Attachment #108552 for bug #115568

Lines 657-671 static inline void transpose4x4(uint8_t Link Here

(-)ffmpeg-old/libavcodec/i386/dsputil_mmx.c (-63 / +62 lines)
657	"punpckhwd %%mm2, %%mm1 \n\t"	657	"punpckhwd %%mm2, %%mm1 \n\t"
658	"movd %%mm0, %0 \n\t"	658	"movd %%mm0, %0 \n\t"
659	"punpckhdq %%mm0, %%mm0 \n\t"	659	"punpckhdq %%mm0, %%mm0 \n\t"
660	"movd %%mm0, %1 \n\t"	660	"movd %%mm0, (%0,%1) \n\t"
661	"movd %%mm1, %2 \n\t"	661	"movd %%mm1, (%0,%1,2) \n\t"
662	"punpckhdq %%mm1, %%mm1 \n\t"	662	"punpckhdq %%mm1, %%mm1 \n\t"
663	"movd %%mm1, %3 \n\t"	663	"lea (%1,%1,2), %1 \n\t"
		664	"movd %%mm1, (%0,%1) \n\t"
664		665
665	: "=m" ((uint32_t)(dst + 0*dst_stride)),	666	: "=r" ((uint32_t)(dst)), "+r" (dst_stride)
666	"=m" ((uint32_t)(dst + 1*dst_stride)),	667	:: "memory"
667	"=m" ((uint32_t)(dst + 2*dst_stride)),
668	"=m" ((uint32_t)(dst + 3*dst_stride))
669	);	668	);
670	}	669	}
671		670
Lines 1745-1751 WARPER8_16_SQ(hadamard8_diff_mmx2, hadam Link Here
1745		1744
1746	#define QPEL_V_LOW(m3,m4,m5,m6, pw_20, pw_3, rnd, in0, in1, in2, in7, out, OP)\	1745	#define QPEL_V_LOW(m3,m4,m5,m6, pw_20, pw_3, rnd, in0, in1, in2, in7, out, OP)\
1747	"paddw " #m4 ", " #m3 " \n\t" /* x1 */\	1746	"paddw " #m4 ", " #m3 " \n\t" /* x1 */\
1748	"movq "MANGLE(ff_pw_20)", %%mm4 \n\t" /* 20 */\	1747	"movq %5, %%mm4 \n\t" /* 20 */\
1749	"pmullw " #m3 ", %%mm4 \n\t" /* 20x1 */\	1748	"pmullw " #m3 ", %%mm4 \n\t" /* 20x1 */\
1750	"movq "#in7", " #m3 " \n\t" /* d */\	1749	"movq "#in7", " #m3 " \n\t" /* d */\
1751	"movq "#in0", %%mm5 \n\t" /* D */\	1750	"movq "#in0", %%mm5 \n\t" /* D */\
Lines 1757-1763 WARPER8_16_SQ(hadamard8_diff_mmx2, hadam Link Here
1757	"paddw " #m5 ", %%mm6 \n\t" /* x2 */\	1756	"paddw " #m5 ", %%mm6 \n\t" /* x2 */\
1758	"paddw %%mm6, %%mm6 \n\t" /* 2x2 */\	1757	"paddw %%mm6, %%mm6 \n\t" /* 2x2 */\
1759	"psubw %%mm6, %%mm5 \n\t" /* -2x2 + x3 */\	1758	"psubw %%mm6, %%mm5 \n\t" /* -2x2 + x3 */\
1760	"pmullw "MANGLE(ff_pw_3)", %%mm5 \n\t" /* -6x2 + 3x3 */\	1759	"pmullw %6, %%mm5 \n\t" /* -6x2 + 3x3 */\
1761	"paddw " #rnd ", %%mm4 \n\t" /* x2 */\	1760	"paddw " #rnd ", %%mm4 \n\t" /* x2 */\
1762	"paddw %%mm4, %%mm5 \n\t" /* 20x1 - 6x2 + 3x3 - x4 */\	1761	"paddw %%mm4, %%mm5 \n\t" /* 20x1 - 6x2 + 3x3 - x4 */\
1763	"psraw $5, %%mm5 \n\t"\	1762	"psraw $5, %%mm5 \n\t"\
Lines 1791-1805 static void OPNAME ## mpeg4_qpel16_h_low Link Here
1791	"paddw %%mm5, %%mm5 \n\t" /* 2b */\	1790	"paddw %%mm5, %%mm5 \n\t" /* 2b */\
1792	"psubw %%mm5, %%mm6 \n\t" /* c - 2b */\	1791	"psubw %%mm5, %%mm6 \n\t" /* c - 2b */\
1793	"pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\	1792	"pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\
1794	"pmullw "MANGLE(ff_pw_3)", %%mm6 \n\t" /* 3c - 6b */\	1793	"pmullw %6, %%mm6 \n\t" /* 3c - 6b */\
1795	"paddw %%mm4, %%mm0 \n\t" /* a */\	1794	"paddw %%mm4, %%mm0 \n\t" /* a */\
1796	"paddw %%mm1, %%mm5 \n\t" /* d */\	1795	"paddw %%mm1, %%mm5 \n\t" /* d */\
1797	"pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\	1796	"pmullw %5, %%mm0 \n\t" /* 20a */\
1798	"psubw %%mm5, %%mm0 \n\t" /* 20a - d */\	1797	"psubw %%mm5, %%mm0 \n\t" /* 20a - d */\
1799	"paddw %6, %%mm6 \n\t"\	1798	"paddw %8, %%mm6 \n\t"\
1800	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\	1799	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\
1801	"psraw $5, %%mm0 \n\t"\	1800	"psraw $5, %%mm0 \n\t"\
1802	"movq %%mm0, %5 \n\t"\	1801	"movq %%mm0, %7 \n\t"\
1803	/* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\	1802	/* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\
1804	\	1803	\
1805	"movq 5(%0), %%mm0 \n\t" /* FGHIJKLM */\	1804	"movq 5(%0), %%mm0 \n\t" /* FGHIJKLM */\
Lines 1817-1831 static void OPNAME ## mpeg4_qpel16_h_low Link Here
1817	"psrlq $24, %%mm6 \n\t" /* IJKLM000 */\	1816	"psrlq $24, %%mm6 \n\t" /* IJKLM000 */\
1818	"punpcklbw %%mm7, %%mm2 \n\t" /* 0F0G0H0I */\	1817	"punpcklbw %%mm7, %%mm2 \n\t" /* 0F0G0H0I */\
1819	"punpcklbw %%mm7, %%mm6 \n\t" /* 0I0J0K0L */\	1818	"punpcklbw %%mm7, %%mm6 \n\t" /* 0I0J0K0L */\
1820	"pmullw "MANGLE(ff_pw_3)", %%mm3 \n\t" /* 3c - 6b */\	1819	"pmullw %6, %%mm3 \n\t" /* 3c - 6b */\
1821	"paddw %%mm2, %%mm1 \n\t" /* a */\	1820	"paddw %%mm2, %%mm1 \n\t" /* a */\
1822	"paddw %%mm6, %%mm4 \n\t" /* d */\	1821	"paddw %%mm6, %%mm4 \n\t" /* d */\
1823	"pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\	1822	"pmullw %5, %%mm1 \n\t" /* 20a */\
1824	"psubw %%mm4, %%mm3 \n\t" /* - 6b +3c - d */\	1823	"psubw %%mm4, %%mm3 \n\t" /* - 6b +3c - d */\
1825	"paddw %6, %%mm1 \n\t"\	1824	"paddw %8, %%mm1 \n\t"\
1826	"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b +3c - d */\	1825	"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b +3c - d */\
1827	"psraw $5, %%mm3 \n\t"\	1826	"psraw $5, %%mm3 \n\t"\
1828	"movq %5, %%mm1 \n\t"\	1827	"movq %7, %%mm1 \n\t"\
1829	"packuswb %%mm3, %%mm1 \n\t"\	1828	"packuswb %%mm3, %%mm1 \n\t"\
1830	OP_MMX2(%%mm1, (%1),%%mm4, q)\	1829	OP_MMX2(%%mm1, (%1),%%mm4, q)\
1831	/* mm0= GHIJ, mm2=FGHI, mm5=HIJK, mm6=IJKL, mm7=0 */\	1830	/* mm0= GHIJ, mm2=FGHI, mm5=HIJK, mm6=IJKL, mm7=0 */\
Lines 1843-1849 static void OPNAME ## mpeg4_qpel16_h_low Link Here
1843	"psubw %%mm5, %%mm0 \n\t" /* c - 2b */\	1842	"psubw %%mm5, %%mm0 \n\t" /* c - 2b */\
1844	"movq %%mm3, %%mm5 \n\t" /* JKLMNOPQ */\	1843	"movq %%mm3, %%mm5 \n\t" /* JKLMNOPQ */\
1845	"psrlq $24, %%mm3 \n\t" /* MNOPQ000 */\	1844	"psrlq $24, %%mm3 \n\t" /* MNOPQ000 */\
1846	"pmullw "MANGLE(ff_pw_3)", %%mm0 \n\t" /* 3c - 6b */\	1845	"pmullw %6, %%mm0 \n\t" /* 3c - 6b */\
1847	"punpcklbw %%mm7, %%mm3 \n\t" /* 0M0N0O0P */\	1846	"punpcklbw %%mm7, %%mm3 \n\t" /* 0M0N0O0P */\
1848	"paddw %%mm3, %%mm2 \n\t" /* d */\	1847	"paddw %%mm3, %%mm2 \n\t" /* d */\
1849	"psubw %%mm2, %%mm0 \n\t" /* -6b + 3c - d */\	1848	"psubw %%mm2, %%mm0 \n\t" /* -6b + 3c - d */\
Lines 1851-1858 static void OPNAME ## mpeg4_qpel16_h_low Link Here
1851	"punpcklbw %%mm7, %%mm2 \n\t" /* 0J0K0L0M */\	1850	"punpcklbw %%mm7, %%mm2 \n\t" /* 0J0K0L0M */\
1852	"punpckhbw %%mm7, %%mm5 \n\t" /* 0N0O0P0Q */\	1851	"punpckhbw %%mm7, %%mm5 \n\t" /* 0N0O0P0Q */\
1853	"paddw %%mm2, %%mm6 \n\t" /* a */\	1852	"paddw %%mm2, %%mm6 \n\t" /* a */\
1854	"pmullw "MANGLE(ff_pw_20)", %%mm6 \n\t" /* 20a */\	1853	"pmullw %5, %%mm6 \n\t" /* 20a */\
1855	"paddw %6, %%mm0 \n\t"\	1854	"paddw %8, %%mm0 \n\t"\
1856	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\	1855	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\
1857	"psraw $5, %%mm0 \n\t"\	1856	"psraw $5, %%mm0 \n\t"\
1858	/* mm1=KLMN, mm2=JKLM, mm3=MNOP, mm4=LMNO, mm5=NOPQ mm7=0 */\	1857	/* mm1=KLMN, mm2=JKLM, mm3=MNOP, mm4=LMNO, mm5=NOPQ mm7=0 */\
Lines 1866-1875 static void OPNAME ## mpeg4_qpel16_h_low Link Here
1866	"paddw %%mm2, %%mm5 \n\t" /* d */\	1865	"paddw %%mm2, %%mm5 \n\t" /* d */\
1867	"paddw %%mm6, %%mm6 \n\t" /* 2b */\	1866	"paddw %%mm6, %%mm6 \n\t" /* 2b */\
1868	"psubw %%mm6, %%mm4 \n\t" /* c - 2b */\	1867	"psubw %%mm6, %%mm4 \n\t" /* c - 2b */\
1869	"pmullw "MANGLE(ff_pw_20)", %%mm3 \n\t" /* 20a */\	1868	"pmullw %5, %%mm3 \n\t" /* 20a */\
1870	"pmullw "MANGLE(ff_pw_3)", %%mm4 \n\t" /* 3c - 6b */\	1869	"pmullw %6, %%mm4 \n\t" /* 3c - 6b */\
1871	"psubw %%mm5, %%mm3 \n\t" /* -6b + 3c - d */\	1870	"psubw %%mm5, %%mm3 \n\t" /* -6b + 3c - d */\
1872	"paddw %6, %%mm4 \n\t"\	1871	"paddw %8, %%mm4 \n\t"\
1873	"paddw %%mm3, %%mm4 \n\t" /* 20a - 6b + 3c - d */\	1872	"paddw %%mm3, %%mm4 \n\t" /* 20a - 6b + 3c - d */\
1874	"psraw $5, %%mm4 \n\t"\	1873	"psraw $5, %%mm4 \n\t"\
1875	"packuswb %%mm4, %%mm0 \n\t"\	1874	"packuswb %%mm4, %%mm0 \n\t"\
Lines 1879-1886 static void OPNAME ## mpeg4_qpel16_h_low Link Here
1879	"add %4, %1 \n\t"\	1878	"add %4, %1 \n\t"\
1880	"decl %2 \n\t"\	1879	"decl %2 \n\t"\
1881	" jnz 1b \n\t"\	1880	" jnz 1b \n\t"\
1882	: "+a"(src), "+c"(dst), "+m"(h)\	1881	: : "a"(src), "c"(dst), "m"(h), \
1883	: "d"((long)srcStride), "S"((long)dstStride), /"m"(ff_pw_20), "m"(ff_pw_3),/ "m"(temp), "m"(ROUNDER)\	1882	"d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(temp), "m"(ROUNDER)\
1884	: "memory"\	1883	: "memory"\
1885	);\	1884	);\
1886	}\	1885	}\
Lines 1958-1969 static void OPNAME ## mpeg4_qpel8_h_lowp Link Here
1958	"paddw %%mm5, %%mm5 \n\t" /* 2b */\	1957	"paddw %%mm5, %%mm5 \n\t" /* 2b */\
