120 lines
3.6 KiB
C++
120 lines
3.6 KiB
C++
/*
|
|
Copyright (C) 2007 Paul sDavis
|
|
Written by Sampo Savolainen
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation; either version 2 of the License, or
|
|
(at your option) any later version.
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program; if not, write to the Free Software
|
|
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
|
|
|
|
*/
|
|
|
|
#include <xmmintrin.h>
|
|
#include <immintrin.h>
|
|
#include <stdint.h>
|
|
|
|
|
|
void
|
|
x86_sse_avx_find_peaks(const float* buf, uint32_t nframes, float *min, float *max)
|
|
{
|
|
__m256 current_max, current_min, work;
|
|
|
|
// Load max and min values into all eight slots of the YMM registers
|
|
current_min = _mm256_set1_ps(*min);
|
|
current_max = _mm256_set1_ps(*max);
|
|
|
|
// Work input until "buf" reaches 16 byte alignment
|
|
while ( ((intptr_t)buf) % 32 != 0 && nframes > 0) {
|
|
|
|
// Load the next float into the work buffer
|
|
work = _mm256_set1_ps(*buf);
|
|
|
|
current_min = _mm256_min_ps(current_min, work);
|
|
current_max = _mm256_max_ps(current_max, work);
|
|
|
|
buf++;
|
|
nframes--;
|
|
}
|
|
|
|
// use 64 byte prefetch for quadruple quads:
|
|
// load each 64 bytes into cash before processing
|
|
while (nframes >= 16) {
|
|
#if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
|
|
_mm_prefetch(((char*)buf+64), _mm_hint(0) );
|
|
#else
|
|
__builtin_prefetch(buf+64,0,0);
|
|
#endif
|
|
work = _mm256_load_ps(buf);
|
|
current_min = _mm256_min_ps(current_min, work);
|
|
current_max = _mm256_max_ps(current_max, work);
|
|
buf+=8;
|
|
work = _mm256_load_ps(buf);
|
|
current_min = _mm256_min_ps(current_min, work);
|
|
current_max = _mm256_max_ps(current_max, work);
|
|
buf+=8;
|
|
|
|
nframes-=16;
|
|
}
|
|
|
|
// work through 32 bytes aligned buffers
|
|
while (nframes >= 8) {
|
|
|
|
work = _mm256_load_ps(buf);
|
|
|
|
current_min = _mm256_min_ps(current_min, work);
|
|
current_max = _mm256_max_ps(current_max, work);
|
|
|
|
buf+=8;
|
|
nframes-=8;
|
|
}
|
|
|
|
// work through the rest < 4 samples
|
|
while ( nframes > 0) {
|
|
|
|
// Load the next float into the work buffer
|
|
work = _mm256_set1_ps(*buf);
|
|
|
|
current_min = _mm256_min_ps(current_min, work);
|
|
current_max = _mm256_max_ps(current_max, work);
|
|
|
|
buf++;
|
|
nframes--;
|
|
}
|
|
|
|
// Find min & max value in current_max through shuffle tricks
|
|
|
|
work = current_min;
|
|
work = _mm256_shuffle_ps (current_min, current_min, _MM_SHUFFLE(2, 3, 0, 1));
|
|
current_min = _mm256_min_ps (work, current_min);
|
|
work = _mm256_shuffle_ps (current_min, current_min, _MM_SHUFFLE(1, 0, 3, 2));
|
|
current_min = _mm256_min_ps (work, current_min);
|
|
work = _mm256_permute2f128_ps( current_min, current_min, 1);
|
|
current_min = _mm256_min_ps (work, current_min);
|
|
|
|
*min = current_min[0];
|
|
|
|
work = current_max;
|
|
work = _mm256_shuffle_ps(current_max, current_max, _MM_SHUFFLE(2, 3, 0, 1));
|
|
current_max = _mm256_max_ps (work, current_max);
|
|
work = _mm256_shuffle_ps(current_max, current_max, _MM_SHUFFLE(1, 0, 3, 2));
|
|
current_max = _mm256_max_ps (work, current_max);
|
|
work = _mm256_permute2f128_ps( current_max, current_max, 1);
|
|
current_max = _mm256_max_ps (work, current_max);
|
|
|
|
*max = current_max[0];
|
|
|
|
// zero upper 128 bit of 256 bit ymm register to avoid penalties using non-AVX instructions
|
|
_mm256_zeroupper ();
|
|
}
|
|
|
|
|