Mercurial > libavcodec.hg

comparison i386/h264dsp_mmx.c @ 6334:bf8c9e82c7ad libavcodec

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
fix aliasing warnings. simpler too.
author lorenm
date Wed, 06 Feb 2008 04:14:07 +0000
parents c57670e07668
children 950811a14eb3
comparison
equal deleted inserted replaced
6333:beb52d4a5efe 6334:bf8c9e82c7ad
1679 static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1679 static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1680 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src+1, stride, stride);\ 1680 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src+1, stride, stride);\
1681 }\ 1681 }\
1682 \ 1682 \
1683 static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1683 static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1684 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*SIZE/8]);\ 1684 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*SIZE]);\
1685 uint8_t * const half= (uint8_t*)temp;\ 1685 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
1686 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(half, src, SIZE, stride);\ 1686 OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, temp, stride, stride, SIZE);\
1687 OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, half, stride, stride, SIZE);\
1688 }\ 1687 }\
1689 \ 1688 \
1690 static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1689 static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1691 OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## MMX(dst, src, stride, stride);\ 1690 OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## MMX(dst, src, stride, stride);\
1692 }\ 1691 }\
1693 \ 1692 \
1694 static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1693 static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1695 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*SIZE/8]);\ 1694 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*SIZE]);\
1696 uint8_t * const half= (uint8_t*)temp;\ 1695 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
1697 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(half, src, SIZE, stride);\ 1696 OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+stride, temp, stride, stride, SIZE);\
1698 OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+stride, half, stride, stride, SIZE);\
1699 }\ 1697 }\
1700 \ 1698 \
1701 static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1699 static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1702 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*SIZE/8]);\ 1700 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*SIZE]);\
1703 uint8_t * const halfV= (uint8_t*)temp;\ 1701 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
1704 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src, SIZE, stride);\ 1702 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
1705 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfV, stride, SIZE);\
1706 }\ 1703 }\
1707 \ 1704 \
1708 static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1705 static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1709 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*SIZE/8]);\ 1706 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*SIZE]);\
1710 uint8_t * const halfV= (uint8_t*)temp;\ 1707 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
1711 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src+1, SIZE, stride);\ 1708 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
1712 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfV, stride, SIZE);\
1713 }\ 1709 }\
1714 \ 1710 \
1715 static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1711 static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1716 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*SIZE/8]);\ 1712 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*SIZE]);\
1717 uint8_t * const halfV= (uint8_t*)temp;\ 1713 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
1718 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src, SIZE, stride);\ 1714 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
1719 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfV, stride, SIZE);\
1720 }\ 1715 }\
1721 \ 1716 \
1722 static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1717 static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1723 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*SIZE/8]);\ 1718 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*SIZE]);\
1724 uint8_t * const halfV= (uint8_t*)temp;\ 1719 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
1725 put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src+1, SIZE, stride);\ 1720 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
1726 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfV, stride, SIZE);\
1727 }\ 1721 }\
1728 \ 1722 \
1729 static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1723 static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1730 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*(SIZE<8?12:24)/4]);\ 1724 DECLARE_ALIGNED(ALIGN, uint16_t, temp[SIZE*(SIZE<8?12:24)]);\
1731 int16_t * const tmp= (int16_t*)temp;\ 1725 OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, temp, src, stride, SIZE, stride);\
1732 OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, tmp, src, stride, SIZE, stride);\
1733 }\ 1726 }\
1734 \ 1727 \
1735 static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1728 static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1736 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*(SIZE<8?12:24)/4 + SIZE*SIZE/8]);\ 1729 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
1737 uint8_t * const halfHV= (uint8_t*)temp;\ 1730 uint8_t * const halfHV= temp;\
1738 int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE/2;\ 1731 int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
1739 assert(((int)temp & 7) == 0);\ 1732 assert(((int)temp & 7) == 0);\
1740 put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\ 1733 put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
1741 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfHV, stride, SIZE);\ 1734 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfHV, stride, SIZE);\
1742 }\ 1735 }\
1743 \ 1736 \
1744 static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1737 static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1745 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*(SIZE<8?12:24)/4 + SIZE*SIZE/8]);\ 1738 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
1746 uint8_t * const halfHV= (uint8_t*)temp;\ 1739 uint8_t * const halfHV= temp;\
1747 int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE/2;\ 1740 int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
1748 assert(((int)temp & 7) == 0);\ 1741 assert(((int)temp & 7) == 0);\
1749 put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\ 1742 put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
1750 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfHV, stride, SIZE);\ 1743 OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfHV, stride, SIZE);\
1751 }\ 1744 }\
1752 \ 1745 \
1753 static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1746 static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1754 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*(SIZE<8?12:24)/4 + SIZE*SIZE/8]);\ 1747 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
1755 int16_t * const halfV= ((int16_t*)temp) + SIZE*SIZE/2;\ 1748 uint8_t * const halfHV= temp;\
1756 uint8_t * const halfHV= ((uint8_t*)temp);\ 1749 int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
1757 assert(((int)temp & 7) == 0);\ 1750 assert(((int)temp & 7) == 0);\
1758 put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\ 1751 put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
1759 OPNAME ## pixels ## SIZE ## _l2_shift5_ ## MMX(dst, halfV+2, halfHV, stride, SIZE, SIZE);\ 1752 OPNAME ## pixels ## SIZE ## _l2_shift5_ ## MMX(dst, halfV+2, halfHV, stride, SIZE, SIZE);\
1760 }\ 1753 }\
1761 \ 1754 \
1762 static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\ 1755 static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
1763 DECLARE_ALIGNED(ALIGN, uint64_t, temp[SIZE*(SIZE<8?12:24)/4 + SIZE*SIZE/8]);\ 1756 DECLARE_ALIGNED(ALIGN, uint8_t, temp[SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
1764 int16_t * const halfV= ((int16_t*)temp) + SIZE*SIZE/2;\ 1757 uint8_t * const halfHV= temp;\
1765 uint8_t * const halfHV= ((uint8_t*)temp);\ 1758 int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
1766 assert(((int)temp & 7) == 0);\ 1759 assert(((int)temp & 7) == 0);\
1767 put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\ 1760 put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
1768 OPNAME ## pixels ## SIZE ## _l2_shift5_ ## MMX(dst, halfV+3, halfHV, stride, SIZE, SIZE);\ 1761 OPNAME ## pixels ## SIZE ## _l2_shift5_ ## MMX(dst, halfV+3, halfHV, stride, SIZE, SIZE);\
1769 }\ 1762 }\
1770 1763