2 * Format Conversion Utils
3 * Copyright (c) 2000, 2001 Fabrice Bellard
4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
6 * This file is part of Libav.
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
25 #include "libavutil/cpu.h"
26 #include "libavutil/x86_cpu.h"
27 #include "libavcodec/fmtconvert.h"
28 #include "libavcodec/dsputil.h"
32 void ff_int32_to_float_fmul_scalar_sse (float *dst, const int *src, float mul, int len);
33 void ff_int32_to_float_fmul_scalar_sse2(float *dst, const int *src, float mul, int len);
35 void ff_float_to_int16_3dnow(int16_t *dst, const float *src, long len);
36 void ff_float_to_int16_sse (int16_t *dst, const float *src, long len);
37 void ff_float_to_int16_sse2 (int16_t *dst, const float *src, long len);
39 void ff_float_to_int16_step_3dnow(int16_t *dst, const float *src, long len, long step);
40 void ff_float_to_int16_step_sse (int16_t *dst, const float *src, long len, long step);
41 void ff_float_to_int16_step_sse2 (int16_t *dst, const float *src, long len, long step);
43 void ff_float_to_int16_interleave2_3dnow(int16_t *dst, const float **src, long len);
44 void ff_float_to_int16_interleave2_sse (int16_t *dst, const float **src, long len);
45 void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const float **src, long len);
47 void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len);
48 void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len);
49 void ff_float_to_int16_interleave6_3dn2(int16_t *dst, const float **src, int len);
51 #define ff_float_to_int16_interleave6_sse2 ff_float_to_int16_interleave6_sse
53 #define FLOAT_TO_INT16_INTERLEAVE(cpu) \
54 /* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
55 static av_noinline void float_to_int16_interleave_misc_##cpu(int16_t *dst, const float **src, long len, int channels){\
57 for(c=0; c<channels; c++){\
58 ff_float_to_int16_step_##cpu(dst+c, src[c], len, channels);\
62 static void float_to_int16_interleave_##cpu(int16_t *dst, const float **src, long len, int channels){\
64 ff_float_to_int16_##cpu(dst, src[0], len);\
65 else if(channels==2){\
66 ff_float_to_int16_interleave2_##cpu(dst, src, len);\
67 }else if(channels==6){\
68 ff_float_to_int16_interleave6_##cpu(dst, src, len);\
70 float_to_int16_interleave_misc_##cpu(dst, src, len, channels);\
73 FLOAT_TO_INT16_INTERLEAVE(3dnow)
74 FLOAT_TO_INT16_INTERLEAVE(sse)
75 FLOAT_TO_INT16_INTERLEAVE(sse2)
77 static void float_to_int16_interleave_3dn2(int16_t *dst, const float **src, long len, int channels){
79 ff_float_to_int16_interleave6_3dn2(dst, src, len);
81 float_to_int16_interleave_3dnow(dst, src, len, channels);
84 void ff_float_interleave2_mmx(float *dst, const float **src, unsigned int len);
85 void ff_float_interleave2_sse(float *dst, const float **src, unsigned int len);
87 void ff_float_interleave6_mmx(float *dst, const float **src, unsigned int len);
88 void ff_float_interleave6_sse(float *dst, const float **src, unsigned int len);
90 static void float_interleave_mmx(float *dst, const float **src,
91 unsigned int len, int channels)
94 ff_float_interleave2_mmx(dst, src, len);
95 } else if (channels == 6)
96 ff_float_interleave6_mmx(dst, src, len);
98 ff_float_interleave_c(dst, src, len, channels);
101 static void float_interleave_sse(float *dst, const float **src,
102 unsigned int len, int channels)
105 ff_float_interleave2_sse(dst, src, len);
106 } else if (channels == 6)
107 ff_float_interleave6_sse(dst, src, len);
109 ff_float_interleave_c(dst, src, len, channels);
113 void ff_fmt_convert_init_x86(FmtConvertContext *c, AVCodecContext *avctx)
116 int mm_flags = av_get_cpu_flags();
118 if (mm_flags & AV_CPU_FLAG_MMX) {
119 c->float_interleave = float_interleave_mmx;
121 if (HAVE_AMD3DNOW && mm_flags & AV_CPU_FLAG_3DNOW) {
122 if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
123 c->float_to_int16 = ff_float_to_int16_3dnow;
124 c->float_to_int16_interleave = float_to_int16_interleave_3dnow;
127 if (HAVE_AMD3DNOWEXT && mm_flags & AV_CPU_FLAG_3DNOWEXT) {
128 if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
129 c->float_to_int16_interleave = float_to_int16_interleave_3dn2;
132 if (HAVE_SSE && mm_flags & AV_CPU_FLAG_SSE) {
133 c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse;
134 c->float_to_int16 = ff_float_to_int16_sse;
135 c->float_to_int16_interleave = float_to_int16_interleave_sse;
136 c->float_interleave = float_interleave_sse;
138 if (HAVE_SSE && mm_flags & AV_CPU_FLAG_SSE2) {
139 c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse2;
140 c->float_to_int16 = ff_float_to_int16_sse2;
141 c->float_to_int16_interleave = float_to_int16_interleave_sse2;