From 8061acd501f1cb6c00a886f4ee5cb9adc6cda39a Mon Sep 17 00:00:00 2001 From: Alexey Bataev Date: Wed, 20 Feb 2019 16:36:22 +0000 Subject: [OPENMP][NVPTX]Use faster teams reduction algorithm. A faster way to reduce the values in teams reductions was found, the codegen is updated to use this faster algorithm and new runtime functions. llvm-svn: 354479 --- clang/lib/Frontend/CompilerInvocation.cpp | 3 +++ 1 file changed, 3 insertions(+) (limited to 'clang/lib/Frontend/CompilerInvocation.cpp') diff --git a/clang/lib/Frontend/CompilerInvocation.cpp b/clang/lib/Frontend/CompilerInvocation.cpp index d4094ac5aea..1a33a00004d 100644 --- a/clang/lib/Frontend/CompilerInvocation.cpp +++ b/clang/lib/Frontend/CompilerInvocation.cpp @@ -2840,6 +2840,9 @@ static void ParseLangArgs(LangOptions &Opts, ArgList &Args, InputKind IK, Opts.OpenMPCUDABlocksPerSM = getLastArgIntValue(Args, options::OPT_fopenmp_cuda_blocks_per_sm_EQ, Opts.OpenMPCUDABlocksPerSM, Diags); + Opts.OpenMPCUDAReductionBufNum = getLastArgIntValue( + Args, options::OPT_fopenmp_cuda_teams_reduction_recs_num_EQ, + Opts.OpenMPCUDAReductionBufNum, Diags); } // Prevent auto-widening the representation of loop counters during an -- cgit v1.2.3