llvm-project

eff183b6 - [AArch64] Wrap integer SCALAR_TO_VECTOR nodes in bitcasts (#172837)

[AArch64] Wrap integer SCALAR_TO_VECTOR nodes in bitcasts (#172837) This patch changes lowering of SCALAR_TO_VECTOR node with i32 or i64 by inserting bitconvert to/from floating point type infront of its operands. This bitconverts serve as a hint to instruction selector and they make it clear that the operand and return value of this node need to be NEON registers. Depends on: #179315 --------- Co-authored-by: Kerry McLaughlin <kerry.mclaughlin@arm.com>