saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-25 19:48:11 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Change _mm512_setzero to _mm512_setzero_ps (#17362)

Browse source 
### Description
_mm512_setzero is just an alias of _mm512_setzero_ps, and it is a wrong
one.
See: https://gcc.gnu.org/legacy-ml/gcc-patches/2018-05/msg00338.html
And
https://github.com/gcc-mirror/gcc/blob/master/gcc/config/i386/avx512fintrin.h
This commit is contained in:
Changming Sun 2023-08-31 13:11:44 -07:00 committed by GitHub
parent 3a53836836
commit ae90b716ff
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
onnxruntime/core/mlas/lib/q4gemm_avx512.cpp
View file

@ -85,10 +85,10 @@ MlasQ4GemmKernelAvx512f(
int64_t nblk = (int64_t)(CountN) - 4;
while (nblk >= 0) {
__m512 acc_lo0 = _mm512_setzero();
__m512 acc_lo1 = _mm512_setzero();
__m512 acc_lo2 = _mm512_setzero();
__m512 acc_lo3 = _mm512_setzero();
__m512 acc_lo0 = _mm512_setzero_ps();
__m512 acc_lo1 = _mm512_setzero_ps();
__m512 acc_lo2 = _mm512_setzero_ps();
__m512 acc_lo3 = _mm512_setzero_ps();
const auto* b = b_col;
for (size_t k = 0; k < CountK; k += Q4Type::BlkLen) {
@ -1092,7 +1092,7 @@ MlasQ80BlkQuantRow(const float* A, void* Qblob, size_t size)
for (size_t k = 0; k < size; k += QType::BlkLen) {
const size_t step = std::min(QType::BlkLen, size - k);
__m512 maxAbs = _mm512_setzero();
__m512 maxAbs = _mm512_setzero_ps();
for (size_t kk = 0; kk < step; kk += 16) {
const size_t klen = std::min(size_t(16), step - kk);