[RISCV] Narrow types of index operand matched pattern (shl (zext), C).

[fakepaper56]

Summary

(shl (zext to iXLenVec), C) is a possible pattern in auto-vectorized code for
indexed loads/stores. But extending to iXLen might be too aggressive, RVV
indexed load/store instructions zero extend their indexed operand to XLEN.
The patch tries to narrow the type of the zero extension. It's benefit to
decrease register pressure.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

10639

extended*

10648

I don't think this should be a TargetConstant.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

10634

Can we guarantee Src has an MVT type here? The intrinsic could be used with an illegal zext input type before type legalization.

10635

Can SrcElen be 1 if the input is a mask type before we legalize zero_extend?

12982

I think this should be getMemIntrinsicSDNode to preserve the mem operand.

[fakepaper56]

Address comment.

[fakepaper56]

Fix typo.

[fakepaper56]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

10634

I replaced MVT with EVT.

10635

I think it is acceptable after not using MVT.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

10637

Can this create element sizes like i4 or i2?

[fakepaper56]

Avoid generating to i2 or i4 type.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

10634

Do we have tests where SrcVT is not an MVT?

10638

Do we have tests for the case where NewElen is less than 8?

[fakepaper56]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

10634

I think test_vloxei6 and test_vloxei7 in llvm/test/CodeGen/RISCV/rvv/narrow-shift-extend.ll are the case.

10638

I think test_vloxei7 in llvm/test/CodeGen/RISCV/rvv/narrow-shift-extend.ll is the case.

[craig.topper]

LGTM

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

   setPrefFunctionAlignment(Subtarget.getPrefFunctionAlignment());
   setPrefLoopAlignment(Subtarget.getPrefLoopAlignment());
 
   setMinimumJumpTableEntries(5);
 
   // Jumps are expensive, compared to logic
   setJumpIsExpensive();
 
-  setTargetDAGCombine({ISD::INTRINSIC_WO_CHAIN, ISD::ADD, ISD::SUB, ISD::AND,
+  setTargetDAGCombine({ISD::INTRINSIC_VOID, ISD::INTRINSIC_W_CHAIN,
+                       ISD::INTRINSIC_WO_CHAIN, ISD::ADD, ISD::SUB, ISD::AND,
                        ISD::OR, ISD::XOR, ISD::SETCC, ISD::SELECT});
   if (Subtarget.is64Bit())
     setTargetDAGCombine(ISD::SRA);
 
   if (Subtarget.hasStdExtFOrZfinx())
     setTargetDAGCombine({ISD::FADD, ISD::FMAXNUM, ISD::FMINNUM});
 
   if (Subtarget.hasStdExtZbb())
static SDValue performXORCombine(SDNode *N, SelectionDAG &DAG,
 
   if (SDValue V = combineBinOpToReduce(N, DAG, Subtarget))
     return V;
   // fold (xor (select cond, 0, y), x) ->
   //      (select cond, x, (xor x, y))
   return combineSelectAndUseCommutative(N, DAG, /*AllOnes*/ false, Subtarget);
 }
 
+// According to the property that indexed load/store instructions
+// zero-extended their indices, \p narrowIndex tries to narrow the type of index
+// operand if it is matched to pattern (shl (zext x to ty), C) and bits(x) + C <
+// bits(ty).
+static SDValue narrowIndex(SDValue N, SelectionDAG &DAG) {
+  if (N.getOpcode() != ISD::SHL || !N->hasOneUse())
+    return SDValue();
+
+  SDValue N0 = N.getOperand(0);
+  if (N0.getOpcode() != ISD::ZERO_EXTEND &&
+      N0.getOpcode() != RISCVISD::VZEXT_VL)
+    return SDValue();
+  if (!N0->hasOneUse())
+    return SDValue();
+
+  APInt ShAmt;
+  SDValue N1 = N.getOperand(1);
+  if (!ISD::isConstantSplatVector(N1.getNode(), ShAmt))
+    return SDValue();
+
+  SDLoc DL(N);
+  SDValue Src = N0.getOperand(0);
+  MVT SrcVT = Src.getSimpleValueType();
+  unsigned SrcElen = SrcVT.getScalarSizeInBits();
+  unsigned ShAmtV = ShAmt.getZExtValue();
+  unsigned NewElen = PowerOf2Ceil(SrcElen + ShAmtV);
+
+  // Skip if NewElen is not narrower than the original extneded type.
+  if (NewElen >= N0.getValueType().getScalarSizeInBits())
+    return SDValue();
+
+  MVT NewEltVT = MVT::getIntegerVT(NewElen);
+  MVT NewVT = SrcVT.changeVectorElementType(NewEltVT);
+
+  SDValue NewExt = DAG.getNode(N0->getOpcode(), DL, NewVT, N0->ops());
+  SDValue NewShAmtVec =
+      DAG.getSplat(NewVT, DL, DAG.getTargetConstant(ShAmtV, DL, NewEltVT));
+  return DAG.getNode(ISD::SHL, DL, NewVT, NewExt, NewShAmtVec);
+}
+
 // Replace (seteq (i64 (and X, 0xffffffff)), C1) with
 // (seteq (i64 (sext_inreg (X, i32)), C1')) where C1' is C1 sign extended from
 // bit 31. Same for setne. C1' may be cheaper to materialize and the sext_inreg
 // can become a sext.w instead of a shift pair.
 static SDValue performSETCCCombine(SDNode *N, SelectionDAG &DAG,
                                    const RISCVSubtarget &Subtarget) {
   SDValue N0 = N->getOperand(0);
   SDValue N1 = N->getOperand(1);
if (N->getOperand(0).isUndef() &&
       assert(SrcVT.getVectorElementType() == VT.getVectorElementType());
       // Widths match, just return the original vector.
       if (SrcVT == VT)
         return Src.getOperand(0);
       // TODO: Use insert_subvector/extract_subvector to change widen/narrow?
     }
     break;
   }
+  case ISD::INTRINSIC_VOID:
+  case ISD::INTRINSIC_W_CHAIN:
   case ISD::INTRINSIC_WO_CHAIN: {
-    unsigned IntNo = N->getConstantOperandVal(0);
+    unsigned IntOpNo = N->getOpcode() == ISD::INTRINSIC_WO_CHAIN ? 0 : 1;
+    unsigned IntNo = N->getConstantOperandVal(IntOpNo);
     switch (IntNo) {
       // By default we do not combine any intrinsic.
     default:
       return SDValue();
     case Intrinsic::riscv_vcpop:
     case Intrinsic::riscv_vcpop_mask:
     case Intrinsic::riscv_vfirst:
     case Intrinsic::riscv_vfirst_mask: {
       SDValue VL = N->getOperand(2);
       if (IntNo == Intrinsic::riscv_vcpop_mask ||
           IntNo == Intrinsic::riscv_vfirst_mask)
         VL = N->getOperand(3);
       if (!isNullConstant(VL))
         return SDValue();
       // If VL is 0, vcpop -> li 0, vfirst -> li -1.
       SDLoc DL(N);
       EVT VT = N->getValueType(0);
       if (IntNo == Intrinsic::riscv_vfirst ||
           IntNo == Intrinsic::riscv_vfirst_mask)
         return DAG.getConstant(-1, DL, VT);
       return DAG.getConstant(0, DL, VT);
     }
+    case Intrinsic::riscv_vloxei:
+    case Intrinsic::riscv_vloxei_mask:
+    case Intrinsic::riscv_vluxei:
+    case Intrinsic::riscv_vluxei_mask:
+    case Intrinsic::riscv_vsoxei:
+    case Intrinsic::riscv_vsoxei_mask:
+    case Intrinsic::riscv_vsuxei:
+    case Intrinsic::riscv_vsuxei_mask:
+      if (SDValue V = narrowIndex(N->getOperand(4), DAG)) {
+        SmallVector<SDValue, 8> Ops(N->ops());
+        Ops[4] = V;
+        return DAG.getNode(N->getOpcode(), SDLoc(N), N->getVTList(), Ops);
+      }
+      return SDValue();
     }
   }
   case ISD::BITCAST: {
     assert(Subtarget.useRVVForFixedLengthVectors());
     SDValue N0 = N->getOperand(0);
     EVT VT = N->getValueType(0);
     EVT SrcVT = N0.getValueType();
     // If this is a bitcast between a MVT::v4i1/v2i1/v1i1 and an illegal integer

llvm/test/CodeGen/RISCV/rvv/mgather-sdnode.ll

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds i16, ptr %base, <vscale x 8 x i16> %eidxs
   %v = call <vscale x 8 x i16> @llvm.masked.gather.nxv8i16.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 2, <vscale x 8 x i1> %m, <vscale x 8 x i16> %passthru)
   ret <vscale x 8 x i16> %v
 }
 
 define <vscale x 8 x i16> @mgather_baseidx_zext_nxv8i8_nxv8i16(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i16> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i16:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vadd.vv v12, v12, v12
-; RV32-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
-; RV32-NEXT:    vluxei32.v v10, (a0), v12, v0.t
-; RV32-NEXT:    vmv.v.v v8, v10
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i16:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vadd.vv v16, v16, v16
-; RV64-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
-; RV64-NEXT:    vluxei64.v v10, (a0), v16, v0.t
-; RV64-NEXT:    vmv.v.v v8, v10
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vzext.vf2 v12, v8
+; CHECK-NEXT:    vsll.vi v8, v12, 1
+; CHECK-NEXT:    vluxei16.v v10, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds i16, ptr %base, <vscale x 8 x i16> %eidxs
   %v = call <vscale x 8 x i16> @llvm.masked.gather.nxv8i16.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 2, <vscale x 8 x i1> %m, <vscale x 8 x i16> %passthru)
   ret <vscale x 8 x i16> %v
 }
 
 define <vscale x 8 x i16> @mgather_baseidx_nxv8i16(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i16> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8i16:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m, <vscale x 8 x i32> %passthru)
   ret <vscale x 8 x i32> %v
 }
 
 define <vscale x 8 x i32> @mgather_baseidx_zext_nxv8i8_nxv8i32(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i32> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, mu
-; RV32-NEXT:    vzext.vf4 v16, v8
-; RV32-NEXT:    vsll.vi v8, v16, 2
-; RV32-NEXT:    vluxei32.v v12, (a0), v8, v0.t
-; RV32-NEXT:    vmv.v.v v8, v12
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vsll.vi v16, v16, 2
-; RV64-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
-; RV64-NEXT:    vluxei64.v v12, (a0), v16, v0.t
-; RV64-NEXT:    vmv.v.v v8, v12
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v10, v8
+; CHECK-NEXT:    vsll.vi v8, v10, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vluxei16.v v12, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v12
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m, <vscale x 8 x i32> %passthru)
   ret <vscale x 8 x i32> %v
 }
 
 define <vscale x 8 x i32> @mgather_baseidx_nxv8i16_nxv8i32(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i32> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8i16_nxv8i32:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m, <vscale x 8 x i32> %passthru)
   ret <vscale x 8 x i32> %v
 }
 
 define <vscale x 8 x i32> @mgather_baseidx_zext_nxv8i16_nxv8i32(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i32> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i16_nxv8i32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, mu
-; RV32-NEXT:    vzext.vf2 v16, v8
-; RV32-NEXT:    vsll.vi v8, v16, 2
-; RV32-NEXT:    vluxei32.v v12, (a0), v8, v0.t
-; RV32-NEXT:    vmv.v.v v8, v12
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i16_nxv8i32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v8
-; RV64-NEXT:    vsll.vi v16, v16, 2
-; RV64-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
-; RV64-NEXT:    vluxei64.v v12, (a0), v16, v0.t
-; RV64-NEXT:    vmv.v.v v8, v12
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i16_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vzext.vf2 v16, v8
+; CHECK-NEXT:    vsll.vi v8, v16, 2
+; CHECK-NEXT:    vluxei32.v v12, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v12
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x i32> @llvm.masked.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m, <vscale x 8 x i32> %passthru)
   ret <vscale x 8 x i32> %v
 }
 
 define <vscale x 8 x i32> @mgather_baseidx_nxv8i32(ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i32> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8i32:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x i64> @llvm.masked.gather.nxv8i64.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m, <vscale x 8 x i64> %passthru)
   ret <vscale x 8 x i64> %v
 }
 
 define <vscale x 8 x i64> @mgather_baseidx_zext_nxv8i8_nxv8i64(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i64> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vsll.vi v8, v12, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
-; RV32-NEXT:    vluxei32.v v16, (a0), v8, v0.t
-; RV32-NEXT:    vmv.v.v v8, v16
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, mu
-; RV64-NEXT:    vzext.vf8 v24, v8
-; RV64-NEXT:    vsll.vi v8, v24, 3
-; RV64-NEXT:    vluxei64.v v16, (a0), v8, v0.t
-; RV64-NEXT:    vmv.v.v v8, v16
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i8_nxv8i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v10, v8
+; CHECK-NEXT:    vsll.vi v8, v10, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vluxei16.v v16, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v16
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x i64> @llvm.masked.gather.nxv8i64.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m, <vscale x 8 x i64> %passthru)
   ret <vscale x 8 x i64> %v
 }
 
 define <vscale x 8 x i64> @mgather_baseidx_nxv8i16_nxv8i64(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i64> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8i16_nxv8i64:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x i64> @llvm.masked.gather.nxv8i64.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m, <vscale x 8 x i64> %passthru)
   ret <vscale x 8 x i64> %v
 }
 
 define <vscale x 8 x i64> @mgather_baseidx_zext_nxv8i16_nxv8i64(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i64> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i16_nxv8i64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf2 v12, v8
-; RV32-NEXT:    vsll.vi v8, v12, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
-; RV32-NEXT:    vluxei32.v v16, (a0), v8, v0.t
-; RV32-NEXT:    vmv.v.v v8, v16
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i16_nxv8i64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, mu
-; RV64-NEXT:    vzext.vf4 v24, v8
-; RV64-NEXT:    vsll.vi v8, v24, 3
-; RV64-NEXT:    vluxei64.v v16, (a0), v8, v0.t
-; RV64-NEXT:    vmv.v.v v8, v16
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i16_nxv8i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vzext.vf2 v12, v8
+; CHECK-NEXT:    vsll.vi v8, v12, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vluxei32.v v16, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v16
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x i64> @llvm.masked.gather.nxv8i64.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m, <vscale x 8 x i64> %passthru)
   ret <vscale x 8 x i64> %v
 }
 
 define <vscale x 8 x i64> @mgather_baseidx_nxv8i32_nxv8i64(ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x i64> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8i32_nxv8i64:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds half, ptr %base, <vscale x 8 x i16> %eidxs
   %v = call <vscale x 8 x half> @llvm.masked.gather.nxv8f16.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 2, <vscale x 8 x i1> %m, <vscale x 8 x half> %passthru)
   ret <vscale x 8 x half> %v
 }
 
 define <vscale x 8 x half> @mgather_baseidx_zext_nxv8i8_nxv8f16(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x half> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f16:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vadd.vv v12, v12, v12
-; RV32-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
-; RV32-NEXT:    vluxei32.v v10, (a0), v12, v0.t
-; RV32-NEXT:    vmv.v.v v8, v10
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f16:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vadd.vv v16, v16, v16
-; RV64-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
-; RV64-NEXT:    vluxei64.v v10, (a0), v16, v0.t
-; RV64-NEXT:    vmv.v.v v8, v10
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vzext.vf2 v12, v8
+; CHECK-NEXT:    vsll.vi v8, v12, 1
+; CHECK-NEXT:    vluxei16.v v10, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds half, ptr %base, <vscale x 8 x i16> %eidxs
   %v = call <vscale x 8 x half> @llvm.masked.gather.nxv8f16.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 2, <vscale x 8 x i1> %m, <vscale x 8 x half> %passthru)
   ret <vscale x 8 x half> %v
 }
 
 define <vscale x 8 x half> @mgather_baseidx_nxv8f16(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x half> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8f16:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x float> @llvm.masked.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m, <vscale x 8 x float> %passthru)
   ret <vscale x 8 x float> %v
 }
 
