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

audioconvert: handle odd writes in delay

Browse source 
Add some padding to the delay buffer. If we wrap around, copy the
spilled samples to the front of the buffer. This makes it possible to
use the more optimized sse delay function in more cases.
This commit is contained in:
Wim Taymans 2024-10-23 12:34:04 +02:00
parent 3309e0b244
commit 662bf68122
2 changed files with 14 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

17
spa/plugins/audioconvert/channelmix-ops-sse.c
View file

@ -287,8 +287,7 @@ static inline void delay_convolve_run_sse(float *buffer, uint32_t *pos,
uint32_t n, unrolled;
if (SPA_IS_ALIGNED(src, 16) &&
SPA_IS_ALIGNED(dst, 16) &&
(w & 3) == 0)
SPA_IS_ALIGNED(dst, 16))
unrolled = n_samples & ~3;
else
unrolled = 0;
@ -301,7 +300,12 @@ static inline void delay_convolve_run_sse(float *buffer, uint32_t *pos,
t[0] = _mm_loadu_ps(&buffer[w+o]);
t[0] = _mm_mul_ps(t[0], v);
_mm_store_ps(&dst[n], t[0]);
w = w + 4 >= n_buffer ? 0 : w + 4;
w += 4;
if (w >= n_buffer) {
w -= n_buffer;
t[0] = _mm_loadu_ps(&buffer[n_buffer]);
_mm_storeu_ps(&buffer[0], t[0]);
}
}
for(; n < n_samples; n++) {
t[0] = _mm_load_ss(&src[n]);
@ -319,7 +323,12 @@ static inline void delay_convolve_run_sse(float *buffer, uint32_t *pos,
_mm_storeu_ps(&buffer[w+n_buffer], t[0]);
for(i = 0; i < 4; i++)
convolver_run(&buffer[w+o+i], &dst[n+i], taps, n_taps, v);
w = w + 4 >= n_buffer ? 0 : w + 4;
w += 4;
if (w >= n_buffer) {
w -= n_buffer;
t[0] = _mm_loadu_ps(&buffer[n_buffer]);
_mm_storeu_ps(&buffer[0], t[0]);
}
}
for(; n < n_samples; n++) {
t[0] = _mm_load_ss(&src[n]);