1959	"psubw %%mm5, %%mm6 \n\t" /* c - 2b */\	1958	"psubw %%mm5, %%mm6 \n\t" /* c - 2b */\
1960	"pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\	1959	"pshufw $0x06, %%mm0, %%mm5 \n\t" /* 0C0B0A0A */\
1961	"pmullw "MANGLE(ff_pw_3)", %%mm6 \n\t" /* 3c - 6b */\	1960	"pmullw %6, %%mm6 \n\t" /* 3c - 6b */\
1962	"paddw %%mm4, %%mm0 \n\t" /* a */\	1961	"paddw %%mm4, %%mm0 \n\t" /* a */\
1963	"paddw %%mm1, %%mm5 \n\t" /* d */\	1962	"paddw %%mm1, %%mm5 \n\t" /* d */\
1964	"pmullw "MANGLE(ff_pw_20)", %%mm0 \n\t" /* 20a */\	1963	"pmullw %5, %%mm0 \n\t" /* 20a */\
1965	"psubw %%mm5, %%mm0 \n\t" /* 20a - d */\	1964	"psubw %%mm5, %%mm0 \n\t" /* 20a - d */\
1966	"paddw %6, %%mm6 \n\t"\	1965	"paddw %8, %%mm6 \n\t"\
1967	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\	1966	"paddw %%mm6, %%mm0 \n\t" /* 20a - 6b + 3c - d */\
1968	"psraw $5, %%mm0 \n\t"\	1967	"psraw $5, %%mm0 \n\t"\
1969	/* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\	1968	/* mm1=EFGH, mm2=DEFG, mm3=CDEF, mm4=BCDE, mm7=0 */\
Lines 1979-1988 static void OPNAME ## mpeg4_qpel8_h_lowp Link Here
1979	"paddw %%mm5, %%mm4 \n\t" /* d */\	1978	"paddw %%mm5, %%mm4 \n\t" /* d */\
1980	"paddw %%mm2, %%mm2 \n\t" /* 2b */\	1979	"paddw %%mm2, %%mm2 \n\t" /* 2b */\
1981	"psubw %%mm2, %%mm3 \n\t" /* c - 2b */\	1980	"psubw %%mm2, %%mm3 \n\t" /* c - 2b */\
1982	"pmullw "MANGLE(ff_pw_20)", %%mm1 \n\t" /* 20a */\	1981	"pmullw %5, %%mm1 \n\t" /* 20a */\
1983	"pmullw "MANGLE(ff_pw_3)", %%mm3 \n\t" /* 3c - 6b */\	1982	"pmullw %6, %%mm3 \n\t" /* 3c - 6b */\
1984	"psubw %%mm4, %%mm3 \n\t" /* -6b + 3c - d */\	1983	"psubw %%mm4, %%mm3 \n\t" /* -6b + 3c - d */\
1985	"paddw %6, %%mm1 \n\t"\	1984	"paddw %8, %%mm1 \n\t"\
1986	"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b + 3c - d */\	1985	"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b + 3c - d */\
1987	"psraw $5, %%mm3 \n\t"\	1986	"psraw $5, %%mm3 \n\t"\
1988	"packuswb %%mm3, %%mm0 \n\t"\	1987	"packuswb %%mm3, %%mm0 \n\t"\
Lines 1992-1999 static void OPNAME ## mpeg4_qpel8_h_lowp Link Here
1992	"add %4, %1 \n\t"\	1991	"add %4, %1 \n\t"\
1993	"decl %2 \n\t"\	1992	"decl %2 \n\t"\
1994	" jnz 1b \n\t"\	1993	" jnz 1b \n\t"\
1995	: "+a"(src), "+c"(dst), "+m"(h)\	1994	: : "a"(src), "c"(dst), "m"(h), \
1996	: "S"((long)srcStride), "D"((long)dstStride), /"m"(ff_pw_20), "m"(ff_pw_3),/ "m"(temp), "m"(ROUNDER)\	1995	"S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(temp), "m"(ROUNDER)\
1997	: "memory"\	1996	: "memory"\
1998	);\	1997	);\
1999	}\	1998	}\
Lines 2072-2110 static void OPNAME ## mpeg4_qpel16_v_low Link Here
2072	"movq 8(%0), %%mm1 \n\t"\	2071	"movq 8(%0), %%mm1 \n\t"\
2073	"movq 16(%0), %%mm2 \n\t"\	2072	"movq 16(%0), %%mm2 \n\t"\
2074	"movq 24(%0), %%mm3 \n\t"\	2073	"movq 24(%0), %%mm3 \n\t"\
2075	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\	2074	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\
2076	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\	2075	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\
2077	"add %4, %1 \n\t"\	2076	"add %4, %1 \n\t"\
2078	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, (%0), (%0), 8(%0), 48(%0), (%1), OP)\	2077	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, (%0), (%0), 8(%0), 48(%0), (%1), OP)\
2079	\	2078	\
2080	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\	2079	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\
2081	"add %4, %1 \n\t"\	2080	"add %4, %1 \n\t"\
2082	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\	2081	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
2083	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\	2082	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\
2084	"add %4, %1 \n\t"\	2083	"add %4, %1 \n\t"\
2085	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\	2084	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\
2086	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\	2085	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\
2087	"add %4, %1 \n\t"\	2086	"add %4, %1 \n\t"\
2088	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\	2087	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\
2089	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\	2088	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\
2090	"add %4, %1 \n\t"\	2089	"add %4, %1 \n\t"\
2091	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\	2090	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\
2092	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\	2091	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\
2093	"add %4, %1 \n\t"\	2092	"add %4, %1 \n\t"\
2094	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\	2093	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\
2095	\	2094	\
2096	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\	2095	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\
2097	"add %4, %1 \n\t" \	2096	"add %4, %1 \n\t" \
2098	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\	2097	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\
2099	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\	2098	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\
2100	\	2099	\
2101	"add $136, %0 \n\t"\	2100	"add $136, %0 \n\t"\
2102	"add %6, %1 \n\t"\	2101	"add %8, %1 \n\t"\
2103	"decl %2 \n\t"\	2102	"decl %2 \n\t"\
2104	" jnz 1b \n\t"\	2103	" jnz 1b \n\t"\
2105	\	2104	\
2106	: "+r"(temp_ptr), "+r"(dst), "+g"(count)\	2105	: : "r"(temp_ptr), "r"(dst), "g"(count), \
2107	: "r"((long)dstStride), "r"(2(long)dstStride), /"m"(ff_pw_20), "m"(ff_pw_3),/ "m"(ROUNDER), "g"(4-14(long)dstStride)\	2106	"r"((long)dstStride), "r"(2(long)dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(ROUNDER), "g"(4-14(long)dstStride)\
2108	:"memory"\	2107	:"memory"\
2109	);\	2108	);\
2110	}\	2109	}\
Lines 2144-2170 static void OPNAME ## mpeg4_qpel8_v_lowp Link Here
2144	"movq 8(%0), %%mm1 \n\t"\	2143	"movq 8(%0), %%mm1 \n\t"\
2145	"movq 16(%0), %%mm2 \n\t"\	2144	"movq 16(%0), %%mm2 \n\t"\
2146	"movq 24(%0), %%mm3 \n\t"\	2145	"movq 24(%0), %%mm3 \n\t"\
2147	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\	2146	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\
2148	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\	2147	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\
2149	"add %4, %1 \n\t"\	2148	"add %4, %1 \n\t"\
2150	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, (%0), (%0), 8(%0), 48(%0), (%1), OP)\	2149	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, (%0), (%0), 8(%0), 48(%0), (%1), OP)\
2151	\	2150	\
2152	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\	2151	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\
2153	"add %4, %1 \n\t"\	2152	"add %4, %1 \n\t"\
2154	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %5, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\	2153	QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %7, %8, %7, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
2155	\	2154	\
2156	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %5, 16(%0), 24(%0), 32(%0), 64(%0), (%1, %3), OP)\	2155	QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %7, %8, %7, 16(%0), 24(%0), 32(%0), 64(%0), (%1, %3), OP)\
2157	"add %4, %1 \n\t"\	2156	"add %4, %1 \n\t"\
2158	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %5, 24(%0), 32(%0), 40(%0), 56(%0), (%1), OP)\	2157	QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %7, %8, %7, 24(%0), 32(%0), 40(%0), 56(%0), (%1), OP)\
2159	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %5, 32(%0), 40(%0), 48(%0), 48(%0), (%1, %3), OP)\	2158	QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %7, %8, %7, 32(%0), 40(%0), 48(%0), 48(%0), (%1, %3), OP)\
2160	\	2159	\
2161	"add $72, %0 \n\t"\	2160	"add $72, %0 \n\t"\
2162	"add %6, %1 \n\t"\	2161	"add %8, %1 \n\t"\
2163	"decl %2 \n\t"\	2162	"decl %2 \n\t"\
2164	" jnz 1b \n\t"\	2163	" jnz 1b \n\t"\
2165	\	2164	\
2166	: "+r"(temp_ptr), "+r"(dst), "+g"(count)\	2165	: : "r"(temp_ptr), "r"(dst), "rm"(count), \
2167	: "r"((long)dstStride), "r"(2(long)dstStride), /"m"(ff_pw_20), "m"(ff_pw_3),/ "m"(ROUNDER), "g"(4-6(long)dstStride)\	2166	"r"((long)dstStride), "r"(2(long)dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(ROUNDER), "g"(4-6(long)dstStride)\
2168	: "memory"\	2167	: "memory"\
2169	);\	2168	);\
2170	}\	2169	}\

