mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2025-12-16 08:56:45 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

fmt: fix and enable f32 -> s16 conversion

Browse source
This commit is contained in:
Wim Taymans 2018-10-30 08:37:07 +00:00
parent 4432db4fa7
commit ef7a111611
2 changed files with 39 additions and 36 deletions
Download patch file Download diff file Expand all files Collapse all files

72
spa/plugins/audioconvert/fmt-ops-sse.c
View file

@ -121,7 +121,7 @@ conv_f32d_to_s32_1_sse(void *data, void *dst, int n_src, const void *src[n_src],
in[0] = _mm_mul_ps(_mm_loadu_ps(&s0[n]), int_max); in[0] = _mm_mul_ps(_mm_loadu_ps(&s0[n]), int_max);
in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min)); in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min));
out[0] = _mm_slli_epi32(_mm_cvttps_epi32(in[0]), 8); out[0] = _mm_slli_epi32(_mm_cvtps_epi32(in[0]), 8);
out[1] = _mm_shuffle_epi32(out[0], _MM_SHUFFLE(0, 3, 2, 1)); 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[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)); out[3] = _mm_shuffle_epi32(out[0], _MM_SHUFFLE(2, 1, 0, 3));
@ -136,7 +136,7 @@ conv_f32d_to_s32_1_sse(void *data, void *dst, int n_src, const void *src[n_src],
in[0] = _mm_load_ss(&s0[n]); in[0] = _mm_load_ss(&s0[n]);
in[0] = _mm_mul_ss(in[0], int_max); in[0] = _mm_mul_ss(in[0], int_max);
in[0] = _mm_min_ss(int_max, _mm_max_ss(in[0], int_min)); in[0] = _mm_min_ss(int_max, _mm_max_ss(in[0], int_min));
*d = _mm_cvttss_si32(in[0]) << 8; *d = _mm_cvtss_si32(in[0]) << 8;
d += n_src; d += n_src;
} }
} }
@ -165,8 +165,8 @@ conv_f32d_to_s32_2_sse(void *data, void *dst, int n_src, const void *src[n_src],
in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min)); in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min));
in[1] = _mm_min_ps(int_max, _mm_max_ps(in[1], int_min)); in[1] = _mm_min_ps(int_max, _mm_max_ps(in[1], int_min));
out[0] = _mm_slli_epi32(_mm_cvttps_epi32(in[0]), 8); out[0] = _mm_slli_epi32(_mm_cvtps_epi32(in[0]), 8);
out[1] = _mm_slli_epi32(_mm_cvttps_epi32(in[1]), 8); out[1] = _mm_slli_epi32(_mm_cvtps_epi32(in[1]), 8);
t[0] = _mm_unpacklo_epi32(out[0], out[1]); t[0] = _mm_unpacklo_epi32(out[0], out[1]);
t[1] = _mm_shuffle_epi32(t[0], _MM_SHUFFLE(0, 0, 2, 2)); t[1] = _mm_shuffle_epi32(t[0], _MM_SHUFFLE(0, 0, 2, 2));
@ -187,7 +187,7 @@ conv_f32d_to_s32_2_sse(void *data, void *dst, int n_src, const void *src[n_src],
in[0] = _mm_mul_ps(in[0], int_max); in[0] = _mm_mul_ps(in[0], int_max);
in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min)); in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min));
out[0] = _mm_slli_epi32(_mm_cvttps_epi32(in[0]), 8); out[0] = _mm_slli_epi32(_mm_cvtps_epi32(in[0]), 8);
_mm_storel_epi64((__m128i*)d, out[0]); _mm_storel_epi64((__m128i*)d, out[0]);
d += n_src; d += n_src;
} }
@ -221,10 +221,10 @@ conv_f32d_to_s32_4_sse(void *data, void *dst, int n_src, const void *src[n_src],
in[2] = _mm_min_ps(int_max, _mm_max_ps(in[2], int_min)); in[2] = _mm_min_ps(int_max, _mm_max_ps(in[2], int_min));
in[3] = _mm_min_ps(int_max, _mm_max_ps(in[3], int_min)); in[3] = _mm_min_ps(int_max, _mm_max_ps(in[3], int_min));
out[0] = _mm_slli_epi32(_mm_cvttps_epi32(in[0]), 8); out[0] = _mm_slli_epi32(_mm_cvtps_epi32(in[0]), 8);
out[1] = _mm_slli_epi32(_mm_cvttps_epi32(in[1]), 8); out[1] = _mm_slli_epi32(_mm_cvtps_epi32(in[1]), 8);
out[2] = _mm_slli_epi32(_mm_cvttps_epi32(in[2]), 8); out[2] = _mm_slli_epi32(_mm_cvtps_epi32(in[2]), 8);
out[3] = _mm_slli_epi32(_mm_cvttps_epi32(in[3]), 8); out[3] = _mm_slli_epi32(_mm_cvtps_epi32(in[3]), 8);
/* transpose */ /* transpose */
t[0] = _mm_unpacklo_epi32(out[0], out[1]); t[0] = _mm_unpacklo_epi32(out[0], out[1]);
@ -254,7 +254,7 @@ conv_f32d_to_s32_4_sse(void *data, void *dst, int n_src, const void *src[n_src],
in[0] = _mm_mul_ps(in[0], int_max); in[0] = _mm_mul_ps(in[0], int_max);
in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min)); in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min));
out[0] = _mm_slli_epi32(_mm_cvttps_epi32(in[0]), 8); out[0] = _mm_slli_epi32(_mm_cvtps_epi32(in[0]), 8);
