[RISCV] Add initial support for converting fixed vectors to scalable vectors during lowering to use RVV instructions.

[craig.topper]

Summary

This is an alternative to D95563.

This is modeled after a similar feature for AArch64's SVE that uses
predicated scalable vector instructions.a

Rather than use predication, this patch uses an explicit VL operand.
I've limited it to always use LMUL=1 for now, but we can improve this
in the future.

This requires a bunch of new ISD opcodes to carry the VL operand.
I think we can probably lower intrinsics to these ISD opcodes to
cut down on the size of the isel table. Which is why I've added
patterns for all integer/float types and not just LMUL=1.

I'm only testing one vector width right now, but the width is
programmable via the command line.

[craig.topper]

Use 0.10 as the version in the new td file

[jrtc27]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

1

Not sure how you managed this diff :)

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

1

That's amazing. I don't either.

[craig.topper]

Remove stray change

[craig.topper]

clang-format

[arcbbb]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

1716

For supporting LMUL > 1,
I was thinking calculating LMUL by

unsigned MinVLen = Subtarget.getMinRVVVectorSizeInBits();
unsigned LMul = (VT.getSizeInBits() + MinVLen - 1) / MinVLen;

and expand the case MVT::i8: by

case MVT::i8:
   if (LMul == 1)
      return EVT(MVT::nxv8i8);
   else if (LMul == 2)
      return EVT(MVT::nxv16i8);
   else if (LMul <= 4)
       return EVT(MVT::nxv32i8);
   else if (LMul <= 8)
       return EVT(MVT::nxv64i8);

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

1716

I guess my concern is that large LMULs increase register pressure so depending on the code it might be better to split the operations than use the increased LMUL if it will cause spills.

[arcbbb]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

1716

I have a use case for [1]

blur_x.store_at(blur_y, y).compute_at(blur_y, yi).vectorize(x, 8);

User can evaluate various VLS choices by changing the number of element in .vectorize() and evaluate which LMul gains most performance.
not sure if it makes sense.
[1] https://github.com/halide/Halide/blob/master/apps/blur/halide_blur_generator.cpp

[HsiangKai]

llvm/lib/Target/RISCV/RISCVSubtarget.cpp

100

V has no such requirement or I misunderstood the specification?

[craig.topper]

llvm/lib/Target/RISCV/RISCVSubtarget.cpp

100

Good point. I blindly copied that from AArch64 and was more focused on getting on to the lowering work.

What restrictions should we have here?

[HsiangKai]

llvm/lib/Target/RISCV/RISCVSubtarget.cpp

100

There are two restrictions in specification. VLEN≥128 and VLEN must be a power of 2.

[khchen]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

1716

I think maybe the vectorizer or users(Halide) need to aware the register pressure (or performance) when they perform transformation. If codegen want to help on avoiding register pressure problem, maybe backend could have an IR pass to find out the best LMUL for scalabe vector type operation?

[frasercrmck]

I think this is a sensible direction to consider. I do think we'll need to think about other LMULs though.

If we know the minimum size, there's presumably a way to successively enable larger vectors as the size increases? Ideally with vector-bits-min=1024 we'd be able to support e.g. <16 x i64> with LMUL=1, with vector-bits-min=512 we could do it with LMUL=2. Is that going to be challenging?

We might have to limit it somewhat sensibly: for vector types that would require LMUL=8 we're not necessarily going to see a huge difference in performance compared with just splitting the vector in two.

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

1737

Copy/paste on SVE?

llvm/lib/Target/RISCV/RISCVISelLowering.h

119

Is this something that is likely to be part of the vector predication support, once that becomes more "first class"?

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.h

119

It looks like the vector predication SD nodes also have a mask operand. So I'm not sure if we should synthesize an all 1s mask to pattern match back out. Or if we should DAG combine an all ones mask on the vector predication SD nodes to these nodes.

It looks like for masking, the vector predication nodes just make the masked out elements undefined, so I think we still need a VSELECT to specify a passthru value? So I'm not sure if we should pattern match that sequence or DAG combine to an ISD node that has the mask, passthru, and VL all together?

[craig.topper]

Add support for different LMuls with a command line option to limit.

[HsiangKai]

llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td

60

Need rebase here.

llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td

65

Use RVVBaseAddr instead of reg_rs1.

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll

2

small case for "experimental-zfh".

3

Do we need different check prefix for riscv32 and riscv64? That is LMULMAX2-RV32 and LMULMAX2-RV64.

[craig.topper]

-Rebase
-Correct the register classes for fixed vectors for LMUL > 1.
-Correct the mapping for udiv/sdiv. I copy pasted the same mistake SDPatterns had before

[craig.topper]

Remove stale FIXME

[HsiangKai]

LGTM.

[frasercrmck]

llvm/lib/Target/RISCV/RISCVISelLowering.h

119

Yeah that seems a shame. I've been wondering how we'd support the vp intrinsics and I guess that this question about the SDNodes is an extension of that. I seem to recall @rogfer01 asking about them on the list: I wonder if he has any ideas? Do we only ever expect an all-ones mask and mandate that the VL is the only "predicate" we use?

Circling back a bit: my original question was about the fact that it seems a shame we have to make our own copies of all of these nodes. Could we lower fixed-length vectors to scalable-vector VP nodes with an all-ones mask, and pattern-match that?

But if we do indeed have to duplicate nodes to account for the passthru value for "full" VP support then we'd probably have to add some operands to these nodes. It would be good to share these nodes for this purpose and for the VP support. Do we do that now to get it out of the way, or later? Maybe the extra VP operands could be optional which means that the "fixed-length" patterns won't have to change when the time comes.

llvm/lib/Target/RISCV/RISCVSubtarget.h

152

typo: beyong

[craig.topper]

-Add mask argument to binary ops. Did not add passthru value which is consistent with VP_*
-Maintaining custom set of nodes since FP nodes ISD::VP_ are missing. Also the VE target is translating all VP_* nodes to custom nodes.

[frasercrmck]

llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td

165

The sdiv/udiv mapping is incorrect. Somehow it sneaked back in?

[frasercrmck]

The "require changes" feels really drastic to me but I thought HsiangKai's LGTM was official acceptance, sorry!

[HsiangKai]

In D95705#2544492, @frasercrmck wrote:

The "require changes" feels really drastic to me but I thought HsiangKai's LGTM was official acceptance, sorry!

I also think this patch needs your acceptance. So, I didn't accept the patch. I think there still are some issues to address. Thanks for your feedback.

[craig.topper]

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll

3

The RV32 and RV64 codegen seem to be the same for all tests for LMULMAX2. We're not allowed to have unused prefixes that don't appear in a function so I can't add LMULMAX2-RV32/RV64 if they aren't used.

[craig.topper]

Fix sdiv/udiv swap again

[craig.topper]

Use SDValue instead of auto in a couple places.
Run clang-format

[craig.topper]

Fix capitalization of +experimental-zfh in test RUN lines.

[frasercrmck]

LGTM. I don't know if others want to review it?

llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td

182

nit: This is 14.12 in v0.10. I'm changing that in the SD pats as part of D96028

[khchen]

LGTM.

[Jim]

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll

257

Hi,

Could I ask you how to update this check label of this function?

I use update_llc_test_checks.py to update this check label of this function.

But It deletes all LMULMAX1-RV32 and LMULMAX1-RV64 label and adds LMULMAX1.

No just update label which is already existed.

[craig.topper]

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll

257

I just ran the script on the test on trunk and nothing change.

llvm/lib/Target/RISCV/RISCVISelLowering.h

 //===-- RISCVISelLowering.h - RISCV DAG Lowering Interface ------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
enum NodeType : unsigned {
   VSLIDEUP,
   VSLIDEDOWN,
   // Matches the semantics of the unmasked vid.v instruction.
   VID,
   // Matches the semantics of the vfcnvt.rod function (Convert double-width
   // float to single-width float, rounding towards odd). Takes a double-width
   // float vector and produces a single-width float vector.
   VFNCVT_ROD,
+
+  // Vector binary and unary ops with VL as a third operand.
+  ADD_VL,
+  AND_VL,
+  MUL_VL,
+  OR_VL,
+  SDIV_VL,
+  SHL_VL,
+  SREM_VL,
+  SRA_VL,
+  SRL_VL,
+  SUB_VL,
+  UDIV_VL,
+  UREM_VL,
+  XOR_VL,
+  FADD_VL,
+  FSUB_VL,
+  FMUL_VL,
+  FDIV_VL,
+  FNEG_VL,
+
+  // Memory opcodes start here.
+  VLE_VL = ISD::FIRST_TARGET_MEMORY_OPCODE,
+  VSE_VL,
+
+  // WARNING: Do not add anything in the end unless you want the node to
+  // have memop! In fact, starting from FIRST_TARGET_MEMORY_OPCODE all
+  // opcodes will be thought as target memory ops!
 };
 } // namespace RISCVISD
 
 class RISCVTargetLowering : public TargetLowering {
   const RISCVSubtarget &Subtarget;
 
 public:
   explicit RISCVTargetLowering(const TargetMachine &TM,
private:
   SDValue lowerSPLATVECTOR(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerVectorMaskExt(SDValue Op, SelectionDAG &DAG,
                              int64_t ExtTrueVal) const;
   SDValue lowerVectorMaskTrunc(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerINSERT_VECTOR_ELT(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerEXTRACT_VECTOR_ELT(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, SelectionDAG &DAG) const;
+  SDValue lowerFixedLengthVectorLoadToRVV(SDValue Op, SelectionDAG &DAG) const;
+  SDValue lowerFixedLengthVectorStoreToRVV(SDValue Op, SelectionDAG &DAG) const;
+  SDValue lowerToScalableOp(SDValue Op, SelectionDAG &DAG,
+                            unsigned NewOpc) const;
 
   bool isEligibleForTailCallOptimization(
       CCState &CCInfo, CallLoweringInfo &CLI, MachineFunction &MF,
       const SmallVector<CCValAssign, 16> &ArgLocs) const;
 
