projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 11cadc3) | patch
[OPENMP][NVPTX]Improved interwarp copy function.
authorAlexey Bataev <a.bataev@hotmail.com>
Fri, 14 Dec 2018 21:00:58 +0000 (21:00 +0000)
committerAlexey Bataev <a.bataev@hotmail.com>
Fri, 14 Dec 2018 21:00:58 +0000 (21:00 +0000)
commitae51b96f99bb0fb4956465633e8b036c57c21007
tree139af5b2763752afaa0baec0a6997e78b17d8085tree | snapshot
parent11cadc3e6b42467f6c5388aa6bf5ee619d8146c5commit | diff
[OPENMP][NVPTX]Improved interwarp copy function.

Inlined runtime with the current implementation of the interwarp copy
function leads to the undefined behavior because of the not quite
correct implementation of the barriers. Start using generic
__kmpc_barier function instead of the custom made barriers.

llvm-svn: 349192
clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp diff | blob | history
clang/test/OpenMP/nvptx_target_parallel_reduction_codegen.cpp diff | blob | history
clang/test/OpenMP/nvptx_teams_reduction_codegen.cpp diff | blob | history
Domain: System / Toolchain;