Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

bugfix of FMA port to FAST_FLOAT: 8 float FPs fit in a single 256bit #1

Merged
merged 1 commit into from
Jul 13, 2021
Merged

bugfix of FMA port to FAST_FLOAT: 8 float FPs fit in a single 256bit #1

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
bugfix of FMA port to FAST_FLOAT: 8 float FPs fit in a single 256bit … 
…vector (8x32) (contrasting 4 double FPs: 4*64)
  • Loading branch information
@GerHobbelt
GerHobbelt committed Jul 13, 2021
commit 30bf2635fd6a31e534e354608052db415ea12ba9
16 changes: 8 additions & 8 deletions src/arch/dotproductfma.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,26 +31,26 @@ namespace tesseract {
// Uses Intel FMA intrinsics to access the SIMD instruction set.
#if defined(FAST_FLOAT)
TFloat DotProductFMA(const TFloat *u, const TFloat *v, int n) {
const unsigned quot = n / 8;
const unsigned rem = n % 8;
const unsigned quot = n / 16;
const unsigned rem = n % 16;
__m256 t0 = _mm256_setzero_ps();
__m256 t1 = _mm256_setzero_ps();
for (unsigned k = 0; k < quot; k++) {
__m256 f0 = _mm256_loadu_ps(u);
__m256 f1 = _mm256_loadu_ps(v);
t0 = _mm256_fmadd_ps(f0, f1, t0);
u += 4;
v += 4;
u += 8;
v += 8;
__m256 f2 = _mm256_loadu_ps(u);
__m256 f3 = _mm256_loadu_ps(v);
t1 = _mm256_fmadd_ps(f2, f3, t1);
u += 4;
v += 4;
u += 8;
v += 8;
}
t0 = _mm256_hadd_ps(t0, t1);
alignas(32) float tmp[4];
alignas(32) TFloat tmp[8];
_mm256_store_ps(tmp, t0);
float result = tmp[0] + tmp[1] + tmp[2] + tmp[3];
TFloat result = tmp[0] + tmp[1] + tmp[2] + tmp[3] + tmp[4] + tmp[5] + tmp[6] + tmp[7];
for (unsigned k = 0; k < rem; k++) {
result += *u++ * *v++;
}
Expand Down