 define <vscale x 8 x float> @mgather_baseidx_zext_nxv8i8_nxv8f32(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x float> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, mu
-; RV32-NEXT:    vzext.vf4 v16, v8
-; RV32-NEXT:    vsll.vi v8, v16, 2
-; RV32-NEXT:    vluxei32.v v12, (a0), v8, v0.t
-; RV32-NEXT:    vmv.v.v v8, v12
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vsll.vi v16, v16, 2
-; RV64-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
-; RV64-NEXT:    vluxei64.v v12, (a0), v16, v0.t
-; RV64-NEXT:    vmv.v.v v8, v12
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v10, v8
+; CHECK-NEXT:    vsll.vi v8, v10, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vluxei16.v v12, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v12
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x float> @llvm.masked.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m, <vscale x 8 x float> %passthru)
   ret <vscale x 8 x float> %v
 }
 
 define <vscale x 8 x float> @mgather_baseidx_nxv8i16_nxv8f32(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x float> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8i16_nxv8f32:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x float> @llvm.masked.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m, <vscale x 8 x float> %passthru)
   ret <vscale x 8 x float> %v
 }
 
 define <vscale x 8 x float> @mgather_baseidx_zext_nxv8i16_nxv8f32(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x float> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i16_nxv8f32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, mu
-; RV32-NEXT:    vzext.vf2 v16, v8
-; RV32-NEXT:    vsll.vi v8, v16, 2
-; RV32-NEXT:    vluxei32.v v12, (a0), v8, v0.t
-; RV32-NEXT:    vmv.v.v v8, v12
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i16_nxv8f32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v8
-; RV64-NEXT:    vsll.vi v16, v16, 2
-; RV64-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
-; RV64-NEXT:    vluxei64.v v12, (a0), v16, v0.t
-; RV64-NEXT:    vmv.v.v v8, v12
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i16_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vzext.vf2 v16, v8
+; CHECK-NEXT:    vsll.vi v8, v16, 2
+; CHECK-NEXT:    vluxei32.v v12, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v12
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x float> @llvm.masked.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m, <vscale x 8 x float> %passthru)
   ret <vscale x 8 x float> %v
 }
 
 define <vscale x 8 x float> @mgather_baseidx_nxv8f32(ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x float> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8f32:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x double> @llvm.masked.gather.nxv8f64.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m, <vscale x 8 x double> %passthru)
   ret <vscale x 8 x double> %v
 }
 
 define <vscale x 8 x double> @mgather_baseidx_zext_nxv8i8_nxv8f64(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x double> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vsll.vi v8, v12, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
-; RV32-NEXT:    vluxei32.v v16, (a0), v8, v0.t
-; RV32-NEXT:    vmv.v.v v8, v16
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, mu
-; RV64-NEXT:    vzext.vf8 v24, v8
-; RV64-NEXT:    vsll.vi v8, v24, 3
-; RV64-NEXT:    vluxei64.v v16, (a0), v8, v0.t
-; RV64-NEXT:    vmv.v.v v8, v16
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i8_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v10, v8
+; CHECK-NEXT:    vsll.vi v8, v10, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vluxei16.v v16, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v16
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x double> @llvm.masked.gather.nxv8f64.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m, <vscale x 8 x double> %passthru)
   ret <vscale x 8 x double> %v
 }
 
 define <vscale x 8 x double> @mgather_baseidx_nxv8i16_nxv8f64(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x double> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8i16_nxv8f64:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x double> @llvm.masked.gather.nxv8f64.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m, <vscale x 8 x double> %passthru)
   ret <vscale x 8 x double> %v
 }
 
 define <vscale x 8 x double> @mgather_baseidx_zext_nxv8i16_nxv8f64(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x double> %passthru) {
-; RV32-LABEL: mgather_baseidx_zext_nxv8i16_nxv8f64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf2 v12, v8
-; RV32-NEXT:    vsll.vi v8, v12, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
-; RV32-NEXT:    vluxei32.v v16, (a0), v8, v0.t
-; RV32-NEXT:    vmv.v.v v8, v16
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mgather_baseidx_zext_nxv8i16_nxv8f64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, mu
-; RV64-NEXT:    vzext.vf4 v24, v8
-; RV64-NEXT:    vsll.vi v8, v24, 3
-; RV64-NEXT:    vluxei64.v v16, (a0), v8, v0.t
-; RV64-NEXT:    vmv.v.v v8, v16
-; RV64-NEXT:    ret
+; CHECK-LABEL: mgather_baseidx_zext_nxv8i16_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vzext.vf2 v12, v8
+; CHECK-NEXT:    vsll.vi v8, v12, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vluxei32.v v16, (a0), v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v16
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x double> @llvm.masked.gather.nxv8f64.nxv8p0(<vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m, <vscale x 8 x double> %passthru)
   ret <vscale x 8 x double> %v
 }
 
 define <vscale x 8 x double> @mgather_baseidx_nxv8i32_nxv8f64(ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, <vscale x 8 x double> %passthru) {
 ; RV32-LABEL: mgather_baseidx_nxv8i32_nxv8f64:

llvm/test/CodeGen/RISCV/rvv/mscatter-sdnode.ll

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds i16, ptr %base, <vscale x 8 x i16> %eidxs
   call void @llvm.masked.scatter.nxv8i16.nxv8p0(<vscale x 8 x i16> %val, <vscale x 8 x ptr> %ptrs, i32 2, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i8_nxv8i16(<vscale x 8 x i16> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i16:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v10
-; RV32-NEXT:    vadd.vv v12, v12, v12
-; RV32-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i16:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v10
-; RV64-NEXT:    vadd.vv v16, v16, v16
-; RV64-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v12, v10
+; CHECK-NEXT:    vsll.vi v10, v12, 1
+; CHECK-NEXT:    vsoxei16.v v8, (a0), v10, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds i16, ptr %base, <vscale x 8 x i16> %eidxs
   call void @llvm.masked.scatter.nxv8i16.nxv8p0(<vscale x 8 x i16> %val, <vscale x 8 x ptr> %ptrs, i32 2, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8i16(<vscale x 8 x i16> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8i16:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.masked.scatter.nxv8i32.nxv8p0(<vscale x 8 x i32> %val, <vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i8_nxv8i32(<vscale x 8 x i32> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v16, v12
-; RV32-NEXT:    vsll.vi v12, v16, 2
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v12
-; RV64-NEXT:    vsll.vi v16, v16, 2
-; RV64-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v14, v12
+; CHECK-NEXT:    vsll.vi v12, v14, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vsoxei16.v v8, (a0), v12, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.masked.scatter.nxv8i32.nxv8p0(<vscale x 8 x i32> %val, <vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8i16_nxv8i32(<vscale x 8 x i32> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8i16_nxv8i32:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.masked.scatter.nxv8i32.nxv8p0(<vscale x 8 x i32> %val, <vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i16_nxv8i32(<vscale x 8 x i32> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8i32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf2 v16, v12
-; RV32-NEXT:    vsll.vi v12, v16, 2
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8i32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v12
-; RV64-NEXT:    vsll.vi v16, v16, 2
-; RV64-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vzext.vf2 v16, v12
+; CHECK-NEXT:    vsll.vi v12, v16, 2
+; CHECK-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.masked.scatter.nxv8i32.nxv8p0(<vscale x 8 x i32> %val, <vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8i32(<vscale x 8 x i32> %val, ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8i32:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.masked.scatter.nxv8i64.nxv8p0(<vscale x 8 x i64> %val, <vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i8_nxv8i64(<vscale x 8 x i64> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v20, v16
-; RV32-NEXT:    vsll.vi v16, v20, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v18, v16
+; CHECK-NEXT:    vsll.vi v16, v18, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vsoxei16.v v8, (a0), v16, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.masked.scatter.nxv8i64.nxv8p0(<vscale x 8 x i64> %val, <vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8i16_nxv8i64(<vscale x 8 x i64> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8i16_nxv8i64:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.masked.scatter.nxv8i64.nxv8p0(<vscale x 8 x i64> %val, <vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i16_nxv8i64(<vscale x 8 x i64> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8i64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf2 v20, v16
-; RV32-NEXT:    vsll.vi v16, v20, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8i64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vzext.vf2 v20, v16
+; CHECK-NEXT:    vsll.vi v16, v20, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.masked.scatter.nxv8i64.nxv8p0(<vscale x 8 x i64> %val, <vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8i32_nxv8i64(<vscale x 8 x i64> %val, ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8i32_nxv8i64:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds half, ptr %base, <vscale x 8 x i16> %eidxs
   call void @llvm.masked.scatter.nxv8f16.nxv8p0(<vscale x 8 x half> %val, <vscale x 8 x ptr> %ptrs, i32 2, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i8_nxv8f16(<vscale x 8 x half> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f16:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v10
-; RV32-NEXT:    vadd.vv v12, v12, v12
-; RV32-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f16:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v10
-; RV64-NEXT:    vadd.vv v16, v16, v16
-; RV64-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v12, v10
+; CHECK-NEXT:    vsll.vi v10, v12, 1
+; CHECK-NEXT:    vsoxei16.v v8, (a0), v10, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds half, ptr %base, <vscale x 8 x i16> %eidxs
   call void @llvm.masked.scatter.nxv8f16.nxv8p0(<vscale x 8 x half> %val, <vscale x 8 x ptr> %ptrs, i32 2, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8f16(<vscale x 8 x half> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8f16:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.masked.scatter.nxv8f32.nxv8p0(<vscale x 8 x float> %val, <vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i8_nxv8f32(<vscale x 8 x float> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v16, v12
-; RV32-NEXT:    vsll.vi v12, v16, 2
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v12
-; RV64-NEXT:    vsll.vi v16, v16, 2
-; RV64-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v14, v12
+; CHECK-NEXT:    vsll.vi v12, v14, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vsoxei16.v v8, (a0), v12, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.masked.scatter.nxv8f32.nxv8p0(<vscale x 8 x float> %val, <vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8i16_nxv8f32(<vscale x 8 x float> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8i16_nxv8f32:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.masked.scatter.nxv8f32.nxv8p0(<vscale x 8 x float> %val, <vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i16_nxv8f32(<vscale x 8 x float> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8f32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf2 v16, v12
-; RV32-NEXT:    vsll.vi v12, v16, 2
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8f32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v12
-; RV64-NEXT:    vsll.vi v16, v16, 2
-; RV64-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vzext.vf2 v16, v12
+; CHECK-NEXT:    vsll.vi v12, v16, 2
+; CHECK-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.masked.scatter.nxv8f32.nxv8p0(<vscale x 8 x float> %val, <vscale x 8 x ptr> %ptrs, i32 4, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8f32(<vscale x 8 x float> %val, ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8f32:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.masked.scatter.nxv8f64.nxv8p0(<vscale x 8 x double> %val, <vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i8_nxv8f64(<vscale x 8 x double> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v20, v16
-; RV32-NEXT:    vsll.vi v16, v20, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i8_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vzext.vf2 v18, v16
+; CHECK-NEXT:    vsll.vi v16, v18, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vsoxei16.v v8, (a0), v16, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.masked.scatter.nxv8f64.nxv8p0(<vscale x 8 x double> %val, <vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8i16_nxv8f64(<vscale x 8 x double> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8i16_nxv8f64:

 ; RV64-NEXT:    ret
   %eidxs = sext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.masked.scatter.nxv8f64.nxv8p0(<vscale x 8 x double> %val, <vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_zext_nxv8i16_nxv8f64(<vscale x 8 x double> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m) {
-; RV32-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8f64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf2 v20, v16
-; RV32-NEXT:    vsll.vi v16, v20, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8f64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    ret
+; CHECK-LABEL: mscatter_baseidx_zext_nxv8i16_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vzext.vf2 v20, v16
+; CHECK-NEXT:    vsll.vi v16, v20, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
+; CHECK-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.masked.scatter.nxv8f64.nxv8p0(<vscale x 8 x double> %val, <vscale x 8 x ptr> %ptrs, i32 8, <vscale x 8 x i1> %m)
   ret void
 }
 
 define void @mscatter_baseidx_nxv8i32_nxv8f64(<vscale x 8 x double> %val, ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m) {
 ; RV32-LABEL: mscatter_baseidx_nxv8i32_nxv8f64:

llvm/test/CodeGen/RISCV/rvv/narrow-shift-extend.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+v -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 4 x i32> @llvm.riscv.vloxei.nxv4i32.nxv4i64(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i64>,
+  i64);
+
+define <vscale x 4 x i32> @test_vloxei(<vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, i64 %vl) {
+; CHECK-LABEL: test_vloxei:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v9, v8
+; CHECK-NEXT:    vsll.vi v10, v9, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vloxei16.v v8, (a0), v10
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  %res = call <vscale x 4 x i32> @llvm.riscv.vloxei.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> undef,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    i64 %vl)
+  ret <vscale x 4 x i32> %res
+}
+
+define <vscale x 4 x i32> @test_vloxei2(<vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, i64 %vl) {
+; CHECK-LABEL: test_vloxei2:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vzext.vf4 v10, v8
+; CHECK-NEXT:    vsll.vi v8, v10, 14
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vloxei32.v v8, (a0), v8
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 14, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  %res = call <vscale x 4 x i32> @llvm.riscv.vloxei.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> undef,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    i64 %vl)
+  ret <vscale x 4 x i32> %res
+}
+
+define <vscale x 4 x i32> @test_vloxei3(<vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, i64 %vl) {
+; CHECK-LABEL: test_vloxei3:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vzext.vf8 v12, v8
+; CHECK-NEXT:    vsll.vi v12, v12, 26
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vloxei64.v v8, (a0), v12
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 26, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  %res = call <vscale x 4 x i32> @llvm.riscv.vloxei.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> undef,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    i64 %vl)
+  ret <vscale x 4 x i32> %res
+}
+
+; Test use vp.zext to extend.
+declare <vscale x 4 x i64> @llvm.vp.zext.nxvi64.nxv1i8(<vscale x 4 x i8>, <vscale x 4 x i1>, i32)
+define <vscale x 4 x i32> @test_vloxei4(<vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, <vscale x 4 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: test_vloxei4:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli zero, a1, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v9, v8, v0.t
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vsll.vi v10, v9, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vloxei16.v v8, (a0), v10
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = call <vscale x 4 x i64> @llvm.vp.zext.nxvi64.nxv1i8(<vscale x 4 x i8> %offset, <vscale x 4 x i1> %m, i32 %vl)
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  %vl.i64 = zext i32 %vl to i64
+  %res = call <vscale x 4 x i32> @llvm.riscv.vloxei.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> undef,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    i64 %vl.i64)
+  ret <vscale x 4 x i32> %res
+}
+
+; Test orignal extnened type is enough narrow.
+declare <vscale x 4 x i32> @llvm.riscv.vloxei.nxv4i32.nxv4i16(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i16>,
+  i64);
+define <vscale x 4 x i32> @test_vloxei5(<vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, i64 %vl) {
+; CHECK-LABEL: test_vloxei5:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v9, v8
+; CHECK-NEXT:    vsll.vi v10, v9, 12
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vloxei16.v v8, (a0), v10
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i16>
+  %shamt = insertelement <vscale x 4 x i16> undef, i16 12, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i16> %shamt, <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i16> %offset.ext, %shamt.vec
+  %res = call <vscale x 4 x i32> @llvm.riscv.vloxei.nxv4i32.nxv4i16(
+    <vscale x 4 x i32> undef,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i16> %shl,
+    i64 %vl)
+  ret <vscale x 4 x i32> %res
+}
+
+declare <vscale x 4 x i32> @llvm.riscv.vloxei.mask.nxv4i32.nxv4i64(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i64>,
+  <vscale x 4 x i1>,
+  i64,
+  i64);
+
+define <vscale x 4 x i32> @test_vloxei_mask(<vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, <vscale x 4 x i1> %m, i64 %vl) {
+; CHECK-LABEL: test_vloxei_mask:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v9, v8
+; CHECK-NEXT:    vsll.vi v10, v9, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vloxei16.v v8, (a0), v10, v0.t
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  %res = call <vscale x 4 x i32> @llvm.riscv.vloxei.mask.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> undef,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    <vscale x 4 x i1> %m,
+    i64 %vl, i64 1)
+  ret <vscale x 4 x i32> %res
+}
+
+declare <vscale x 4 x i32> @llvm.riscv.vluxei.nxv4i32.nxv4i64(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i64>,
+  i64);
+
+define <vscale x 4 x i32> @test_vluxei(<vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, i64 %vl) {
+; CHECK-LABEL: test_vluxei:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v9, v8
+; CHECK-NEXT:    vsll.vi v10, v9, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vluxei16.v v8, (a0), v10
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  %res = call <vscale x 4 x i32> @llvm.riscv.vluxei.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> undef,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    i64 %vl)
+  ret <vscale x 4 x i32> %res
+}
+
+declare <vscale x 4 x i32> @llvm.riscv.vluxei.mask.nxv4i32.nxv4i64(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i64>,
+  <vscale x 4 x i1>,
+  i64,
+  i64);
+
+define <vscale x 4 x i32> @test_vluxei_mask(<vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, <vscale x 4 x i1> %m, i64 %vl) {
+; CHECK-LABEL: test_vluxei_mask:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v9, v8
+; CHECK-NEXT:    vsll.vi v10, v9, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vluxei16.v v8, (a0), v10, v0.t
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  %res = call <vscale x 4 x i32> @llvm.riscv.vluxei.mask.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> undef,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    <vscale x 4 x i1> %m,
+    i64 %vl, i64 1)
+  ret <vscale x 4 x i32> %res
+}
+
+declare void @llvm.riscv.vsoxei.nxv4i32.nxv4i64(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i64>,
+  i64);
+
+define void @test_vsoxei(<vscale x 4 x i32> %val, <vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, i64 %vl) {
+; CHECK-LABEL: test_vsoxei:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v11, v10
+; CHECK-NEXT:    vsll.vi v10, v11, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vsoxei16.v v8, (a0), v10
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  call void @llvm.riscv.vsoxei.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> %val,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    i64 %vl)
+  ret void
+}
+
+declare void @llvm.riscv.vsoxei.mask.nxv4i32.nxv4i64(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i64>,
+  <vscale x 4 x i1>,
+  i64);
+
+define void @test_vsoxei_mask(<vscale x 4 x i32> %val, <vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, <vscale x 4 x i1> %m, i64 %vl) {
+; CHECK-LABEL: test_vsoxei_mask:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v11, v10
+; CHECK-NEXT:    vsll.vi v10, v11, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vsoxei16.v v8, (a0), v10, v0.t
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  call void @llvm.riscv.vsoxei.mask.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> %val,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    <vscale x 4 x i1> %m,
+    i64 %vl)
+  ret void
+}
+
+declare void @llvm.riscv.vsuxei.nxv4i32.nxv4i64(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i64>,
+  i64);
+
+define void @test_vsuxei(<vscale x 4 x i32> %val, <vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, i64 %vl) {
+; CHECK-LABEL: test_vsuxei:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v11, v10
+; CHECK-NEXT:    vsll.vi v10, v11, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vsuxei16.v v8, (a0), v10
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  call void @llvm.riscv.vsuxei.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> %val,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    i64 %vl)
+  ret void
+}
+
+declare void @llvm.riscv.vsuxei.mask.nxv4i32.nxv4i64(
+  <vscale x 4 x i32>,
+  <vscale x 4 x i32>*,
+  <vscale x 4 x i64>,
+  <vscale x 4 x i1>,
+  i64);
+
+define void @test_vsuxei_mask(<vscale x 4 x i32> %val, <vscale x 4 x i32>* %ptr, <vscale x 4 x i8> %offset, <vscale x 4 x i1> %m, i64 %vl) {
+; CHECK-LABEL: test_vsuxei_mask:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vzext.vf2 v11, v10
+; CHECK-NEXT:    vsll.vi v10, v11, 4
+; CHECK-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; CHECK-NEXT:    vsuxei16.v v8, (a0), v10, v0.t
+; CHECK-NEXT:    ret
+entry:
+  %offset.ext = zext <vscale x 4 x i8> %offset to <vscale x 4 x i64>
+  %shamt = insertelement <vscale x 4 x i64> undef, i64 4, i32 0
+  %shamt.vec = shufflevector <vscale x 4 x i64> %shamt, <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+  %shl = shl <vscale x 4 x i64> %offset.ext, %shamt.vec
+  call void @llvm.riscv.vsuxei.mask.nxv4i32.nxv4i64(
+    <vscale x 4 x i32> %val,
+    <vscale x 4 x i32>* %ptr,
+    <vscale x 4 x i64> %shl,
+    <vscale x 4 x i1> %m,
+    i64 %vl)
+  ret void
+}

llvm/test/CodeGen/RISCV/rvv/vpgather-sdnode.ll

; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds i16, ptr %base, <vscale x 8 x i16> %eidxs
   %v = call <vscale x 8 x i16> @llvm.vp.gather.nxv8i16.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x i16> %v
 }
 
 define <vscale x 8 x i16> @vpgather_baseidx_zext_nxv8i8_nxv8i16(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8i16:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vadd.vv v12, v12, v12
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v10, v8
+; RV32-NEXT:    vsll.vi v8, v10, 1
 ; RV32-NEXT:    vsetvli zero, a1, e16, m2, ta, ma
-; RV32-NEXT:    vluxei32.v v8, (a0), v12, v0.t
+; RV32-NEXT:    vluxei16.v v8, (a0), v8, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8i16:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vadd.vv v16, v16, v16
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v10, v8
+; RV64-NEXT:    vsll.vi v8, v10, 1
 ; RV64-NEXT:    vsetvli zero, a1, e16, m2, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vluxei16.v v8, (a0), v8, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds i16, ptr %base, <vscale x 8 x i16> %eidxs
   %v = call <vscale x 8 x i16> @llvm.vp.gather.nxv8i16.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x i16> %v
 }
 
 define <vscale x 8 x i16> @vpgather_baseidx_nxv8i16(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x i32> @llvm.vp.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x i32> %v
 }
 
 define <vscale x 8 x i32> @vpgather_baseidx_zext_nxv8i8_nxv8i32(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8i32:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vsll.vi v8, v12, 2
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v10, v8
+; RV32-NEXT:    vsll.vi v12, v10, 2
 ; RV32-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV32-NEXT:    vluxei32.v v8, (a0), v8, v0.t
+; RV32-NEXT:    vluxei16.v v8, (a0), v12, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8i32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vsll.vi v16, v16, 2
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v10, v8
+; RV64-NEXT:    vsll.vi v12, v10, 2
 ; RV64-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vluxei16.v v8, (a0), v12, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x i32> @llvm.vp.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x i32> %v
 }
 
 define <vscale x 8 x i32> @vpgather_baseidx_nxv8i16_nxv8i32(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v12, v8
 ; RV32-NEXT:    vsll.vi v8, v12, 2
 ; RV32-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
 ; RV32-NEXT:    vluxei32.v v8, (a0), v8, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i16_nxv8i32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v8
-; RV64-NEXT:    vsll.vi v16, v16, 2
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v12, v8
+; RV64-NEXT:    vsll.vi v8, v12, 2
 ; RV64-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vluxei32.v v8, (a0), v8, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x i32> @llvm.vp.gather.nxv8i32.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x i32> %v
 }
 
 define <vscale x 8 x i32> @vpgather_baseidx_nxv8i32(ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x i64> @llvm.vp.gather.nxv8i64.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x i64> %v
 }
 
 define <vscale x 8 x i64> @vpgather_baseidx_zext_nxv8i8_nxv8i64(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8i64:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vsll.vi v16, v12, 3
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v10, v8
+; RV32-NEXT:    vsll.vi v16, v10, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV32-NEXT:    vluxei32.v v8, (a0), v16, v0.t
+; RV32-NEXT:    vluxei16.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8i64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vsll.vi v8, v16, 3
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v10, v8
+; RV64-NEXT:    vsll.vi v16, v10, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v8, v0.t
+; RV64-NEXT:    vluxei16.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x i64> @llvm.vp.gather.nxv8i64.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x i64> %v
 }
 
 define <vscale x 8 x i64> @vpgather_baseidx_nxv8i16_nxv8i64(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v12, v8
 ; RV32-NEXT:    vsll.vi v16, v12, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
 ; RV32-NEXT:    vluxei32.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i16_nxv8i64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v8
-; RV64-NEXT:    vsll.vi v8, v16, 3
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v12, v8
+; RV64-NEXT:    vsll.vi v16, v12, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v8, v0.t
+; RV64-NEXT:    vluxei32.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x i64> @llvm.vp.gather.nxv8i64.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x i64> %v
 }
 
 define <vscale x 8 x i64> @vpgather_baseidx_nxv8i32_nxv8i64(ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds half, ptr %base, <vscale x 8 x i16> %eidxs
   %v = call <vscale x 8 x half> @llvm.vp.gather.nxv8f16.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x half> %v
 }
 
 define <vscale x 8 x half> @vpgather_baseidx_zext_nxv8i8_nxv8f16(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8f16:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vadd.vv v12, v12, v12
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v10, v8
+; RV32-NEXT:    vsll.vi v8, v10, 1
 ; RV32-NEXT:    vsetvli zero, a1, e16, m2, ta, ma
-; RV32-NEXT:    vluxei32.v v8, (a0), v12, v0.t
+; RV32-NEXT:    vluxei16.v v8, (a0), v8, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8f16:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vadd.vv v16, v16, v16
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v10, v8
+; RV64-NEXT:    vsll.vi v8, v10, 1
 ; RV64-NEXT:    vsetvli zero, a1, e16, m2, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vluxei16.v v8, (a0), v8, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds half, ptr %base, <vscale x 8 x i16> %eidxs
   %v = call <vscale x 8 x half> @llvm.vp.gather.nxv8f16.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x half> %v
 }
 
 define <vscale x 8 x half> @vpgather_baseidx_nxv8f16(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x float> @llvm.vp.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x float> %v
 }
 
 define <vscale x 8 x float> @vpgather_baseidx_zext_nxv8i8_nxv8f32(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8f32:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vsll.vi v8, v12, 2
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v10, v8
+; RV32-NEXT:    vsll.vi v12, v10, 2
 ; RV32-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV32-NEXT:    vluxei32.v v8, (a0), v8, v0.t
+; RV32-NEXT:    vluxei16.v v8, (a0), v12, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8f32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vsll.vi v16, v16, 2
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v10, v8
+; RV64-NEXT:    vsll.vi v12, v10, 2
 ; RV64-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vluxei16.v v8, (a0), v12, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x float> @llvm.vp.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x float> %v
 }
 
 define <vscale x 8 x float> @vpgather_baseidx_nxv8i16_nxv8f32(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v12, v8
 ; RV32-NEXT:    vsll.vi v8, v12, 2
 ; RV32-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
 ; RV32-NEXT:    vluxei32.v v8, (a0), v8, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i16_nxv8f32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v8
-; RV64-NEXT:    vsll.vi v16, v16, 2
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v12, v8
+; RV64-NEXT:    vsll.vi v8, v12, 2
 ; RV64-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vluxei32.v v8, (a0), v8, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   %v = call <vscale x 8 x float> @llvm.vp.gather.nxv8f32.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x float> %v
 }
 
 define <vscale x 8 x float> @vpgather_baseidx_nxv8f32(ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 6 x i64> %eidxs
   %v = call <vscale x 6 x double> @llvm.vp.gather.nxv6f64.nxv6p0(<vscale x 6 x ptr> %ptrs, <vscale x 6 x i1> %m, i32 %evl)
   ret <vscale x 6 x double> %v
 }
 
 define <vscale x 6 x double> @vpgather_baseidx_zext_nxv6i8_nxv6f64(ptr %base, <vscale x 6 x i8> %idxs, <vscale x 6 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpgather_baseidx_zext_nxv6i8_nxv6f64:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vsll.vi v16, v12, 3
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v10, v8
+; RV32-NEXT:    vsll.vi v16, v10, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV32-NEXT:    vluxei32.v v8, (a0), v16, v0.t
+; RV32-NEXT:    vluxei16.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv6i8_nxv6f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vsll.vi v8, v16, 3
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v10, v8
+; RV64-NEXT:    vsll.vi v16, v10, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v8, v0.t
+; RV64-NEXT:    vluxei16.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 6 x i8> %idxs to <vscale x 6 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 6 x i64> %eidxs
   %v = call <vscale x 6 x double> @llvm.vp.gather.nxv6f64.nxv6p0(<vscale x 6 x ptr> %ptrs, <vscale x 6 x i1> %m, i32 %evl)
   ret <vscale x 6 x double> %v
 }
 
 define <vscale x 6 x double> @vpgather_baseidx_nxv6i16_nxv6f64(ptr %base, <vscale x 6 x i16> %idxs, <vscale x 6 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v12, v8
 ; RV32-NEXT:    vsll.vi v16, v12, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
 ; RV32-NEXT:    vluxei32.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv6i16_nxv6f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v8
-; RV64-NEXT:    vsll.vi v8, v16, 3
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v12, v8
+; RV64-NEXT:    vsll.vi v16, v12, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v8, v0.t
+; RV64-NEXT:    vluxei32.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 6 x i16> %idxs to <vscale x 6 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 6 x i64> %eidxs
   %v = call <vscale x 6 x double> @llvm.vp.gather.nxv6f64.nxv6p0(<vscale x 6 x ptr> %ptrs, <vscale x 6 x i1> %m, i32 %evl)
   ret <vscale x 6 x double> %v
 }
 
 define <vscale x 6 x double> @vpgather_baseidx_nxv6i32_nxv6f64(ptr %base, <vscale x 6 x i32> %idxs, <vscale x 6 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x double> @llvm.vp.gather.nxv8f64.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x double> %v
 }
 
 define <vscale x 8 x double> @vpgather_baseidx_zext_nxv8i8_nxv8f64(ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8f64:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v8
-; RV32-NEXT:    vsll.vi v16, v12, 3
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v10, v8
+; RV32-NEXT:    vsll.vi v16, v10, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV32-NEXT:    vluxei32.v v8, (a0), v16, v0.t
+; RV32-NEXT:    vluxei16.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i8_nxv8f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v8
-; RV64-NEXT:    vsll.vi v8, v16, 3
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v10, v8
+; RV64-NEXT:    vsll.vi v16, v10, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v8, v0.t
+; RV64-NEXT:    vluxei16.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x double> @llvm.vp.gather.nxv8f64.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x double> %v
 }
 
 define <vscale x 8 x double> @vpgather_baseidx_nxv8i16_nxv8f64(ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v12, v8
 ; RV32-NEXT:    vsll.vi v16, v12, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
 ; RV32-NEXT:    vluxei32.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv8i16_nxv8f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v8
-; RV64-NEXT:    vsll.vi v8, v16, 3
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v12, v8
+; RV64-NEXT:    vsll.vi v16, v12, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vluxei64.v v8, (a0), v8, v0.t
+; RV64-NEXT:    vluxei32.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   %v = call <vscale x 8 x double> @llvm.vp.gather.nxv8f64.nxv8p0(<vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret <vscale x 8 x double> %v
 }
 