   /// Generate error diagnostics if any register used by CC has been marked
   /// reserved.
   void validateCCReservedRegs(
       const SmallVectorImpl<std::pair<llvm::Register, llvm::SDValue>> &Regs,
       MachineFunction &MF) const;
+
+  bool useRVVForFixedLengthVectorVT(EVT VT) const;
 };
 
 namespace RISCVVIntrinsicsTable {
 
 struct RISCVVIntrinsicInfo {
   unsigned IntrinsicID;
   uint8_t ExtendedOperand;
 };

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

 //===-- RISCVISelLowering.cpp - RISCV DAG Lowering Implementation  --------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file defines the interfaces that RISCV uses to lower LLVM code into a
if (Subtarget.hasStdExtV()) {
 
     if (Subtarget.hasStdExtF())
       for (MVT VT : F32VecVTs)
         addRegClassForRVV(VT);
 
     if (Subtarget.hasStdExtD())
       for (MVT VT : F64VecVTs)
         addRegClassForRVV(VT);
+
+    if (Subtarget.useRVVForFixedLengthVectors()) {
+      // FIXME: How should we support larger LMUL values?
+      for (MVT VT : MVT::integer_fixedlen_vector_valuetypes())
+        if (useRVVForFixedLengthVectorVT(VT))
+          addRegisterClass(VT, &RISCV::VRRegClass);
+
+      for (MVT VT : MVT::fp_fixedlen_vector_valuetypes())
+        if (useRVVForFixedLengthVectorVT(VT))
+          addRegisterClass(VT, &RISCV::VRRegClass);
+    }
   }
 
   // Compute derived properties from the register classes.
   computeRegisterProperties(STI.getRegisterInfo());
 
   setStackPointerRegisterToSaveRestore(RISCV::X2);
 
   for (auto N : {ISD::EXTLOAD, ISD::SEXTLOAD, ISD::ZEXTLOAD})
if (Subtarget.hasStdExtV()) {
 
     if (Subtarget.hasStdExtF())
       for (MVT VT : F32VecVTs)
         SetCommonVFPActions(VT);
 
     if (Subtarget.hasStdExtD())
       for (MVT VT : F64VecVTs)
         SetCommonVFPActions(VT);
+
+    if (Subtarget.useRVVForFixedLengthVectors()) {
+      for (MVT VT : MVT::integer_fixedlen_vector_valuetypes()) {
+        if (!useRVVForFixedLengthVectorVT(VT))
+          continue;
+
+        // By default everything must be expanded.
+        for (unsigned Op = 0; Op < ISD::BUILTIN_OP_END; ++Op)
+          setOperationAction(Op, VT, Expand);
+
+        // We use EXTRACT_SUBVECTOR as a "cast" from scalable to fixed.
+        setOperationAction(ISD::EXTRACT_SUBVECTOR, VT, Legal);
+
+        setOperationAction(ISD::LOAD, VT, Custom);
+        setOperationAction(ISD::STORE, VT, Custom);
+        setOperationAction(ISD::ADD, VT, Custom);
+        setOperationAction(ISD::MUL, VT, Custom);
+        setOperationAction(ISD::SUB, VT, Custom);
+        setOperationAction(ISD::AND, VT, Custom);
+        setOperationAction(ISD::OR, VT, Custom);
+        setOperationAction(ISD::XOR, VT, Custom);
+        setOperationAction(ISD::SDIV, VT, Custom);
+        setOperationAction(ISD::SREM, VT, Custom);
+        setOperationAction(ISD::UDIV, VT, Custom);
+        setOperationAction(ISD::UREM, VT, Custom);
+        setOperationAction(ISD::SHL, VT, Custom);
+        setOperationAction(ISD::SRA, VT, Custom);
+        setOperationAction(ISD::SRL, VT, Custom);
+      }
+
+      for (MVT VT : MVT::fp_fixedlen_vector_valuetypes()) {
+        if (!useRVVForFixedLengthVectorVT(VT))
+          continue;
+
+        // By default everything must be expanded.
+        for (unsigned Op = 0; Op < ISD::BUILTIN_OP_END; ++Op)
+          setOperationAction(Op, VT, Expand);
+
+        // We use EXTRACT_SUBVECTOR as a "cast" from scalable to fixed.
+        setOperationAction(ISD::EXTRACT_SUBVECTOR, VT, Legal);
+
+        setOperationAction(ISD::LOAD, VT, Custom);
+        setOperationAction(ISD::STORE, VT, Custom);
+        setOperationAction(ISD::FADD, VT, Custom);
+        setOperationAction(ISD::FSUB, VT, Custom);
+        setOperationAction(ISD::FMUL, VT, Custom);
+        setOperationAction(ISD::FDIV, VT, Custom);
+        setOperationAction(ISD::FNEG, VT, Custom);
+      }
+    }
   }
 
   // Function alignments.
   const Align FunctionAlignment(Subtarget.hasStdExtC() ? 2 : 4);
   setMinFunctionAlignment(FunctionAlignment);
   setPrefFunctionAlignment(FunctionAlignment);
 
   setMinimumJumpTableEntries(5);
if (SrcEltSize > EltSize && (SrcEltSize / EltSize >= 4)) {
           MVT::getVectorVT(MVT::getIntegerVT(SrcEltVT.getSizeInBits() / 2),
                            VT.getVectorElementCount());
       SDValue FP2Int = DAG.getNode(Op.getOpcode(), DL, IVecVT, Src);
       return DAG.getNode(ISD::TRUNCATE, DL, VT, FP2Int);
     }
 
     return Op;
   }
+  case ISD::LOAD:
+    return lowerFixedLengthVectorLoadToRVV(Op, DAG);
+  case ISD::STORE:
+    return lowerFixedLengthVectorStoreToRVV(Op, DAG);
+  case ISD::ADD:
+    return lowerToScalableOp(Op, DAG, RISCVISD::ADD_VL);
+  case ISD::SUB:
+    return lowerToScalableOp(Op, DAG, RISCVISD::SUB_VL);
+  case ISD::MUL:
+    return lowerToScalableOp(Op, DAG, RISCVISD::MUL_VL);
+  case ISD::AND:
+    return lowerToScalableOp(Op, DAG, RISCVISD::AND_VL);
+  case ISD::OR:
+    return lowerToScalableOp(Op, DAG, RISCVISD::OR_VL);
+  case ISD::XOR:
+    return lowerToScalableOp(Op, DAG, RISCVISD::XOR_VL);
+  case ISD::SDIV:
+    return lowerToScalableOp(Op, DAG, RISCVISD::SDIV_VL);
+  case ISD::SREM:
+    return lowerToScalableOp(Op, DAG, RISCVISD::SREM_VL);
+  case ISD::UDIV:
+    return lowerToScalableOp(Op, DAG, RISCVISD::UDIV_VL);
+  case ISD::UREM:
+    return lowerToScalableOp(Op, DAG, RISCVISD::UREM_VL);
+  case ISD::SHL:
+    return lowerToScalableOp(Op, DAG, RISCVISD::SHL_VL);
+  case ISD::SRA:
+    return lowerToScalableOp(Op, DAG, RISCVISD::SRA_VL);
+  case ISD::SRL:
+    return lowerToScalableOp(Op, DAG, RISCVISD::SRL_VL);
+  case ISD::FADD:
+    return lowerToScalableOp(Op, DAG, RISCVISD::FADD_VL);
+  case ISD::FSUB:
+    return lowerToScalableOp(Op, DAG, RISCVISD::FSUB_VL);
+  case ISD::FMUL:
+    return lowerToScalableOp(Op, DAG, RISCVISD::FMUL_VL);
+  case ISD::FDIV:
+    return lowerToScalableOp(Op, DAG, RISCVISD::FDIV_VL);
+  case ISD::FNEG:
+    return lowerToScalableOp(Op, DAG, RISCVISD::FNEG_VL);
   }
 }
 
 static SDValue getTargetNode(GlobalAddressSDNode *N, SDLoc DL, EVT Ty,
                              SelectionDAG &DAG, unsigned Flags) {
   return DAG.getTargetGlobalAddress(N->getGlobal(), DL, Ty, 0, Flags);
 }
 
SDValue ReadVL =
         SDValue(DAG.getMachineNode(RISCV::PseudoReadVL, DL, Op->getValueType(1),
                                    Load.getValue(2)),
                 0);
     return DAG.getMergeValues({Load, ReadVL, Load.getValue(1)}, DL);
   }
   }
 }
 