_mm_storeu_si128((__m128i*)d, out[0]); _mm_storeu_si128((__m128i*)d, out[0]);
d += n_src; d += n_src;
} }
@ -283,7 +283,8 @@ conv_f32d_to_s16_1_sse(void *data, void *dst, int n_src, const void *src[n_src],
int n, n_samples, unrolled; int n, n_samples, unrolled;
__m128 in[1]; __m128 in[1];
__m128i out[4]; __m128i out[4];
__m128i scale = _mm_set1_epi32(15 << 23); __m128 int_max = _mm_set1_ps(S16_MAX_F);
__m128 int_min = _mm_sub_ps(_mm_setzero_ps(), int_max);
n_samples = n_bytes / sizeof(float); n_samples = n_bytes / sizeof(float);
@ -291,10 +292,11 @@ conv_f32d_to_s16_1_sse(void *data, void *dst, int n_src, const void *src[n_src],
n_samples = n_samples & 3; n_samples = n_samples & 3;
for(n = 0; unrolled--; n += 4) { for(n = 0; unrolled--; n += 4) {
in[0] = _mm_loadu_ps(&s0[n]); in[0] = _mm_mul_ps(_mm_loadu_ps(&s0[n]), int_max);
in[0] = (__m128)_mm_adds_epi16((__m128i)in[0], scale); in[0] = _mm_min_ps(int_max, _mm_max_ps(in[0], int_min));
out[0] = _mm_cvtps_epi32(in[0]); out[0] = _mm_cvtps_epi32(in[0]);
out[0] = _mm_packs_epi32(out[0], out[0]); out[0] = _mm_packs_epi32(out[0], out[0]);
d[0*n_src] = _mm_extract_pi16(*(__m64*)out, 0); d[0*n_src] = _mm_extract_pi16(*(__m64*)out, 0);
d[1*n_src] = _mm_extract_pi16(*(__m64*)out, 1); d[1*n_src] = _mm_extract_pi16(*(__m64*)out, 1);
d[2*n_src] = _mm_extract_pi16(*(__m64*)out, 2); d[2*n_src] = _mm_extract_pi16(*(__m64*)out, 2);
@ -302,11 +304,9 @@ conv_f32d_to_s16_1_sse(void *data, void *dst, int n_src, const void *src[n_src],
d += 4*n_src; d += 4*n_src;
} }
for(; n_samples--; n++) { for(; n_samples--; n++) {
in[0] = _mm_load_ss(&s0[n]); in[0] = _mm_mul_ss(_mm_load_ss(&s0[n]), int_max);
in[0] = (__m128)_mm_adds_epi16((__m128i)in[0], scale); in[0] = _mm_min_ss(int_max, _mm_max_ss(in[0], int_min));
out[0] = _mm_cvtps_epi32(in[0]); *d = _mm_cvtss_si32(in[0]);
out[0] = _mm_packs_epi32(out[0], out[0]);
*d = _mm_extract_pi16(*(__m64*)out, 0);
d += n_src; d += n_src;
} }
} }
@ -318,9 +318,10 @@ conv_f32d_to_s16_2_sse(void *data, void *dst, int n_src, const void *src[n_src],
const float *s0 = s[0], *s1 = s[1]; const float *s0 = s[0], *s1 = s[1];
int16_t *d = dst; int16_t *d = dst;
int n = 0, n_samples, unrolled; int n = 0, n_samples, unrolled;
__m128 in[2], t[2]; __m128 in[2];
__m128i out[4]; __m128i out[4], t[2];
__m128i scale = _mm_set1_epi32(15 << 23); __m128 int_max = _mm_set1_ps(S16_MAX_F);
__m128 int_min = _mm_sub_ps(_mm_setzero_ps(), int_max);
n_samples = n_bytes / sizeof(float); n_samples = n_bytes / sizeof(float);
@ -328,16 +329,16 @@ conv_f32d_to_s16_2_sse(void *data, void *dst, int n_src, const void *src[n_src],
n_samples = n_samples & 3; n_samples = n_samples & 3;
for(; unrolled--; n += 4) { for(; unrolled--; n += 4) {
in[0] = _mm_loadu_ps(&s0[n]); in[0] = _mm_mul_ps(_mm_loadu_ps(&s0[n]), int_max);
in[1] = _mm_loadu_ps(&s1[n]); in[1] = _mm_mul_ps(_mm_loadu_ps(&s1[n]), int_max);
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]);
t[0] = _mm_cvtps_epi32(in[0]);
t[1] = _mm_cvtps_epi32(in[1]);
t[0] = _mm_packs_epi32(t[0], t[1]);
t[1] = _mm_packs_epi32(t[1], t[1]);
out[0] = _mm_unpacklo_epi16(t[0], t[1]);
out[1] = _mm_shuffle_epi32(out[0], _MM_SHUFFLE(0, 3, 2, 1)); 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[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)); out[3] = _mm_shuffle_epi32(out[0], _MM_SHUFFLE(2, 1, 0, 3));
@ -349,11 +350,12 @@ conv_f32d_to_s16_2_sse(void *data, void *dst, int n_src, const void *src[n_src],
d += 4*n_src; d += 4*n_src;
} }
for(; n_samples--; n++) { for(; n_samples--; n++) {
in[0] = _mm_set_ps(0, 0, s0[n], s1[n]); in[0] = _mm_mul_ss(_mm_load_ss(&s0[n]), int_max);
in[0] = (__m128)_mm_adds_epi16((__m128i)in[0], scale); in[1] = _mm_mul_ss(_mm_load_ss(&s1[n]), int_max);
out[0] = _mm_cvtps_epi32(in[0]); in[0] = _mm_min_ss(int_max, _mm_max_ss(in[0], int_min));
out[0] = _mm_packs_epi32(out[0], out[0]); in[1] = _mm_min_ss(int_max, _mm_max_ss(in[1], int_min));
*(uint32_t*)d = _mm_cvtsi128_si32(out[0]); d[0] = _mm_cvtss_si32(in[0]);
d[1] = _mm_cvtss_si32(in[1]);
d += n_src; d += n_src;
} }
} }

