Enable trsm test on NVIDIA GPUs when using DEFAULT TUNING_TARGET

codeplaysoftware / portBLAS

An implementation of BLAS using the SYCL open standard.

Apache License 2.0

250 stars 48 forks source link

Enable trsm test on NVIDIA GPUs when using DEFAULT TUNING_TARGET #518

Closed s-Nick closed 3 months ago

s-Nick commented 3 months ago

This PR fixes the error in trsm that led to disable its tests. Adding special icpx flag -mllvm=loopopt=0 to the linker solves the issue. A minor change was necessary to compile gemm with mixed precision for default TUNING_TARGET.