Lines 589-610 static void h264_loop_filter_strength_mm Link Here

(-)ffmpeg-old/libavcodec/i386/h264dsp_mmx.c (-7 / +11 lines)
589	"paddb %%mm6, %%mm1 \n\t"	589	"paddb %%mm6, %%mm1 \n\t"
590	"punpckhbw %%mm7, %%mm1 \n\t" // ref[b] != ref[bn]	590	"punpckhbw %%mm7, %%mm1 \n\t" // ref[b] != ref[bn]
591	"por %%mm1, %%mm0 \n\t"	591	"por %%mm1, %%mm0 \n\t"
		592	::"m"(ref[l][b_idx]),
		593	"m"(ref[l][b_idx+d_idx])
		594	: "memory"
		595	);
592		596
593	"movq %2, %%mm1 \n\t"	597	asm volatile(
594	"movq %3, %%mm2 \n\t"	598	"movq %0, %%mm1 \n\t"
595	"psubw %4, %%mm1 \n\t"	599	"movq %1, %%mm2 \n\t"
596	"psubw %5, %%mm2 \n\t"	600	"psubw %2, %%mm1 \n\t"
		601	"psubw %3, %%mm2 \n\t"
597	"packsswb %%mm2, %%mm1 \n\t"	602	"packsswb %%mm2, %%mm1 \n\t"
598	"paddb %%mm5, %%mm1 \n\t"	603	"paddb %%mm5, %%mm1 \n\t"
599	"pminub %%mm4, %%mm1 \n\t"	604	"pminub %%mm4, %%mm1 \n\t"
600	"pcmpeqb %%mm4, %%mm1 \n\t" // abs(mv[b] - mv[bn]) >= limit	605	"pcmpeqb %%mm4, %%mm1 \n\t" // abs(mv[b] - mv[bn]) >= limit
601	"por %%mm1, %%mm0 \n\t"	606	"por %%mm1, %%mm0 \n\t"
602	::"m"(ref[l][b_idx]),	607	::"m"(mv[l][b_idx][0]),
603	"m"(ref[l][b_idx+d_idx]),
604	"m"(mv[l][b_idx][0]),
605	"m"(mv[l][b_idx+2][0]),	608	"m"(mv[l][b_idx+2][0]),
606	"m"(mv[l][b_idx+d_idx][0]),	609	"m"(mv[l][b_idx+d_idx][0]),
607	"m"(mv[l][b_idx+d_idx+2][0])	610	"m"(mv[l][b_idx+d_idx+2][0])
		611	: "memory"
608	);	612	);
609	}	613	}
610	}	614	}

Lines 121-127 static inline void sad8_4_mmx2(uint8_t * Link Here

(-)ffmpeg-old/libavcodec/i386/motion_est_mmx.c (-2 / +2 lines)
121	long len= -(stride*h);	121	long len= -(stride*h);
122	asm volatile(	122	asm volatile(
123	ASMALIGN(4)	123	ASMALIGN(4)
124	"movq "MANGLE(bone)", %%mm5 \n\t"	124	"movq %5, %%mm5 \n\t"
125	"1: \n\t"	125	"1: \n\t"
126	"movq (%1, %%"REG_a"), %%mm0 \n\t"	126	"movq (%1, %%"REG_a"), %%mm0 \n\t"
127	"movq (%2, %%"REG_a"), %%mm2 \n\t"	127	"movq (%2, %%"REG_a"), %%mm2 \n\t"
Lines 149-155 static inline void sad8_4_mmx2(uint8_t * Link Here
149	"add %4, %%"REG_a" \n\t"	149	"add %4, %%"REG_a" \n\t"
150	" js 1b \n\t"	150	" js 1b \n\t"
151	: "+a" (len)	151	: "+a" (len)
152	: "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), "r" ((long)stride)	152	: "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), "r" ((long)stride), "m" (bone)
153	);	153	);
154	}	154	}
155		155




            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"

            "movd %%mm3, %%"REG_a"              \n\t"
            "movzb %%al, %%"REG_a"              \n\t" // last_non_zero_p1
            : "+a" (last_non_zero_p1)
            : "r" (block+64), "m" (qmat), "r" (bias),
              "r" (inv_zigzag_direct16+64), "r" (temp_block+64)
        );
        // note the asm is split cuz gcc doesnt like that many operands ...

        );
    }else{ // FMT_H263
        asm volatile(
            "movd %0, %%mm3                     \n\t" // last_non_zero_p1
            SPREADW(%%mm3)
            "pxor %%mm7, %%mm7                  \n\t" // 0
            "pxor %%mm4, %%mm4                  \n\t" // 0
            "push %%"REG_a"                     \n\t"
            "mov $-128, %%"REG_a"               \n\t"
            ASMALIGN(4)
            "1:                                 \n\t"
            "pxor %%mm1, %%mm1                  \n\t" // 0
            "push %%"REG_a"                     \n\t"
            "add %c6(%%"REG_SP"), %%"REG_a"     \n\t"
            "movq (%%"REG_a"), %%mm0            \n\t" // block[i]
            "pop %%"REG_a"                      \n\t"
            "pcmpgtw %%mm0, %%mm1               \n\t" // block[i] <= 0 ? 0xFF : 0x00
            "pxor %%mm1, %%mm0                  \n\t"
            "psubw %%mm1, %%mm0                 \n\t" // ABS(block[i])

            PMAX(%%mm3, %%mm0)
            "movd %%mm3, %%"REG_a"              \n\t"
            "movzb %%al, %%"REG_a"              \n\t" // last_non_zero_p1
            "mov %%"REG_a", %0                  \n\t"
            "pop %%"REG_a"                      \n\t"
            : "+m" (last_non_zero_p1)
            : "a" (block+64), "r" (qmat+64), "r" (bias+64),
              "r" (inv_zigzag_direct16+64), "r" (temp_block+64),
              "i" (sizeof(long))
        );
        // note the asm is split cuz gcc doesnt like that many operands ...
        asm volatile(

Lines 363-369 static inline void idct(int16_t *block) Link Here

(-)ffmpeg-old/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 COL_IDCT( 24(%1), 88(%1), 56(%1), 120(% 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 COL_IDCT( 24(%1), 88(%1), 56(%1), 120(% 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 1270-1276 Temp Link Here
1270	*/	1270	*/
1271		1271
1272	"9: \n\t"	1272	"9: \n\t"
1273	:: "r" (block), "r" (temp), "r" (coeffs)	1273	:: "r" (block), "r" (temp), "r" (coeffs), "m" (wm1010), "m" (d40000)
1274	: "%eax"	1274	: "%eax"
1275	);	1275	);
1276	}	1276	}





#define snow_inner_add_yblock_sse2_header \
    DWTELEM * * dst_array = sb->line + src_y;\
    long tmp = b_h;\
    asm volatile(\
             "mov  %6, %%"REG_c"             \n\t"\
             "mov  %6, %2                    \n\t"\
             "mov  %4, %%"REG_S"             \n\t"\
             "pxor %%xmm7, %%xmm7            \n\t" /* 0 */\
             "pcmpeqd %%xmm3, %%xmm3         \n\t"\


#define snow_inner_add_yblock_sse2_end_common2\
             "jnz 1b                         \n\t"\
             :"+m"(dst8),"+m"(dst_array),"=m"(tmp)\
             :\
             "rm"((long)(src_x<<2)),"m"(obmc),"a"(block),"m"((long)src_stride):\
             "%"REG_c"","%"REG_S"","%"REG_D"","%"REG_d"");

#define snow_inner_add_yblock_sse2_end_8\

#define snow_inner_add_yblock_sse2_end_16\
             "add $"PTR_SIZE"*1, %1          \n\t"\
             snow_inner_add_yblock_sse2_end_common1\
             "sub $1, %2                     \n\t"\
             snow_inner_add_yblock_sse2_end_common2

static void inner_add_yblock_bw_8_obmc_16_bh_even_sse2(uint8_t *obmc, const long obmc_stride, uint8_t * * block, int b_w, long b_h,


#define snow_inner_add_yblock_mmx_header \
    DWTELEM * * dst_array = sb->line + src_y;\
    long tmp = b_h;\
    asm volatile(\
             "mov  %6, %%"REG_c"             \n\t"\
             "mov  %6, %2                    \n\t"\
             "mov  %4, %%"REG_S"             \n\t"\
             "pxor %%mm7, %%mm7              \n\t" /* 0 */\
             "pcmpeqd %%mm3, %%mm3           \n\t"\

             "add %%"REG_c", (%%"REG_a")     \n\t"\
             "add $"PTR_SIZE"*1, %1          \n\t"\
             "add %%"REG_c", %0              \n\t"\
             "sub $1, %2                     \n\t"\
             "jnz 1b                         \n\t"\
             :"+m"(dst8),"+m"(dst_array),"=m"(tmp)\
             :\
             "rm"((long)(src_x<<2)),"m"(obmc),"a"(block),"m"((long)src_stride):\
             "%"REG_c"","%"REG_S"","%"REG_D"","%"REG_d"");

static void inner_add_yblock_bw_8_obmc_16_mmx(uint8_t *obmc, const long obmc_stride, uint8_t * * block, int b_w, long b_h,

