mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2025-10-29 05:40:27 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

fmt: add sse2 optimized s32->f32d

Browse source
This commit is contained in:
Wim Taymans 2019-10-24 11:06:04 +02:00
parent aab72a70e6
commit 3c9aa2e96c
3 changed files with 51 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

47
spa/plugins/audioconvert/fmt-ops-sse2.c
View file

@ -334,6 +334,53 @@ conv_s24_to_f32d_sse2(struct convert *conv, void * SPA_RESTRICT dst[], const voi
conv_s24_to_f32d_1s_sse2(conv, &dst[i], &s[3*i], n_channels, n_samples);
}
void
conv_s32_to_f32d_1s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
uint32_t n_channels, uint32_t n_samples)
{
const uint32_t *s = src;
float **d = (float **) dst;
float *d0 = d[0];
uint32_t n, unrolled;
__m128i in;
__m128 out, factor = _mm_set1_ps(1.0f / S24_SCALE);
if (SPA_IS_ALIGNED(d0, 16))
unrolled = n_samples & ~3;
else
unrolled = 0;
for(n = 0; n < unrolled; n += 4) {
in = _mm_setr_epi32(s[0*n_channels],
s[1*n_channels],
s[2*n_channels],
s[3*n_channels]);
in = _mm_srai_epi32(in, 8);
out = _mm_cvtepi32_ps(in);
out = _mm_mul_ps(out, factor);
_mm_store_ps(&d0[n], out);
s += 4*n_channels;
}
for(; n < n_samples; n++) {
out = _mm_cvtsi32_ss(out, s[0]);
out = _mm_mul_ss(out, factor);
_mm_store_ss(&d0[n], out);
s += n_channels;
}
}
void
conv_s32_to_f32d_sse2(struct convert *conv, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[],
uint32_t n_samples)
{
const int32_t *s = src[0];
uint32_t i = 0, n_channels = conv->n_channels;
for(; i < n_channels; i++)
conv_s32_to_f32d_1s_sse2(conv, &dst[i], &s[i], n_channels, n_samples);
}
static void
conv_f32d_to_s32_1s_sse2(void *data, void * SPA_RESTRICT dst, const void * SPA_RESTRICT src[],
uint32_t n_channels, uint32_t n_samples)

3
spa/plugins/audioconvert/fmt-ops.c
View file

@ -67,6 +67,9 @@ static struct conv_info conv_table[] =
{ SPA_AUDIO_FORMAT_F32, SPA_AUDIO_FORMAT_F32P, 0, 0, conv_deinterleave_32_c },
{ SPA_AUDIO_FORMAT_F32P, SPA_AUDIO_FORMAT_F32, 0, 0, conv_interleave_32_c },
#if defined (HAVE_SSE2)
{ SPA_AUDIO_FORMAT_S32, SPA_AUDIO_FORMAT_F32P, 0, SPA_CPU_FLAG_SSE2, conv_s32_to_f32d_sse2 },
#endif
{ SPA_AUDIO_FORMAT_S32, SPA_AUDIO_FORMAT_F32, 0, 0, conv_s32_to_f32_c },
{ SPA_AUDIO_FORMAT_S32P, SPA_AUDIO_FORMAT_F32P, 0, 0, conv_s32d_to_f32d_c },
{ SPA_AUDIO_FORMAT_S32, SPA_AUDIO_FORMAT_F32P, 0, 0, conv_s32_to_f32d_c },

1
spa/plugins/audioconvert/fmt-ops.h
View file

@ -165,6 +165,7 @@ DEFINE_FUNCTION(interleave_32, c);
DEFINE_FUNCTION(s16_to_f32d_2, sse2);
DEFINE_FUNCTION(s16_to_f32d, sse2);
DEFINE_FUNCTION(s24_to_f32d, sse2);
DEFINE_FUNCTION(s32_to_f32d, sse2);
DEFINE_FUNCTION(f32d_to_s32, sse2);
DEFINE_FUNCTION(f32d_to_s16, sse2);
#endif