mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2025-11-01 22:58:50 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

fmt-ops: fix conversion

Browse source
This commit is contained in:
Wim Taymans 2018-10-29 18:38:14 +00:00
parent 57d9f597d9
commit 4432db4fa7
1 changed files with 17 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

26
spa/plugins/audioconvert/fmt-ops-sse.c
View file

@ -293,7 +293,7 @@ conv_f32d_to_s16_1_sse(void *data, void *dst, int n_src, const void *src[n_src],
for(n = 0; unrolled--; n += 4) {
in[0] = _mm_loadu_ps(&s0[n]);
in[0] = (__m128)_mm_adds_epi16((__m128i)in[0], scale);
out[0] = _mm_cvttps_epi32(in[0]);
out[0] = _mm_cvtps_epi32(in[0]);
out[0] = _mm_packs_epi32(out[0], out[0]);
d[0*n_src] = _mm_extract_pi16(*(__m64*)out, 0);
d[1*n_src] = _mm_extract_pi16(*(__m64*)out, 1);
@ -304,7 +304,7 @@ conv_f32d_to_s16_1_sse(void *data, void *dst, int n_src, const void *src[n_src],
for(; n_samples--; n++) {
in[0] = _mm_load_ss(&s0[n]);
in[0] = (__m128)_mm_adds_epi16((__m128i)in[0], scale);
out[0] = _mm_cvttps_epi32(in[0]);
out[0] = _mm_cvtps_epi32(in[0]);
out[0] = _mm_packs_epi32(out[0], out[0]);
*d = _mm_extract_pi16(*(__m64*)out, 0);
d += n_src;
@ -317,7 +317,7 @@ conv_f32d_to_s16_2_sse(void *data, void *dst, int n_src, const void *src[n_src],
const float **s = (const float **) src;
const float *s0 = s[0], *s1 = s[1];
int16_t *d = dst;
int n, n_samples, unrolled;
int n = 0, n_samples, unrolled;
__m128 in[2], t[2];
__m128i out[4];
__m128i scale = _mm_set1_epi32(15 << 23);
@ -327,25 +327,33 @@ conv_f32d_to_s16_2_sse(void *data, void *dst, int n_src, const void *src[n_src],
unrolled = n_samples / 4;
n_samples = n_samples & 3;
for(n = 0; unrolled--; n += 4) {
for(; unrolled--; n += 4) {
in[0] = _mm_loadu_ps(&s0[n]);
in[1] = _mm_loadu_ps(&s1[n]);
t[0] = _mm_unpacklo_ps(in[0], in[1]);
t[1] = _mm_unpackhi_ps(in[0], in[1]);
t[0] = _mm_unpacklo_ps(in[1], in[0]);
t[1] = _mm_unpackhi_ps(in[1], in[0]);
in[0] = (__m128)_mm_adds_epi16((__m128i)t[0], scale);
in[1] = (__m128)_mm_adds_epi16((__m128i)t[1], scale);
out[0] = _mm_cvtps_epi32(in[0]);
out[1] = _mm_cvtps_epi32(in[1]);
out[0] = _mm_packs_epi32(out[0], out[1]);
_mm_storeu_si128((__m128i*)d, out[0]);
out[1] = _mm_shuffle_epi32(out[0], _MM_SHUFFLE(0, 3, 2, 1));
out[2] = _mm_shuffle_epi32(out[0], _MM_SHUFFLE(1, 0, 3, 2));
out[3] = _mm_shuffle_epi32(out[0], _MM_SHUFFLE(2, 1, 0, 3));
*((uint32_t*)d + 0*n_src) = _mm_cvtsi128_si32(out[0]);
*((uint32_t*)d + 1*n_src) = _mm_cvtsi128_si32(out[1]);
*((uint32_t*)d + 2*n_src) = _mm_cvtsi128_si32(out[2]);
*((uint32_t*)d + 3*n_src) = _mm_cvtsi128_si32(out[3]);
d += 4*n_src;
}
for(; n_samples--; n++) {
in[0] = _mm_set_ps(s0[n], s1[n], 0, 0);
in[0] = _mm_set_ps(0, 0, s0[n], s1[n]);
in[0] = (__m128)_mm_adds_epi16((__m128i)in[0], scale);
out[0] = _mm_cvtps_epi32(in[0]);
out[0] = _mm_packs_epi32(out[0], out[0]);
_mm_store_ss((float*)d, (__m128)out[0]);
*(uint32_t*)d = _mm_cvtsi128_si32(out[0]);
d += n_src;
}
}