Devin Matthews

Results 264 comments of Devin Matthews

The Power9/10 kernels use a "broadcast packing" format, which messed up a lot of the older code, or rather, required some bespoke code which interacted poorly with `[cz]gemm1m`. I rewrote...

@RajalakshmiSR there might be a bug where data is written off the end of the output matrix. Please try the following test program: ```C #include "blis.h" #include #include int main(int...

And hemm/symm too right?

> Yes, I could see the same result. @RajalakshmiSR strange, that is exactly the test that is supposed to be failing in zblat3. I might write a Fortran version just...

Here's a Fortran version. It can't get much more similar to the failing test. ```f90 program main double complex A(2,2), B(2,2), C(2,2) double complex alpha, beta A(1,1) = (1.0,0.0) A(2,1)...

Linux seems unaffected. Windows is also unaffected according to the internet.

I'll leave this open as a warning to others.

I would have phrased it nicer but yes, the code in question is not currently in the FLAME BLIS repo. It is part of #567 which has not been merged...

I would say the we general do not claim to have full compatibility with the reference BLAS APIs, but that we strive to correct any inconsistencies when they are brought...

@ct-clmsn some of the things I have in mind can be used as kernels for various combinations of einsum parameters. Providing einsum itself is almost certainly out-of-scope, but yes we'd...