yading@10: /*
yading@10:  * Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt
yading@10:  * Copyright (c) 2011 Daniel Kang
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: #include "libavutil/attributes.h"
yading@10: #include "libavutil/cpu.h"
yading@10: #include "libavutil/x86/asm.h"
yading@10: #include "libavutil/x86/cpu.h"
yading@10: #include "libavcodec/h264qpel.h"
yading@10: #include "libavcodec/mpegvideo.h"
yading@10: #include "dsputil_mmx.h"
yading@10: 
yading@10: #if HAVE_YASM
yading@10: void ff_put_pixels4_mmxext(uint8_t *block, const uint8_t *pixels,
yading@10:                            ptrdiff_t line_size, int h);
yading@10: void ff_avg_pixels4_mmxext(uint8_t *block, const uint8_t *pixels,
yading@10:                            ptrdiff_t line_size, int h);
yading@10: static void ff_put_pixels16_mmxext(uint8_t *block, const uint8_t *pixels,
yading@10:                                    ptrdiff_t line_size, int h)
yading@10: {
yading@10:     ff_put_pixels8_mmxext(block,     pixels,     line_size, h);
yading@10:     ff_put_pixels8_mmxext(block + 8, pixels + 8, line_size, h);
yading@10: }
yading@10: static void ff_avg_pixels16_mmxext(uint8_t *block, const uint8_t *pixels,
yading@10:                                    ptrdiff_t line_size, int h)
yading@10: {
yading@10:     ff_avg_pixels8_mmxext(block,     pixels,     line_size, h);
yading@10:     ff_avg_pixels8_mmxext(block + 8, pixels + 8, line_size, h);
yading@10: }
yading@10: void ff_put_pixels4_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
yading@10:                               int dstStride, int src1Stride, int h);
yading@10: void ff_avg_pixels4_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
yading@10:                               int dstStride, int src1Stride, int h);
yading@10: void ff_put_pixels8_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
yading@10:                               int dstStride, int src1Stride, int h);
yading@10: void ff_avg_pixels8_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
yading@10:                               int dstStride, int src1Stride, int h);
yading@10: void ff_put_pixels16_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
yading@10:                                int dstStride, int src1Stride, int h);
yading@10: void ff_avg_pixels16_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2,
yading@10:                                int dstStride, int src1Stride, int h);
yading@10: #define ff_put_pixels8_l2_sse2  ff_put_pixels8_l2_mmxext
yading@10: #define ff_avg_pixels8_l2_sse2  ff_avg_pixels8_l2_mmxext
yading@10: #define ff_put_pixels16_l2_sse2 ff_put_pixels16_l2_mmxext
yading@10: #define ff_avg_pixels16_l2_sse2 ff_avg_pixels16_l2_mmxext
yading@10: 
yading@10: #define DEF_QPEL(OPNAME)\
yading@10: void ff_ ## OPNAME ## _h264_qpel4_h_lowpass_mmxext(uint8_t *dst, uint8_t *src, int dstStride, int srcStride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_mmxext(uint8_t *dst, uint8_t *src, int dstStride, int srcStride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_ssse3(uint8_t *dst, uint8_t *src, int dstStride, int srcStride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel4_h_lowpass_l2_mmxext(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_l2_mmxext(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_l2_ssse3(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel4_v_lowpass_mmxext(uint8_t *dst, uint8_t *src, int dstStride, int srcStride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8or16_v_lowpass_op_mmxext(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8or16_v_lowpass_sse2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h);\
yading@10: void ff_ ## OPNAME ## _h264_qpel4_hv_lowpass_v_mmxext(uint8_t *src, int16_t *tmp, int srcStride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel4_hv_lowpass_h_mmxext(int16_t *tmp, uint8_t *dst, int dstStride);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8or16_hv1_lowpass_op_mmxext(uint8_t *src, int16_t *tmp, int srcStride, int size);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8or16_hv1_lowpass_op_sse2(uint8_t *src, int16_t *tmp, int srcStride, int size);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8or16_hv2_lowpass_op_mmxext(uint8_t *dst, int16_t *tmp, int dstStride, int unused, int h);\
yading@10: void ff_ ## OPNAME ## _h264_qpel8or16_hv2_lowpass_ssse3(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size);\
yading@10: void ff_ ## OPNAME ## _pixels4_l2_shift5_mmxext(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h);\
yading@10: void ff_ ## OPNAME ## _pixels8_l2_shift5_mmxext(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h);
yading@10: 
yading@10: DEF_QPEL(avg)
yading@10: DEF_QPEL(put)
yading@10: 
yading@10: #define QPEL_H264(OPNAME, OP, MMX)\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
yading@10:     int w=3;\
yading@10:     src -= 2*srcStride+2;\
yading@10:     while(w--){\
yading@10:         ff_ ## OPNAME ## h264_qpel4_hv_lowpass_v_mmxext(src, tmp, srcStride);\
yading@10:         tmp += 4;\
yading@10:         src += 4;\
yading@10:     }\
yading@10:     tmp -= 3*4;\
yading@10:     ff_ ## OPNAME ## h264_qpel4_hv_lowpass_h_mmxext(tmp, dst, dstStride);\
yading@10: }\
yading@10: \
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
yading@10:     src -= 2*srcStride;\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_op_mmxext(dst, src, dstStride, srcStride, h);\
yading@10:     src += 4;\
yading@10:     dst += 4;\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_op_mmxext(dst, src, dstStride, srcStride, h);\
yading@10: }\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv1_lowpass_ ## MMX(int16_t *tmp, uint8_t *src, int tmpStride, int srcStride, int size){\
yading@10:     int w = (size+8)>>2;\
yading@10:     src -= 2*srcStride+2;\
yading@10:     while(w--){\
yading@10:         ff_ ## OPNAME ## h264_qpel8or16_hv1_lowpass_op_mmxext(src, tmp, srcStride, size);\
yading@10:         tmp += 4;\
yading@10:         src += 4;\
yading@10:     }\
yading@10: }\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size){\
yading@10:     int w = size>>4;\
yading@10:     do{\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_op_mmxext(dst, tmp, dstStride, 0, size);\
yading@10:     tmp += 8;\
yading@10:     dst += 8;\
yading@10:     }while(w--);\
yading@10: }\
yading@10: \
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
yading@10: }\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
yading@10: }\
yading@10: \
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
yading@10:     src += 8*srcStride;\
yading@10:     dst += 8*dstStride;\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
yading@10: }\
yading@10: \
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
yading@10:     src += 8*dstStride;\
yading@10:     dst += 8*dstStride;\
yading@10:     src2 += 8*src2Stride;\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
yading@10: }\
yading@10: \
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
yading@10:     ff_put_h264_qpel8or16_hv1_lowpass_ ## MMX(tmp, src, tmpStride, srcStride, size);\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
yading@10: }\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 8);\
yading@10: }\
yading@10: \
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 16);\
yading@10: }\
yading@10: \
yading@10: static av_always_inline void ff_ ## OPNAME ## pixels16_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
yading@10: {\
yading@10:     ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst  , src16  , src8  , dstStride, src8Stride, h);\
yading@10:     ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst+8, src16+8, src8+8, dstStride, src8Stride, h);\
yading@10: }\
yading@10: 
yading@10: 
yading@10: #if ARCH_X86_64
yading@10: #define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
yading@10: 
yading@10: void ff_avg_h264_qpel16_h_lowpass_l2_ssse3(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);
yading@10: void ff_put_h264_qpel16_h_lowpass_l2_ssse3(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride);
yading@10: 
yading@10: #else // ARCH_X86_64
yading@10: #define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
yading@10:     src += 8*dstStride;\
yading@10:     dst += 8*dstStride;\
yading@10:     src2 += 8*src2Stride;\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
yading@10: }
yading@10: #endif // ARCH_X86_64
yading@10: 
yading@10: #define QPEL_H264_H_XMM(OPNAME, OP, MMX)\
yading@10: QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
yading@10:     src += 8*srcStride;\
yading@10:     dst += 8*dstStride;\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
yading@10:     ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
yading@10: }\
yading@10: 
yading@10: #define QPEL_H264_V_XMM(OPNAME, OP, MMX)\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
yading@10: }\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
yading@10: }
yading@10: 
yading@10: static av_always_inline void ff_put_h264_qpel8or16_hv1_lowpass_sse2(int16_t *tmp, uint8_t *src, int tmpStride, int srcStride, int size){
yading@10:     int w = (size+8)>>3;
yading@10:     src -= 2*srcStride+2;
yading@10:     while(w--){
yading@10:         ff_put_h264_qpel8or16_hv1_lowpass_op_sse2(src, tmp, srcStride, size);
yading@10:         tmp += 8;
yading@10:         src += 8;
yading@10:     }
yading@10: }
yading@10: 
yading@10: #define QPEL_H264_HV_XMM(OPNAME, OP, MMX)\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
yading@10:     ff_put_h264_qpel8or16_hv1_lowpass_sse2(tmp, src, tmpStride, srcStride, size);\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
yading@10: }\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 8);\
yading@10: }\
yading@10: static av_always_inline void ff_ ## OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
yading@10:     ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 16);\
yading@10: }\
yading@10: 
yading@10: #define ff_put_h264_qpel8_h_lowpass_l2_sse2  ff_put_h264_qpel8_h_lowpass_l2_mmxext
yading@10: #define ff_avg_h264_qpel8_h_lowpass_l2_sse2  ff_avg_h264_qpel8_h_lowpass_l2_mmxext
yading@10: #define ff_put_h264_qpel16_h_lowpass_l2_sse2 ff_put_h264_qpel16_h_lowpass_l2_mmxext
yading@10: #define ff_avg_h264_qpel16_h_lowpass_l2_sse2 ff_avg_h264_qpel16_h_lowpass_l2_mmxext
yading@10: 
yading@10: #define ff_put_h264_qpel8_v_lowpass_ssse3  ff_put_h264_qpel8_v_lowpass_sse2
yading@10: #define ff_avg_h264_qpel8_v_lowpass_ssse3  ff_avg_h264_qpel8_v_lowpass_sse2
yading@10: #define ff_put_h264_qpel16_v_lowpass_ssse3 ff_put_h264_qpel16_v_lowpass_sse2
yading@10: #define ff_avg_h264_qpel16_v_lowpass_ssse3 ff_avg_h264_qpel16_v_lowpass_sse2
yading@10: 
yading@10: #define ff_put_h264_qpel8or16_hv2_lowpass_sse2 ff_put_h264_qpel8or16_hv2_lowpass_mmxext
yading@10: #define ff_avg_h264_qpel8or16_hv2_lowpass_sse2 ff_avg_h264_qpel8or16_hv2_lowpass_mmxext
yading@10: 
yading@10: #define H264_MC(OPNAME, SIZE, MMX, ALIGN) \
yading@10: H264_MC_C(OPNAME, SIZE, MMX, ALIGN)\
yading@10: H264_MC_V(OPNAME, SIZE, MMX, ALIGN)\
yading@10: H264_MC_H(OPNAME, SIZE, MMX, ALIGN)\
yading@10: H264_MC_HV(OPNAME, SIZE, MMX, ALIGN)\
yading@10: 
yading@10: static void put_h264_qpel16_mc00_sse2 (uint8_t *dst, uint8_t *src,
yading@10:                                        ptrdiff_t stride)
yading@10: {
yading@10:     ff_put_pixels16_sse2(dst, src, stride, 16);
yading@10: }
yading@10: static void avg_h264_qpel16_mc00_sse2 (uint8_t *dst, uint8_t *src,
yading@10:                                        ptrdiff_t stride)
yading@10: {
yading@10:     ff_avg_pixels16_sse2(dst, src, stride, 16);
yading@10: }
yading@10: #define put_h264_qpel8_mc00_sse2 put_h264_qpel8_mc00_mmxext
yading@10: #define avg_h264_qpel8_mc00_sse2 avg_h264_qpel8_mc00_mmxext
yading@10: 
yading@10: #define H264_MC_C(OPNAME, SIZE, MMX, ALIGN) \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## MMX (uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     ff_ ## OPNAME ## pixels ## SIZE ## _ ## MMX(dst, src, stride, SIZE);\
yading@10: }\
yading@10: 
