X32-packw AVX GIO remove maskload for remainder handling

google / XNNPACK

High-efficiency floating-point neural network inference operators for mobile, server, and Web

Other

1.89k stars 376 forks source link

Closed copybara-service[bot] closed 2 weeks ago

copybara-service[bot] commented 2 weeks ago

X32-packw AVX GIO remove maskload for remainder handling

do not rely on extrabytes for NC remainder
remove maskload for remainder handler to be asan friendly Co-authored-by: jun1.xu@intel.com