[RISCV] Combine (vmv.s.x a, (vmv.x.s b)) to nothing.

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

if (Val.getOpcode() == RISCVISD::VMV_X_S) {
             DAG.getConstant(1, DL, Subtarget.getXLenVT()), MemVT,
             Store->getMemOperand(), Store->getAddressingMode(),
             Store->isTruncatingStore(), /*IsCompress*/ false);
       }
     }
 
     break;
   }
+  case RISCVISD::VMV_S_X_VL: {
+    SDValue Src = N->getOperand(1);
+    if (Src.getOpcode() == RISCVISD::VMV_X_S &&
+        Src.getOperand(0).getValueType() == N->getValueType(0) &&
+        isa<ConstantSDNode>(N->getOperand(2))) {
+      if (N->getOperand(0).isUndef())
+        return Src.getOperand(0);
+      if (cast<ConstantSDNode>(N->getOperand(2))->getZExtValue() == 0)
+        return N->getOperand(0);
+      SDLoc DL(N);
+      if (Src.getOperand(0).getOpcode() == RISCVISD::VSLIDEDOWN_VL &&
+          Src.getOperand(0).getOperand(0).isUndef()) {
+        auto Slidedown = Src.getOperand(0);
+        if (isa<ConstantSDNode>(Slidedown.getOperand(4)) &&
+            cast<ConstantSDNode>(Slidedown.getOperand(4))->getZExtValue() == 0)
+          return N->getOperand(0);
+        if ((!isa<ConstantSDNode>(Slidedown.getOperand(4)) ||
+             1 < cast<ConstantSDNode>(Slidedown.getOperand(4))
+                     ->getZExtValue()) &&
+            !Slidedown.hasOneUse())
+          return SDValue();
+        return DAG.getNode(Slidedown.getOpcode(), DL,
+                           Slidedown.getSimpleValueType(), N->getOperand(0),
+                           Slidedown.getOperand(1), Slidedown.getOperand(2),
+                           Slidedown.getOperand(3), Slidedown.getOperand(4));
+      }
+      MVT VecVT = N->getSimpleValueType(0);
+      MVT XLenVT = Subtarget.getXLenVT();
+      SDValue Mask, VL;
+      std::tie(Mask, VL) = getDefaultScalableVLOps(VecVT, DL, DAG, Subtarget);
+      SDValue Zero = DAG.getConstant(0, DL, XLenVT);
+      SDValue OneVL = DAG.getConstant(1, DL, XLenVT);
+      return DAG.getNode(RISCVISD::VSLIDEUP_VL, DL, VecVT, N->getOperand(0),
+                         Src.getOperand(0), Zero, Mask, OneVL);
+    }
+    break;
+  }
   }
 
   return SDValue();
 }
 
 bool RISCVTargetLowering::isDesirableToCommuteWithShift(
     const SDNode *N, CombineLevel Level) const {
   // The following folds are only desirable if `(OP _, c1 << c2)` can be

llvm/test/CodeGen/RISCV/rvv/vmv.s.x-vmv.x.s.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v -riscv-v-vector-bits-min=128 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v -riscv-v-vector-bits-min=128 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK
+
+define <4 x i32> @test_0_0(<4 x i32> %0, <4 x i32> %1) {
+; CHECK-LABEL: test_0_0:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e32, m1, tu, mu
+; CHECK-NEXT:    vslideup.vi v8, v9, 0
+; CHECK-NEXT:    ret
+  %3 = extractelement <4 x i32> %1, i64 0
+  %4 = insertelement <4 x i32> %0, i32 %3, i64 0
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @test_0_1(<4 x i32> %0, <4 x i32> %1) {
+; CHECK-LABEL: test_0_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e32, m1, tu, mu
+; CHECK-NEXT:    vslidedown.vi v8, v9, 1
+; CHECK-NEXT:    ret
+  %3 = extractelement <4 x i32> %1, i64 1
+  %4 = insertelement <4 x i32> %0, i32 %3, i64 0
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @test_1_0(<4 x i32> %0, <4 x i32> %1) {
+; CHECK-LABEL: test_1_0:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e32, m1, tu, mu
+; CHECK-NEXT:    vslideup.vi v8, v9, 1
+; CHECK-NEXT:    ret
+  %3 = extractelement <4 x i32> %1, i64 0
+  %4 = insertelement <4 x i32> %0, i32 %3, i64 1
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @test_1_1(<4 x i32> %0, <4 x i32> %1) {
+; CHECK-LABEL: test_1_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e32, m1, ta, mu
+; CHECK-NEXT:    vslidedown.vi v9, v9, 1
+; CHECK-NEXT:    vsetivli zero, 2, e32, m1, tu, mu
+; CHECK-NEXT:    vslideup.vi v8, v9, 1
+; CHECK-NEXT:    ret
+  %3 = extractelement <4 x i32> %1, i64 1
+  %4 = insertelement <4 x i32> %0, i32 %3, i64 1
+  ret <4 x i32> %4
+}