llvm-project

4295cf1a - [msan][test] Add tests for llvm.vector.partial.reduce.add/fadd (#229196)

[msan][test] Add tests for llvm.vector.partial.reduce.add/fadd (#229196) Mostly, the first and second arguments are the same (e.g., `<4 x i32> @llvm.vector.partial.reduce.add(<4 x i32>, <4 x i32>)`) which can be handled heuristically, but sometimes the second argument is a larger multiple of the first argument (e.g., `<4 x i32> @llvm.vector.partial.reduce.add(<4 x i32>, <8 x i32>)`), which is handled strictly. N.B. although these intrinsics are LLVM cross-platform, we keep the tests in the AArch64 directory to maintain the correspondence with the directory (llvm/test/CodeGen/AArch64/) that the tests are forked from.