Lines 35-44 static int a52_resample_MONO_to_5_MMX(fl Link Here

(-)ffmpeg-old/libavcodec/liba52/resample_mmx.c (-23 / +23 lines)
35	int32_t * f = (int32_t *) _f;	35	int32_t * f = (int32_t *) _f;
36	asm volatile(	36	asm volatile(
37	"movl $-512, %%esi \n\t"	37	"movl $-512, %%esi \n\t"
38	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	38	"movq %2, %%mm7 \n\t"
39	"movq "MANGLE(wm1100)", %%mm3 \n\t"	39	"movq %3, %%mm3 \n\t"
40	"movq "MANGLE(wm0101)", %%mm4 \n\t"	40	"movq %4, %%mm4 \n\t"
41	"movq "MANGLE(wm1010)", %%mm5 \n\t"	41	"movq %5, %%mm5 \n\t"
42	"pxor %%mm6, %%mm6 \n\t"	42	"pxor %%mm6, %%mm6 \n\t"
43	"1: \n\t"	43	"1: \n\t"
44	"movq (%1, %%esi, 2), %%mm0 \n\t"	44	"movq (%1, %%esi, 2), %%mm0 \n\t"
Lines 62-68 static int a52_resample_MONO_to_5_MMX(fl Link Here
62	"addl $8, %%esi \n\t"	62	"addl $8, %%esi \n\t"
63	" jnz 1b \n\t"	63	" jnz 1b \n\t"
64	"emms \n\t"	64	"emms \n\t"
65	:: "r" (s16+1280), "r" (f+256)	65	:: "r" (s16+1280), "r" (f+256), "m" (magicF2W), "m" (wm1100), "m" (wm0101), "m" (wm1010)
66	:"%esi", "%edi", "memory"	66	:"%esi", "%edi", "memory"
67	);	67	);
68	return 5*256;	68	return 5*256;
Lines 90-96 static int a52_resample_STEREO_to_2_MMX( Link Here
90	);*/	90	);*/
91	asm volatile(	91	asm volatile(
92	"movl $-1024, %%esi \n\t"	92	"movl $-1024, %%esi \n\t"
93	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	93	"movq %2, %%mm7 \n\t"
94	"1: \n\t"	94	"1: \n\t"
95	"movq (%1, %%esi), %%mm0 \n\t"	95	"movq (%1, %%esi), %%mm0 \n\t"
96	"movq 8(%1, %%esi), %%mm1 \n\t"	96	"movq 8(%1, %%esi), %%mm1 \n\t"
Lines 110-116 static int a52_resample_STEREO_to_2_MMX( Link Here
110	"addl $16, %%esi \n\t"	110	"addl $16, %%esi \n\t"
111	" jnz 1b \n\t"	111	" jnz 1b \n\t"
112	"emms \n\t"	112	"emms \n\t"
113	:: "r" (s16+512), "r" (f+256)	113	:: "r" (s16+512), "r" (f+256), "m" (magicF2W)
114	:"%esi", "memory"	114	:"%esi", "memory"
115	);	115	);
116	return 2*256;	116	return 2*256;
Lines 120-126 static int a52_resample_3F_to_5_MMX(floa Link Here
120	int32_t * f = (int32_t *) _f;	120	int32_t * f = (int32_t *) _f;
121	asm volatile(	121	asm volatile(
122	"movl $-1024, %%esi \n\t"	122	"movl $-1024, %%esi \n\t"
123	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	123	"movq %2, %%mm7 \n\t"
124	"pxor %%mm6, %%mm6 \n\t"	124	"pxor %%mm6, %%mm6 \n\t"
125	"movq %%mm7, %%mm5 \n\t"	125	"movq %%mm7, %%mm5 \n\t"
126	"punpckldq %%mm6, %%mm5 \n\t"	126	"punpckldq %%mm6, %%mm5 \n\t"
Lines 165-171 static int a52_resample_3F_to_5_MMX(floa Link Here
165	"addl $16, %%esi \n\t"	165	"addl $16, %%esi \n\t"
166	" jnz 1b \n\t"	166	" jnz 1b \n\t"
167	"emms \n\t"	167	"emms \n\t"
168	:: "r" (s16+1280), "r" (f+256)	168	:: "r" (s16+1280), "r" (f+256), "m" (magicF2W)
169	:"%esi", "%edi", "memory"	169	:"%esi", "%edi", "memory"
170	);	170	);
171	return 5*256;	171	return 5*256;
Lines 175-181 static int a52_resample_2F_2R_to_4_MMX(f Link Here
175	int32_t * f = (int32_t *) _f;	175	int32_t * f = (int32_t *) _f;
176	asm volatile(	176	asm volatile(
177	"movl $-1024, %%esi \n\t"	177	"movl $-1024, %%esi \n\t"
178	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	178	"movq %2, %%mm7 \n\t"
179	"1: \n\t"	179	"1: \n\t"
180	"movq (%1, %%esi), %%mm0 \n\t"	180	"movq (%1, %%esi), %%mm0 \n\t"
181	"movq 8(%1, %%esi), %%mm1 \n\t"	181	"movq 8(%1, %%esi), %%mm1 \n\t"
Lines 216-222 static int a52_resample_2F_2R_to_4_MMX(f Link Here
216	"addl $16, %%esi \n\t"	216	"addl $16, %%esi \n\t"
217	" jnz 1b \n\t"	217	" jnz 1b \n\t"
218	"emms \n\t"	218	"emms \n\t"
219	:: "r" (s16+1024), "r" (f+256)	219	:: "r" (s16+1024), "r" (f+256), "m" (magicF2W)
220	:"%esi", "memory"	220	:"%esi", "memory"
221	);	221	);
222	return 4*256;	222	return 4*256;
Lines 226-232 static int a52_resample_3F_2R_to_5_MMX(f Link Here
226	int32_t * f = (int32_t *) _f;	226	int32_t * f = (int32_t *) _f;
227	asm volatile(	227	asm volatile(
228	"movl $-1024, %%esi \n\t"	228	"movl $-1024, %%esi \n\t"
229	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	229	"movq %2, %%mm7 \n\t"
230	"1: \n\t"	230	"1: \n\t"
231	"movd (%1, %%esi), %%mm0 \n\t"	231	"movd (%1, %%esi), %%mm0 \n\t"
232	"punpckldq 2048(%1, %%esi), %%mm0\n\t"	232	"punpckldq 2048(%1, %%esi), %%mm0\n\t"
Lines 275-281 static int a52_resample_3F_2R_to_5_MMX(f Link Here
275	"addl $16, %%esi \n\t"	275	"addl $16, %%esi \n\t"
276	" jnz 1b \n\t"	276	" jnz 1b \n\t"
277	"emms \n\t"	277	"emms \n\t"
278	:: "r" (s16+1280), "r" (f+256)	278	:: "r" (s16+1280), "r" (f+256), "m" (magicF2W)
279	:"%esi", "%edi", "memory"	279	:"%esi", "%edi", "memory"
280	);	280	);
281	return 5*256;	281	return 5*256;
Lines 285-291 static int a52_resample_MONO_LFE_to_6_MM Link Here
285	int32_t * f = (int32_t *) _f;	285	int32_t * f = (int32_t *) _f;
286	asm volatile(	286	asm volatile(
287	"movl $-1024, %%esi \n\t"	287	"movl $-1024, %%esi \n\t"
288	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	288	"movq %2, %%mm7 \n\t"
289	"pxor %%mm6, %%mm6 \n\t"	289	"pxor %%mm6, %%mm6 \n\t"
290	"1: \n\t"	290	"1: \n\t"
291	"movq 1024(%1, %%esi), %%mm0 \n\t"	291	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 315-321 static int a52_resample_MONO_LFE_to_6_MM Link Here
315	"addl $16, %%esi \n\t"	315	"addl $16, %%esi \n\t"
316	" jnz 1b \n\t"	316	" jnz 1b \n\t"
317	"emms \n\t"	317	"emms \n\t"
318	:: "r" (s16+1536), "r" (f+256)	318	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
319	:"%esi", "%edi", "memory"	319	:"%esi", "%edi", "memory"
320	);	320	);
321	return 6*256;	321	return 6*256;
Lines 325-331 static int a52_resample_STEREO_LFE_to_6_ Link Here
325	int32_t * f = (int32_t *) _f;	325	int32_t * f = (int32_t *) _f;
326	asm volatile(	326	asm volatile(
327	"movl $-1024, %%esi \n\t"	327	"movl $-1024, %%esi \n\t"
328	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	328	"movq %2, %%mm7 \n\t"
329	"pxor %%mm6, %%mm6 \n\t"	329	"pxor %%mm6, %%mm6 \n\t"
330	"1: \n\t"	330	"1: \n\t"
331	"movq 1024(%1, %%esi), %%mm0 \n\t"	331	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 353-359 static int a52_resample_STEREO_LFE_to_6_ Link Here
353	"addl $8, %%esi \n\t"	353	"addl $8, %%esi \n\t"
354	" jnz 1b \n\t"	354	" jnz 1b \n\t"
355	"emms \n\t"	355	"emms \n\t"
356	:: "r" (s16+1536), "r" (f+256)	356	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
357	:"%esi", "%edi", "memory"	357	:"%esi", "%edi", "memory"
358	);	358	);
359	return 6*256;	359	return 6*256;
Lines 363-369 static int a52_resample_3F_LFE_to_6_MMX( Link Here
363	int32_t * f = (int32_t *) _f;	363	int32_t * f = (int32_t *) _f;
364	asm volatile(	364	asm volatile(
365	"movl $-1024, %%esi \n\t"	365	"movl $-1024, %%esi \n\t"
366	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	366	"movq %2, %%mm7 \n\t"
367	"pxor %%mm6, %%mm6 \n\t"	367	"pxor %%mm6, %%mm6 \n\t"
368	"1: \n\t"	368	"1: \n\t"
369	"movq 1024(%1, %%esi), %%mm0 \n\t"	369	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 393-399 static int a52_resample_3F_LFE_to_6_MMX( Link Here
393	"addl $8, %%esi \n\t"	393	"addl $8, %%esi \n\t"
394	" jnz 1b \n\t"	394	" jnz 1b \n\t"
395	"emms \n\t"	395	"emms \n\t"
396	:: "r" (s16+1536), "r" (f+256)	396	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
397	:"%esi", "%edi", "memory"	397	:"%esi", "%edi", "memory"
398	);	398	);
399	return 6*256;	399	return 6*256;
Lines 403-409 static int a52_resample_2F_2R_LFE_to_6_M Link Here
403	int32_t * f = (int32_t *) _f;	403	int32_t * f = (int32_t *) _f;
404	asm volatile(	404	asm volatile(
405	"movl $-1024, %%esi \n\t"	405	"movl $-1024, %%esi \n\t"
406	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	406	"movq %2, %%mm7 \n\t"
407	// "pxor %%mm6, %%mm6 \n\t"	407	// "pxor %%mm6, %%mm6 \n\t"
408	"1: \n\t"	408	"1: \n\t"
409	"movq 1024(%1, %%esi), %%mm0 \n\t"	409	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 439-445 static int a52_resample_2F_2R_LFE_to_6_M Link Here
439	"addl $8, %%esi \n\t"	439	"addl $8, %%esi \n\t"
440	" jnz 1b \n\t"	440	" jnz 1b \n\t"
441	"emms \n\t"	441	"emms \n\t"
442	:: "r" (s16+1536), "r" (f+256)	442	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
443	:"%esi", "%edi", "memory"	443	:"%esi", "%edi", "memory"
444	);	444	);
445	return 6*256;	445	return 6*256;
Lines 449-455 static int a52_resample_3F_2R_LFE_to_6_M Link Here
449	int32_t * f = (int32_t *) _f;	449	int32_t * f = (int32_t *) _f;
450	asm volatile(	450	asm volatile(
451	"movl $-1024, %%esi \n\t"	451	"movl $-1024, %%esi \n\t"
452	"movq "MANGLE(magicF2W)", %%mm7 \n\t"	452	"movq %2, %%mm7 \n\t"
453	// "pxor %%mm6, %%mm6 \n\t"	453	// "pxor %%mm6, %%mm6 \n\t"
454	"1: \n\t"	454	"1: \n\t"
455	"movq 1024(%1, %%esi), %%mm0 \n\t"	455	"movq 1024(%1, %%esi), %%mm0 \n\t"
Lines 487-493 static int a52_resample_3F_2R_LFE_to_6_M Link Here
487	"addl $8, %%esi \n\t"	487	"addl $8, %%esi \n\t"
488	" jnz 1b \n\t"	488	" jnz 1b \n\t"
489	"emms \n\t"	489	"emms \n\t"
490	:: "r" (s16+1536), "r" (f+256)	490	:: "r" (s16+1536), "r" (f+256), "m" (magicF2W)
491	:"%esi", "%edi", "memory"	491	:"%esi", "%edi", "memory"
492	);	492	);
493	return 6*256;	493	return 6*256;

