MrBesen/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
faa1471ffc
ffmpeg/libswresample/x86/resample_mmx.h
Ronald S. Bultje faa1471ffc swr: rewrite resample_common/linear_float_sse/avx in yasm. 
Linear interpolation goes from 63 (llvm) or 58 (gcc) to 48 (yasm)
cycles/sample on 64bit, or from 66 (llvm/gcc) to 52 (yasm) cycles/
sample on 32bit. Bon-linear goes from 43 (llvm) or 38 (gcc) to
32 (yasm) cycles/sample on 64bit, or from 46 (llvm) or 44 (gcc) to
38 (yasm) cycles/sample on 32bit (all testing on OSX 10.9.2, llvm
5.1 and gcc 4.8/9).

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2014-06-28 17:06:47 +02:00

183 lines
6.5 KiB
C
Raw Blame History

/*
* Copyright (c) 2012 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/x86/asm.h"
#include "libavutil/cpu.h"
#include "libswresample/swresample_internal.h"
DECLARE_ALIGNED(16, const uint64_t, ff_resample_int16_rounder)[2] = { 0x0000000000004000ULL, 0x0000000000000000ULL};
#define COMMON_CORE_INT16_MMX2 \
x86_reg len= -2*c->filter_length;\
__asm__ volatile(\
"movq "MANGLE(ff_resample_int16_rounder)", %%mm0 \n\t"\
"1: \n\t"\
"movq (%1, %0), %%mm1 \n\t"\
"pmaddwd (%2, %0), %%mm1 \n\t"\
"paddd %%mm1, %%mm0 \n\t"\
"add $8, %0 \n\t"\
" js 1b \n\t"\
"pshufw $0x0E, %%mm0, %%mm1 \n\t"\
"paddd %%mm1, %%mm0 \n\t"\
"psrad $15, %%mm0 \n\t"\
"packssdw %%mm0, %%mm0 \n\t"\
"movd %%mm0, (%3) \n\t"\
: "+r" (len)\
: "r" (((uint8_t*)(src+sample_index))-len),\
"r" (((uint8_t*)filter)-len),\
"r" (dst+dst_index)\
NAMED_CONSTRAINTS_ARRAY_ADD(ff_resample_int16_rounder)\
);
#define LINEAR_CORE_INT16_MMX2 \
x86_reg len= -2*c->filter_length;\
__asm__ volatile(\
"pxor %%mm0, %%mm0 \n\t"\
"pxor %%mm2, %%mm2 \n\t"\
"1: \n\t"\
"movq (%3, %0), %%mm1 \n\t"\
"movq %%mm1, %%mm3 \n\t"\
"pmaddwd (%4, %0), %%mm1 \n\t"\
"pmaddwd (%5, %0), %%mm3 \n\t"\
"paddd %%mm1, %%mm0 \n\t"\
"paddd %%mm3, %%mm2 \n\t"\
"add $8, %0 \n\t"\
" js 1b \n\t"\
"pshufw $0x0E, %%mm0, %%mm1 \n\t"\
"pshufw $0x0E, %%mm2, %%mm3 \n\t"\
"paddd %%mm1, %%mm0 \n\t"\
"paddd %%mm3, %%mm2 \n\t"\
"movd %%mm0, %1 \n\t"\
"movd %%mm2, %2 \n\t"\
: "+r" (len),\
"=r" (val),\
"=r" (v2)\
: "r" (((uint8_t*)(src+sample_index))-len),\
"r" (((uint8_t*)filter)-len),\
"r" (((uint8_t*)(filter+c->filter_alloc))-len)\
);
#define COMMON_CORE_INT16_SSE2 \
x86_reg len= -2*c->filter_length;\
__asm__ volatile(\
"movdqa "MANGLE(ff_resample_int16_rounder)", %%xmm0 \n\t"\
"1: \n\t"\
"movdqu (%1, %0), %%xmm1 \n\t"\
"pmaddwd (%2, %0), %%xmm1 \n\t"\
"paddd %%xmm1, %%xmm0 \n\t"\
"add $16, %0 \n\t"\
" js 1b \n\t"\
"pshufd $0x0E, %%xmm0, %%xmm1 \n\t"\
"paddd %%xmm1, %%xmm0 \n\t"\
