ardour/ardour
ardour/ardour
13
1
Fork 0
Code Pull Requests Releases Activity

Tune performance for AVX optimized find_peaks

Browse Source 
Propagating the changes from compute_peaks to find_peaks
This commit is contained in:
Ayan Shafqat 2022-01-02 15:58:48 -05:00 committed by Robin Gareus
parent a07ae169a3
commit 5fc3ae79ae
Signed by: rgareus
GPG Key ID: A090BCE02CF57F04
1 changed files with 31 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

68
libs/ardour/sse_functions_avx_linux.cc
View File

@ -29,11 +29,11 @@
#define IS_ALIGNED_TO(ptr, bytes) (((uintptr_t)ptr) % (bytes) == 0)
#if defined(__GNUC__)
#define IS_NOT_ALIGNED_TO(ptr, bytes) \
__builtin_expect(!!(reinterpret_cast<intptr_t>(ptr) % (bytes)), 0)
#define IS_NOT_ALIGNED_TO(ptr, bytes) \
__builtin_expect(!!(reinterpret_cast<intptr_t>(ptr) % (bytes)), 0)
#else
#define IS_NOT_ALIGNED_TO(ptr, bytes) \
(!!(reinterpret_cast<intptr_t>(ptr) % (bytes)))
#define IS_NOT_ALIGNED_TO(ptr, bytes) \
(!!(reinterpret_cast<intptr_t>(ptr) % (bytes)))
#endif
#ifdef __cplusplus
@ -155,9 +155,7 @@ x86_sse_avx_compute_peak(const float *src, uint32_t nframes, float current)
--nframes;
}
vmax = _mm256_max_ps(vmax, _mm256_permute2f128_ps(vmax, vmax, 1));
vmax = _mm256_max_ps(vmax, _mm256_permute_ps(vmax, _MM_SHUFFLE(0, 0, 3, 2)));
vmax = _mm256_max_ps(vmax, _mm256_permute_ps(vmax, _MM_SHUFFLE(0, 0, 0, 1)));
vmax = avx_getmax_ps(vmax);
// zero upper 128 bit of 256 bit ymm register to avoid penalties using non-AVX
// instructions.
@ -191,7 +189,7 @@ x86_sse_avx_find_peaks(const float *src, uint32_t nframes, float *minf, float *m
__m256 vmax = _mm256_broadcast_ss(maxf);
// Compute single min/max of unaligned portion until alignment is reached
while ((((intptr_t)src) % 32 != 0) && nframes > 0) {
while (IS_NOT_ALIGNED_TO(src, sizeof(__m256)) && nframes > 0) {
__m256 vsrc;
vsrc = _mm256_broadcast_ss(src);
@ -202,27 +200,31 @@ x86_sse_avx_find_peaks(const float *src, uint32_t nframes, float *minf, float *m
--nframes;
}
// Process the remaining samples 16 at a time
while (nframes >= 16)
// Process the aligned portion 32 samples at a time
while (nframes >= 32)
{
#if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
_mm_prefetch(((char *)src + 64), _mm_hint(0));
#ifdef _WIN32
_mm_prefetch(reinterpret_cast<char const *>(src + 32), _mm_hint(0));
#else
__builtin_prefetch(src + 64, 0, 0);
__builtin_prefetch(reinterpret_cast<void const *>(src + 32), 0, 0);
#endif
__m256 t0 = _mm256_load_ps(src + 0);
__m256 t1 = _mm256_load_ps(src + 8);
__m256 t2 = _mm256_load_ps(src + 16);
__m256 t3 = _mm256_load_ps(src + 24);
__m256 vsrc1, vsrc2;
vsrc1 = _mm256_load_ps(src + 0);
vsrc2 = _mm256_load_ps(src + 8);
vmax = _mm256_max_ps(vmax, t0);
vmax = _mm256_max_ps(vmax, t1);
vmax = _mm256_max_ps(vmax, t2);
vmax = _mm256_max_ps(vmax, t3);
vmax = _mm256_max_ps(vmax, vsrc1);
vmin = _mm256_min_ps(vmin, vsrc1);
vmin = _mm256_min_ps(vmin, t0);
vmin = _mm256_min_ps(vmin, t1);
vmin = _mm256_min_ps(vmin, t2);
vmin = _mm256_min_ps(vmin, t3);
vmax = _mm256_max_ps(vmax, vsrc2);
vmin = _mm256_min_ps(vmin, vsrc2);
src += 16;
nframes -= 16;
src += 32;
nframes -= 32;
}
// Process the remaining samples 8 at a time
@ -833,13 +835,9 @@ x86_sse_avx_mix_buffers_no_gain_aligned(float *dst, const float *src, uint32_t n
*/
static inline __m256 avx_getmax_ps(__m256 vmax)
{
__m256 tmp;
tmp = _mm256_shuffle_ps(vmax, vmax, _MM_SHUFFLE(2, 3, 0, 1));
vmax = _mm256_max_ps(tmp, vmax);
tmp = _mm256_shuffle_ps(vmax, vmax, _MM_SHUFFLE(1, 0, 3, 2));
vmax = _mm256_max_ps(tmp, vmax);
tmp = _mm256_permute2f128_ps(vmax, vmax, 1);
vmax = _mm256_max_ps(tmp, vmax);
vmax = _mm256_max_ps(vmax, _mm256_permute2f128_ps(vmax, vmax, 1));
vmax = _mm256_max_ps(vmax, _mm256_permute_ps(vmax, _MM_SHUFFLE(0, 0, 3, 2)));
vmax = _mm256_max_ps(vmax, _mm256_permute_ps(vmax, _MM_SHUFFLE(0, 0, 0, 1)));
return vmax;
}
@ -850,12 +848,8 @@ static inline __m256 avx_getmax_ps(__m256 vmax)
*/
static inline __m256 avx_getmin_ps(__m256 vmin)
{
__m256 tmp;
tmp = _mm256_shuffle_ps(vmin, vmin, _MM_SHUFFLE(2, 3, 0, 1));
vmin = _mm256_min_ps(tmp, vmin);
tmp = _mm256_shuffle_ps(vmin, vmin, _MM_SHUFFLE(1, 0, 3, 2));
vmin = _mm256_min_ps(tmp, vmin);
tmp = _mm256_permute2f128_ps(vmin, vmin, 1);
vmin = _mm256_min_ps(tmp, vmin);
vmin = _mm256_min_ps(vmin, _mm256_permute2f128_ps(vmin, vmin, 1));
vmin = _mm256_min_ps(vmin, _mm256_permute_ps(vmin, _MM_SHUFFLE(0, 0, 3, 2)));
vmin = _mm256_min_ps(vmin, _mm256_permute_ps(vmin, _MM_SHUFFLE(0, 0, 0, 1)));
return vmin;
}