Lines 387-402 static inline void RENAME(vertRK1Filter) Link Here

(-)ffmpeg-old/libpostproc/postprocess_template.c (-78 / +80 lines)
387	// FIXME rounding	387	// FIXME rounding
388	asm volatile(	388	asm volatile(
389	"pxor %%mm7, %%mm7 \n\t" // 0	389	"pxor %%mm7, %%mm7 \n\t" // 0
390	"movq "MANGLE(b80)", %%mm6 \n\t" // MIN_SIGNED_BYTE	390	"movq %2, %%mm6 \n\t" // MIN_SIGNED_BYTE
391	"leal (%0, %1), %%"REG_a" \n\t"	391	"leal (%0, %1), %%"REG_a" \n\t"
392	"leal (%%"REG_a", %1, 4), %%"REG_c" \n\t"	392	"leal (%%"REG_a", %1, 4), %%"REG_c" \n\t"
393	// 0 1 2 3 4 5 6 7 8 9	393	// 0 1 2 3 4 5 6 7 8 9
394	// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1 %0+8%1 ecx+4%1	394	// %0 eax eax+%1 eax+2%1 %0+4%1 ecx ecx+%1 ecx+2%1 %0+8%1 ecx+4%1
395	"movq "MANGLE(pQPb)", %%mm0 \n\t" // QP,..., QP	395	"movq %3, %%mm0 \n\t" // QP,..., QP
396	"movq %%mm0, %%mm1 \n\t" // QP,..., QP	396	"movq %%mm0, %%mm1 \n\t" // QP,..., QP
397	"paddusb "MANGLE(b02)", %%mm0 \n\t"	397	"paddusb %4, %%mm0 \n\t"
398	"psrlw $2, %%mm0 \n\t"	398	"psrlw $2, %%mm0 \n\t"
399	"pand "MANGLE(b3F)", %%mm0 \n\t" // QP/4,..., QP/4	399	"pand %5, %%mm0 \n\t" // QP/4,..., QP/4
400	"paddusb %%mm1, %%mm0 \n\t" // QP*1.25 ...	400	"paddusb %%mm1, %%mm0 \n\t" // QP*1.25 ...
401	"movq (%0, %1, 4), %%mm2 \n\t" // line 4	401	"movq (%0, %1, 4), %%mm2 \n\t" // line 4
402	"movq (%%"REG_c"), %%mm3 \n\t" // line 5	402	"movq (%%"REG_c"), %%mm3 \n\t" // line 5
Lines 425-432 static inline void RENAME(vertRK1Filter) Link Here
425		425
426	"paddb %%mm6, %%mm5 \n\t"	426	"paddb %%mm6, %%mm5 \n\t"
427	"psrlw $2, %%mm5 \n\t"	427	"psrlw $2, %%mm5 \n\t"
428	"pand "MANGLE(b3F)", %%mm5 \n\t"	428	"pand %5, %%mm5 \n\t"
429	"psubb "MANGLE(b20)", %%mm5 \n\t" // (l5-l4)/8	429	"psubb %6, %%mm5 \n\t" // (l5-l4)/8
430		430
431	"movq (%%"REG_a", %1, 2), %%mm2 \n\t"	431	"movq (%%"REG_a", %1, 2), %%mm2 \n\t"
432	"paddb %%mm6, %%mm2 \n\t" // line 3 + 0x80	432	"paddb %%mm6, %%mm2 \n\t" // line 3 + 0x80
Lines 441-447 static inline void RENAME(vertRK1Filter) Link Here
441	"movq %%mm2, (%%"REG_c", %1) \n\t"	441	"movq %%mm2, (%%"REG_c", %1) \n\t"
442		442
443	:	443	:
444	: "r" (src), "r" ((long)stride)	444	: "r" (src), "r" ((long)stride), "m" (b80), "m" (pQPb), "m" (b02), "m" (b3F), "m" (b20)
445	: "%"REG_a, "%"REG_c	445	: "%"REG_a, "%"REG_c
446	);	446	);
447	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)	447	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)
Lines 517-523 static inline void RENAME(vertX1Filter)( Link Here
517	"paddusb %%mm0, %%mm0 \n\t"	517	"paddusb %%mm0, %%mm0 \n\t"
518	"psubusb %%mm0, %%mm4 \n\t"	518	"psubusb %%mm0, %%mm4 \n\t"
519	"pcmpeqb %%mm7, %%mm4 \n\t" // d <= QP ? -1 : 0	519	"pcmpeqb %%mm7, %%mm4 \n\t" // d <= QP ? -1 : 0
520	"psubusb "MANGLE(b01)", %%mm3 \n\t"	520	"psubusb %3, %%mm3 \n\t"
521	"pand %%mm4, %%mm3 \n\t" // d <= QP ? d : 0	521	"pand %%mm4, %%mm3 \n\t" // d <= QP ? d : 0
522		522
523	PAVGB(%%mm7, %%mm3) // d/2	523	PAVGB(%%mm7, %%mm3) // d/2
Lines 566-572 static inline void RENAME(vertX1Filter)( Link Here
566	"movq %%mm0, (%%"REG_c", %1, 2) \n\t" // line 7	566	"movq %%mm0, (%%"REG_c", %1, 2) \n\t" // line 7
567		567
568	:	568	:
569	: "r" (src), "r" ((long)stride), "m" (co->pQPb)	569	: "r" (src), "r" ((long)stride), "m" (co->pQPb), "m" (b01)
570	: "%"REG_a, "%"REG_c	570	: "%"REG_a, "%"REG_c
571	);	571	);
572	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)	572	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)
Lines 699-715 static inline void RENAME(doVertDefFilte Link Here
699		699
700	PMINUB(%%mm2, %%mm1, %%mm4) // MIN(\|lenergy\|,\|renergy\|)/8	700	PMINUB(%%mm2, %%mm1, %%mm4) // MIN(\|lenergy\|,\|renergy\|)/8
701	"movq %2, %%mm4 \n\t" // QP //FIXME QP+1 ?	701	"movq %2, %%mm4 \n\t" // QP //FIXME QP+1 ?
702	"paddusb "MANGLE(b01)", %%mm4 \n\t"	702	"paddusb %3, %%mm4 \n\t"
703	"pcmpgtb %%mm3, %%mm4 \n\t" // \|menergy\|/8 < QP	703	"pcmpgtb %%mm3, %%mm4 \n\t" // \|menergy\|/8 < QP
704	"psubusb %%mm1, %%mm3 \n\t" // d=\|menergy\|/8-MIN(\|lenergy\|,\|renergy\|)/8	704	"psubusb %%mm1, %%mm3 \n\t" // d=\|menergy\|/8-MIN(\|lenergy\|,\|renergy\|)/8
705	"pand %%mm4, %%mm3 \n\t"	705	"pand %%mm4, %%mm3 \n\t"
706		706
707	"movq %%mm3, %%mm1 \n\t"	707	"movq %%mm3, %%mm1 \n\t"
708	// "psubusb "MANGLE(b01)", %%mm3 \n\t"	708	// "psubusb %3, %%mm3 \n\t"
709	PAVGB(%%mm7, %%mm3)	709	PAVGB(%%mm7, %%mm3)
710	PAVGB(%%mm7, %%mm3)	710	PAVGB(%%mm7, %%mm3)
711	"paddusb %%mm1, %%mm3 \n\t"	711	"paddusb %%mm1, %%mm3 \n\t"
712	// "paddusb "MANGLE(b01)", %%mm3 \n\t"	712	// "paddusb %3, %%mm3 \n\t"
713		713
714	"movq (%%"REG_a", %1, 2), %%mm6 \n\t" //l3	714	"movq (%%"REG_a", %1, 2), %%mm6 \n\t" //l3
715	"movq (%0, %1, 4), %%mm5 \n\t" //l4	715	"movq (%0, %1, 4), %%mm5 \n\t" //l4
Lines 722-728 static inline void RENAME(doVertDefFilte Link Here
722	"pand %%mm0, %%mm3 \n\t"	722	"pand %%mm0, %%mm3 \n\t"
723	PMINUB(%%mm5, %%mm3, %%mm0)	723	PMINUB(%%mm5, %%mm3, %%mm0)
724		724
725	"psubusb "MANGLE(b01)", %%mm3 \n\t"	725	"psubusb %3, %%mm3 \n\t"
726	PAVGB(%%mm7, %%mm3)	726	PAVGB(%%mm7, %%mm3)
727		727
728	"movq (%%"REG_a", %1, 2), %%mm0 \n\t"	728	"movq (%%"REG_a", %1, 2), %%mm0 \n\t"
Lines 754-760 static inline void RENAME(doVertDefFilte Link Here
754	"movq (%%"REG_a", %1), %%mm3 \n\t" // l2	754	"movq (%%"REG_a", %1), %%mm3 \n\t" // l2
755	"pxor %%mm6, %%mm2 \n\t" // -l5-1	755	"pxor %%mm6, %%mm2 \n\t" // -l5-1
756	"movq %%mm2, %%mm5 \n\t" // -l5-1	756	"movq %%mm2, %%mm5 \n\t" // -l5-1
757	"movq "MANGLE(b80)", %%mm4 \n\t" // 128	757	"movq %4, %%mm4 \n\t" // 128
758	"lea (%%"REG_a", %1, 4), %%"REG_c" \n\t"	758	"lea (%%"REG_a", %1, 4), %%"REG_c" \n\t"
759	PAVGB(%%mm3, %%mm2) // (l2-l5+256)/2	759	PAVGB(%%mm3, %%mm2) // (l2-l5+256)/2
760	PAVGB(%%mm0, %%mm4) // ~(l4-l3)/4 + 128	760	PAVGB(%%mm0, %%mm4) // ~(l4-l3)/4 + 128
Lines 766-772 static inline void RENAME(doVertDefFilte Link Here
766	"pxor %%mm6, %%mm2 \n\t" // -l1-1	766	"pxor %%mm6, %%mm2 \n\t" // -l1-1
767	PAVGB(%%mm3, %%mm2) // (l2-l1+256)/2	767	PAVGB(%%mm3, %%mm2) // (l2-l1+256)/2
768	PAVGB((%0), %%mm1) // (l0-l3+256)/2	768	PAVGB((%0), %%mm1) // (l0-l3+256)/2
769	"movq "MANGLE(b80)", %%mm3 \n\t" // 128	769	"movq %4, %%mm3 \n\t" // 128
770	PAVGB(%%mm2, %%mm3) // ~(l2-l1)/4 + 128	770	PAVGB(%%mm2, %%mm3) // ~(l2-l1)/4 + 128
771	PAVGB(%%mm1, %%mm3) // ~(l0-l3)/4 +(l2-l1)/8 + 128	771	PAVGB(%%mm1, %%mm3) // ~(l0-l3)/4 +(l2-l1)/8 + 128
