We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 78b2caf commit 30de7afCopy full SHA for 30de7af
NeoMathEngine/src/CPU/CPUInfo.h
@@ -149,6 +149,9 @@ struct CCPUInfo final {
149
#error "Platform isn't supported!"
150
#endif // ERROR FINE_PLATFORM
151
}
152
+ if( HasAvx512And2Fma ) {
153
+ floatAlignment = 16;
154
+ }
155
156
return floatAlignment;
157
#endif // !NEOML_USE_NEON
NeoMathEngine/src/CPU/x86/avx512/Avx512Functions.h
@@ -24,7 +24,7 @@ namespace NeoML {
24
namespace Avx512 {
25
26
// The minimum vector size recommended for using AVX512 vector functions
27
-static constexpr int VectorMathMinSize = 32;
+static constexpr int VectorMathMinSize = 16;
28
29
void dataCopy( float* dst, const float* src, int vectorSize );
30
0 commit comments