yading@11: /* yading@11: * Copyright (C) 2012 Michael Niedermayer (michaelni@gmx.at) yading@11: * yading@11: * This file is part of libswresample yading@11: * yading@11: * libswresample is free software; you can redistribute it and/or yading@11: * modify it under the terms of the GNU Lesser General Public yading@11: * License as published by the Free Software Foundation; either yading@11: * version 2.1 of the License, or (at your option) any later version. yading@11: * yading@11: * libswresample is distributed in the hope that it will be useful, yading@11: * but WITHOUT ANY WARRANTY; without even the implied warranty of yading@11: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU yading@11: * Lesser General Public License for more details. yading@11: * yading@11: * You should have received a copy of the GNU Lesser General Public yading@11: * License along with libswresample; if not, write to the Free Software yading@11: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA yading@11: */ yading@11: yading@11: #include "libswresample/swresample_internal.h" yading@11: #include "libswresample/audioconvert.h" yading@11: yading@11: #define PROTO(pre, in, out, cap) void ff ## pre ## _ ##in## _to_ ##out## _a_ ##cap(uint8_t **dst, const uint8_t **src, int len); yading@11: #define PROTO2(pre, out, cap) PROTO(pre, int16, out, cap) PROTO(pre, int32, out, cap) PROTO(pre, float, out, cap) yading@11: #define PROTO3(pre, cap) PROTO2(pre, int16, cap) PROTO2(pre, int32, cap) PROTO2(pre, float, cap) yading@11: #define PROTO4(pre) PROTO3(pre, mmx) PROTO3(pre, sse) PROTO3(pre, sse2) PROTO3(pre, ssse3) PROTO3(pre, sse4) PROTO3(pre, avx) yading@11: PROTO4() yading@11: PROTO4(_pack_2ch) yading@11: PROTO4(_pack_6ch) yading@11: PROTO4(_unpack_2ch) yading@11: yading@11: av_cold void swri_audio_convert_init_x86(struct AudioConvert *ac, yading@11: enum AVSampleFormat out_fmt, yading@11: enum AVSampleFormat in_fmt, yading@11: int channels){ yading@11: int mm_flags = av_get_cpu_flags(); yading@11: yading@11: ac->simd_f= NULL; yading@11: yading@11: //FIXME add memcpy case yading@11: yading@11: #define MULTI_CAPS_FUNC(flag, cap) \ yading@11: if (mm_flags & flag) {\ yading@11: if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S16 || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S16P)\ yading@11: ac->simd_f = ff_int16_to_int32_a_ ## cap;\ yading@11: if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S32P)\ yading@11: ac->simd_f = ff_int32_to_int16_a_ ## cap;\ yading@11: } yading@11: yading@11: MULTI_CAPS_FUNC(AV_CPU_FLAG_MMX, mmx) yading@11: MULTI_CAPS_FUNC(AV_CPU_FLAG_SSE2, sse2) yading@11: yading@11: if(mm_flags & AV_CPU_FLAG_MMX) { yading@11: if(channels == 6) { yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_pack_6ch_float_to_float_a_mmx; yading@11: } yading@11: } yading@11: yading@11: if(mm_flags & AV_CPU_FLAG_SSE2) { yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_int32_to_float_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S16 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S16P) yading@11: ac->simd_f = ff_int16_to_float_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_FLTP) yading@11: ac->simd_f = ff_float_to_int32_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_FLTP) yading@11: ac->simd_f = ff_float_to_int16_a_sse2; yading@11: yading@11: if(channels == 2) { yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_pack_2ch_int32_to_int32_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_S16P) yading@11: ac->simd_f = ff_pack_2ch_int16_to_int16_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S16P) yading@11: ac->simd_f = ff_pack_2ch_int16_to_int32_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_pack_2ch_int32_to_int16_a_sse2; yading@11: yading@11: if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S32) yading@11: ac->simd_f = ff_unpack_2ch_int32_to_int32_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S16) yading@11: ac->simd_f = ff_unpack_2ch_int16_to_int16_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S16) yading@11: ac->simd_f = ff_unpack_2ch_int16_to_int32_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S32) yading@11: ac->simd_f = ff_unpack_2ch_int32_to_int16_a_sse2; yading@11: yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_pack_2ch_int32_to_float_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLTP) yading@11: ac->simd_f = ff_pack_2ch_float_to_int32_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S16P) yading@11: ac->simd_f = ff_pack_2ch_int16_to_float_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_FLTP) yading@11: ac->simd_f = ff_pack_2ch_float_to_int16_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32) yading@11: ac->simd_f = ff_unpack_2ch_int32_to_float_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_FLT) yading@11: ac->simd_f = ff_unpack_2ch_float_to_int32_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S16) yading@11: ac->simd_f = ff_unpack_2ch_int16_to_float_a_sse2; yading@11: if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_FLT) yading@11: ac->simd_f = ff_unpack_2ch_float_to_int16_a_sse2; yading@11: } yading@11: } yading@11: if(mm_flags & AV_CPU_FLAG_SSSE3) { yading@11: if(channels == 2) { yading@11: if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S16) yading@11: ac->simd_f = ff_unpack_2ch_int16_to_int16_a_ssse3; yading@11: if( out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S16) yading@11: ac->simd_f = ff_unpack_2ch_int16_to_int32_a_ssse3; yading@11: if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S16) yading@11: ac->simd_f = ff_unpack_2ch_int16_to_float_a_ssse3; yading@11: } yading@11: } yading@11: if(mm_flags & AV_CPU_FLAG_SSE4) { yading@11: if(channels == 6) { yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_pack_6ch_float_to_float_a_sse4; yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_pack_6ch_int32_to_float_a_sse4; yading@11: if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLTP) yading@11: ac->simd_f = ff_pack_6ch_float_to_int32_a_sse4; yading@11: } yading@11: } yading@11: if(HAVE_AVX_EXTERNAL && mm_flags & AV_CPU_FLAG_AVX) { yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_int32_to_float_a_avx; yading@11: if(channels == 6) { yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_pack_6ch_float_to_float_a_avx; yading@11: if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32P) yading@11: ac->simd_f = ff_pack_6ch_int32_to_float_a_avx; yading@11: if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLTP) yading@11: ac->simd_f = ff_pack_6ch_float_to_int32_a_avx; yading@11: } yading@11: } yading@11: } yading@11: yading@11: #define D(type, simd) \ yading@11: mix_1_1_func_type ff_mix_1_1_a_## type ## _ ## simd;\ yading@11: mix_2_1_func_type ff_mix_2_1_a_## type ## _ ## simd; yading@11: yading@11: D(float, sse) yading@11: D(float, avx) yading@11: D(int16, mmx) yading@11: D(int16, sse2) yading@11: yading@11: yading@11: av_cold void swri_rematrix_init_x86(struct SwrContext *s){ yading@11: int mm_flags = av_get_cpu_flags(); yading@11: int nb_in = av_get_channel_layout_nb_channels(s->in_ch_layout); yading@11: int nb_out = av_get_channel_layout_nb_channels(s->out_ch_layout); yading@11: int num = nb_in * nb_out; yading@11: int i,j; yading@11: yading@11: s->mix_1_1_simd = NULL; yading@11: s->mix_2_1_simd = NULL; yading@11: yading@11: if (s->midbuf.fmt == AV_SAMPLE_FMT_S16P){ yading@11: if(mm_flags & AV_CPU_FLAG_MMX) { yading@11: s->mix_1_1_simd = ff_mix_1_1_a_int16_mmx; yading@11: s->mix_2_1_simd = ff_mix_2_1_a_int16_mmx; yading@11: } yading@11: if(mm_flags & AV_CPU_FLAG_SSE2) { yading@11: s->mix_1_1_simd = ff_mix_1_1_a_int16_sse2; yading@11: s->mix_2_1_simd = ff_mix_2_1_a_int16_sse2; yading@11: } yading@11: s->native_simd_matrix = av_mallocz(2 * num * sizeof(int16_t)); yading@11: for(i=0; inative_matrix)[i * nb_in + j])); yading@11: sh = FFMAX(av_log2(sh) - 14, 0); yading@11: for(j=0; jnative_simd_matrix)[2*(i * nb_in + j)+1] = 15 - sh; yading@11: ((int16_t*)s->native_simd_matrix)[2*(i * nb_in + j)] = yading@11: ((((int*)s->native_matrix)[i * nb_in + j]) + (1<>1)) >> sh; yading@11: } yading@11: } yading@11: } else if(s->midbuf.fmt == AV_SAMPLE_FMT_FLTP){ yading@11: if(mm_flags & AV_CPU_FLAG_SSE) { yading@11: s->mix_1_1_simd = ff_mix_1_1_a_float_sse; yading@11: s->mix_2_1_simd = ff_mix_2_1_a_float_sse; yading@11: } yading@11: if(HAVE_AVX_EXTERNAL && mm_flags & AV_CPU_FLAG_AVX) { yading@11: s->mix_1_1_simd = ff_mix_1_1_a_float_avx; yading@11: s->mix_2_1_simd = ff_mix_2_1_a_float_avx; yading@11: } yading@11: s->native_simd_matrix = av_mallocz(num * sizeof(float)); yading@11: memcpy(s->native_simd_matrix, s->native_matrix, num * sizeof(float)); yading@11: } yading@11: }