+// Return the largest legal scalable vector type that matches VT's element type.
+static MVT getContainerForFixedLengthVector(SelectionDAG &DAG, MVT VT,
+                                            const RISCVSubtarget &Subtarget) {
+  assert(VT.isFixedLengthVector() &&
+         DAG.getTargetLoweringInfo().isTypeLegal(VT) &&
+         "Expected legal fixed length vector!");
+
+  unsigned MinVLen = Subtarget.getMinRVVVectorSizeInBits();
+  unsigned LMul = divideCeil(VT.getSizeInBits(), MinVLen);
+  assert(LMul <= 8 && isPowerOf2_32(LMul) && "Unexpected LMUL!");
+
+  // We use the types that will be mapped to LMUL=1 here.
+  // FIXME: Should we have a way to support larger LMULs?
+  switch (VT.getVectorElementType().SimpleTy) {
+  default:
+    llvm_unreachable("unexpected element type for RVV container");
+  case MVT::i8:
+    return MVT::getScalableVectorVT(MVT::i8, LMul * 8);
+  case MVT::i16:
+    return MVT::getScalableVectorVT(MVT::i16, LMul * 4);
+  case MVT::i32:
+    return MVT::getScalableVectorVT(MVT::i32, LMul * 2);
+  case MVT::i64:
+    return MVT::getScalableVectorVT(MVT::i64, LMul);
+  case MVT::f16:
+    return MVT::getScalableVectorVT(MVT::f16, LMul * 4);
+  case MVT::f32:
+    return MVT::getScalableVectorVT(MVT::f32, LMul * 2);
+  case MVT::f64:
+    return MVT::getScalableVectorVT(MVT::f64, LMul);
+  }
+}
+
+// Grow V to consume an entire RVV register.
+static SDValue convertToScalableVector(EVT VT, SDValue V, SelectionDAG &DAG,
+                                       const RISCVSubtarget &Subtarget) {
+  assert(VT.isScalableVector() &&
+         "Expected to convert into a scalable vector!");
+  assert(V.getValueType().isFixedLengthVector() &&
+         "Expected a fixed length vector operand!");
+  SDLoc DL(V);
+  SDValue Zero = DAG.getConstant(0, DL, Subtarget.getXLenVT());
+  return DAG.getNode(ISD::INSERT_SUBVECTOR, DL, VT, DAG.getUNDEF(VT), V, Zero);
+}
+
+// Shrink V so it's just big enough to maintain a VT's worth of data.
+static SDValue convertFromScalableVector(EVT VT, SDValue V, SelectionDAG &DAG,
+                                         const RISCVSubtarget &Subtarget) {
+  assert(VT.isFixedLengthVector() &&
+         "Expected to convert into a fixed length vector!");
+  assert(V.getValueType().isScalableVector() &&
+         "Expected a scalable vector operand!");
+  SDLoc DL(V);
+  SDValue Zero = DAG.getConstant(0, DL, Subtarget.getXLenVT());
+  return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, VT, V, Zero);
+}
+
+SDValue
+RISCVTargetLowering::lowerFixedLengthVectorLoadToRVV(SDValue Op,
+                                                     SelectionDAG &DAG) const {
+  auto *Load = cast<LoadSDNode>(Op);
+
+  SDLoc DL(Op);
+  MVT VT = Op.getSimpleValueType();
+  MVT ContainerVT = getContainerForFixedLengthVector(DAG, VT, Subtarget);
+
+  SDValue VL =
+      DAG.getConstant(VT.getVectorNumElements(), DL, Subtarget.getXLenVT());
+
+  SDVTList VTs = DAG.getVTList({ContainerVT, MVT::Other});
+  auto NewLoad = DAG.getMemIntrinsicNode(
+      RISCVISD::VLE_VL, DL, VTs, {Load->getChain(), Load->getBasePtr(), VL},
+      Load->getMemoryVT(), Load->getMemOperand());
+
+  auto Result = convertFromScalableVector(VT, NewLoad, DAG, Subtarget);
+  return DAG.getMergeValues({Result, Load->getChain()}, DL);
+}
+
+SDValue
+RISCVTargetLowering::lowerFixedLengthVectorStoreToRVV(SDValue Op,
+                                                      SelectionDAG &DAG) const {
+  auto *Store = cast<StoreSDNode>(Op);
+
+  SDLoc DL(Op);
+  MVT VT = Store->getValue().getSimpleValueType();
+  MVT ContainerVT = getContainerForFixedLengthVector(DAG, VT, Subtarget);
+
+  SDValue VL =
+      DAG.getConstant(VT.getVectorNumElements(), DL, Subtarget.getXLenVT());
+
+  SDValue NewValue =
+      convertToScalableVector(ContainerVT, Store->getValue(), DAG, Subtarget);
+  return DAG.getMemIntrinsicNode(
+      RISCVISD::VSE_VL, DL, DAG.getVTList(MVT::Other),
+      {Store->getChain(), NewValue, Store->getBasePtr(), VL},
+      Store->getMemoryVT(), Store->getMemOperand());
+}
+
+SDValue RISCVTargetLowering::lowerToScalableOp(SDValue Op, SelectionDAG &DAG,
+                                               unsigned NewOpc) const {
+  MVT VT = Op.getSimpleValueType();
+  assert(useRVVForFixedLengthVectorVT(VT) &&
+         "Only expected to lower fixed length vector operation!");
+  MVT ContainerVT = getContainerForFixedLengthVector(DAG, VT, Subtarget);
+
+  // Create list of operands by converting existing ones to scalable types.
+  SmallVector<SDValue, 5> Ops;
+  for (const SDValue &V : Op->op_values()) {
+    assert(!isa<VTSDNode>(V) && "Unexpected VTSDNode node!");
+
+    // Pass through non-vector operands.
+    if (!V.getValueType().isVector()) {
+      Ops.push_back(V);
+      continue;
+    }
+
+    // "cast" fixed length vector to a scalable vector.
+    assert(useRVVForFixedLengthVectorVT(V.getValueType()) &&
+           "Only fixed length vectors are supported!");
+    Ops.push_back(convertToScalableVector(ContainerVT, V, DAG, Subtarget));
+  }
+
+  SDLoc DL(Op);
+  SDValue VL =
+      DAG.getConstant(VT.getVectorNumElements(), DL, Subtarget.getXLenVT());
+  Ops.push_back(VL);
+
+  SDValue ScalableRes = DAG.getNode(NewOpc, DL, ContainerVT, Ops);
+  return convertFromScalableVector(VT, ScalableRes, DAG, Subtarget);
+}
+
 // Returns the opcode of the target-specific SDNode that implements the 32-bit
 // form of the given Opcode.
 static RISCVISD::NodeType getRISCVWOpcode(unsigned Opcode) {
   switch (Opcode) {
   default:
     llvm_unreachable("Unexpected opcode");
   case ISD::SHL:
     return RISCVISD::SLLW;
#define NODE_NAME_CASE(NODE)                                                   \
   NODE_NAME_CASE(READ_VLENB)
   NODE_NAME_CASE(TRUNCATE_VECTOR)
   NODE_NAME_CASE(VLEFF)
   NODE_NAME_CASE(VLEFF_MASK)
   NODE_NAME_CASE(VSLIDEUP)
   NODE_NAME_CASE(VSLIDEDOWN)
   NODE_NAME_CASE(VID)
   NODE_NAME_CASE(VFNCVT_ROD)
+  NODE_NAME_CASE(ADD_VL)
+  NODE_NAME_CASE(AND_VL)
+  NODE_NAME_CASE(MUL_VL)
+  NODE_NAME_CASE(OR_VL)
+  NODE_NAME_CASE(SDIV_VL)
+  NODE_NAME_CASE(SHL_VL)
+  NODE_NAME_CASE(SREM_VL)
+  NODE_NAME_CASE(SRA_VL)
+  NODE_NAME_CASE(SRL_VL)
+  NODE_NAME_CASE(SUB_VL)
+  NODE_NAME_CASE(UDIV_VL)
+  NODE_NAME_CASE(UREM_VL)
+  NODE_NAME_CASE(XOR_VL)
+  NODE_NAME_CASE(FADD_VL)
+  NODE_NAME_CASE(FSUB_VL)
+  NODE_NAME_CASE(FMUL_VL)
+  NODE_NAME_CASE(FDIV_VL)
+  NODE_NAME_CASE(FNEG_VL)
+  NODE_NAME_CASE(VLE_VL)
+  NODE_NAME_CASE(VSE_VL)
   }
   // clang-format on
   return nullptr;
 #undef NODE_NAME_CASE
 }
 
 /// getConstraintType - Given a constraint letter, return the type of
 /// constraint it is for this target.
if (auto *ConstNode = dyn_cast<ConstantSDNode>(C.getNode())) {
         return true;
       }
     }
   }
 
   return false;
 }
 
+bool RISCVTargetLowering::useRVVForFixedLengthVectorVT(EVT VT) const {
+  if (!Subtarget.useRVVForFixedLengthVectors())
+    return false;
+
+  if (!VT.isFixedLengthVector())
+    return false;
+
+  // Don't use RVV for vectors we cannot scalarize if required.
+  switch (VT.getVectorElementType().getSimpleVT().SimpleTy) {
+  default:
+    return false;
+  case MVT::i1:
+  case MVT::i8:
+  case MVT::i16:
+  case MVT::i32:
+  case MVT::i64:
+    break;
+  case MVT::f16:
+    if (!Subtarget.hasStdExtZfh())
+      return false;
+    break;
+  case MVT::f32:
+    if (!Subtarget.hasStdExtF())
+      return false;
+    break;
+  case MVT::f64:
+    if (!Subtarget.hasStdExtD())
+      return false;
+    break;
+  }
+
+  unsigned MinVLen = Subtarget.getMinRVVVectorSizeInBits();
+  unsigned LMul = divideCeil(VT.getSizeInBits(), MinVLen);
+  // Don't use RVV for types that don't fit.
+  if (LMul > Subtarget.getMaxLMULForFixedLengthVectors())
+    return false;
+
+  // TODO: Perhaps an artificial restriction, but worth having whilst getting
+  // the base fixed length RVV support in place.
+  if (!VT.isPow2VectorType())
+    return false;
+
+  return true;
+}
+
 #define GET_REGISTER_MATCHER
 #include "RISCVGenAsmMatcher.inc"
 
 Register
 RISCVTargetLowering::getRegisterByName(const char *RegName, LLT VT,
                                        const MachineFunction &MF) const {
   Register Reg = MatchRegisterAltName(RegName);
   if (Reg == RISCV::NoRegister)

llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td

 } // Predicates = [HasStdExtV]
 
 let Predicates = [HasStdExtV, HasStdExtF] in {
   defm "" : VPatUnaryV_V_AnyMask<"int_riscv_vcompress", "PseudoVCOMPRESS", AllFloatVectors>;
 } // Predicates = [HasStdExtV, HasStdExtF]
 
 // Include the non-intrinsic ISel patterns
 include "RISCVInstrInfoVSDPatterns.td"
+include "RISCVInstrInfoVVLPatterns.td"

llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td

multiclass VPatUSLoadStoreSDNode<LLVMType type,
                                  VReg reg_class>
 {
   defvar load_instr = !cast<Instruction>("PseudoVLE"#sew#"_V_"#vlmul.MX);
   defvar store_instr = !cast<Instruction>("PseudoVSE"#sew#"_V_"#vlmul.MX);
   // Load
   def : Pat<(type (load reg_rs1:$rs1)),
             (load_instr reg_rs1:$rs1, avl, sew)>;
   // Store
-  def : Pat<(store type:$rs2, reg_rs1:$rs1),
+  def : Pat<(store (type reg_class:$rs2), reg_rs1:$rs1),
             (store_instr reg_class:$rs2, reg_rs1:$rs1, avl, sew)>;
 }
 
 multiclass VPatUSLoadStoreSDNodes<RegisterClass reg_rs1> {
   foreach vti = AllVectors in
     defm "" : VPatUSLoadStoreSDNode<vti.Vector, vti.Mask, vti.SEW, vti.LMul,
                                     vti.AVL, reg_rs1, vti.RegClass>;
 }

llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td

+//===- RISCVInstrInfoVVLPatterns.td - RVV VL patterns ------*- tablegen -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+///
+/// This file contains the required infrastructure and VL patterns to
+/// support code generation for the standard 'V' (Vector) extension, version
+/// 0.10.  This version is still experimental as the 'V' extension hasn't been
+/// ratified yet.
+///
+/// This file is included from and depends upon RISCVInstrInfoVPseudos.td
+///
+/// Note: the patterns for RVV intrinsics are found in
+/// RISCVInstrInfoVPseudos.td.
+///
+//===----------------------------------------------------------------------===//
+
+//===----------------------------------------------------------------------===//
+// Helpers to define the VL patterns.
+//===----------------------------------------------------------------------===//
+
+def SDT_RISCVVLE_VL : SDTypeProfile<1, 2, [SDTCisVec<0>, SDTCisPtrTy<1>, SDTCisVT<2, XLenVT>]>;
+def SDT_RISCVVSE_VL : SDTypeProfile<0, 3, [SDTCisVec<0>, SDTCisPtrTy<1>, SDTCisVT<2, XLenVT>]>;
+
+def SDT_RISCVVecUnOp_VL : SDTypeProfile<1, 2, [SDTCisSameAs<0, 1>,
+                                               SDTCisVec<0>,
+                                               SDTCisVT<2, XLenVT>]>;
+def SDT_RISCVVecBinOp_VL : SDTypeProfile<1, 3, [SDTCisSameAs<0, 1>,
+                                                SDTCisSameAs<0, 2>,
+                                                SDTCisVec<0>,
+                                                SDTCisVT<3, XLenVT>]>;
+
+def riscv_vle_vl : SDNode<"RISCVISD::VLE_VL", SDT_RISCVVLE_VL,
+                          [SDNPHasChain, SDNPMayLoad, SDNPMemOperand]>;
+def riscv_vse_vl : SDNode<"RISCVISD::VSE_VL", SDT_RISCVVSE_VL,
+                          [SDNPHasChain, SDNPMayStore, SDNPMemOperand]>;
+
+def riscv_add_vl  : SDNode<"RISCVISD::ADD_VL",  SDT_RISCVVecBinOp_VL>;
+def riscv_sub_vl  : SDNode<"RISCVISD::SUB_VL",  SDT_RISCVVecBinOp_VL>;
+def riscv_mul_vl  : SDNode<"RISCVISD::MUL_VL",  SDT_RISCVVecBinOp_VL>;
+def riscv_and_vl  : SDNode<"RISCVISD::AND_VL",  SDT_RISCVVecBinOp_VL>;
+def riscv_or_vl   : SDNode<"RISCVISD::OR_VL",   SDT_RISCVVecBinOp_VL>;
+def riscv_xor_vl  : SDNode<"RISCVISD::XOR_VL",  SDT_RISCVVecBinOp_VL>;
+def riscv_sdiv_vl : SDNode<"RISCVISD::SDIV_VL", SDT_RISCVVecBinOp_VL>;
+def riscv_srem_vl : SDNode<"RISCVISD::SREM_VL", SDT_RISCVVecBinOp_VL>;
+def riscv_udiv_vl : SDNode<"RISCVISD::UDIV_VL", SDT_RISCVVecBinOp_VL>;
+def riscv_urem_vl : SDNode<"RISCVISD::UREM_VL", SDT_RISCVVecBinOp_VL>;
+def riscv_shl_vl  : SDNode<"RISCVISD::SHL_VL",  SDT_RISCVVecBinOp_VL>;
+def riscv_sra_vl  : SDNode<"RISCVISD::SRA_VL",  SDT_RISCVVecBinOp_VL>;
+def riscv_srl_vl  : SDNode<"RISCVISD::SRL_VL",  SDT_RISCVVecBinOp_VL>;
+def riscv_fadd_vl : SDNode<"RISCVISD::FADD_VL", SDT_RISCVVecBinOp_VL>;
+def riscv_fsub_vl : SDNode<"RISCVISD::FSUB_VL", SDT_RISCVVecBinOp_VL>;
+def riscv_fmul_vl : SDNode<"RISCVISD::FMUL_VL", SDT_RISCVVecBinOp_VL>;
+def riscv_fdiv_vl : SDNode<"RISCVISD::FDIV_VL", SDT_RISCVVecBinOp_VL>;
+def riscv_fneg_vl : SDNode<"RISCVISD::FNEG_VL", SDT_RISCVVecUnOp_VL>;
+
+multiclass VPatUSLoadStoreVL<RegisterClass reg_rs1> {
+  foreach vti = AllVectors in {
+    defvar load_instr = !cast<Instruction>("PseudoVLE"#vti.SEW#"_V_"#vti.LMul.MX);
+    defvar store_instr = !cast<Instruction>("PseudoVSE"#vti.SEW#"_V_"#vti.LMul.MX);
+    // Load
+    def : Pat<(vti.Vector (riscv_vle_vl reg_rs1:$rs1, (XLenVT GPR:$vl))),
+              (load_instr reg_rs1:$rs1, (NoX0 GPR:$vl), vti.SEW)>;
+    // Store
+    def : Pat<(riscv_vse_vl (vti.Vector vti.RegClass:$rs2), reg_rs1:$rs1, (XLenVT GPR:$vl)),
+              (store_instr vti.RegClass:$rs2, reg_rs1:$rs1, (NoX0 GPR:$vl), vti.SEW)>;
+  }
+}
+
+class VPatBinaryVL_VV<SDNode vop,
+                      string instruction_name,
+                      ValueType result_type,
+                      ValueType op_type,
+                      ValueType mask_type,
+                      int sew,
+                      LMULInfo vlmul,
+                      VReg RetClass,
+                      VReg op_reg_class> :
+    Pat<(result_type (vop
+                     (op_type op_reg_class:$rs1),
+                     (op_type op_reg_class:$rs2),
+                     (XLenVT GPR:$vl))),
+        (!cast<Instruction>(instruction_name#"_VV_"# vlmul.MX)
+                     op_reg_class:$rs1,
+                     op_reg_class:$rs2,
+                     (NoX0 GPR:$vl), sew)>;
+
+multiclass VPatBinaryVL_VV_VX<SDNode vop, string instruction_name> {
+  foreach vti = AllIntegerVectors in {
+    def : VPatBinaryVL_VV<vop, instruction_name,
+                          vti.Vector, vti.Vector, vti.Mask, vti.SEW,
+                          vti.LMul, vti.RegClass, vti.RegClass>;
+    // FIXME: Support splats.
+  }
+}
+
+multiclass VPatBinaryVL_VV_VX_VI<SDNode vop, string instruction_name,
+                                 Operand ImmType = simm5> {
+  foreach vti = AllIntegerVectors in {
+    def : VPatBinaryVL_VV<vop, instruction_name,
+                          vti.Vector, vti.Vector, vti.Mask, vti.SEW,
+                          vti.LMul, vti.RegClass, vti.RegClass>;
+    // FIXME: Support splats.
+  }
+}
+
+multiclass VPatBinaryFPVL_VV_VF<SDNode vop, string instruction_name> {
+  foreach vti = AllFloatVectors in {
+    def : VPatBinaryVL_VV<vop, instruction_name,
+                          vti.Vector, vti.Vector, vti.Mask, vti.SEW,
+                          vti.LMul, vti.RegClass, vti.RegClass>;
+    // FIXME: Support splats.
+  }
+}
+
+//===----------------------------------------------------------------------===//
+// Patterns.
+//===----------------------------------------------------------------------===//
+
+let Predicates = [HasStdExtV] in {
+
+// 7.4. Vector Unit-Stride Instructions
+defm "" : VPatUSLoadStoreVL<GPR>;
+defm "" : VPatUSLoadStoreVL<AddrFI>;
+
+// 12.1. Vector Single-Width Integer Add and Subtract
+defm "" : VPatBinaryVL_VV_VX_VI<riscv_add_vl, "PseudoVADD">;
+defm "" : VPatBinaryVL_VV_VX<riscv_sub_vl, "PseudoVSUB">;
+
+// 12.5. Vector Bitwise Logical Instructions
+defm "" : VPatBinaryVL_VV_VX_VI<riscv_and_vl, "PseudoVAND">;
+defm "" : VPatBinaryVL_VV_VX_VI<riscv_or_vl,  "PseudoVOR">;
+defm "" : VPatBinaryVL_VV_VX_VI<riscv_xor_vl, "PseudoVXOR">;
+
+// 12.6. Vector Single-Width Bit Shift Instructions
+defm "" : VPatBinaryVL_VV_VX_VI<riscv_shl_vl, "PseudoVSLL", uimm5>;
+defm "" : VPatBinaryVL_VV_VX_VI<riscv_srl_vl, "PseudoVSRL", uimm5>;
+defm "" : VPatBinaryVL_VV_VX_VI<riscv_sra_vl, "PseudoVSRA", uimm5>;
+
+// 12.10. Vector Single-Width Integer Multiply Instructions
+defm "" : VPatBinaryVL_VV_VX<riscv_mul_vl, "PseudoVMUL">;
+
+// 12.11. Vector Integer Divide Instructions
+defm "" : VPatBinaryVL_VV_VX<riscv_sdiv_vl, "PseudoVDIVU">;
+defm "" : VPatBinaryVL_VV_VX<riscv_udiv_vl, "PseudoVDIV">;
+defm "" : VPatBinaryVL_VV_VX<riscv_urem_vl, "PseudoVREMU">;
+defm "" : VPatBinaryVL_VV_VX<riscv_srem_vl, "PseudoVREM">;
+
+} // Predicates = [HasStdExtV]
+
+let Predicates = [HasStdExtV, HasStdExtF] in {
+
+// 14.2. Vector Single-Width Floating-Point Add/Subtract Instructions
+defm "" : VPatBinaryFPVL_VV_VF<riscv_fadd_vl, "PseudoVFADD">;
+defm "" : VPatBinaryFPVL_VV_VF<riscv_fsub_vl, "PseudoVFSUB">;
+
+// 14.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
+defm "" : VPatBinaryFPVL_VV_VF<riscv_fmul_vl, "PseudoVFMUL">;
+defm "" : VPatBinaryFPVL_VV_VF<riscv_fdiv_vl, "PseudoVFDIV">;
+
+// 14.10. Vector Floating-Point Sign-Injection Instructions
+// Handle fneg with VFSGNJN using the same input for both operands.
+foreach vti = AllFloatVectors in {
+  def : Pat<(riscv_fneg_vl (vti.Vector vti.RegClass:$rs), (XLenVT GPR:$vl)),
+            (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX)
+                 vti.RegClass:$rs, vti.RegClass:$rs, (NoX0 GPR:$vl), vti.SEW)>;
+}
+
+} // Predicates = [HasStdExtV, HasStdExtF]

llvm/lib/Target/RISCV/RISCVSubtarget.h

 //===-- RISCVSubtarget.h - Define Subtarget for the RISCV -------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
protected:
   std::unique_ptr<LegalizerInfo> Legalizer;
   std::unique_ptr<RegisterBankInfo> RegBankInfo;
 
 public:
   const CallLowering *getCallLowering() const override;
   InstructionSelector *getInstructionSelector() const override;
   const LegalizerInfo *getLegalizerInfo() const override;
   const RegisterBankInfo *getRegBankInfo() const override;
+
+  // Return the known range for the bit length of RVV data registers. A value
+  // of 0 means nothing is known about that particular limit beyong what's
+  // implied by the architecture.
+  unsigned getMinRVVVectorSizeInBits() const;
+  unsigned getMaxLMULForFixedLengthVectors() const;
+  bool useRVVForFixedLengthVectors() const;
 };
 } // End llvm namespace
 
 #endif

llvm/lib/Target/RISCV/RISCVSubtarget.cpp

 //===-- RISCVSubtarget.cpp - RISCV Subtarget Information ------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //

 using namespace llvm;
 
 #define DEBUG_TYPE "riscv-subtarget"
 
 #define GET_SUBTARGETINFO_TARGET_DESC
 #define GET_SUBTARGETINFO_CTOR
 #include "RISCVGenSubtargetInfo.inc"
 
+static cl::opt<unsigned> RVVVectorBitsMin(
+    "riscv-v-vector-bits-min",
+    cl::desc("Assume V extension vector registers are at least this big, "
+             "with zero meaning no minimum size is assumed."),
+    cl::init(0), cl::Hidden);
+
+static cl::opt<unsigned> RVVVectorLMULMax(
+    "riscv-v-fixed-length-vector-lmul-max",
+    cl::desc("The maximum LMUL value to use for fixed length vectors. "
+             "Fractional LMUL values are not supported."),
+    cl::init(8), cl::Hidden);
+
 void RISCVSubtarget::anchor() {}
 
 RISCVSubtarget &RISCVSubtarget::initializeSubtargetDependencies(
     const Triple &TT, StringRef CPU, StringRef TuneCPU, StringRef FS, StringRef ABIName) {
   // Determine default and user-specified characteristics
   bool Is64Bit = TT.isArch64Bit();
   std::string CPUName = std::string(CPU);
   std::string TuneCPUName = std::string(TuneCPU);

 
 const LegalizerInfo *RISCVSubtarget::getLegalizerInfo() const {
   return Legalizer.get();
 }
 
 const RegisterBankInfo *RISCVSubtarget::getRegBankInfo() const {
   return RegBankInfo.get();
 }
+
+unsigned RISCVSubtarget::getMinRVVVectorSizeInBits() const {
+  assert(hasStdExtV() &&
+         "Tried to get vector length without V extension support!");
+  assert((RVVVectorBitsMin == 0 ||
+          (RVVVectorBitsMin >= 128 && isPowerOf2_32(RVVVectorBitsMin))) &&
+         "V extension requires vector length to be at least 128 and a power of "
+         "2!");
+  return PowerOf2Floor(RVVVectorBitsMin < 128 ? 0 : RVVVectorBitsMin);
+}
+
+unsigned RISCVSubtarget::getMaxLMULForFixedLengthVectors() const {
+  assert(hasStdExtV() &&
+         "Tried to get maximum LMUL without V extension support!");
+  assert(RVVVectorLMULMax <= 8 && isPowerOf2_32(RVVVectorLMULMax) &&
+         "V extension requires a LMUL to be at most 8 and a power of 2!");
+  return PowerOf2Floor(std::max<unsigned>(RVVVectorLMULMax, 1));
+}
+
+bool RISCVSubtarget::useRVVForFixedLengthVectors() const {
+  return hasStdExtV() && getMinRVVVectorSizeInBits() != 0;
+}

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v,+experimental-Zfh,+f,+d -verify-machineinstrs -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v,+experimental-Zfh,+f,+d -verify-machineinstrs -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v,+experimental-Zfh,+f,+d -verify-machineinstrs -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=1 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX1,LMULMAX1-RV32
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v,+experimental-Zfh,+f,+d -verify-machineinstrs -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=1 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX1,LMULMAX1-RV64
+
+define void @fadd_v8f16(<8 x half>* %x, <8 x half>* %y) {
+; CHECK-LABEL: fadd_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vfadd.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = load <8 x half>, <8 x half>* %y
+  %c = fadd <8 x half> %a, %b
+  store <8 x half> %c, <8 x half>* %x
+  ret void
+}
+
+define void @fadd_v4f32(<4 x float>* %x, <4 x float>* %y) {
+; CHECK-LABEL: fadd_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vfadd.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = load <4 x float>, <4 x float>* %y
+  %c = fadd <4 x float> %a, %b
+  store <4 x float> %c, <4 x float>* %x
+  ret void
+}
+
+define void @fadd_v2f64(<2 x double>* %x, <2 x double>* %y) {
+; CHECK-LABEL: fadd_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vfadd.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = load <2 x double>, <2 x double>* %y
+  %c = fadd <2 x double> %a, %b
+  store <2 x double> %c, <2 x double>* %x
+  ret void
+}
+
+define void @fsub_v8f16(<8 x half>* %x, <8 x half>* %y) {
+; CHECK-LABEL: fsub_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vfsub.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = load <8 x half>, <8 x half>* %y
+  %c = fsub <8 x half> %a, %b
+  store <8 x half> %c, <8 x half>* %x
+  ret void
+}
+
+define void @fsub_v4f32(<4 x float>* %x, <4 x float>* %y) {
+; CHECK-LABEL: fsub_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vfsub.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = load <4 x float>, <4 x float>* %y
+  %c = fsub <4 x float> %a, %b
+  store <4 x float> %c, <4 x float>* %x
+  ret void
+}
+
+define void @fsub_v2f64(<2 x double>* %x, <2 x double>* %y) {
+; CHECK-LABEL: fsub_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vfsub.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = load <2 x double>, <2 x double>* %y
+  %c = fsub <2 x double> %a, %b
+  store <2 x double> %c, <2 x double>* %x
+  ret void
+}
+
+define void @fmul_v8f16(<8 x half>* %x, <8 x half>* %y) {
+; CHECK-LABEL: fmul_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vfmul.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = load <8 x half>, <8 x half>* %y
+  %c = fmul <8 x half> %a, %b
+  store <8 x half> %c, <8 x half>* %x
+  ret void
+}
+
+define void @fmul_v4f32(<4 x float>* %x, <4 x float>* %y) {
+; CHECK-LABEL: fmul_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vfmul.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = load <4 x float>, <4 x float>* %y
+  %c = fmul <4 x float> %a, %b
+  store <4 x float> %c, <4 x float>* %x
+  ret void
+}
+
+define void @fmul_v2f64(<2 x double>* %x, <2 x double>* %y) {
+; CHECK-LABEL: fmul_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vfmul.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = load <2 x double>, <2 x double>* %y
+  %c = fmul <2 x double> %a, %b
+  store <2 x double> %c, <2 x double>* %x
+  ret void
+}
+
+define void @fdiv_v8f16(<8 x half>* %x, <8 x half>* %y) {
+; CHECK-LABEL: fdiv_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vfdiv.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = load <8 x half>, <8 x half>* %y
+  %c = fdiv <8 x half> %a, %b
+  store <8 x half> %c, <8 x half>* %x
+  ret void
+}
+
+define void @fdiv_v4f32(<4 x float>* %x, <4 x float>* %y) {
+; CHECK-LABEL: fdiv_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vfdiv.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = load <4 x float>, <4 x float>* %y
+  %c = fdiv <4 x float> %a, %b
+  store <4 x float> %c, <4 x float>* %x
+  ret void
+}
+
+define void @fdiv_v2f64(<2 x double>* %x, <2 x double>* %y) {
+; CHECK-LABEL: fdiv_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vfdiv.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = load <2 x double>, <2 x double>* %y
+  %c = fdiv <2 x double> %a, %b
+  store <2 x double> %c, <2 x double>* %x
+  ret void
+}
+
+define void @fneg_v8f16(<8 x half>* %x) {
+; CHECK-LABEL: fneg_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a1, zero, 8
+; CHECK-NEXT:    vsetvli a1, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vfsgnjn.vv v25, v25, v25
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = fneg <8 x half> %a
+  store <8 x half> %b, <8 x half>* %x
+  ret void
+}
+
+define void @fneg_v4f32(<4 x float>* %x) {
+; CHECK-LABEL: fneg_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a1, zero, 4
+; CHECK-NEXT:    vsetvli a1, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vfsgnjn.vv v25, v25, v25
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = fneg <4 x float> %a
+  store <4 x float> %b, <4 x float>* %x
+  ret void
+}
+
+define void @fneg_v2f64(<2 x double>* %x) {
+; CHECK-LABEL: fneg_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a1, zero, 2
+; CHECK-NEXT:    vsetvli a1, a1, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vfsgnjn.vv v25, v25, v25
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = fneg <2 x double> %a
+  store <2 x double> %b, <2 x double>* %x
+  ret void
+}
+
+define void @fadd_v16f16(<16 x half>* %x, <16 x half>* %y) {
+; LMULMAX2-LABEL: fadd_v16f16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vfadd.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fadd_v16f16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfadd.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfadd.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fadd_v16f16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfadd.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfadd.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = load <16 x half>, <16 x half>* %y
+  %c = fadd <16 x half> %a, %b
+  store <16 x half> %c, <16 x half>* %x
+  ret void
+}
+
+define void @fadd_v8f32(<8 x float>* %x, <8 x float>* %y) {
+; LMULMAX2-LABEL: fadd_v8f32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vfadd.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fadd_v8f32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfadd.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfadd.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fadd_v8f32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfadd.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfadd.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = load <8 x float>, <8 x float>* %y
+  %c = fadd <8 x float> %a, %b
+  store <8 x float> %c, <8 x float>* %x
+  ret void
+}
+
+define void @fadd_v4f64(<4 x double>* %x, <4 x double>* %y) {
+; LMULMAX2-LABEL: fadd_v4f64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vfadd.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fadd_v4f64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfadd.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfadd.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fadd_v4f64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfadd.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfadd.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = load <4 x double>, <4 x double>* %y
+  %c = fadd <4 x double> %a, %b
+  store <4 x double> %c, <4 x double>* %x
+  ret void
+}
+
+define void @fsub_v16f16(<16 x half>* %x, <16 x half>* %y) {
+; LMULMAX2-LABEL: fsub_v16f16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vfsub.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fsub_v16f16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfsub.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfsub.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fsub_v16f16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfsub.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfsub.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = load <16 x half>, <16 x half>* %y
+  %c = fsub <16 x half> %a, %b
+  store <16 x half> %c, <16 x half>* %x
+  ret void
+}
+
+define void @fsub_v8f32(<8 x float>* %x, <8 x float>* %y) {
+; LMULMAX2-LABEL: fsub_v8f32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vfsub.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fsub_v8f32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfsub.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfsub.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fsub_v8f32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfsub.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfsub.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = load <8 x float>, <8 x float>* %y
+  %c = fsub <8 x float> %a, %b
+  store <8 x float> %c, <8 x float>* %x
+  ret void
+}
+
+define void @fsub_v4f64(<4 x double>* %x, <4 x double>* %y) {
+; LMULMAX2-LABEL: fsub_v4f64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vfsub.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fsub_v4f64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfsub.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfsub.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fsub_v4f64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfsub.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfsub.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = load <4 x double>, <4 x double>* %y
+  %c = fsub <4 x double> %a, %b
+  store <4 x double> %c, <4 x double>* %x
+  ret void
+}
+
+define void @fmul_v16f16(<16 x half>* %x, <16 x half>* %y) {
+; LMULMAX2-LABEL: fmul_v16f16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vfmul.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fmul_v16f16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfmul.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfmul.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fmul_v16f16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfmul.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfmul.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = load <16 x half>, <16 x half>* %y
+  %c = fmul <16 x half> %a, %b
+  store <16 x half> %c, <16 x half>* %x
+  ret void
+}
+
+define void @fmul_v8f32(<8 x float>* %x, <8 x float>* %y) {
+; LMULMAX2-LABEL: fmul_v8f32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vfmul.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fmul_v8f32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfmul.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfmul.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fmul_v8f32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfmul.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfmul.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = load <8 x float>, <8 x float>* %y
+  %c = fmul <8 x float> %a, %b
+  store <8 x float> %c, <8 x float>* %x
+  ret void
+}
+
+define void @fmul_v4f64(<4 x double>* %x, <4 x double>* %y) {
+; LMULMAX2-LABEL: fmul_v4f64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vfmul.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fmul_v4f64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfmul.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfmul.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fmul_v4f64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfmul.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfmul.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = load <4 x double>, <4 x double>* %y
+  %c = fmul <4 x double> %a, %b
+  store <4 x double> %c, <4 x double>* %x
+  ret void
+}
+
+define void @fdiv_v16f16(<16 x half>* %x, <16 x half>* %y) {
+; LMULMAX2-LABEL: fdiv_v16f16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vfdiv.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fdiv_v16f16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfdiv.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfdiv.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fdiv_v16f16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfdiv.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfdiv.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = load <16 x half>, <16 x half>* %y
+  %c = fdiv <16 x half> %a, %b
+  store <16 x half> %c, <16 x half>* %x
+  ret void
+}
+
+define void @fdiv_v8f32(<8 x float>* %x, <8 x float>* %y) {
+; LMULMAX2-LABEL: fdiv_v8f32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vfdiv.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fdiv_v8f32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfdiv.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfdiv.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fdiv_v8f32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfdiv.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfdiv.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = load <8 x float>, <8 x float>* %y
+  %c = fdiv <8 x float> %a, %b
+  store <8 x float> %c, <8 x float>* %x
+  ret void
+}
+
+define void @fdiv_v4f64(<4 x double>* %x, <4 x double>* %y) {
+; LMULMAX2-LABEL: fdiv_v4f64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vfdiv.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: fdiv_v4f64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vfdiv.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vfdiv.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: fdiv_v4f64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vfdiv.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vfdiv.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = load <4 x double>, <4 x double>* %y
+  %c = fdiv <4 x double> %a, %b
+  store <4 x double> %c, <4 x double>* %x
+  ret void
+}
+
+define void @fneg_v16f16(<16 x half>* %x) {
+; LMULMAX2-LABEL: fneg_v16f16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a1, zero, 16
+; LMULMAX2-NEXT:    vsetvli a1, a1, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vfsgnjn.vv v26, v26, v26
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-LABEL: fneg_v16f16:
+; LMULMAX1:       # %bb.0:
+; LMULMAX1-NEXT:    addi a1, zero, 8
+; LMULMAX1-NEXT:    vsetvli a1, a1, e16,m1,ta,mu
+; LMULMAX1-NEXT:    addi a1, a0, 16
+; LMULMAX1-NEXT:    vle16.v v25, (a1)
+; LMULMAX1-NEXT:    vle16.v v26, (a0)
+; LMULMAX1-NEXT:    vfsgnjn.vv v25, v25, v25
+; LMULMAX1-NEXT:    vfsgnjn.vv v26, v26, v26
+; LMULMAX1-NEXT:    vse16.v v26, (a0)
+; LMULMAX1-NEXT:    vse16.v v25, (a1)
+; LMULMAX1-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = fneg <16 x half> %a
+  store <16 x half> %b, <16 x half>* %x
+  ret void
+}
+
+define void @fneg_v8f32(<8 x float>* %x) {
+; LMULMAX2-LABEL: fneg_v8f32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a1, zero, 8
+; LMULMAX2-NEXT:    vsetvli a1, a1, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vfsgnjn.vv v26, v26, v26
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-LABEL: fneg_v8f32:
+; LMULMAX1:       # %bb.0:
+; LMULMAX1-NEXT:    addi a1, zero, 4
+; LMULMAX1-NEXT:    vsetvli a1, a1, e32,m1,ta,mu
+; LMULMAX1-NEXT:    addi a1, a0, 16
+; LMULMAX1-NEXT:    vle32.v v25, (a1)
+; LMULMAX1-NEXT:    vle32.v v26, (a0)
+; LMULMAX1-NEXT:    vfsgnjn.vv v25, v25, v25
+; LMULMAX1-NEXT:    vfsgnjn.vv v26, v26, v26
+; LMULMAX1-NEXT:    vse32.v v26, (a0)
+; LMULMAX1-NEXT:    vse32.v v25, (a1)
+; LMULMAX1-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = fneg <8 x float> %a
+  store <8 x float> %b, <8 x float>* %x
+  ret void
+}
+
+define void @fneg_v4f64(<4 x double>* %x) {
+; LMULMAX2-LABEL: fneg_v4f64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a1, zero, 4
+; LMULMAX2-NEXT:    vsetvli a1, a1, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vfsgnjn.vv v26, v26, v26
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-LABEL: fneg_v4f64:
+; LMULMAX1:       # %bb.0:
+; LMULMAX1-NEXT:    addi a1, zero, 2
+; LMULMAX1-NEXT:    vsetvli a1, a1, e64,m1,ta,mu
+; LMULMAX1-NEXT:    addi a1, a0, 16
+; LMULMAX1-NEXT:    vle64.v v25, (a1)
+; LMULMAX1-NEXT:    vle64.v v26, (a0)
+; LMULMAX1-NEXT:    vfsgnjn.vv v25, v25, v25
+; LMULMAX1-NEXT:    vfsgnjn.vv v26, v26, v26
+; LMULMAX1-NEXT:    vse64.v v26, (a0)
+; LMULMAX1-NEXT:    vse64.v v25, (a1)
+; LMULMAX1-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = fneg <4 x double> %a
+  store <4 x double> %b, <4 x double>* %x
+  ret void
+}

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v -verify-machineinstrs -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v -verify-machineinstrs -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v -verify-machineinstrs -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=1 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX1-RV32
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v -verify-machineinstrs -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=1 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX1-RV64
+
+define void @add_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: add_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vadd.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = add <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @add_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: add_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vadd.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = add <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @add_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: add_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vadd.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = add <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @add_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: add_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vadd.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = add <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @sub_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: sub_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vsub.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = sub <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @sub_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: sub_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vsub.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = sub <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @sub_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: sub_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vsub.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = sub <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @sub_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: sub_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vsub.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = sub <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @mul_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: mul_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vmul.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = mul <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @mul_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: mul_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vmul.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = mul <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @mul_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: mul_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vmul.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = mul <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @mul_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: mul_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vmul.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = mul <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @and_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: and_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vand.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = and <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @and_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: and_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vand.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = and <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @and_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: and_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vand.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = and <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @and_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: and_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vand.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = and <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @or_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: or_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vor.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = or <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @or_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: or_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vor.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = or <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @or_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: or_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vor.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = or <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @or_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: or_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vor.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = or <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @xor_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: xor_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vxor.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = xor <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @xor_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: xor_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vxor.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = xor <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @xor_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: xor_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vxor.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = xor <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @xor_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: xor_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vxor.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = xor <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @lshr_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: lshr_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vsrl.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = lshr <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @lshr_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: lshr_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vsrl.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = lshr <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @lshr_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: lshr_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vsrl.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = lshr <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @lshr_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: lshr_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vsrl.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = lshr <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @ashr_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: ashr_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vsra.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = ashr <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @ashr_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: ashr_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vsra.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = ashr <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @ashr_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: ashr_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vsra.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = ashr <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @ashr_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: ashr_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vsra.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = ashr <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @shl_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: shl_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vsll.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = shl <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @shl_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: shl_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vsll.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = shl <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @shl_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: shl_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vsll.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = shl <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @shl_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: shl_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vsll.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = shl <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @sdiv_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: sdiv_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vdivu.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = sdiv <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @sdiv_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: sdiv_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vdivu.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = sdiv <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @sdiv_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: sdiv_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vdivu.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = sdiv <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @sdiv_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: sdiv_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vdivu.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = sdiv <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @srem_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: srem_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vrem.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = srem <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @srem_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: srem_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vrem.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = srem <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @srem_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: srem_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vrem.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = srem <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @srem_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: srem_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vrem.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = srem <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @udiv_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: udiv_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vdiv.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = udiv <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @udiv_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: udiv_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vdiv.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = udiv <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @udiv_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: udiv_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vdiv.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = udiv <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @udiv_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: udiv_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vdiv.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = udiv <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @urem_v16i8(<16 x i8>* %x, <16 x i8>* %y) {
+; CHECK-LABEL: urem_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vle8.v v25, (a0)
+; CHECK-NEXT:    vle8.v v26, (a1)
+; CHECK-NEXT:    vremu.vv v25, v25, v26
+; CHECK-NEXT:    vse8.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <16 x i8>, <16 x i8>* %x
+  %b = load <16 x i8>, <16 x i8>* %y
+  %c = urem <16 x i8> %a, %b
+  store <16 x i8> %c, <16 x i8>* %x
+  ret void
+}
+
+define void @urem_v8i16(<8 x i16>* %x, <8 x i16>* %y) {
+; CHECK-LABEL: urem_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vremu.vv v25, v25, v26
+; CHECK-NEXT:    vse16.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <8 x i16>, <8 x i16>* %x
+  %b = load <8 x i16>, <8 x i16>* %y
+  %c = urem <8 x i16> %a, %b
+  store <8 x i16> %c, <8 x i16>* %x
+  ret void
+}
+
+define void @urem_v4i32(<4 x i32>* %x, <4 x i32>* %y) {
+; CHECK-LABEL: urem_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vremu.vv v25, v25, v26
+; CHECK-NEXT:    vse32.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <4 x i32>, <4 x i32>* %x
+  %b = load <4 x i32>, <4 x i32>* %y
+  %c = urem <4 x i32> %a, %b
+  store <4 x i32> %c, <4 x i32>* %x
+  ret void
+}
+
+define void @urem_v2i64(<2 x i64>* %x, <2 x i64>* %y) {
+; CHECK-LABEL: urem_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vremu.vv v25, v25, v26
+; CHECK-NEXT:    vse64.v v25, (a0)
+; CHECK-NEXT:    ret
+  %a = load <2 x i64>, <2 x i64>* %x
+  %b = load <2 x i64>, <2 x i64>* %y
+  %c = urem <2 x i64> %a, %b
+  store <2 x i64> %c, <2 x i64>* %x
+  ret void
+}
+
+define void @add_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: add_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: add_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vadd.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vadd.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: add_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vadd.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vadd.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = add <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @add_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: add_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: add_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vadd.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vadd.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: add_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vadd.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vadd.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = add <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @add_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: add_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: add_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vadd.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vadd.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: add_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vadd.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vadd.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = add <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @add_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: add_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: add_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vadd.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vadd.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: add_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vadd.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vadd.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = add <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @sub_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: sub_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vsub.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: sub_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsub.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsub.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: sub_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsub.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsub.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = sub <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @sub_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: sub_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vsub.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: sub_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsub.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsub.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: sub_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsub.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsub.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = sub <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @sub_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: sub_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vsub.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: sub_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsub.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsub.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: sub_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsub.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsub.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = sub <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @sub_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: sub_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vsub.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: sub_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsub.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsub.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: sub_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsub.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsub.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = sub <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @mul_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: mul_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vmul.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: mul_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vmul.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vmul.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: mul_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vmul.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vmul.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = mul <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @mul_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: mul_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vmul.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: mul_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vmul.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vmul.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: mul_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vmul.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vmul.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = mul <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @mul_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: mul_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vmul.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: mul_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vmul.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vmul.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: mul_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vmul.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vmul.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = mul <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @mul_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: mul_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vmul.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: mul_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vmul.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vmul.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: mul_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vmul.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vmul.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = mul <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @and_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: and_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vand.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: and_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vand.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vand.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: and_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vand.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vand.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = and <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @and_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: and_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vand.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: and_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vand.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vand.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: and_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vand.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vand.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = and <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @and_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: and_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vand.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: and_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vand.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vand.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: and_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vand.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vand.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = and <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @and_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: and_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vand.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: and_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vand.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vand.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: and_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vand.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vand.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = and <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @or_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: or_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vor.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: or_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vor.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vor.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: or_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vor.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vor.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = or <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @or_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: or_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vor.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: or_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vor.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vor.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: or_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vor.