ronlieb

[AMD Public Use] Hi Torsten Thanks for reporting this issue. I can reproduce the issue here. Looking into it a bit more, and adding it to our regularly run tests....

[AMD Public Use] I modified the example and can get it to run. I think the issues is that the mapper is accessing host memory on device. I changed it...

Hi Ron, Is this example being used on a unified shared memory (USM) machine. If not USM The idea of directly allocating on the device and storing the device pointer...

we have a test case in aomp/test/omp5/mapper_prob this compiles and runs with amd-stg-open , aomp 11.8-0 11.9-2 11.11-* but not aomp_13.0-2

unable to access this link: https://github.com/FabioLuporini/hpc-bugs/tree/main/omp-off-leak/c

@Lynd98 could you grab this testcase and valgrind it

the script is fixed in upcoming 5.4 release. the change is fairly straightforward update your copy from here: https://github.com/ROCm-Developer-Tools/aomp-extras/blob/aomp-dev/utils/bin/aompcc to move to clang/clang++/amdclang/amdclang++ explicitly add -v to your aompcc and...

did this deepbot issue get resolved ? ie can we close this issue?

a recent PR shows this is still an issue

the rpm can be manually unpacked, and the resultant tarfile, can be untarred anywhere you like to use: export PATH=/bin:$PATH