"pshufd $0x01, %%xmm0, %%xmm1 \n\t"\
"paddd %%xmm1, %%xmm0 \n\t"\
"psrad $15, %%xmm0 \n\t"\
"packssdw %%xmm0, %%xmm0 \n\t"\
"movd %%xmm0, (%3) \n\t"\
: "+r" (len)\
: "r" (((uint8_t*)(src+sample_index))-len),\
"r" (((uint8_t*)filter)-len),\
"r" (dst+dst_index)\
NAMED_CONSTRAINTS_ARRAY_ADD(ff_resample_int16_rounder)\
XMM_CLOBBERS_ONLY("%xmm0", "%xmm1")\
);
#define LINEAR_CORE_INT16_SSE2 \
x86_reg len= -2*c->filter_length;\
__asm__ volatile(\
"pxor %%xmm0, %%xmm0 \n\t"\
"pxor %%xmm2, %%xmm2 \n\t"\
"1: \n\t"\
"movdqu (%3, %0), %%xmm1 \n\t"\
"movdqa %%xmm1, %%xmm3 \n\t"\
"pmaddwd (%4, %0), %%xmm1 \n\t"\
"pmaddwd (%5, %0), %%xmm3 \n\t"\
"paddd %%xmm1, %%xmm0 \n\t"\
"paddd %%xmm3, %%xmm2 \n\t"\
"add $16, %0 \n\t"\
" js 1b \n\t"\
"pshufd $0x0E, %%xmm0, %%xmm1 \n\t"\
"pshufd $0x0E, %%xmm2, %%xmm3 \n\t"\
"paddd %%xmm1, %%xmm0 \n\t"\
"paddd %%xmm3, %%xmm2 \n\t"\
"pshufd $0x01, %%xmm0, %%xmm1 \n\t"\
"pshufd $0x01, %%xmm2, %%xmm3 \n\t"\
"paddd %%xmm1, %%xmm0 \n\t"\
"paddd %%xmm3, %%xmm2 \n\t"\
"movd %%xmm0, %1 \n\t"\
"movd %%xmm2, %2 \n\t"\
: "+r" (len),\
"=r" (val),\
"=r" (v2)\
: "r" (((uint8_t*)(src+sample_index))-len),\
"r" (((uint8_t*)filter)-len),\
"r" (((uint8_t*)(filter+c->filter_alloc))-len)\
XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3")\
);
#define COMMON_CORE_DBL_SSE2 \
x86_reg len= -8*c->filter_length;\
__asm__ volatile(\
"xorpd %%xmm0, %%xmm0 \n\t"\
"1: \n\t"\
"movupd (%1, %0), %%xmm1 \n\t"\
"mulpd (%2, %0), %%xmm1 \n\t"\
"addpd %%xmm1, %%xmm0 \n\t"\
"add $16, %0 \n\t"\
" js 1b \n\t"\
"movhlps %%xmm0, %%xmm1 \n\t"\
"addpd %%xmm1, %%xmm0 \n\t"\
"movsd %%xmm0, (%3) \n\t"\
: "+r" (len)\
: "r" (((uint8_t*)(src+sample_index))-len),\
"r" (((uint8_t*)filter)-len),\
"r" (dst+dst_index)\
XMM_CLOBBERS_ONLY("%xmm0", "%xmm1")\
);
#define LINEAR_CORE_DBL_SSE2 \
x86_reg len= -8*c->filter_length;\
__asm__ volatile(\
"xorpd %%xmm0, %%xmm0 \n\t"\
"xorpd %%xmm2, %%xmm2 \n\t"\
"1: \n\t"\
"movupd (%3, %0), %%xmm1 \n\t"\
"movapd %%xmm1, %%xmm3 \n\t"\
"mulpd (%4, %0), %%xmm1 \n\t"\
"mulpd (%5, %0), %%xmm3 \n\t"\
"addpd %%xmm1, %%xmm0 \n\t"\
"addpd %%xmm3, %%xmm2 \n\t"\
"add $16, %0 \n\t"\
" js 1b \n\t"\
"movhlps %%xmm0, %%xmm1 \n\t"\
"movhlps %%xmm2, %%xmm3 \n\t"\
"addpd %%xmm1, %%xmm0 \n\t"\
"addpd %%xmm3, %%xmm2 \n\t"\
"movsd %%xmm0, %1 \n\t"\
"movsd %%xmm2, %2 \n\t"\
: "+r" (len),\
"=m" (val),\
"=m" (v2)\
: "r" (((uint8_t*)(src+sample_index))-len),\
"r" (((uint8_t*)filter)-len),\
"r" (((uint8_t*)(filter+c->filter_alloc))-len)\
XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3")\
);
Reference in New Issue View Git Blame Copy Permalink