[all-commits] [llvm/llvm-project] 8f668c: [MLIR][OpenMP] Prevent teams reductions from deadl...

[Sergio Afonso via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 8f668cec471355b1d04692f2483679dc1cb35990
      https://github.com/llvm/llvm-project/commit/8f668cec471355b1d04692f2483679dc1cb35990
  Author: Sergio Afonso <safonsof at amd.com>
  Date:   2026-03-09 (Mon, 09 Mar 2026)

  Changed paths:
    M mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
    M mlir/test/Target/LLVMIR/omptarget-teams-distribute-reduction.mlir
    M mlir/test/Target/LLVMIR/omptarget-teams-reduction.mlir
    M mlir/test/Target/LLVMIR/openmp-teams-distribute-reduction.mlir
    M mlir/test/Target/LLVMIR/openmp-teams-reduction.mlir

  Log Message:
  -----------
  [MLIR][OpenMP] Prevent teams reductions from deadlocking (#184625)

Currently, simple Fortran reductions like the example below cause a
deadlock at runtime:

```f90
integer :: i, x

!$omp teams distribute reduction(+:x)
do i=1, 10
  x = x + 1
end do
```

Preventing a redundant barrier from being added in that case addresses
this issue. Synchronization is already being handled by the
`__kmpc_reduce` and `__kmpc_end_reduce` runtime calls for the host, and
by the OMPIRBuilder-generated `_omp_reduction_inter_warp_copy_func`
function for GPUs.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications