Fix an issue where fast math flags were dropped during scalarization.
Most portions of InstCombine properly propagate fast math flags, but apparently the vector scalarization section was overlooked. llvm-svn: 262376
Loading
Please sign in to comment