772	PAVGB(%%mm2, %%mm3) // ~(l0-l3)/8 +5(l2-l1)/16 + 128	772	PAVGB(%%mm2, %%mm3) // ~(l0-l3)/8 +5(l2-l1)/16 + 128
Lines 776-789 static inline void RENAME(doVertDefFilte Link Here
776	"movq (%%"REG_c", %1, 2), %%mm1 \n\t" // l7	776	"movq (%%"REG_c", %1, 2), %%mm1 \n\t" // l7
777	"pxor %%mm6, %%mm1 \n\t" // -l7-1	777	"pxor %%mm6, %%mm1 \n\t" // -l7-1
778	PAVGB((%0, %1, 4), %%mm1) // (l4-l7+256)/2	778	PAVGB((%0, %1, 4), %%mm1) // (l4-l7+256)/2
779	"movq "MANGLE(b80)", %%mm2 \n\t" // 128	779	"movq %4, %%mm2 \n\t" // 128
780	PAVGB(%%mm5, %%mm2) // ~(l6-l5)/4 + 128	780	PAVGB(%%mm5, %%mm2) // ~(l6-l5)/4 + 128
781	PAVGB(%%mm1, %%mm2) // ~(l4-l7)/4 +(l6-l5)/8 + 128	781	PAVGB(%%mm1, %%mm2) // ~(l4-l7)/4 +(l6-l5)/8 + 128
782	PAVGB(%%mm5, %%mm2) // ~(l4-l7)/8 +5(l6-l5)/16 + 128	782	PAVGB(%%mm5, %%mm2) // ~(l4-l7)/8 +5(l6-l5)/16 + 128
783	// mm0=128-q, mm2=renergy/16 + 128, mm3=lenergy/16 + 128, mm4= menergy/16 + 128	783	// mm0=128-q, mm2=renergy/16 + 128, mm3=lenergy/16 + 128, mm4= menergy/16 + 128
784		784
785	"movq "MANGLE(b00)", %%mm1 \n\t" // 0	785	"movq %5, %%mm1 \n\t" // 0
786	"movq "MANGLE(b00)", %%mm5 \n\t" // 0	786	"movq %5, %%mm5 \n\t" // 0
787	"psubb %%mm2, %%mm1 \n\t" // 128 - renergy/16	787	"psubb %%mm2, %%mm1 \n\t" // 128 - renergy/16
788	"psubb %%mm3, %%mm5 \n\t" // 128 - lenergy/16	788	"psubb %%mm3, %%mm5 \n\t" // 128 - lenergy/16
789	PMAXUB(%%mm1, %%mm2) // 128 + \|renergy/16\|	789	PMAXUB(%%mm1, %%mm2) // 128 + \|renergy/16\|
Lines 792-798 static inline void RENAME(doVertDefFilte Link Here
792		792
793	// mm0=128-q, mm3=128 + MIN(\|lenergy\|,\|renergy\|)/16, mm4= menergy/16 + 128	793	// mm0=128-q, mm3=128 + MIN(\|lenergy\|,\|renergy\|)/16, mm4= menergy/16 + 128
794		794
795	"movq "MANGLE(b00)", %%mm7 \n\t" // 0	795	"movq %5, %%mm7 \n\t" // 0
796	"movq %2, %%mm2 \n\t" // QP	796	"movq %2, %%mm2 \n\t" // QP
797	PAVGB(%%mm6, %%mm2) // 128 + QP/2	797	PAVGB(%%mm6, %%mm2) // 128 + QP/2
798	"psubb %%mm6, %%mm2 \n\t"	798	"psubb %%mm6, %%mm2 \n\t"
Lines 806-818 static inline void RENAME(doVertDefFilte Link Here
806	// mm0=128-q, mm1= SIGN(menergy), mm2= \|menergy\|/16 < QP/2, mm4= d/16	806	// mm0=128-q, mm1= SIGN(menergy), mm2= \|menergy\|/16 < QP/2, mm4= d/16
807		807
808	"movq %%mm4, %%mm3 \n\t" // d	808	"movq %%mm4, %%mm3 \n\t" // d
809	"psubusb "MANGLE(b01)", %%mm4 \n\t"	809	"psubusb %3, %%mm4 \n\t"
810	PAVGB(%%mm7, %%mm4) // d/32	810	PAVGB(%%mm7, %%mm4) // d/32
811	PAVGB(%%mm7, %%mm4) // (d + 32)/64	811	PAVGB(%%mm7, %%mm4) // (d + 32)/64
812	"paddb %%mm3, %%mm4 \n\t" // 5d/64	812	"paddb %%mm3, %%mm4 \n\t" // 5d/64
813	"pand %%mm2, %%mm4 \n\t"	813	"pand %%mm2, %%mm4 \n\t"
814		814
815	"movq "MANGLE(b80)", %%mm5 \n\t" // 128	815	"movq %4, %%mm5 \n\t" // 128
816	"psubb %%mm0, %%mm5 \n\t" // q	816	"psubb %%mm0, %%mm5 \n\t" // q
817	"paddsb %%mm6, %%mm5 \n\t" // fix bad rounding	817	"paddsb %%mm6, %%mm5 \n\t" // fix bad rounding
818	"pcmpgtb %%mm5, %%mm7 \n\t" // SIGN(q)	818	"pcmpgtb %%mm5, %%mm7 \n\t" // SIGN(q)
Lines 834-840 static inline void RENAME(doVertDefFilte Link Here
834	"movq %%mm2, (%0, %1, 4) \n\t"	834	"movq %%mm2, (%0, %1, 4) \n\t"
835		835
836	:	836	:
837	: "r" (src), "r" ((long)stride), "m" (c->pQPb)	837	: "r" (src), "r" ((long)stride), "m" (c->pQPb), "m" (b01), "m" (b80), "m" (b00)
838	: "%"REG_a, "%"REG_c	838	: "%"REG_a, "%"REG_c
839	);	839	);
840		840
Lines 1078-1087 src-=8; Link Here
1078	"psubusw %%mm1, %%mm5 \n\t" // ld	1078	"psubusw %%mm1, %%mm5 \n\t" // ld
1079		1079
1080		1080
1081	"movq "MANGLE(w05)", %%mm2 \n\t" // 5	1081	"movq %3, %%mm2 \n\t" // 5
1082	"pmullw %%mm2, %%mm4 \n\t"	1082	"pmullw %%mm2, %%mm4 \n\t"
1083	"pmullw %%mm2, %%mm5 \n\t"	1083	"pmullw %%mm2, %%mm5 \n\t"
1084	"movq "MANGLE(w20)", %%mm2 \n\t" // 32	1084	"movq %4, %%mm2 \n\t" // 32
1085	"paddw %%mm2, %%mm4 \n\t"	1085	"paddw %%mm2, %%mm4 \n\t"
1086	"paddw %%mm2, %%mm5 \n\t"	1086	"paddw %%mm2, %%mm5 \n\t"
1087	"psrlw $6, %%mm4 \n\t"	1087	"psrlw $6, %%mm4 \n\t"
Lines 1131-1137 src-=8; Link Here
1131	"movq %%mm0, (%0, %1) \n\t"	1131	"movq %%mm0, (%0, %1) \n\t"
1132		1132
1133	: "+r" (src)	1133	: "+r" (src)
1134	: "r" ((long)stride), "m" (c->pQPb)	1134	: "r" ((long)stride), "m" (c->pQPb), "m" (w05), "m" (w20)
1135	: "%"REG_a, "%"REG_c	1135	: "%"REG_a, "%"REG_c
1136	);	1136	);
1137	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)	1137	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)
Lines 1275-1281 FIND_MIN_MAX((%0, %1, 8)) Link Here
1275	"movq %%mm6, %%mm0 \n\t" // max	1275	"movq %%mm6, %%mm0 \n\t" // max
1276	"psubb %%mm7, %%mm6 \n\t" // max - min	1276	"psubb %%mm7, %%mm6 \n\t" // max - min
1277	"movd %%mm6, %%ecx \n\t"	1277	"movd %%mm6, %%ecx \n\t"
1278	"cmpb "MANGLE(deringThreshold)", %%cl \n\t"	1278	"cmpb %4, %%cl \n\t"
1279	" jb 1f \n\t"	1279	" jb 1f \n\t"
1280	"lea -24(%%"REG_SP"), %%"REG_c" \n\t"	1280	"lea -24(%%"REG_SP"), %%"REG_c" \n\t"
1281	"and "ALIGN_MASK", %%"REG_c" \n\t"	1281	"and "ALIGN_MASK", %%"REG_c" \n\t"
Lines 1302-1310 FIND_MIN_MAX((%0, %1, 8)) Link Here
1302	"psubusb %%mm7, %%mm0 \n\t"	1302	"psubusb %%mm7, %%mm0 \n\t"
1303	"psubusb %%mm7, %%mm2 \n\t"	1303	"psubusb %%mm7, %%mm2 \n\t"
1304	"psubusb %%mm7, %%mm3 \n\t"	1304	"psubusb %%mm7, %%mm3 \n\t"
1305	"pcmpeqb "MANGLE(b00)", %%mm0 \n\t" // L10 > a ? 0 : -1	1305	"pcmpeqb %5, %%mm0 \n\t" // L10 > a ? 0 : -1
1306	"pcmpeqb "MANGLE(b00)", %%mm2 \n\t" // L20 > a ? 0 : -1	1306	"pcmpeqb %5, %%mm2 \n\t" // L20 > a ? 0 : -1
1307	"pcmpeqb "MANGLE(b00)", %%mm3 \n\t" // L00 > a ? 0 : -1	1307	"pcmpeqb %5, %%mm3 \n\t" // L00 > a ? 0 : -1
1308	"paddb %%mm2, %%mm0 \n\t"	1308	"paddb %%mm2, %%mm0 \n\t"
1309	"paddb %%mm3, %%mm0 \n\t"	1309	"paddb %%mm3, %%mm0 \n\t"
1310		1310
Lines 1325-1333 FIND_MIN_MAX((%0, %1, 8)) Link Here
1325	"psubusb %%mm7, %%mm2 \n\t"	1325	"psubusb %%mm7, %%mm2 \n\t"
1326	"psubusb %%mm7, %%mm4 \n\t"	1326	"psubusb %%mm7, %%mm4 \n\t"
1327	"psubusb %%mm7, %%mm5 \n\t"	1327	"psubusb %%mm7, %%mm5 \n\t"
1328	"pcmpeqb "MANGLE(b00)", %%mm2 \n\t" // L11 > a ? 0 : -1	1328	"pcmpeqb %5, %%mm2 \n\t" // L11 > a ? 0 : -1
1329	"pcmpeqb "MANGLE(b00)", %%mm4 \n\t" // L21 > a ? 0 : -1	1329	"pcmpeqb %5, %%mm4 \n\t" // L21 > a ? 0 : -1
1330	"pcmpeqb "MANGLE(b00)", %%mm5 \n\t" // L01 > a ? 0 : -1	1330	"pcmpeqb %5, %%mm5 \n\t" // L01 > a ? 0 : -1
1331	"paddb %%mm4, %%mm2 \n\t"	1331	"paddb %%mm4, %%mm2 \n\t"
1332	"paddb %%mm5, %%mm2 \n\t"	1332	"paddb %%mm5, %%mm2 \n\t"
1333	// 0, 2, 3, 1	1333	// 0, 2, 3, 1
Lines 1352-1358 FIND_MIN_MAX((%0, %1, 8)) Link Here