yading@10: #define H264_MC_H(OPNAME, SIZE, MMX, ALIGN) \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src, stride, stride);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## MMX(dst, src, stride, stride);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src+1, stride, stride);\
yading@10: }\
yading@10: 
yading@10: #define H264_MC_V(OPNAME, SIZE, MMX, ALIGN) \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
yading@10:     ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, temp, stride, stride, SIZE);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## MMX(dst, src, stride, stride);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
yading@10:     ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+stride, temp, stride, stride, SIZE);\
yading@10: }\
yading@10: 
yading@10: #define H264_MC_HV(OPNAME, SIZE, MMX, ALIGN) \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
yading@10:     ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
yading@10:     ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
yading@10:     ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*SIZE];\
yading@10:     ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint16_t, temp)[SIZE*(SIZE<8?12:24)];\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, temp, src, stride, SIZE, stride);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE];\
yading@10:     uint8_t * const halfHV= temp;\
yading@10:     int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
yading@10:     av_assert2(((int)temp & 7) == 0);\
yading@10:     ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfHV, stride, SIZE);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE];\
yading@10:     uint8_t * const halfHV= temp;\
yading@10:     int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
yading@10:     av_assert2(((int)temp & 7) == 0);\
yading@10:     ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfHV, stride, SIZE);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE];\
yading@10:     uint8_t * const halfHV= temp;\
yading@10:     int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
yading@10:     av_assert2(((int)temp & 7) == 0);\
yading@10:     ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## pixels ## SIZE ## _l2_shift5_mmxext(dst, halfV+2, halfHV, stride, SIZE, SIZE);\
yading@10: }\
yading@10: \
yading@10: static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
yading@10: {\
yading@10:     DECLARE_ALIGNED(ALIGN, uint8_t, temp)[SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE];\
yading@10:     uint8_t * const halfHV= temp;\
yading@10:     int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
yading@10:     av_assert2(((int)temp & 7) == 0);\
yading@10:     ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
yading@10:     ff_ ## OPNAME ## pixels ## SIZE ## _l2_shift5_mmxext(dst, halfV+3, halfHV, stride, SIZE, SIZE);\
yading@10: }\
yading@10: 
yading@10: #define H264_MC_4816(MMX)\
yading@10: H264_MC(put_, 4, MMX, 8)\
yading@10: H264_MC(put_, 8, MMX, 8)\
yading@10: H264_MC(put_, 16,MMX, 8)\
yading@10: H264_MC(avg_, 4, MMX, 8)\
yading@10: H264_MC(avg_, 8, MMX, 8)\
yading@10: H264_MC(avg_, 16,MMX, 8)\
yading@10: 
yading@10: #define H264_MC_816(QPEL, XMM)\
yading@10: QPEL(put_, 8, XMM, 16)\
yading@10: QPEL(put_, 16,XMM, 16)\
yading@10: QPEL(avg_, 8, XMM, 16)\
yading@10: QPEL(avg_, 16,XMM, 16)\
yading@10: 
yading@10: QPEL_H264(put_,        PUT_OP, mmxext)
yading@10: QPEL_H264(avg_, AVG_MMXEXT_OP, mmxext)
yading@10: QPEL_H264_V_XMM(put_,       PUT_OP, sse2)
yading@10: QPEL_H264_V_XMM(avg_,AVG_MMXEXT_OP, sse2)
yading@10: QPEL_H264_HV_XMM(put_,       PUT_OP, sse2)
yading@10: QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, sse2)
yading@10: QPEL_H264_H_XMM(put_,       PUT_OP, ssse3)
yading@10: QPEL_H264_H_XMM(avg_,AVG_MMXEXT_OP, ssse3)
yading@10: QPEL_H264_HV_XMM(put_,       PUT_OP, ssse3)
yading@10: QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, ssse3)
yading@10: 
yading@10: H264_MC_4816(mmxext)
yading@10: H264_MC_816(H264_MC_V, sse2)
yading@10: H264_MC_816(H264_MC_HV, sse2)
yading@10: H264_MC_816(H264_MC_H, ssse3)
yading@10: H264_MC_816(H264_MC_HV, ssse3)
yading@10: 
yading@10: 
yading@10: //10bit
yading@10: #define LUMA_MC_OP(OP, NUM, DEPTH, TYPE, OPT) \
yading@10: void ff_ ## OP ## _h264_qpel ## NUM ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT \
yading@10:     (uint8_t *dst, uint8_t *src, ptrdiff_t stride);
yading@10: 
yading@10: #define LUMA_MC_ALL(DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(put,  4, DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(avg,  4, DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(put,  8, DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(avg,  8, DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(put, 16, DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(avg, 16, DEPTH, TYPE, OPT)
yading@10: 
yading@10: #define LUMA_MC_816(DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(put,  8, DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(avg,  8, DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(put, 16, DEPTH, TYPE, OPT) \
yading@10:     LUMA_MC_OP(avg, 16, DEPTH, TYPE, OPT)
yading@10: 
yading@10: LUMA_MC_ALL(10, mc00, mmxext)
yading@10: LUMA_MC_ALL(10, mc10, mmxext)
yading@10: LUMA_MC_ALL(10, mc20, mmxext)
yading@10: LUMA_MC_ALL(10, mc30, mmxext)
yading@10: LUMA_MC_ALL(10, mc01, mmxext)
yading@10: LUMA_MC_ALL(10, mc11, mmxext)
yading@10: LUMA_MC_ALL(10, mc21, mmxext)
yading@10: LUMA_MC_ALL(10, mc31, mmxext)
yading@10: LUMA_MC_ALL(10, mc02, mmxext)
yading@10: LUMA_MC_ALL(10, mc12, mmxext)
yading@10: LUMA_MC_ALL(10, mc22, mmxext)
yading@10: LUMA_MC_ALL(10, mc32, mmxext)
yading@10: LUMA_MC_ALL(10, mc03, mmxext)
yading@10: LUMA_MC_ALL(10, mc13, mmxext)
yading@10: LUMA_MC_ALL(10, mc23, mmxext)
yading@10: LUMA_MC_ALL(10, mc33, mmxext)
yading@10: 
yading@10: LUMA_MC_816(10, mc00, sse2)
yading@10: LUMA_MC_816(10, mc10, sse2)
yading@10: LUMA_MC_816(10, mc10, sse2_cache64)
yading@10: LUMA_MC_816(10, mc10, ssse3_cache64)
yading@10: LUMA_MC_816(10, mc20, sse2)
yading@10: LUMA_MC_816(10, mc20, sse2_cache64)
yading@10: LUMA_MC_816(10, mc20, ssse3_cache64)
yading@10: LUMA_MC_816(10, mc30, sse2)
yading@10: LUMA_MC_816(10, mc30, sse2_cache64)
yading@10: LUMA_MC_816(10, mc30, ssse3_cache64)
yading@10: LUMA_MC_816(10, mc01, sse2)
yading@10: LUMA_MC_816(10, mc11, sse2)
yading@10: LUMA_MC_816(10, mc21, sse2)
yading@10: LUMA_MC_816(10, mc31, sse2)
yading@10: LUMA_MC_816(10, mc02, sse2)
yading@10: LUMA_MC_816(10, mc12, sse2)
yading@10: LUMA_MC_816(10, mc22, sse2)
yading@10: LUMA_MC_816(10, mc32, sse2)
yading@10: LUMA_MC_816(10, mc03, sse2)
yading@10: LUMA_MC_816(10, mc13, sse2)
yading@10: LUMA_MC_816(10, mc23, sse2)
yading@10: LUMA_MC_816(10, mc33, sse2)
yading@10: 
yading@10: #define QPEL16_OPMC(OP, MC, MMX)\
yading@10: void ff_ ## OP ## _h264_qpel16_ ## MC ## _10_ ## MMX(uint8_t *dst, uint8_t *src, ptrdiff_t stride){\
yading@10:     ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst   , src   , stride);\
yading@10:     ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst+16, src+16, stride);\
yading@10:     src += 8*stride;\
yading@10:     dst += 8*stride;\
yading@10:     ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst   , src   , stride);\
yading@10:     ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst+16, src+16, stride);\
yading@10: }
yading@10: 
yading@10: #define QPEL16_OP(MC, MMX)\
yading@10: QPEL16_OPMC(put, MC, MMX)\
yading@10: QPEL16_OPMC(avg, MC, MMX)
yading@10: 
yading@10: #define QPEL16(MMX)\
yading@10: QPEL16_OP(mc00, MMX)\
yading@10: QPEL16_OP(mc01, MMX)\
yading@10: QPEL16_OP(mc02, MMX)\
yading@10: QPEL16_OP(mc03, MMX)\
yading@10: QPEL16_OP(mc10, MMX)\
yading@10: QPEL16_OP(mc11, MMX)\
yading@10: QPEL16_OP(mc12, MMX)\
yading@10: QPEL16_OP(mc13, MMX)\
yading@10: QPEL16_OP(mc20, MMX)\
yading@10: QPEL16_OP(mc21, MMX)\
yading@10: QPEL16_OP(mc22, MMX)\
yading@10: QPEL16_OP(mc23, MMX)\
yading@10: QPEL16_OP(mc30, MMX)\
yading@10: QPEL16_OP(mc31, MMX)\
yading@10: QPEL16_OP(mc32, MMX)\
