[llvm] [RISCV] Remove vmv.s.x and vmv.x.s lmul pseudo variants (PR #71501)

[Craig Topper via llvm-commits]

================
@@ -7400,34 +7394,13 @@ defm : VPatNullaryV<"int_riscv_vid", "PseudoVID">;
 // 16.1. Integer Scalar Move Instructions
 //===----------------------------------------------------------------------===//
 
-foreach vti = AllIntegerVectors in {
+foreach vti = NoGroupIntegerVectors in {
   let Predicates = GetVTypePredicates<vti>.Predicates in
   def : Pat<(XLenVT (riscv_vmv_x_s (vti.Vector vti.RegClass:$rs2))),
-            (!cast<Instruction>("PseudoVMV_X_S_" # vti.LMul.MX) $rs2, vti.Log2SEW)>;
+            (PseudoVMV_X_S $rs2, vti.Log2SEW)>;
   // vmv.s.x is handled with a custom node in RISCVInstrInfoVVLPatterns.td
 }
 
-//===----------------------------------------------------------------------===//
-// 16.2. Floating-Point Scalar Move Instructions
-//===----------------------------------------------------------------------===//
-
-foreach fvti = AllFloatVectors in {
-  let Predicates = GetVTypePredicates<fvti>.Predicates in {
-    def : Pat<(fvti.Vector (int_riscv_vfmv_s_f (fvti.Vector fvti.RegClass:$rs1),
-                           (fvti.Scalar fvti.ScalarRegClass:$rs2), VLOpFrag)),
-              (!cast<Instruction>("PseudoVFMV_S_"#fvti.ScalarSuffix#"_" #
----------------
topperc wrote:

What's the advantage of having a dead pattern? Or are you suggesting to not remap int_riscv_vfmv_s_f to RISCVISD::VFMV_S_F_VL during lowering?

https://github.com/llvm/llvm-project/pull/71501