mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2025-11-02 09:01:50 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

fmt-ops: fix 32 bit compilation

Browse source 
_mm256_extract_epi64 is only for 64 bits, add workaround for 32
bits.

Fixes #220
This commit is contained in:
Wim Taymans 2020-03-23 14:16:36 +01:00
parent c4cf5e6629
commit b8a1ea1d3a
1 changed files with 22 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

23
spa/plugins/audioconvert/fmt-ops-avx2.c
View file

@ -603,6 +603,7 @@ conv_f32d_to_s32_2s_avx2(void *data, void * SPA_RESTRICT dst, const void * SPA_R
t[0] = _mm256_unpacklo_epi32(out[0], out[1]); /* a0 b0 a1 b1 a4 b4 a5 b5 */
t[1] = _mm256_unpackhi_epi32(out[0], out[1]); /* a2 b2 a3 b3 a6 b6 a7 b7 */
#ifdef __x86_64__
*((int64_t*)(d + 0*n_channels)) = _mm256_extract_epi64(t[0], 0);
*((int64_t*)(d + 1*n_channels)) = _mm256_extract_epi64(t[0], 1);
*((int64_t*)(d + 2*n_channels)) = _mm256_extract_epi64(t[1], 0);
@ -611,7 +612,16 @@ conv_f32d_to_s32_2s_avx2(void *data, void * SPA_RESTRICT dst, const void * SPA_R
*((int64_t*)(d + 5*n_channels)) = _mm256_extract_epi64(t[0], 3);
*((int64_t*)(d + 6*n_channels)) = _mm256_extract_epi64(t[1], 2);
*((int64_t*)(d + 7*n_channels)) = _mm256_extract_epi64(t[1], 3);
#else
_mm_storel_pi((__m64*)(d + 0*n_channels), (__m128)_mm256_extracti128_si256(t[0], 0));
_mm_storeh_pi((__m64*)(d + 1*n_channels), (__m128)_mm256_extracti128_si256(t[0], 0));
_mm_storel_pi((__m64*)(d + 2*n_channels), (__m128)_mm256_extracti128_si256(t[1], 0));
_mm_storeh_pi((__m64*)(d + 3*n_channels), (__m128)_mm256_extracti128_si256(t[1], 0));
_mm_storel_pi((__m64*)(d + 4*n_channels), (__m128)_mm256_extracti128_si256(t[0], 1));
_mm_storeh_pi((__m64*)(d + 5*n_channels), (__m128)_mm256_extracti128_si256(t[0], 1));
_mm_storel_pi((__m64*)(d + 6*n_channels), (__m128)_mm256_extracti128_si256(t[1], 1));
_mm_storeh_pi((__m64*)(d + 7*n_channels), (__m128)_mm256_extracti128_si256(t[1], 1));
#endif
d += 8*n_channels;
}
for(; n < n_samples; n++) {
@ -863,6 +873,7 @@ conv_f32d_to_s16_4s_avx2(void *data, void * SPA_RESTRICT dst, const void * SPA_R
out[2] = _mm256_unpacklo_epi32(out[0], out[1]); /* a0 b0 c0 d0 a1 b1 c1 d1 a4 b4 c4 d4 a5 b5 c5 d5 */
out[3] = _mm256_unpackhi_epi32(out[0], out[1]); /* a2 b2 c2 d2 a3 b3 c3 d3 a6 b6 c6 d6 a7 b7 c7 d7 */
#ifdef __x86_64__
*(int64_t*)(d + 0*n_channels) = _mm256_extract_epi64(out[2], 0); /* a0 b0 c0 d0 */
*(int64_t*)(d + 1*n_channels) = _mm256_extract_epi64(out[2], 1); /* a1 b1 c1 d1 */
*(int64_t*)(d + 2*n_channels) = _mm256_extract_epi64(out[3], 0); /* a2 b2 c2 d2 */
@ -871,6 +882,16 @@ conv_f32d_to_s16_4s_avx2(void *data, void * SPA_RESTRICT dst, const void * SPA_R
*(int64_t*)(d + 5*n_channels) = _mm256_extract_epi64(out[2], 3); /* a5 b5 c5 d5 */
*(int64_t*)(d + 6*n_channels) = _mm256_extract_epi64(out[3], 2); /* a6 b6 c6 d6 */
*(int64_t*)(d + 7*n_channels) = _mm256_extract_epi64(out[3], 3); /* a7 b7 c7 d7 */
#else
_mm_storel_pi((__m64*)(d + 0*n_channels), (__m128)_mm256_extracti128_si256(out[2], 0));
_mm_storeh_pi((__m64*)(d + 1*n_channels), (__m128)_mm256_extracti128_si256(out[2], 0));
_mm_storel_pi((__m64*)(d + 2*n_channels), (__m128)_mm256_extracti128_si256(out[3], 0));
_mm_storeh_pi((__m64*)(d + 3*n_channels), (__m128)_mm256_extracti128_si256(out[3], 0));
_mm_storel_pi((__m64*)(d + 4*n_channels), (__m128)_mm256_extracti128_si256(out[2], 1));
_mm_storeh_pi((__m64*)(d + 5*n_channels), (__m128)_mm256_extracti128_si256(out[2], 1));
_mm_storel_pi((__m64*)(d + 6*n_channels), (__m128)_mm256_extracti128_si256(out[3], 1));
_mm_storeh_pi((__m64*)(d + 7*n_channels), (__m128)_mm256_extracti128_si256(out[3], 1));
#endif
d += 8*n_channels;
}