yading@10: QPEL16_OP(mc33, MMX)
yading@10: 
yading@10: #if ARCH_X86_32 && HAVE_YASM && CONFIG_H264QPEL // ARCH_X86_64 implies SSE2+
yading@10: QPEL16(mmxext)
yading@10: #endif
yading@10: 
yading@10: #endif /* HAVE_YASM */
yading@10: 
yading@10: #define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX)                          \
yading@10:     do {                                                                     \
yading@10:     c->PFX ## _pixels_tab[IDX][ 0] = PREFIX ## PFX ## SIZE ## _mc00_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 1] = PREFIX ## PFX ## SIZE ## _mc10_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 2] = PREFIX ## PFX ## SIZE ## _mc20_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 3] = PREFIX ## PFX ## SIZE ## _mc30_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 4] = PREFIX ## PFX ## SIZE ## _mc01_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 5] = PREFIX ## PFX ## SIZE ## _mc11_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 6] = PREFIX ## PFX ## SIZE ## _mc21_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 7] = PREFIX ## PFX ## SIZE ## _mc31_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 8] = PREFIX ## PFX ## SIZE ## _mc02_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][ 9] = PREFIX ## PFX ## SIZE ## _mc12_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][10] = PREFIX ## PFX ## SIZE ## _mc22_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][11] = PREFIX ## PFX ## SIZE ## _mc32_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][12] = PREFIX ## PFX ## SIZE ## _mc03_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][13] = PREFIX ## PFX ## SIZE ## _mc13_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][14] = PREFIX ## PFX ## SIZE ## _mc23_ ## CPU; \
yading@10:     c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU; \
yading@10:     } while (0)
yading@10: 
yading@10: #define H264_QPEL_FUNCS(x, y, CPU)                                                            \
yading@10:     do {                                                                                      \
yading@10:         c->put_h264_qpel_pixels_tab[0][x + y * 4] = put_h264_qpel16_mc ## x ## y ## _ ## CPU; \
yading@10:         c->put_h264_qpel_pixels_tab[1][x + y * 4] = put_h264_qpel8_mc  ## x ## y ## _ ## CPU; \
yading@10:         c->avg_h264_qpel_pixels_tab[0][x + y * 4] = avg_h264_qpel16_mc ## x ## y ## _ ## CPU; \
yading@10:         c->avg_h264_qpel_pixels_tab[1][x + y * 4] = avg_h264_qpel8_mc  ## x ## y ## _ ## CPU; \
yading@10:     } while (0)
yading@10: 
yading@10: #define H264_QPEL_FUNCS_10(x, y, CPU)                                                               \
yading@10:     do {                                                                                            \
yading@10:         c->put_h264_qpel_pixels_tab[0][x + y * 4] = ff_put_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \
yading@10:         c->put_h264_qpel_pixels_tab[1][x + y * 4] = ff_put_h264_qpel8_mc  ## x ## y ## _10_ ## CPU; \
yading@10:         c->avg_h264_qpel_pixels_tab[0][x + y * 4] = ff_avg_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \
yading@10:         c->avg_h264_qpel_pixels_tab[1][x + y * 4] = ff_avg_h264_qpel8_mc  ## x ## y ## _10_ ## CPU; \
yading@10:     } while (0)
yading@10: 
yading@10: av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
yading@10: {
yading@10: #if HAVE_YASM
yading@10:     int high_bit_depth = bit_depth > 8;
yading@10:     int mm_flags = av_get_cpu_flags();
yading@10: 
yading@10:     if (EXTERNAL_MMXEXT(mm_flags)) {
yading@10:         if (!high_bit_depth) {
yading@10:             SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmxext, );
yading@10:             SET_QPEL_FUNCS(put_h264_qpel, 1,  8, mmxext, );
yading@10:             SET_QPEL_FUNCS(put_h264_qpel, 2,  4, mmxext, );
yading@10:             SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmxext, );
yading@10:             SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, mmxext, );
