-
Notifications
You must be signed in to change notification settings - Fork 16
Known Issues
Matt Norman edited this page Apr 20, 2023
·
8 revisions
For YAKL current main branch testing status, please see the CDash:
https://my.cdash.org/index.php?project=YAKL
- The AMD AOCC 3.2 compiler has a bug in the
yakl::intrinsics::matinv
routine forFSArray
objects when the optimization is higher than-O1
. Please use AMD AOCC 4.0 or newer when using this routine, or use-O1
optimizations or lower when using AOCC3.2 with this routine. - When using the
nvhpc
compiler suite withYAKL_ARCH="CUDA"
, please set theYAKL_CUDA_FLAGS
CMake variable to flags that are valid fornvc++
(e.g.,-gpu=cc86
) rather than flags that are valid fornvcc
(e.g.,-arch sm_86
). YAKL detects theCMAKE_CXX_COMPILER
ID, and if it isNVHPC
orPGI
, it compiles the CUDA code with that compiler rather thannvcc
.