[llvm] [RISCV] Lower VP_SELECT constant false to use vmerge.vxm/vmerge.vim (PR #144461)

[Pengcheng Wang via llvm-commits]

================
@@ -34,10 +34,10 @@ define <vscale x 1 x i8> @masked_load_passthru_nxv1i8(ptr %a, <vscale x 1 x i1>
 ; ZVE32:       # %bb.0:
 ; ZVE32-NEXT:    csrr a1, vlenb
 ; ZVE32-NEXT:    srli a1, a1, 3
-; ZVE32-NEXT:    vsetvli a2, zero, e8, mf4, ta, ma
-; ZVE32-NEXT:    vmv.v.i v8, 0
-; ZVE32-NEXT:    vsetvli zero, a1, e8, mf4, ta, mu
+; ZVE32-NEXT:    vsetvli zero, a1, e8, mf4, ta, ma
 ; ZVE32-NEXT:    vle8.v v8, (a0), v0.t
+; ZVE32-NEXT:    vmnot.m v0, v0
+; ZVE32-NEXT:    vmerge.vim v8, v8, 0, v0
----------------
wangpc-pp wrote:

There can be some microarchitecture optimizations like mv/zero idioms elimination, that is my point. And for this test case, it is a regression since we generate `vmnot.m+vmerge.vim` instead of a `vmv.vi`, the latency is doubled.

https://github.com/llvm/llvm-project/pull/144461