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vor.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = or <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @or_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: or_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vor.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: or_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vor.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vor.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: or_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vor.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vor.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = or <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @or_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: or_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vor.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: or_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vor.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vor.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: or_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vor.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vor.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = or <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @xor_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: xor_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vxor.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: xor_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vxor.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vxor.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: xor_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vxor.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vxor.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = xor <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @xor_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: xor_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vxor.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: xor_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vxor.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vxor.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: xor_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vxor.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vxor.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = xor <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @xor_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: xor_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vxor.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: xor_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vxor.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vxor.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: xor_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vxor.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vxor.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = xor <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @xor_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: xor_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vxor.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: xor_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vxor.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vxor.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: xor_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vxor.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vxor.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = xor <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @lshr_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: lshr_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vsrl.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: lshr_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsrl.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsrl.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: lshr_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsrl.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsrl.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = lshr <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @lshr_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: lshr_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vsrl.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: lshr_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsrl.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsrl.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: lshr_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsrl.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsrl.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = lshr <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @lshr_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: lshr_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vsrl.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: lshr_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsrl.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsrl.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: lshr_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsrl.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsrl.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = lshr <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @lshr_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: lshr_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vsrl.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: lshr_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsrl.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsrl.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: lshr_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsrl.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsrl.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = lshr <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @ashr_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: ashr_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vsra.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: ashr_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsra.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsra.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: ashr_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsra.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsra.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = ashr <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @ashr_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: ashr_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vsra.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: ashr_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsra.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsra.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: ashr_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsra.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsra.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = ashr <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @ashr_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: ashr_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vsra.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: ashr_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsra.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsra.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: ashr_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsra.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsra.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = ashr <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @ashr_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: ashr_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vsra.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: ashr_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsra.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsra.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: ashr_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsra.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsra.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = ashr <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @shl_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: shl_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vsll.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: shl_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsll.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsll.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: shl_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsll.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsll.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = shl <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @shl_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: shl_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vsll.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: shl_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsll.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsll.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: shl_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsll.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsll.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = shl <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @shl_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: shl_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vsll.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: shl_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsll.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsll.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: shl_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsll.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsll.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = shl <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @shl_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: shl_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vsll.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: shl_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vsll.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vsll.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: shl_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vsll.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vsll.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = shl <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @sdiv_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: sdiv_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vdivu.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: sdiv_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vdivu.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vdivu.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: sdiv_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vdivu.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vdivu.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = sdiv <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @sdiv_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: sdiv_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vdivu.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: sdiv_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vdivu.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vdivu.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: sdiv_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vdivu.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vdivu.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = sdiv <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @sdiv_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: sdiv_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vdivu.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: sdiv_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vdivu.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vdivu.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: sdiv_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vdivu.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vdivu.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = sdiv <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @sdiv_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: sdiv_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vdivu.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: sdiv_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vdivu.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vdivu.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: sdiv_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vdivu.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vdivu.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = sdiv <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @srem_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: srem_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vrem.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: srem_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vrem.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vrem.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: srem_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vrem.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vrem.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = srem <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @srem_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: srem_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vrem.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: srem_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vrem.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vrem.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: srem_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vrem.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vrem.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = srem <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @srem_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: srem_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vrem.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: srem_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vrem.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vrem.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: srem_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vrem.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vrem.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = srem <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @srem_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: srem_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vrem.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: srem_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vrem.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vrem.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: srem_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vrem.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vrem.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = srem <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @udiv_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: udiv_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vdiv.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: udiv_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vdiv.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vdiv.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: udiv_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vdiv.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vdiv.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = udiv <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @udiv_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: udiv_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vdiv.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: udiv_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vdiv.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vdiv.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: udiv_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vdiv.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vdiv.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = udiv <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @udiv_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: udiv_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vdiv.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: udiv_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vdiv.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vdiv.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: udiv_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vdiv.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vdiv.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = udiv <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @udiv_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: udiv_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vdiv.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: udiv_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vdiv.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vdiv.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: udiv_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vdiv.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vdiv.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = udiv <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}
+
+define void @urem_v32i8(<32 x i8>* %x, <32 x i8>* %y) {
+; LMULMAX2-LABEL: urem_v32i8:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 32
+; LMULMAX2-NEXT:    vsetvli a2, a2, e8,m2,ta,mu
+; LMULMAX2-NEXT:    vle8.v v26, (a0)
+; LMULMAX2-NEXT:    vle8.v v28, (a1)
+; LMULMAX2-NEXT:    vremu.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse8.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: urem_v32i8:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle8.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vremu.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vremu.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: urem_v32i8:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 16
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e8,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle8.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle8.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vremu.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vremu.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse8.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse8.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <32 x i8>, <32 x i8>* %x
+  %b = load <32 x i8>, <32 x i8>* %y
+  %c = urem <32 x i8> %a, %b
+  store <32 x i8> %c, <32 x i8>* %x
+  ret void
+}
+
+define void @urem_v16i16(<16 x i16>* %x, <16 x i16>* %y) {
+; LMULMAX2-LABEL: urem_v16i16:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 16
+; LMULMAX2-NEXT:    vsetvli a2, a2, e16,m2,ta,mu
+; LMULMAX2-NEXT:    vle16.v v26, (a0)
+; LMULMAX2-NEXT:    vle16.v v28, (a1)
+; LMULMAX2-NEXT:    vremu.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse16.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: urem_v16i16:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle16.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vremu.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vremu.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: urem_v16i16:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 8
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e16,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle16.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle16.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vremu.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vremu.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse16.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse16.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <16 x i16>, <16 x i16>* %x
+  %b = load <16 x i16>, <16 x i16>* %y
+  %c = urem <16 x i16> %a, %b
+  store <16 x i16> %c, <16 x i16>* %x
+  ret void
+}
+
+define void @urem_v8i32(<8 x i32>* %x, <8 x i32>* %y) {
+; LMULMAX2-LABEL: urem_v8i32:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 8
+; LMULMAX2-NEXT:    vsetvli a2, a2, e32,m2,ta,mu
+; LMULMAX2-NEXT:    vle32.v v26, (a0)
+; LMULMAX2-NEXT:    vle32.v v28, (a1)
+; LMULMAX2-NEXT:    vremu.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse32.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: urem_v8i32:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle32.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vremu.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vremu.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: urem_v8i32:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 4
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e32,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle32.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle32.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vremu.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vremu.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse32.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse32.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <8 x i32>, <8 x i32>* %x
+  %b = load <8 x i32>, <8 x i32>* %y
+  %c = urem <8 x i32> %a, %b
+  store <8 x i32> %c, <8 x i32>* %x
+  ret void
+}
+
+define void @urem_v4i64(<4 x i64>* %x, <4 x i64>* %y) {
+; LMULMAX2-LABEL: urem_v4i64:
+; LMULMAX2:       # %bb.0:
+; LMULMAX2-NEXT:    addi a2, zero, 4
+; LMULMAX2-NEXT:    vsetvli a2, a2, e64,m2,ta,mu
+; LMULMAX2-NEXT:    vle64.v v26, (a0)
+; LMULMAX2-NEXT:    vle64.v v28, (a1)
+; LMULMAX2-NEXT:    vremu.vv v26, v26, v28
+; LMULMAX2-NEXT:    vse64.v v26, (a0)
+; LMULMAX2-NEXT:    ret
+;
+; LMULMAX1-RV32-LABEL: urem_v4i64:
+; LMULMAX1-RV32:       # %bb.0:
+; LMULMAX1-RV32-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV32-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV32-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    addi a3, a1, 16
+; LMULMAX1-RV32-NEXT:    vle64.v v27, (a3)
+; LMULMAX1-RV32-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV32-NEXT:    vremu.vv v26, v26, v27
+; LMULMAX1-RV32-NEXT:    vremu.vv v25, v25, v28
+; LMULMAX1-RV32-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV32-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV32-NEXT:    ret
+;
+; LMULMAX1-RV64-LABEL: urem_v4i64:
+; LMULMAX1-RV64:       # %bb.0:
+; LMULMAX1-RV64-NEXT:    addi a2, zero, 2
+; LMULMAX1-RV64-NEXT:    vsetvli a2, a2, e64,m1,ta,mu
+; LMULMAX1-RV64-NEXT:    vle64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    addi a2, a1, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    addi a2, a0, 16
+; LMULMAX1-RV64-NEXT:    vle64.v v27, (a2)
+; LMULMAX1-RV64-NEXT:    vle64.v v28, (a1)
+; LMULMAX1-RV64-NEXT:    vremu.vv v26, v27, v26
+; LMULMAX1-RV64-NEXT:    vremu.vv v25, v25, v28
+; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
+; LMULMAX1-RV64-NEXT:    vse64.v v26, (a2)
+; LMULMAX1-RV64-NEXT:    ret
+  %a = load <4 x i64>, <4 x i64>* %x
+  %b = load <4 x i64>, <4 x i64>* %y
+  %c = urem <4 x i64> %a, %b
+  store <4 x i64> %c, <4 x i64>* %x
+  ret void
+}