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

@ -29,6 +29,7 @@
#define S16_MIN -32767 #define S16_MIN -32767
#define S16_MAX 32767 #define S16_MAX 32767
#define S16_MAX_F 32767.0f
#define S16_SCALE 32767 #define S16_SCALE 32767
#define S24_MIN -8388607 #define S24_MIN -8388607
@ -716,7 +717,7 @@ static const struct conv_info {
{ SPA_AUDIO_FORMAT_F32, SPA_AUDIO_FORMAT_S16, 0, conv_f32_to_s16 }, { SPA_AUDIO_FORMAT_F32, SPA_AUDIO_FORMAT_S16, 0, conv_f32_to_s16 },
{ SPA_AUDIO_FORMAT_F32, SPA_AUDIO_FORMAT_S16P, 0, conv_f32_to_s16d }, { SPA_AUDIO_FORMAT_F32, SPA_AUDIO_FORMAT_S16P, 0, conv_f32_to_s16d },
#if defined (__SSE2__) #if defined (__SSE2__)
// { SPA_AUDIO_FORMAT_F32P, SPA_AUDIO_FORMAT_S16, FEATURE_SSE, conv_f32d_to_s16_sse }, { SPA_AUDIO_FORMAT_F32P, SPA_AUDIO_FORMAT_S16, FEATURE_SSE, conv_f32d_to_s16_sse },
#endif #endif
{ SPA_AUDIO_FORMAT_F32P, SPA_AUDIO_FORMAT_S16, 0, conv_f32d_to_s16 }, { SPA_AUDIO_FORMAT_F32P, SPA_AUDIO_FORMAT_S16, 0, conv_f32d_to_s16 },