1352	"psubusb " #lx ", " #t1 " \n\t"\	1352	"psubusb " #lx ", " #t1 " \n\t"\
1353	"psubusb " #lx ", " #t0 " \n\t"\	1353	"psubusb " #lx ", " #t0 " \n\t"\
1354	"psubusb " #lx ", " #sx " \n\t"\	1354	"psubusb " #lx ", " #sx " \n\t"\
1355	"movq "MANGLE(b00)", " #lx " \n\t"\	1355	"movq %5, " #lx " \n\t"\
1356	"pcmpeqb " #lx ", " #t1 " \n\t" /* src[-1] > a ? 0 : -1*/\	1356	"pcmpeqb " #lx ", " #t1 " \n\t" /* src[-1] > a ? 0 : -1*/\
1357	"pcmpeqb " #lx ", " #t0 " \n\t" /* src[+1] > a ? 0 : -1*/\	1357	"pcmpeqb " #lx ", " #t0 " \n\t" /* src[+1] > a ? 0 : -1*/\
1358	"pcmpeqb " #lx ", " #sx " \n\t" /* src[0] > a ? 0 : -1*/\	1358	"pcmpeqb " #lx ", " #sx " \n\t" /* src[0] > a ? 0 : -1*/\
Lines 1368-1375 FIND_MIN_MAX((%0, %1, 8)) Link Here
1368	PMINUB(t1, pplx, t0)\	1368	PMINUB(t1, pplx, t0)\
1369	"paddb " #sx ", " #ppsx " \n\t"\	1369	"paddb " #sx ", " #ppsx " \n\t"\
1370	"paddb " #psx ", " #ppsx " \n\t"\	1370	"paddb " #psx ", " #ppsx " \n\t"\
1371	"#paddb "MANGLE(b02)", " #ppsx " \n\t"\	1371	"#paddb %6, " #ppsx " \n\t"\
1372	"pand "MANGLE(b08)", " #ppsx " \n\t"\	1372	"pand %7, " #ppsx " \n\t"\
1373	"pcmpeqb " #lx ", " #ppsx " \n\t"\	1373	"pcmpeqb " #lx ", " #ppsx " \n\t"\
1374	"pand " #ppsx ", " #pplx " \n\t"\	1374	"pand " #ppsx ", " #pplx " \n\t"\
1375	"pandn " #dst ", " #ppsx " \n\t"\	1375	"pandn " #dst ", " #ppsx " \n\t"\
Lines 1405-1411 DERING_CORE((%%REGd, %1, 2),(%0, %1, 8) Link Here
1405	DERING_CORE((%0, %1, 8) ,(%%REGd, %1, 4),%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)	1405	DERING_CORE((%0, %1, 8) ,(%%REGd, %1, 4),%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
1406		1406
1407	"1: \n\t"	1407	"1: \n\t"
1408	: : "r" (src), "r" ((long)stride), "m" (c->pQPb), "m"(c->pQPb2)	1408	: : "r" (src), "r" ((long)stride), "m" (c->pQPb), "m"(c->pQPb2), "m" (deringThreshold), "m" (b00), "m" (b02), "m" (b08)
1409	: "%"REG_a, "%"REG_d, "%"REG_c	1409	: "%"REG_a, "%"REG_d, "%"REG_c
1410	);	1410	);
1411	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)	1411	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)
Lines 2283-2289 static inline void RENAME(tempNoiseReduc Link Here
2283	#else //L1_DIFF	2283	#else //L1_DIFF
2284	#if defined (FAST_L2_DIFF)	2284	#if defined (FAST_L2_DIFF)
2285	"pcmpeqb %%mm7, %%mm7 \n\t"	2285	"pcmpeqb %%mm7, %%mm7 \n\t"
2286	"movq "MANGLE(b80)", %%mm6 \n\t"	2286	"movq %4, %%mm6 \n\t"
2287	"pxor %%mm0, %%mm0 \n\t"	2287	"pxor %%mm0, %%mm0 \n\t"
2288	#define REAL_L2_DIFF_CORE(a, b)\	2288	#define REAL_L2_DIFF_CORE(a, b)\
2289	"movq " #a ", %%mm5 \n\t"\	2289	"movq " #a ", %%mm5 \n\t"\
Lines 2532-2538 L2_DIFF_CORE((%0, %%REGc) , (%1, %%REGc Link Here
2532		2532
2533	"4: \n\t"	2533	"4: \n\t"
2534		2534
2535	:: "r" (src), "r" (tempBlured), "r"((long)stride), "m" (tempBluredPast)	2535	:: "r" (src), "r" (tempBlured), "r"((long)stride), "m" (tempBluredPast), "m" (b80)
2536	: "%"REG_a, "%"REG_d, "%"REG_c, "memory"	2536	: "%"REG_a, "%"REG_d, "%"REG_c, "memory"
2537	);	2537	);
2538	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)	2538	#else //defined (HAVE_MMX2) \|\| defined (HAVE_3DNOW)
Lines 2805-2812 asm volatile( 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 asm volatile( 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 asm volatile( 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 asm volatile( 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 static inline void RENAME(blockCopy)(uin 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 3257-3278 static inline void RENAME(blockCopy)(uin 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	"lea (%1,%3,2), %1 \n\t"
3262	SCALED_CPY((%2, %4, 4), (%%REGa, %4, 4), (%3, %5, 4), (%%REGd, %5, 4))	3260	"lea (%2,%4,2), %2 \n\t"
3263	"lea (%%"REG_a",%4,4), %%"REG_a" \n\t"	3261	SCALED_CPY((%1), (%1, %3), (%2), (%2, %4))
3264	"lea (%%"REG_d",%5,4), %%"REG_d" \n\t"	3262	"lea (%1,%3,2), %1 \n\t"
3265	SCALED_CPY((%%REGa, %4), (%%REGa, %4, 2), (%%REGd, %5), (%%REGd, %5, 2))	3263	"lea (%2,%4,2), %2 \n\t"
3266		3264	SCALED_CPY((%1), (%1, %3), (%2), (%2, %4))
3267		3265	"lea (%1,%3,2), %1 \n\t"
3268	: "=&a" (packedOffsetAndScale)	3266	"lea (%2,%4,2), %2 \n\t"
3269	: "0" (packedOffsetAndScale),	3267	SCALED_CPY((%1), (%1, %3), (%2), (%2, %4))
3270	"r"(src),	3268
3271	"r"(dst),	3269	: "+r" (packedOffsetAndScale),
3272	"r" ((long)srcStride),	3270	"+r"(src),
3273	"r" ((long)dstStride)	3271	"+r"(dst)
3274	: "%"REG_d	3272	: "r" ((long)srcStride),
3275	);	3273	"r" ((long)dstStride)
		3274	: "memory"
		3275	);
3276	#else //HAVE_MMX	3276	#else //HAVE_MMX
3277	for(i=0; i<8; i++)	3277	for(i=0; i<8; i++)
3278	memcpy( &(dst[dstStride*i]),	3278	memcpy( &(dst[dstStride*i]),
Lines 3283-3290 SCALED_CPY((%%REGa, %4), (%%REGa, %4, 2) Link Here
3283	{	3283	{
3284	#ifdef HAVE_MMX	3284	#ifdef HAVE_MMX
3285	asm volatile(	3285	asm volatile(
3286	"lea (%0,%2), %%"REG_a" \n\t"
3287	"lea (%1,%3), %%"REG_d" \n\t"
3288		3286
3289	#define REAL_SIMPLE_CPY(src1, src2, dst1, dst2) \	3287	#define REAL_SIMPLE_CPY(src1, src2, dst1, dst2) \
3290	"movq " #src1 ", %%mm0 \n\t"\	3288	"movq " #src1 ", %%mm0 \n\t"\
Lines 3295-3312 SCALED_CPY((%%REGa, %4), (%%REGa, %4, 2) Link Here
3295	#define SIMPLE_CPY(src1, src2, dst1, dst2)\	3293	#define SIMPLE_CPY(src1, src2, dst1, dst2)\
3296	REAL_SIMPLE_CPY(src1, src2, dst1, dst2)	3294	REAL_SIMPLE_CPY(src1, src2, dst1, dst2)
3297		3295
3298	SIMPLE_CPY((%0) , (%0, %2) , (%1) , (%1, %3))	3296	SIMPLE_CPY((%0), (%0, %2), (%1), (%1, %3))
3299	SIMPLE_CPY((%0, %2, 2), (%%REGa, %2, 2), (%1, %3, 2), (%%REGd, %3, 2))	3297	"lea (%0,%2,2), %0 \n\t"
3300	SIMPLE_CPY((%0, %2, 4), (%%REGa, %2, 4), (%1, %3, 4), (%%REGd, %3, 4))	3298	"lea (%1,%3,2), %1 \n\t"
3301	"lea (%%"REG_a",%2,4), %%"REG_a" \n\t"	3299	SIMPLE_CPY((%0), (%0, %2), (%1), (%1, %3))
3302	"lea (%%"REG_d",%3,4), %%"REG_d" \n\t"	3300	"lea (%0,%2,2), %0 \n\t"
3303	SIMPLE_CPY((%%REGa, %2), (%%REGa, %2, 2), (%%REGd, %3), (%%REGd, %3, 2))	3301	"lea (%1,%3,2), %1 \n\t"
3304		3302	SIMPLE_CPY((%0), (%0, %2), (%1), (%1, %3))
3305	: : "r" (src),	3303	"lea (%0,%2), %0 \n\t"
3306	"r" (dst),	3304	"lea (%1,%3), %1 \n\t"
3307	"r" ((long)srcStride),	3305	SIMPLE_CPY((%0), (%0, %2), (%1), (%1, %3))
3308	"r" ((long)dstStride)	3306
3309	: "%"REG_a, "%"REG_d	3307	: "+r" (src),
		3308	"+r" (dst)
		3309	: "r" ((long)srcStride),
		3310	"r" ((long)dstStride)
		3311	: "memory"
3310	);	3312	);
3311	#else //HAVE_MMX	3313	#else //HAVE_MMX
3312	for(i=0; i<8; i++)	3314	for(i=0; i<8; i++)