 define <vscale x 8 x double> @vpgather_baseidx_nxv8i32_nxv8f64(ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
 ; RV32-NEXT:    vmv1r.v v0, v12
 ; RV32-NEXT:    vluxei32.v v8, (a0), v24, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpgather_baseidx_zext_nxv16i16_nxv16f64:
 ; RV64:       # %bb.0:
 ; RV64-NEXT:    vmv1r.v v12, v0
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v10
-; RV64-NEXT:    vzext.vf4 v24, v8
-; RV64-NEXT:    vsll.vi v24, v24, 3
-; RV64-NEXT:    vsll.vi v16, v16, 3
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v16, v8
+; RV64-NEXT:    vsll.vi v24, v16, 3
+; RV64-NEXT:    vzext.vf2 v16, v10
+; RV64-NEXT:    vsll.vi v8, v16, 3
 ; RV64-NEXT:    csrr a2, vlenb
 ; RV64-NEXT:    sub a3, a1, a2
 ; RV64-NEXT:    sltu a4, a1, a3
 ; RV64-NEXT:    addi a4, a4, -1
 ; RV64-NEXT:    and a3, a4, a3
 ; RV64-NEXT:    srli a4, a2, 3
 ; RV64-NEXT:    vsetvli a5, zero, e8, mf4, ta, ma
 ; RV64-NEXT:    vslidedown.vx v0, v0, a4
 ; RV64-NEXT:    vsetvli zero, a3, e64, m8, ta, ma
-; RV64-NEXT:    vluxei64.v v16, (a0), v16, v0.t
+; RV64-NEXT:    vluxei32.v v16, (a0), v8, v0.t
 ; RV64-NEXT:    bltu a1, a2, .LBB105_2
 ; RV64-NEXT:  # %bb.1:
 ; RV64-NEXT:    mv a1, a2
 ; RV64-NEXT:  .LBB105_2:
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
 ; RV64-NEXT:    vmv1r.v v0, v12
-; RV64-NEXT:    vluxei64.v v8, (a0), v24, v0.t
+; RV64-NEXT:    vluxei32.v v8, (a0), v24, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 16 x i16> %idxs to <vscale x 16 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 16 x i64> %eidxs
   %v = call <vscale x 16 x double> @llvm.vp.gather.nxv16f64.nxv16p0(<vscale x 16 x ptr> %ptrs, <vscale x 16 x i1> %m, i32 %evl)
   ret <vscale x 16 x double> %v
 }

llvm/test/CodeGen/RISCV/rvv/vpscatter-sdnode.ll

; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds i16, ptr %base, <vscale x 8 x i16> %eidxs
   call void @llvm.vp.scatter.nxv8i16.nxv8p0(<vscale x 8 x i16> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_zext_nxv8i8_nxv8i16(<vscale x 8 x i16> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8i16:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v10
-; RV32-NEXT:    vadd.vv v12, v12, v12
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v12, v10
+; RV32-NEXT:    vsll.vi v10, v12, 1
 ; RV32-NEXT:    vsetvli zero, a1, e16, m2, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
+; RV32-NEXT:    vsoxei16.v v8, (a0), v10, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8i16:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v10
-; RV64-NEXT:    vadd.vv v16, v16, v16
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v12, v10
+; RV64-NEXT:    vsll.vi v10, v12, 1
 ; RV64-NEXT:    vsetvli zero, a1, e16, m2, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei16.v v8, (a0), v10, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds i16, ptr %base, <vscale x 8 x i16> %eidxs
   call void @llvm.vp.scatter.nxv8i16.nxv8p0(<vscale x 8 x i16> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8i16(<vscale x 8 x i16> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.vp.scatter.nxv8i32.nxv8p0(<vscale x 8 x i32> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_zext_nxv8i8_nxv8i32(<vscale x 8 x i32> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8i32:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v16, v12
-; RV32-NEXT:    vsll.vi v12, v16, 2
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v14, v12
+; RV32-NEXT:    vsll.vi v12, v14, 2
 ; RV32-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
+; RV32-NEXT:    vsoxei16.v v8, (a0), v12, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8i32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v12
-; RV64-NEXT:    vsll.vi v16, v16, 2
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v14, v12
+; RV64-NEXT:    vsll.vi v12, v14, 2
 ; RV64-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei16.v v8, (a0), v12, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.vp.scatter.nxv8i32.nxv8p0(<vscale x 8 x i32> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8i16_nxv8i32(<vscale x 8 x i32> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v16, v12
 ; RV32-NEXT:    vsll.vi v12, v16, 2
 ; RV32-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
 ; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i16_nxv8i32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v12
-; RV64-NEXT:    vsll.vi v16, v16, 2
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v16, v12
+; RV64-NEXT:    vsll.vi v12, v16, 2
 ; RV64-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds i32, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.vp.scatter.nxv8i32.nxv8p0(<vscale x 8 x i32> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8i32(<vscale x 8 x i32> %val, ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.vp.scatter.nxv8i64.nxv8p0(<vscale x 8 x i64> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_zext_nxv8i8_nxv8i64(<vscale x 8 x i64> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8i64:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v20, v16
-; RV32-NEXT:    vsll.vi v16, v20, 3
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v18, v16
+; RV32-NEXT:    vsll.vi v16, v18, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
+; RV32-NEXT:    vsoxei16.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8i64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v18, v16
+; RV64-NEXT:    vsll.vi v16, v18, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei16.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.vp.scatter.nxv8i64.nxv8p0(<vscale x 8 x i64> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8i16_nxv8i64(<vscale x 8 x i64> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v20, v16
 ; RV32-NEXT:    vsll.vi v16, v20, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
 ; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i16_nxv8i64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v20, v16
+; RV64-NEXT:    vsll.vi v16, v20, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds i64, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.vp.scatter.nxv8i64.nxv8p0(<vscale x 8 x i64> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8i32_nxv8i64(<vscale x 8 x i64> %val, ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds half, ptr %base, <vscale x 8 x i16> %eidxs
   call void @llvm.vp.scatter.nxv8f16.nxv8p0(<vscale x 8 x half> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_zext_nxv8i8_nxv8f16(<vscale x 8 x half> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8f16:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v12, v10
-; RV32-NEXT:    vadd.vv v12, v12, v12
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v12, v10
+; RV32-NEXT:    vsll.vi v10, v12, 1
 ; RV32-NEXT:    vsetvli zero, a1, e16, m2, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
+; RV32-NEXT:    vsoxei16.v v8, (a0), v10, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8f16:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v10
-; RV64-NEXT:    vadd.vv v16, v16, v16
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v12, v10
+; RV64-NEXT:    vsll.vi v10, v12, 1
 ; RV64-NEXT:    vsetvli zero, a1, e16, m2, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei16.v v8, (a0), v10, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i16>
   %ptrs = getelementptr inbounds half, ptr %base, <vscale x 8 x i16> %eidxs
   call void @llvm.vp.scatter.nxv8f16.nxv8p0(<vscale x 8 x half> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8f16(<vscale x 8 x half> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.vp.scatter.nxv8f32.nxv8p0(<vscale x 8 x float> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_zext_nxv8i8_nxv8f32(<vscale x 8 x float> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8f32:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v16, v12
-; RV32-NEXT:    vsll.vi v12, v16, 2
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v14, v12
+; RV32-NEXT:    vsll.vi v12, v14, 2
 ; RV32-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
+; RV32-NEXT:    vsoxei16.v v8, (a0), v12, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8f32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v16, v12
-; RV64-NEXT:    vsll.vi v16, v16, 2
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v14, v12
+; RV64-NEXT:    vsll.vi v12, v14, 2
 ; RV64-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei16.v v8, (a0), v12, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.vp.scatter.nxv8f32.nxv8p0(<vscale x 8 x float> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8i16_nxv8f32(<vscale x 8 x float> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v16, v12
 ; RV32-NEXT:    vsll.vi v12, v16, 2
 ; RV32-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
 ; RV32-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i16_nxv8f32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v12
-; RV64-NEXT:    vsll.vi v16, v16, 2
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v16, v12
+; RV64-NEXT:    vsll.vi v12, v16, 2
 ; RV64-NEXT:    vsetvli zero, a1, e32, m4, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei32.v v8, (a0), v12, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i32>
   %ptrs = getelementptr inbounds float, ptr %base, <vscale x 8 x i32> %eidxs
   call void @llvm.vp.scatter.nxv8f32.nxv8p0(<vscale x 8 x float> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8f32(<vscale x 8 x float> %val, ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 6 x i64> %eidxs
   call void @llvm.vp.scatter.nxv6f64.nxv6p0(<vscale x 6 x double> %val, <vscale x 6 x ptr> %ptrs, <vscale x 6 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_zext_nxv6i8_nxv6f64(<vscale x 6 x double> %val, ptr %base, <vscale x 6 x i8> %idxs, <vscale x 6 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpscatter_baseidx_zext_nxv6i8_nxv6f64:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v20, v16
-; RV32-NEXT:    vsll.vi v16, v20, 3
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v18, v16
+; RV32-NEXT:    vsll.vi v16, v18, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
+; RV32-NEXT:    vsoxei16.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv6i8_nxv6f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v18, v16
+; RV64-NEXT:    vsll.vi v16, v18, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei16.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 6 x i8> %idxs to <vscale x 6 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 6 x i64> %eidxs
   call void @llvm.vp.scatter.nxv6f64.nxv6p0(<vscale x 6 x double> %val, <vscale x 6 x ptr> %ptrs, <vscale x 6 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv6i16_nxv6f64(<vscale x 6 x double> %val, ptr %base, <vscale x 6 x i16> %idxs, <vscale x 6 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v20, v16
 ; RV32-NEXT:    vsll.vi v16, v20, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
 ; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv6i16_nxv6f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v20, v16
+; RV64-NEXT:    vsll.vi v16, v20, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 6 x i16> %idxs to <vscale x 6 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 6 x i64> %eidxs
   call void @llvm.vp.scatter.nxv6f64.nxv6p0(<vscale x 6 x double> %val, <vscale x 6 x ptr> %ptrs, <vscale x 6 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv6i32_nxv6f64(<vscale x 6 x double> %val, ptr %base, <vscale x 6 x i32> %idxs, <vscale x 6 x i1> %m, i32 zeroext %evl) {
; RV64-NEXT:    ret
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.vp.scatter.nxv8f64.nxv8p0(<vscale x 8 x double> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_zext_nxv8i8_nxv8f64(<vscale x 8 x double> %val, ptr %base, <vscale x 8 x i8> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {
 ; RV32-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8f64:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
-; RV32-NEXT:    vzext.vf4 v20, v16
-; RV32-NEXT:    vsll.vi v16, v20, 3
+; RV32-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV32-NEXT:    vzext.vf2 v18, v16
+; RV32-NEXT:    vsll.vi v16, v18, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
+; RV32-NEXT:    vsoxei16.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i8_nxv8f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf8 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
+; RV64-NEXT:    vsetvli a2, zero, e16, m2, ta, ma
+; RV64-NEXT:    vzext.vf2 v18, v16
+; RV64-NEXT:    vsll.vi v16, v18, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei16.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i8> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.vp.scatter.nxv8f64.nxv8p0(<vscale x 8 x double> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8i16_nxv8f64(<vscale x 8 x double> %val, ptr %base, <vscale x 8 x i16> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vzext.vf2 v20, v16
 ; RV32-NEXT:    vsll.vi v16, v20, 3
 ; RV32-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
 ; RV32-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv8i16_nxv8f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vsetvli a2, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v24, v16
-; RV64-NEXT:    vsll.vi v16, v24, 3
+; RV64-NEXT:    vsetvli a2, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v20, v16
+; RV64-NEXT:    vsll.vi v16, v20, 3
 ; RV64-NEXT:    vsetvli zero, a1, e64, m8, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
+; RV64-NEXT:    vsoxei32.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 8 x i16> %idxs to <vscale x 8 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 8 x i64> %eidxs
   call void @llvm.vp.scatter.nxv8f64.nxv8p0(<vscale x 8 x double> %val, <vscale x 8 x ptr> %ptrs, <vscale x 8 x i1> %m, i32 %evl)
   ret void
 }
 
