[llvm] [RISCV][VLOPT] Added support for vmv and vfmv (PR #149562)

[Luke Lau via llvm-commits]

================
@@ -129,9 +129,8 @@ define <vscale x 8 x i8> @insertelt_nxv8i8_idx(<vscale x 8 x i8> %v, i8 signext
 ; CHECK-LABEL: insertelt_nxv8i8_idx:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    addi a2, a1, 1
-; CHECK-NEXT:    vsetvli a3, zero, e8, m1, ta, ma
-; CHECK-NEXT:    vmv.s.x v9, a0
 ; CHECK-NEXT:    vsetvli zero, a2, e8, m1, tu, ma
+; CHECK-NEXT:    vmv.s.x v9, a0
----------------
lukel97 wrote:

That's interesting that we now perform the vmv.s.x in tu policy. I thought that we didn't go from ta -> tu generally? Is the passthru in this vmv.s.x not poison? Either way I'm going out on a limb here and assuming this is profitable given we're avoiding a VL toggle

https://github.com/llvm/llvm-project/pull/149562