (block+64), "r" (qmat+64), "r" (bias+64),
185	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)	185	"r" (inv_zigzag_direct16+64), "r" (temp_block+64), "m" (last_non_zero_p1)
186	);	186	);
187	// note the asm is split cuz gcc doesnt like that many operands ...	187	// note the asm is split cuz gcc doesnt like that many operands ...
188	asm volatile(	188	asm volatile(