OpenMP_VV

OpenMP_VV copied to clipboard

The problem I'm seeing is a seg fault on the AMDGPU. I think it is happening because the target region is referencing a pointer (as if it's an array) without...

ddpagan

Break possible deadlock on lockstep execution

1

Some of the target_requires_atomic tests get on infinite loop with LLVM/Clang compiler because lockstep execution on the target device and the `if-else` order of execution. These tests are written assuming...

fel-cab

The OMP_CANCELLATION env variable has to be set to either true or false, other values like 1 are implemenation defined. Hence, the file name is changed from ...env_1.c to ..._env_true.c...

tob2

**Taken from Version 4.5 to 5.0 Differences Section of the 5.0 Specification** - [x] Iterators (see Section 2.1.6 on page 47) were added to support expressions in a list that...

andrewkallai

**Taken from Version 4.5 to 5.0 Differences Section of the 5.0 Specification** - Various changes throughout the specification were made to provide initial support of C11, C++11, C++14, C++17 and...

andrewkallai

Avoid stack allocation for test_requires_unified_shared_memory_heap*

6

Some (pseuo)USM implementations allocate memory in a special memory section that permits either direct memory access (DMA) from the device or on-access migration. In that case, stack and static memory...

tob2

Tests in C that are missing Fortran counterparts

2

We have a number of tests that are in C but still need to be written in Fortran: For 4.5 - [x] test_target_enter_exit_data_depend.c (PR Merged) - [ ] test_target_enter_exit_data_map_global_array.c -...

jhdavis8

https://github.com/SOLLVE/sollve_vv/blob/b628bffba66d6b5dc19f83d12dd083a6c0070f6f/tests/5.1/target/test_target_memcpy_rect_async_no_obj.c#L26 https://github.com/SOLLVE/sollve_vv/blob/4a432b4859c5e149b7f3296b830b926b8bbf6753/tests/5.1/target/test_target_memcpy_rect_async_depobj.c#L26 Change to: const size_t volume[2] = {N, M}; A similar suggestion was made for the Fortran tests in PR725, 726

andrewkallai

No target is specified in the test, so should OMPVV_TEST_OFFLOADING macro for offloading be removed or left alone?

2

https://github.com/SOLLVE/sollve_vv/blob/3d38296009168344babf88fdf91a89e6b3a6e129/tests/5.1/scope/test_scope_nowait_construct.c#L61

andrewkallai

test_target_update_mapper_from_discontiguous.F90 mapper not strided

1

**Describe the bug** The Fortran version of this test has mapper ``` !$omp declare mapper(newvec :: v)& !$omp& map(v, v%data(1:v%len)) ``` whereas the C version has mapper ``` #pragma omp...

bcornille