This patch Shuffles FMAs with x86vector operations as operands such that FMAs are grouped with respect to odd/even packed index. Continuation to PR: https://github.com/llvm/llvm-project/pull/170267 to manage register allocation efficiently.
