if (EXTERNAL_SSE2(cpu_flags)) {
s->synth_filter_float = synth_filter_sse2;
}
- if (EXTERNAL_AVX(cpu_flags)) {
+ if (EXTERNAL_AVX_FAST(cpu_flags)) {
s->synth_filter_float = synth_filter_avx;
}
- if (EXTERNAL_FMA3(cpu_flags)) {
+ if (EXTERNAL_FMA3(cpu_flags) && !(cpu_flags & AV_CPU_FLAG_AVXSLOW)) {
s->synth_filter_float = synth_filter_fma3;
}
#endif /* HAVE_YASM */
s->dct32 = ff_dct32_float_sse;
if (EXTERNAL_SSE2(cpu_flags))
s->dct32 = ff_dct32_float_sse2;
- if (EXTERNAL_AVX(cpu_flags))
+ if (EXTERNAL_AVX_FAST(cpu_flags))
s->dct32 = ff_dct32_float_avx;
}
s->fft_calc = ff_fft_calc_sse;
s->fft_permutation = FF_FFT_PERM_SWAP_LSBS;
}
- if (EXTERNAL_AVX(cpu_flags) && s->nbits >= 5) {
+ if (EXTERNAL_AVX_FAST(cpu_flags) && s->nbits >= 5) {
/* AVX for SB */
s->imdct_half = ff_imdct_half_avx;
s->fft_calc = ff_fft_calc_avx;
dsp->itxfm_add[TX_32X32][ADST_DCT] =
dsp->itxfm_add[TX_32X32][DCT_ADST] =
dsp->itxfm_add[TX_32X32][DCT_DCT] = ff_vp9_idct_idct_32x32_add_avx;
- init_fpel(1, 0, 32, put, avx);
- init_fpel(0, 0, 64, put, avx);
init_lpf(avx);
init_dir_tm_h_ipred(8, avx);
init_dir_tm_h_ipred(16, avx);
init_dir_tm_h_ipred(32, avx);
+ }
+ if (EXTERNAL_AVX_FAST(cpu_flags)) {
+ init_fpel(1, 0, 32, put, avx);
+ init_fpel(0, 0, 64, put, avx);
init_ipred(32, avx, v, VERT);
}
vol->scale_samples = ff_scale_samples_s32_ssse3_atom;
vol->samples_align = 4;
}
- if (EXTERNAL_AVX(cpu_flags)) {
+ if (EXTERNAL_AVX_FAST(cpu_flags)) {
vol->scale_samples = ff_scale_samples_s32_avx;
vol->samples_align = 8;
}
if (EXTERNAL_SSE2(cpu_flags)) {
fdsp->vector_dmul_scalar = ff_vector_dmul_scalar_sse2;
}
- if (EXTERNAL_AVX(cpu_flags)) {
+ if (EXTERNAL_AVX_FAST(cpu_flags)) {
fdsp->vector_fmul = ff_vector_fmul_avx;
fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_avx;
fdsp->vector_dmul_scalar = ff_vector_dmul_scalar_avx;
fdsp->vector_fmul_add = ff_vector_fmul_add_avx;
fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_avx;
}
- if (EXTERNAL_FMA3(cpu_flags)) {
+ if (EXTERNAL_FMA3(cpu_flags) && !(cpu_flags & AV_CPU_FLAG_AVXSLOW)) {
fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_fma3;
fdsp->vector_fmul_add = ff_vector_fmul_add_fma3;
}
if (m->indep_count >= 4)
m->evaluate_lls = ff_evaluate_lls_sse2;
}
- if (EXTERNAL_AVX(cpu_flags)) {
+ if (EXTERNAL_AVX_FAST(cpu_flags)) {
m->update_lls = ff_update_lls_avx;
}
}
ac->simd_f = ff_unpack_2ch_int16_to_float_a_ssse3;
}
}
- if(EXTERNAL_AVX(mm_flags)) {
+ if(EXTERNAL_AVX_FAST(mm_flags)) {
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)
ac->simd_f = ff_int32_to_float_a_avx;
+ }
+ if(EXTERNAL_AVX(mm_flags)) {
if(channels == 6) {
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)
ac->simd_f = ff_pack_6ch_float_to_float_a_avx;
s->mix_1_1_simd = ff_mix_1_1_a_float_sse;
s->mix_2_1_simd = ff_mix_2_1_a_float_sse;
}
- if(EXTERNAL_AVX(mm_flags)) {
+ if(EXTERNAL_AVX_FAST(mm_flags)) {
s->mix_1_1_simd = ff_mix_1_1_a_float_avx;
s->mix_2_1_simd = ff_mix_2_1_a_float_avx;
}
c->dsp.resample = c->linear ? ff_resample_linear_float_sse
: ff_resample_common_float_sse;
}
- if (EXTERNAL_AVX(mm_flags)) {
+ if (EXTERNAL_AVX_FAST(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_float_avx
: ff_resample_common_float_avx;
}
- if (EXTERNAL_FMA3(mm_flags)) {
+ if (EXTERNAL_FMA3(mm_flags) && !(mm_flags & AV_CPU_FLAG_AVXSLOW)) {
c->dsp.resample = c->linear ? ff_resample_linear_float_fma3
: ff_resample_common_float_fma3;
}