Mercurial > mplayer.hg
comparison libswscale/rgb2rgb_template.c @ 32137:b7d5e57af959
swscale: avoid reading prior to the source buffer in planar2x() MMX2
| author | ramiro |
|---|---|
| date | Mon, 13 Sep 2010 14:25:18 +0000 |
| parents | 67f44db4fee9 |
| children | 355ca1a08ea2 |
comparison
equal
deleted
inserted
replaced
| 32136:effad4f941ff | 32137:b7d5e57af959 |
|---|---|
| 1771 for (y=1; y<srcHeight; y++) { | 1771 for (y=1; y<srcHeight; y++) { |
| 1772 #if HAVE_MMX2 || HAVE_AMD3DNOW | 1772 #if HAVE_MMX2 || HAVE_AMD3DNOW |
| 1773 const x86_reg mmxSize= srcWidth&~15; | 1773 const x86_reg mmxSize= srcWidth&~15; |
| 1774 __asm__ volatile( | 1774 __asm__ volatile( |
| 1775 "mov %4, %%"REG_a" \n\t" | 1775 "mov %4, %%"REG_a" \n\t" |
| 1776 "movq "MANGLE(mmx_ff)", %%mm0 \n\t" | |
| 1777 "movq (%0, %%"REG_a"), %%mm4 \n\t" | |
| 1778 "movq %%mm4, %%mm2 \n\t" | |
| 1779 "psllq $8, %%mm4 \n\t" | |
| 1780 "pand %%mm0, %%mm2 \n\t" | |
| 1781 "por %%mm2, %%mm4 \n\t" | |
| 1782 "movq (%1, %%"REG_a"), %%mm5 \n\t" | |
| 1783 "movq %%mm5, %%mm3 \n\t" | |
| 1784 "psllq $8, %%mm5 \n\t" | |
| 1785 "pand %%mm0, %%mm3 \n\t" | |
| 1786 "por %%mm3, %%mm5 \n\t" | |
| 1776 "1: \n\t" | 1787 "1: \n\t" |
| 1777 "movq (%0, %%"REG_a"), %%mm0 \n\t" | 1788 "movq (%0, %%"REG_a"), %%mm0 \n\t" |
| 1778 "movq (%1, %%"REG_a"), %%mm1 \n\t" | 1789 "movq (%1, %%"REG_a"), %%mm1 \n\t" |
| 1779 "movq 1(%0, %%"REG_a"), %%mm2 \n\t" | 1790 "movq 1(%0, %%"REG_a"), %%mm2 \n\t" |
| 1780 "movq 1(%1, %%"REG_a"), %%mm3 \n\t" | 1791 "movq 1(%1, %%"REG_a"), %%mm3 \n\t" |
| 1781 "movq -1(%0, %%"REG_a"), %%mm4 \n\t" | |
| 1782 "movq -1(%1, %%"REG_a"), %%mm5 \n\t" | |
| 1783 PAVGB" %%mm0, %%mm5 \n\t" | 1792 PAVGB" %%mm0, %%mm5 \n\t" |
| 1784 PAVGB" %%mm0, %%mm3 \n\t" | 1793 PAVGB" %%mm0, %%mm3 \n\t" |
| 1785 PAVGB" %%mm0, %%mm5 \n\t" | 1794 PAVGB" %%mm0, %%mm5 \n\t" |
| 1786 PAVGB" %%mm0, %%mm3 \n\t" | 1795 PAVGB" %%mm0, %%mm3 \n\t" |
| 1787 PAVGB" %%mm1, %%mm4 \n\t" | 1796 PAVGB" %%mm1, %%mm4 \n\t" |
| 1804 "movq %%mm7, 8(%2, %%"REG_a", 2) \n\t" | 1813 "movq %%mm7, 8(%2, %%"REG_a", 2) \n\t" |
| 1805 "movq %%mm4, (%3, %%"REG_a", 2) \n\t" | 1814 "movq %%mm4, (%3, %%"REG_a", 2) \n\t" |
| 1806 "movq %%mm6, 8(%3, %%"REG_a", 2) \n\t" | 1815 "movq %%mm6, 8(%3, %%"REG_a", 2) \n\t" |
| 1807 #endif | 1816 #endif |
| 1808 "add $8, %%"REG_a" \n\t" | 1817 "add $8, %%"REG_a" \n\t" |
| 1818 "movq -1(%0, %%"REG_a"), %%mm4 \n\t" | |
| 1819 "movq -1(%1, %%"REG_a"), %%mm5 \n\t" | |
| 1809 " js 1b \n\t" | 1820 " js 1b \n\t" |
| 1810 :: "r" (src + mmxSize ), "r" (src + srcStride + mmxSize ), | 1821 :: "r" (src + mmxSize ), "r" (src + srcStride + mmxSize ), |
| 1811 "r" (dst + mmxSize*2), "r" (dst + dstStride + mmxSize*2), | 1822 "r" (dst + mmxSize*2), "r" (dst + dstStride + mmxSize*2), |
| 1812 "g" (-mmxSize) | 1823 "g" (-mmxSize) |
| 1813 : "%"REG_a | 1824 : "%"REG_a |
| 1814 | 1825 |
| 1815 ); | 1826 ); |
| 1816 #else | 1827 #else |
| 1817 const x86_reg mmxSize=1; | 1828 const x86_reg mmxSize=1; |
| 1818 #endif | |
| 1819 dst[0 ]= (3*src[0] + src[srcStride])>>2; | 1829 dst[0 ]= (3*src[0] + src[srcStride])>>2; |
| 1820 dst[dstStride]= ( src[0] + 3*src[srcStride])>>2; | 1830 dst[dstStride]= ( src[0] + 3*src[srcStride])>>2; |
| 1831 #endif | |
| 1821 | 1832 |
| 1822 for (x=mmxSize-1; x<srcWidth-1; x++) { | 1833 for (x=mmxSize-1; x<srcWidth-1; x++) { |
| 1823 dst[2*x +1]= (3*src[x+0] + src[x+srcStride+1])>>2; | 1834 dst[2*x +1]= (3*src[x+0] + src[x+srcStride+1])>>2; |
| 1824 dst[2*x+dstStride+2]= ( src[x+0] + 3*src[x+srcStride+1])>>2; | 1835 dst[2*x+dstStride+2]= ( src[x+0] + 3*src[x+srcStride+1])>>2; |
| 1825 dst[2*x+dstStride+1]= ( src[x+1] + 3*src[x+srcStride ])>>2; | 1836 dst[2*x+dstStride+1]= ( src[x+1] + 3*src[x+srcStride ])>>2; |