Return to bug 115568

Lines 110-116 static int RENAME(dct_quantize)(MpegEncC Link Here

(-)ffmpeg-old/libavcodec/i386/mpegvideo_mmx_template.c (-7 / +14 lines)
110	SPREADW(%%mm3)	110	SPREADW(%%mm3)
111	"pxor %%mm7, %%mm7 \n\t" // 0	111	"pxor %%mm7, %%mm7 \n\t" // 0
112	"pxor %%mm4, %%mm4 \n\t" // 0	112	"pxor %%mm4, %%mm4 \n\t" // 0
113	"movq (%2), %%mm5 \n\t" // qmat[0]	113	"movq %2, %%mm5 \n\t" // qmat[0]
114	"pxor %%mm6, %%mm6 \n\t"	114	"pxor %%mm6, %%mm6 \n\t"
115	"psubw (%3), %%mm6 \n\t" // -bias[0]	115	"psubw (%3), %%mm6 \n\t" // -bias[0]
116	"mov $-128, %%"REG_a" \n\t"	116	"mov $-128, %%"REG_a" \n\t"
Lines 138-144 static int RENAME(dct_quantize)(MpegEncC Link Here
138	"movd %%mm3, %%"REG_a" \n\t"	138	"movd %%mm3, %%"REG_a" \n\t"
139	"movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1	139	"movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1
140	: "+a" (last_non_zero_p1)	140	: "+a" (last_non_zero_p1)
141	: "r" (block+64), "r" (qmat), "r" (bias),	141	: "r" (block+64), "m" (qmat), "r" (bias),
142	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)	142	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)
143	);	143	);
144	// note the asm is split cuz gcc doesnt like that many operands ...	144	// note the asm is split cuz gcc doesnt like that many operands ...
Lines 153-167 static int RENAME(dct_quantize)(MpegEncC Link Here
153	);	153	);
154	}else{ // FMT_H263	154	}else{ // FMT_H263
155	asm volatile(	155	asm volatile(
156	"movd %%"REG_a", %%mm3 \n\t" // last_non_zero_p1	156	"movd %0, %%mm3 \n\t" // last_non_zero_p1
157	SPREADW(%%mm3)	157	SPREADW(%%mm3)
158	"pxor %%mm7, %%mm7 \n\t" // 0	158	"pxor %%mm7, %%mm7 \n\t" // 0
159	"pxor %%mm4, %%mm4 \n\t" // 0	159	"pxor %%mm4, %%mm4 \n\t" // 0
		160	"push %%"REG_a" \n\t"
160	"mov $-128, %%"REG_a" \n\t"	161	"mov $-128, %%"REG_a" \n\t"
161	ASMALIGN(4)	162	ASMALIGN(4)
162	"1: \n\t"	163	"1: \n\t"
163	"pxor %%mm1, %%mm1 \n\t" // 0	164	"pxor %%mm1, %%mm1 \n\t" // 0
164	"movq (%1, %%"REG_a"), %%mm0 \n\t" // block[i]	165	"push %%"REG_a" \n\t"
		166	"add %c6(%%"REG_SP"), %%"REG_a" \n\t"
		167	"movq (%%"REG_a"), %%mm0 \n\t" // block[i]
		168	"pop %%"REG_a" \n\t"
165	"pcmpgtw %%mm0, %%mm1 \n\t" // block[i] <= 0 ? 0xFF : 0x00	169	"pcmpgtw %%mm0, %%mm1 \n\t" // block[i] <= 0 ? 0xFF : 0x00
166	"pxor %%mm1, %%mm0 \n\t"	170	"pxor %%mm1, %%mm0 \n\t"
167	"psubw %%mm1, %%mm0 \n\t" // ABS(block[i])	171	"psubw %%mm1, %%mm0 \n\t" // ABS(block[i])
Lines 183-191 static int RENAME(dct_quantize)(MpegEncC Link Here
183	PMAX(%%mm3, %%mm0)	187	PMAX(%%mm3, %%mm0)
184	"movd %%mm3, %%"REG_a" \n\t"	188	"movd %%mm3, %%"REG_a" \n\t"
185	"movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1	189	"movzb %%al, %%"REG_a" \n\t" // last_non_zero_p1
186	: "+a" (last_non_zero_p1)	190	"mov %%"REG_a", %0 \n\t"
187	: "r" (block+64), "r" (qmat+64), "r" (bias+64),	191	"pop %%"REG_a" \n\t"
188	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)	192	: "+m" (last_non_zero_p1)
		193	: "a" (block+64), "r" (qmat+64), "r" (bias+64),
		194	"r" (inv_zigzag_direct16+64), "r" (temp_block+64),
		195	"i" (sizeof(long))
189	);	196	);
190	// note the asm is split cuz gcc doesnt like that many operands ...	197	// note the asm is split cuz gcc doesnt like that many operands ...
191	asm volatile(	198	asm volatile(

Lines 629-638 void ff_snow_vertical_compose97i_mmx(DWT Link Here

(-)ffmpeg-old/libavcodec/i386/snowdsp_mmx.c (-12 / +10 lines)
629		629
630	#define snow_inner_add_yblock_sse2_header \	630	#define snow_inner_add_yblock_sse2_header \
631	DWTELEM * * dst_array = sb->line + src_y;\	631	DWTELEM * * dst_array = sb->line + src_y;\
632	long tmp;\	632	long tmp = b_h;\
633	asm volatile(\	633	asm volatile(\
634	"mov %7, %%"REG_c" \n\t"\	634	"mov %6, %%"REG_c" \n\t"\
635	"mov %6, %2 \n\t"\
636	"mov %4, %%"REG_S" \n\t"\	635	"mov %4, %%"REG_S" \n\t"\
637	"pxor %%xmm7, %%xmm7 \n\t" /* 0 */\	636	"pxor %%xmm7, %%xmm7 \n\t" /* 0 */\
638	"pcmpeqd %%xmm3, %%xmm3 \n\t"\	637	"pcmpeqd %%xmm3, %%xmm3 \n\t"\
Lines 689-697 void ff_snow_vertical_compose97i_mmx(DWT Link Here
689		688
690	#define snow_inner_add_yblock_sse2_end_common2\	689	#define snow_inner_add_yblock_sse2_end_common2\
691	"jnz 1b \n\t"\	690	"jnz 1b \n\t"\
692	:"+m"(dst8),"+m"(dst_array),"=&r"(tmp)\	691	:"+m"(dst8),"+m"(dst_array),"=m"(tmp)\
693	:\	692	:\
694	"rm"((long)(src_x<<2)),"m"(obmc),"a"(block),"m"((long)b_h),"m"((long)src_stride):\	693	"rm"((long)(src_x<<2)),"m"(obmc),"a"(block),"m"((long)src_stride):\
695	"%"REG_c"","%"REG_S"","%"REG_D"","%"REG_d"");	694	"%"REG_c"","%"REG_S"","%"REG_D"","%"REG_d"");
696		695
697	#define snow_inner_add_yblock_sse2_end_8\	696	#define snow_inner_add_yblock_sse2_end_8\
Lines 705-711 void ff_snow_vertical_compose97i_mmx(DWT Link Here
705	#define snow_inner_add_yblock_sse2_end_16\	704	#define snow_inner_add_yblock_sse2_end_16\
706	"add $"PTR_SIZE"*1, %1 \n\t"\	705	"add $"PTR_SIZE"*1, %1 \n\t"\
707	snow_inner_add_yblock_sse2_end_common1\	706	snow_inner_add_yblock_sse2_end_common1\
708	"dec %2 \n\t"\	707	"sub $1, %2 \n\t"\
709	snow_inner_add_yblock_sse2_end_common2	708	snow_inner_add_yblock_sse2_end_common2
710		709
711	static void inner_add_yblock_bw_8_obmc_16_bh_even_sse2(uint8_t obmc, const long obmc_stride, uint8_t * block, int b_w, long b_h,	710	static void inner_add_yblock_bw_8_obmc_16_bh_even_sse2(uint8_t obmc, const long obmc_stride, uint8_t * block, int b_w, long b_h,
Lines 795-804 snow_inner_add_yblock_sse2_end_16 Link Here
795		794
796	#define snow_inner_add_yblock_mmx_header \	795	#define snow_inner_add_yblock_mmx_header \
797	DWTELEM * * dst_array = sb->line + src_y;\	796	DWTELEM * * dst_array = sb->line + src_y;\
798	long tmp;\	797	long tmp = b_h;\
799	asm volatile(\	798	asm volatile(\
800	"mov %7, %%"REG_c" \n\t"\	799	"mov %6, %%"REG_c" \n\t"\
801	"mov %6, %2 \n\t"\
802	"mov %4, %%"REG_S" \n\t"\	800	"mov %4, %%"REG_S" \n\t"\
803	"pxor %%mm7, %%mm7 \n\t" /* 0 */\	801	"pxor %%mm7, %%mm7 \n\t" /* 0 */\
804	"pcmpeqd %%mm3, %%mm3 \n\t"\	802	"pcmpeqd %%mm3, %%mm3 \n\t"\
Lines 861-871 snow_inner_add_yblock_sse2_end_16 Link Here
861	"add %%"REG_c", (%%"REG_a") \n\t"\	859	"add %%"REG_c", (%%"REG_a") \n\t"\
862	"add $"PTR_SIZE"*1, %1 \n\t"\	860	"add $"PTR_SIZE"*1, %1 \n\t"\
863	"add %%"REG_c", %0 \n\t"\	861	"add %%"REG_c", %0 \n\t"\
864	"dec %2 \n\t"\	862	"sub $1, %2 \n\t"\
865	"jnz 1b \n\t"\	863	"jnz 1b \n\t"\
866	:"+m"(dst8),"+m"(dst_array),"=&r"(tmp)\	864	:"+m"(dst8),"+m"(dst_array),"=m"(tmp)\
867	:\	865	:\
868	"rm"((long)(src_x<<2)),"m"(obmc),"a"(block),"m"((long)b_h),"m"((long)src_stride):\	866	"rm"((long)(src_x<<2)),"m"(obmc),"a"(block),"m"((long)src_stride):\
869	"%"REG_c"","%"REG_S"","%"REG_D"","%"REG_d"");	867	"%"REG_c"","%"REG_S"","%"REG_D"","%"REG_d"");
870		868
871	static void inner_add_yblock_bw_8_obmc_16_mmx(uint8_t obmc, const long obmc_stride, uint8_t * block, int b_w, long b_h,	869	static void inner_add_yblock_bw_8_obmc_16_mmx(uint8_t obmc, const long obmc_stride, uint8_t * block, int b_w, long b_h,