[Openmp-commits] [compiler-rt] [llvm] [openmp] [TSan] Add instrumentation of AVX2 and AVX512 instructions (PR #74636)

[via Openmp-commits]

================
@@ -213,6 +219,26 @@ void ThreadSanitizer::initialize(Module &M, const TargetLibraryInfo &TLI) {
   TsanIgnoreEnd =
       M.getOrInsertFunction("__tsan_ignore_thread_end", Attr, IRB.getVoidTy());
   IntegerType *OrdTy = IRB.getInt32Ty();
+
----------------
jprotze wrote:

Would it be ok to use `#pragma omp simd` and the `-fopenmp-simd` flag to convince clang to generate vector instructions? Or should we use clang-specific pragmas?

https://github.com/llvm/llvm-project/pull/74636