[ARM] Fix i1 shuffle lowering with multiple operands.
The existing lowering of i1 vector shuffle was only considering single-source shuffles, always assuming the second was undef. This extends that to properly handle both operands.
Loading
Please sign in to comment