Re: RFS: hipsolver/5.5.1-1 [ITP] -- portable interface for GPU LAPACK routines
Hi Cory,
On 2023-10-13 11:16, Cordell Bloor wrote:
> The hipSOLVER library was relatively straightforward to package. I
> notice that PyTorch now depends on hipSOLVER, which was not true at the
> time I wrote the ITP.
looked great, built fine, and tested successfully on gfx1030+gfx1034.
Thanks!
Side note: there was an issue last time with NEW packages not running
correctly in CI once they were accepted into experimental. I think this
was a mirror thing. This should be fixed, but I'm monitoring it.
Best,
Christian
Reply to: