yading@10: /*
yading@10:  * Format Conversion Utils
yading@10:  * Copyright (c) 2000, 2001 Fabrice Bellard
yading@10:  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
yading@10:  *
yading@10:  * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
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/fmtconvert.h"
yading@10: 
yading@10: #if HAVE_YASM
yading@10: 
yading@10: void ff_int32_to_float_fmul_scalar_sse (float *dst, const int *src, float mul, int len);
yading@10: void ff_int32_to_float_fmul_scalar_sse2(float *dst, const int *src, float mul, int len);
yading@10: 
yading@10: void ff_float_to_int16_3dnow(int16_t *dst, const float *src, long len);
yading@10: void ff_float_to_int16_sse  (int16_t *dst, const float *src, long len);
yading@10: void ff_float_to_int16_sse2 (int16_t *dst, const float *src, long len);
yading@10: 
yading@10: void ff_float_to_int16_step_3dnow(int16_t *dst, const float *src, long len, long step);
yading@10: void ff_float_to_int16_step_sse  (int16_t *dst, const float *src, long len, long step);
yading@10: void ff_float_to_int16_step_sse2 (int16_t *dst, const float *src, long len, long step);
yading@10: 
yading@10: void ff_float_to_int16_interleave2_3dnow(int16_t *dst, const float **src, long len);
yading@10: void ff_float_to_int16_interleave2_sse  (int16_t *dst, const float **src, long len);
yading@10: void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const float **src, long len);
yading@10: 
yading@10: void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len);
yading@10: void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len);
yading@10: void ff_float_to_int16_interleave6_3dnowext(int16_t *dst, const float **src, int len);
yading@10: 
yading@10: #define ff_float_to_int16_interleave6_sse2 ff_float_to_int16_interleave6_sse
yading@10: 
yading@10: #define FLOAT_TO_INT16_INTERLEAVE(cpu) \
yading@10: /* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
yading@10: static av_noinline void float_to_int16_interleave_misc_##cpu(int16_t *dst, const float **src, long len, int channels){\
yading@10:     int c;\
yading@10:     for(c=0; c<channels; c++){\
yading@10:         ff_float_to_int16_step_##cpu(dst+c, src[c], len, channels);\
yading@10:     }\
yading@10: }\
yading@10: \
yading@10: static void float_to_int16_interleave_##cpu(int16_t *dst, const float **src, long len, int channels){\
yading@10:     if(channels==1)\
yading@10:         ff_float_to_int16_##cpu(dst, src[0], len);\
yading@10:     else if(channels==2){\
yading@10:         ff_float_to_int16_interleave2_##cpu(dst, src, len);\
yading@10:     }else if(channels==6){\
yading@10:         ff_float_to_int16_interleave6_##cpu(dst, src, len);\
yading@10:     }else\
yading@10:         float_to_int16_interleave_misc_##cpu(dst, src, len, channels);\
yading@10: }
yading@10: 
yading@10: FLOAT_TO_INT16_INTERLEAVE(3dnow)
yading@10: FLOAT_TO_INT16_INTERLEAVE(sse)
yading@10: FLOAT_TO_INT16_INTERLEAVE(sse2)
yading@10: 
yading@10: static void float_to_int16_interleave_3dnowext(int16_t *dst, const float **src,
yading@10:                                                long len, int channels)
yading@10: {
yading@10:     if(channels==6)
yading@10:         ff_float_to_int16_interleave6_3dnowext(dst, src, len);
yading@10:     else
yading@10:         float_to_int16_interleave_3dnow(dst, src, len, channels);
yading@10: }
yading@10: 
yading@10: void ff_float_interleave2_mmx(float *dst, const float **src, unsigned int len);
yading@10: void ff_float_interleave2_sse(float *dst, const float **src, unsigned int len);
yading@10: 
yading@10: void ff_float_interleave6_mmx(float *dst, const float **src, unsigned int len);
yading@10: void ff_float_interleave6_sse(float *dst, const float **src, unsigned int len);
yading@10: 
yading@10: static void float_interleave_mmx(float *dst, const float **src,
yading@10:                                  unsigned int len, int channels)
yading@10: {
yading@10:     if (channels == 2) {
yading@10:         ff_float_interleave2_mmx(dst, src, len);
yading@10:     } else if (channels == 6)
yading@10:         ff_float_interleave6_mmx(dst, src, len);
yading@10:     else
yading@10:         ff_float_interleave_c(dst, src, len, channels);
yading@10: }
yading@10: 
yading@10: static void float_interleave_sse(float *dst, const float **src,
yading@10:                                  unsigned int len, int channels)
yading@10: {
yading@10:     if (channels == 2) {
yading@10:         ff_float_interleave2_sse(dst, src, len);
yading@10:     } else if (channels == 6)
yading@10:         ff_float_interleave6_sse(dst, src, len);
yading@10:     else
yading@10:         ff_float_interleave_c(dst, src, len, channels);
yading@10: }
yading@10: #endif /* HAVE_YASM */
yading@10: 
yading@10: av_cold void ff_fmt_convert_init_x86(FmtConvertContext *c, AVCodecContext *avctx)
yading@10: {
yading@10: #if HAVE_YASM
yading@10:     int mm_flags = av_get_cpu_flags();
yading@10: 
yading@10:     if (EXTERNAL_MMX(mm_flags)) {
yading@10:         c->float_interleave = float_interleave_mmx;
yading@10: 
yading@10:         if (EXTERNAL_AMD3DNOW(mm_flags)) {
yading@10:             if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
yading@10:                 c->float_to_int16 = ff_float_to_int16_3dnow;
yading@10:                 c->float_to_int16_interleave = float_to_int16_interleave_3dnow;
yading@10:             }
yading@10:         }
yading@10:         if (EXTERNAL_AMD3DNOWEXT(mm_flags)) {
yading@10:             if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
yading@10:                 c->float_to_int16_interleave = float_to_int16_interleave_3dnowext;
yading@10:             }
yading@10:         }
yading@10:         if (EXTERNAL_SSE(mm_flags)) {
yading@10:             c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse;
yading@10:             c->float_to_int16 = ff_float_to_int16_sse;
yading@10:             c->float_to_int16_interleave = float_to_int16_interleave_sse;
yading@10:             c->float_interleave = float_interleave_sse;
yading@10:         }
yading@10:         if (EXTERNAL_SSE2(mm_flags)) {
yading@10:             c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse2;
yading@10:             c->float_to_int16 = ff_float_to_int16_sse2;
yading@10:             c->float_to_int16_interleave = float_to_int16_interleave_sse2;
yading@10:         }
yading@10:     }
yading@10: #endif /* HAVE_YASM */
yading@10: }