Youngsung Kim
Youngsung Kim
> I will try this and share its result here.
@abbotts Yes, this issue is the same to OLCFDEV-977.
@sarats , I created a simple test case with two simple source files and a long directory path below. This test confirms that the issue is related to amdflang, not...
@dmcdougall, yes, testmod.mod files exist in both directory paths pointed by -I flags in the two test cases. I pre-compiled testmod.F90 in the both directories so that each directory contains...
@arghdos , yes, I am still seeing the same error with rocm 5.1.0 and 4.5.0 on Crusher. I think that the different test outcome comes from the directory path where...
@sarats , the only known issue of the failed test cases is this issue. The other failed test cases might be related to this issue too but that could be...
@sarats, @arghdos , @ronlieb, I checked that rocm 5.3.0 fixed this issue. There are another issues that I found during testing and the new issues will be reported separately. I...
> $!omp loop works for me too with Cray compiler on Crusher. But, it did not work with Amd compiler on Crusher. I think that I need something that works...
> I am still learning openmp target and can not tell what is valid or not. But your suggestion (!$omp parallel do simd collapse(2)) seems to me reasonable. In the...
This issue is resolved by providing proper omp target clauses as explained above.