yading@10:             SET_QPEL_FUNCS(avg_h264_qpel, 2,  4, mmxext, );
yading@10:         } else if (bit_depth == 10) {
yading@10: #if ARCH_X86_32
yading@10:             SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_);
yading@10:             SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_);
yading@10:             SET_QPEL_FUNCS(put_h264_qpel, 1,  8, 10_mmxext, ff_);
yading@10:             SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, 10_mmxext, ff_);
yading@10: #endif
yading@10:             SET_QPEL_FUNCS(put_h264_qpel, 2, 4,  10_mmxext, ff_);
yading@10:             SET_QPEL_FUNCS(avg_h264_qpel, 2, 4,  10_mmxext, ff_);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     if (EXTERNAL_SSE2(mm_flags)) {
yading@10:         if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW) && !high_bit_depth) {
yading@10:             // these functions are slower than mmx on AMD, but faster on Intel
yading@10:             H264_QPEL_FUNCS(0, 0, sse2);
yading@10:         }
yading@10: 
yading@10:         if (!high_bit_depth) {
yading@10:             H264_QPEL_FUNCS(0, 1, sse2);
yading@10:             H264_QPEL_FUNCS(0, 2, sse2);
yading@10:             H264_QPEL_FUNCS(0, 3, sse2);
yading@10:             H264_QPEL_FUNCS(1, 1, sse2);
yading@10:             H264_QPEL_FUNCS(1, 2, sse2);
yading@10:             H264_QPEL_FUNCS(1, 3, sse2);
yading@10:             H264_QPEL_FUNCS(2, 1, sse2);
yading@10:             H264_QPEL_FUNCS(2, 2, sse2);
yading@10:             H264_QPEL_FUNCS(2, 3, sse2);
yading@10:             H264_QPEL_FUNCS(3, 1, sse2);
yading@10:             H264_QPEL_FUNCS(3, 2, sse2);
yading@10:             H264_QPEL_FUNCS(3, 3, sse2);
yading@10:         }
yading@10: 
yading@10:         if (bit_depth == 10) {
yading@10:             SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_);
yading@10:             SET_QPEL_FUNCS(put_h264_qpel, 1,  8, 10_sse2, ff_);
yading@10:             SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_sse2, ff_);
yading@10:             SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, 10_sse2, ff_);
yading@10:             H264_QPEL_FUNCS_10(1, 0, sse2_cache64);
yading@10:             H264_QPEL_FUNCS_10(2, 0, sse2_cache64);
yading@10:             H264_QPEL_FUNCS_10(3, 0, sse2_cache64);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     if (EXTERNAL_SSSE3(mm_flags)) {
yading@10:         if (!high_bit_depth) {
yading@10:             H264_QPEL_FUNCS(1, 0, ssse3);
yading@10:             H264_QPEL_FUNCS(1, 1, ssse3);
yading@10:             H264_QPEL_FUNCS(1, 2, ssse3);
yading@10:             H264_QPEL_FUNCS(1, 3, ssse3);
yading@10:             H264_QPEL_FUNCS(2, 0, ssse3);
yading@10:             H264_QPEL_FUNCS(2, 1, ssse3);
yading@10:             H264_QPEL_FUNCS(2, 2, ssse3);
yading@10:             H264_QPEL_FUNCS(2, 3, ssse3);
yading@10:             H264_QPEL_FUNCS(3, 0, ssse3);
yading@10:             H264_QPEL_FUNCS(3, 1, ssse3);
yading@10:             H264_QPEL_FUNCS(3, 2, ssse3);
yading@10:             H264_QPEL_FUNCS(3, 3, ssse3);
yading@10:         }
yading@10: 
yading@10:         if (bit_depth == 10) {
yading@10:             H264_QPEL_FUNCS_10(1, 0, ssse3_cache64);
yading@10:             H264_QPEL_FUNCS_10(2, 0, ssse3_cache64);
yading@10:             H264_QPEL_FUNCS_10(3, 0, ssse3_cache64);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     if (EXTERNAL_AVX(mm_flags)) {
yading@10:         /* AVX implies 64 byte cache lines without the need to avoid unaligned
yading@10:          * memory accesses that cross the boundary between two cache lines.
yading@10:          * TODO: Port X264_CPU_CACHELINE_32/64 detection from x264 to avoid
yading@10:          * having to treat SSE2 functions with such properties as AVX. */
yading@10:         if (bit_depth == 10) {
yading@10:             H264_QPEL_FUNCS_10(1, 0, sse2);
yading@10:             H264_QPEL_FUNCS_10(2, 0, sse2);
yading@10:             H264_QPEL_FUNCS_10(3, 0, sse2);
yading@10:         }
yading@10:     }
yading@10: #endif
yading@10: }