[Openmp-commits] [PATCH] D60918: [OPENMP][NVPTX]Correctly handle L2 parallelism in SPMD mode.

[Alexey Bataev via Phabricator via Openmp-commits]

ABataev created this revision.
ABataev added reviewers: gtbercea, grokos.
Herald added subscribers: jdoerfert, guansong.
Herald added a project: OpenMP.

The parallelLevel counter must be on per-thread basis to fully support
L2+ parallelism, otherwise we may end up with undefined behavior.
Introduce the parallelLevel on per-warp basis using shared memory. It
allows to avoid the problems with the synchronization and allows fully
support L2+ parallelism in SPMD mode with no runtime.


Repository:
  rOMP OpenMP

https://reviews.llvm.org/D60918

Files:
  libomptarget/deviceRTLs/nvptx/src/libcall.cu
  libomptarget/deviceRTLs/nvptx/src/omp_data.cu
  libomptarget/deviceRTLs/nvptx/src/omptarget-nvptx.cu
  libomptarget/deviceRTLs/nvptx/src/omptarget-nvptx.h
  libomptarget/deviceRTLs/nvptx/src/parallel.cu
  libomptarget/deviceRTLs/nvptx/src/supporti.h
  libomptarget/deviceRTLs/nvptx/test/parallel/spmd_parallel_regions.cpp

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D60918.195904.patch
Type: text/x-patch
Size: 6812 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/openmp-commits/attachments/20190419/ec843f34/attachment.bin>