 define void @vpscatter_baseidx_nxv8i32_nxv8f64(<vscale x 8 x double> %val, ptr %base, <vscale x 8 x i32> %idxs, <vscale x 8 x i1> %m, i32 zeroext %evl) {

 ; RV32-NEXT:    vsetvli a3, zero, e8, mf4, ta, ma
 ; RV32-NEXT:    vslidedown.vx v0, v0, a1
 ; RV32-NEXT:    vsetvli zero, a2, e64, m8, ta, ma
 ; RV32-NEXT:    vsoxei32.v v16, (a0), v28, v0.t
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: vpscatter_baseidx_zext_nxv16i16_nxv16f64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    addi sp, sp, -16
-; RV64-NEXT:    .cfi_def_cfa_offset 16
-; RV64-NEXT:    csrr a3, vlenb
-; RV64-NEXT:    slli a3, a3, 4
-; RV64-NEXT:    sub sp, sp, a3
-; RV64-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 16 * vlenb
-; RV64-NEXT:    vl4re16.v v24, (a1)
+; RV64-NEXT:    vl4re16.v v28, (a1)
+; RV64-NEXT:    vsetvli a1, zero, e32, m4, ta, ma
+; RV64-NEXT:    vzext.vf2 v24, v30
+; RV64-NEXT:    vsll.vi v24, v24, 3
+; RV64-NEXT:    vzext.vf2 v4, v28
 ; RV64-NEXT:    csrr a1, vlenb
-; RV64-NEXT:    slli a1, a1, 3
-; RV64-NEXT:    add a1, sp, a1
-; RV64-NEXT:    addi a1, a1, 16
-; RV64-NEXT:    vs8r.v v16, (a1) # Unknown-size Folded Spill
-; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
-; RV64-NEXT:    vzext.vf4 v16, v26
-; RV64-NEXT:    vsll.vi v16, v16, 3
-; RV64-NEXT:    addi a1, sp, 16
-; RV64-NEXT:    vs8r.v v16, (a1) # Unknown-size Folded Spill
-; RV64-NEXT:    csrr a1, vlenb
-; RV64-NEXT:    vzext.vf4 v16, v24
-; RV64-NEXT:    vsll.vi v24, v16, 3
+; RV64-NEXT:    vsll.vi v28, v4, 3
 ; RV64-NEXT:    mv a3, a2
 ; RV64-NEXT:    bltu a2, a1, .LBB98_2
 ; RV64-NEXT:  # %bb.1:
 ; RV64-NEXT:    mv a3, a1
 ; RV64-NEXT:  .LBB98_2:
 ; RV64-NEXT:    vsetvli zero, a3, e64, m8, ta, ma
-; RV64-NEXT:    vsoxei64.v v8, (a0), v24, v0.t
+; RV64-NEXT:    vsoxei32.v v8, (a0), v28, v0.t
 ; RV64-NEXT:    sub a3, a2, a1
 ; RV64-NEXT:    sltu a2, a2, a3
 ; RV64-NEXT:    addi a2, a2, -1
 ; RV64-NEXT:    and a2, a2, a3
 ; RV64-NEXT:    srli a1, a1, 3
 ; RV64-NEXT:    vsetvli a3, zero, e8, mf4, ta, ma
 ; RV64-NEXT:    vslidedown.vx v0, v0, a1
 ; RV64-NEXT:    vsetvli zero, a2, e64, m8, ta, ma
-; RV64-NEXT:    csrr a1, vlenb
-; RV64-NEXT:    slli a1, a1, 3
-; RV64-NEXT:    add a1, sp, a1
-; RV64-NEXT:    addi a1, a1, 16
-; RV64-NEXT:    vl8r.v v8, (a1) # Unknown-size Folded Reload
-; RV64-NEXT:    addi a1, sp, 16
-; RV64-NEXT:    vl8r.v v16, (a1) # Unknown-size Folded Reload
-; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
-; RV64-NEXT:    csrr a0, vlenb
-; RV64-NEXT:    slli a0, a0, 4
-; RV64-NEXT:    add sp, sp, a0
-; RV64-NEXT:    addi sp, sp, 16
+; RV64-NEXT:    vsoxei32.v v16, (a0), v24, v0.t
 ; RV64-NEXT:    ret
   %eidxs = zext <vscale x 16 x i16> %idxs to <vscale x 16 x i64>
   %ptrs = getelementptr inbounds double, ptr %base, <vscale x 16 x i64> %eidxs
   call void @llvm.vp.scatter.nxv16f64.nxv16p0(<vscale x 16 x double> %val, <vscale x 16 x ptr> %ptrs, <vscale x 16 x i1> %m, i32 %evl)
   ret void
 }