[RISCV] Custom lower vector ISD::CTLZ to the minimum of ISD::CTLZ_ZERO_UNDEF and element size.

[fakepaper56]

Summary

D111904 made RISC-V customized lower ISD::CTLZ_ZERO_UNDEF by converting to float
and using the float result. The expected value of CTLZ with zero input is the
element size of input type. Since the result of above method with zero input
must be greater than the element size, for ISD::CTLZ, we could use the minimum
of element size and the result of CTLZ_ZERO_UNDER with same input.

[craig.topper]

LGTM with those comments fixed.

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

3606

equaled -> equal

"which greater" -> "which is greater"

[fakepaper56]

Rebase and fix typos.

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

for (MVT VT : IntVecVTs) {
 
       // Splice
       setOperationAction(ISD::VECTOR_SPLICE, VT, Custom);
 
       // Lower CTLZ_ZERO_UNDEF and CTTZ_ZERO_UNDEF if element of VT in the range
       // of f32.
       EVT FloatVT = MVT::getVectorVT(MVT::f32, VT.getVectorElementCount());
       if (isTypeLegal(FloatVT)) {
-        setOperationAction({ISD::CTLZ_ZERO_UNDEF, ISD::CTTZ_ZERO_UNDEF}, VT,
-                           Custom);
+        setOperationAction(
+            {ISD::CTLZ, ISD::CTLZ_ZERO_UNDEF, ISD::CTTZ_ZERO_UNDEF}, VT,
       }
     }
 
     // Expand various CCs to best match the RVV ISA, which natively supports UNE
     // but no other unordered comparisons, and supports all ordered comparisons
     // except ONE. Additionally, we expand GT,OGT,GE,OGE for optimization
     // purposes; they are expanded to their swapped-operand CCs (LT,OLT,LE,OLE),
     // and we pattern-match those back to the "original", swapping operands once
if (Subtarget.useRVVForFixedLengthVectors()) {
                            VT, Custom);
 
         setOperationAction(IntegerVPOps, VT, Custom);
 
         // Lower CTLZ_ZERO_UNDEF and CTTZ_ZERO_UNDEF if element of VT in the
         // range of f32.
         EVT FloatVT = MVT::getVectorVT(MVT::f32, VT.getVectorElementCount());
         if (isTypeLegal(FloatVT))
-          setOperationAction({ISD::CTLZ_ZERO_UNDEF, ISD::CTTZ_ZERO_UNDEF}, VT,
-                             Custom);
+          setOperationAction(
+              {ISD::CTLZ, ISD::CTLZ_ZERO_UNDEF, ISD::CTTZ_ZERO_UNDEF}, VT,
       }
 
       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)
RISCVTargetLowering::lowerCTLZ_CTTZ_ZERO_UNDEF(SDValue Op,
   // For trailing zeros, we just need to subtract the bias.
   if (Op.getOpcode() == ISD::CTTZ_ZERO_UNDEF)
     return DAG.getNode(ISD::SUB, DL, VT, Exp,
                        DAG.getConstant(ExponentBias, DL, VT));
 
   // For leading zeros, we need to remove the bias and convert from log2 to
   // leading zeros. We can do this by subtracting from (Bias + (EltSize - 1)).
   unsigned Adjust = ExponentBias + (EltSize - 1);
-  return DAG.getNode(ISD::SUB, DL, VT, DAG.getConstant(Adjust, DL, VT), Exp);
+  SDValue Res =
+      DAG.getNode(ISD::SUB, DL, VT, DAG.getConstant(Adjust, DL, VT), Exp);
+  // The above result with zero input equals to Adjust which is greater than
+  // EltSize. Hence, we can do min(Res, EltSize) for CTLZ.
+  if (Op.getOpcode() == ISD::CTLZ)
+    Res = DAG.getNode(ISD::UMIN, DL, VT, Res, DAG.getConstant(EltSize, DL, VT));
+  return Res;
 }
 
 // While RVV has alignment restrictions, we should always be able to load as a
 // legal equivalently-sized byte-typed vector instead. This method is
 // responsible for re-expressing a ISD::LOAD via a correctly-aligned type. If
 // the load is already correctly-aligned, it returns SDValue().
 SDValue RISCVTargetLowering::expandUnalignedRVVLoad(SDValue Op,
                                                     SelectionDAG &DAG) const {
case ISD::FMINNUM:
     return lowerToScalableOp(Op, DAG, RISCVISD::FMINNUM_VL,
                              /*HasMergeOp*/ true);
   case ISD::FMAXNUM:
     return lowerToScalableOp(Op, DAG, RISCVISD::FMAXNUM_VL,
                              /*HasMergeOp*/ true);
   case ISD::ABS:
   case ISD::VP_ABS:
     return lowerABS(Op, DAG);
+  case ISD::CTLZ:
   case ISD::CTLZ_ZERO_UNDEF:
   case ISD::CTTZ_ZERO_UNDEF:
     return lowerCTLZ_CTTZ_ZERO_UNDEF(Op, DAG);
   case ISD::VSELECT:
     return lowerFixedLengthVectorSelectToRVV(Op, DAG);
   case ISD::FCOPYSIGN:
     return lowerFixedLengthVectorFCOPYSIGNToRVV(Op, DAG);
   case ISD::MGATHER:

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

 ; CHECK-ZVE64X-NEXT:    vadd.vv v8, v8, v9
 ; CHECK-ZVE64X-NEXT:    vand.vi v8, v8, 15
 ; CHECK-ZVE64X-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv1i8:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
 ; CHECK-F-NEXT:    vzext.vf2 v9, v8
-; CHECK-F-NEXT:    vfwcvt.f.xu.v v10, v9
-; CHECK-F-NEXT:    vnsrl.wi v9, v10, 23
+; CHECK-F-NEXT:    vfwcvt.f.xu.v v8, v9
+; CHECK-F-NEXT:    vnsrl.wi v8, v8, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e8, mf8, ta, ma
-; CHECK-F-NEXT:    vnsrl.wi v9, v9, 0
+; CHECK-F-NEXT:    vnsrl.wi v8, v8, 0
 ; CHECK-F-NEXT:    li a0, 134
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-F-NEXT:    vrsub.vx v8, v9, a0
-; CHECK-F-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a0
+; CHECK-F-NEXT:    li a0, 8
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv1i8:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
 ; CHECK-D-NEXT:    vzext.vf2 v9, v8
-; CHECK-D-NEXT:    vfwcvt.f.xu.v v10, v9
-; CHECK-D-NEXT:    vnsrl.wi v9, v10, 23
+; CHECK-D-NEXT:    vfwcvt.f.xu.v v8, v9
+; CHECK-D-NEXT:    vnsrl.wi v8, v8, 23
 ; CHECK-D-NEXT:    vsetvli zero, zero, e8, mf8, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v9, v9, 0
+; CHECK-D-NEXT:    vnsrl.wi v8, v8, 0
 ; CHECK-D-NEXT:    li a0, 134
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-D-NEXT:    vrsub.vx v8, v9, a0
-; CHECK-D-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
+; CHECK-D-NEXT:    li a0, 8
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 1 x i8> @llvm.ctlz.nxv1i8(<vscale x 1 x i8> %va, i1 false)
   ret <vscale x 1 x i8> %a
 }
 declare <vscale x 1 x i8> @llvm.ctlz.nxv1i8(<vscale x 1 x i8>, i1)
 
 define <vscale x 2 x i8> @ctlz_nxv2i8(<vscale x 2 x i8> %va) {
 ; CHECK-ZVE64X-LABEL: ctlz_nxv2i8:

 ; CHECK-ZVE64X-NEXT:    vadd.vv v8, v8, v9
 ; CHECK-ZVE64X-NEXT:    vand.vi v8, v8, 15
 ; CHECK-ZVE64X-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv2i8:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
 ; CHECK-F-NEXT:    vzext.vf2 v9, v8
-; CHECK-F-NEXT:    vfwcvt.f.xu.v v10, v9
-; CHECK-F-NEXT:    vnsrl.wi v9, v10, 23
+; CHECK-F-NEXT:    vfwcvt.f.xu.v v8, v9
+; CHECK-F-NEXT:    vnsrl.wi v8, v8, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e8, mf4, ta, ma
-; CHECK-F-NEXT:    vnsrl.wi v9, v9, 0
+; CHECK-F-NEXT:    vnsrl.wi v8, v8, 0
 ; CHECK-F-NEXT:    li a0, 134
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-F-NEXT:    vrsub.vx v8, v9, a0
-; CHECK-F-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a0
+; CHECK-F-NEXT:    li a0, 8
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv2i8:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
 ; CHECK-D-NEXT:    vzext.vf2 v9, v8
-; CHECK-D-NEXT:    vfwcvt.f.xu.v v10, v9
-; CHECK-D-NEXT:    vnsrl.wi v9, v10, 23
+; CHECK-D-NEXT:    vfwcvt.f.xu.v v8, v9
+; CHECK-D-NEXT:    vnsrl.wi v8, v8, 23
 ; CHECK-D-NEXT:    vsetvli zero, zero, e8, mf4, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v9, v9, 0
+; CHECK-D-NEXT:    vnsrl.wi v8, v8, 0
 ; CHECK-D-NEXT:    li a0, 134
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-D-NEXT:    vrsub.vx v8, v9, a0
-; CHECK-D-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
+; CHECK-D-NEXT:    li a0, 8
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 2 x i8> @llvm.ctlz.nxv2i8(<vscale x 2 x i8> %va, i1 false)
   ret <vscale x 2 x i8> %a
 }
 declare <vscale x 2 x i8> @llvm.ctlz.nxv2i8(<vscale x 2 x i8>, i1)
 
 define <vscale x 4 x i8> @ctlz_nxv4i8(<vscale x 4 x i8> %va) {
 ; CHECK-ZVE64X-LABEL: ctlz_nxv4i8:

 ; CHECK-ZVE64X-NEXT:    vand.vi v8, v8, 15
 ; CHECK-ZVE64X-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv4i8:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
 ; CHECK-F-NEXT:    vzext.vf2 v9, v8
 ; CHECK-F-NEXT:    vfwcvt.f.xu.v v10, v9
-; CHECK-F-NEXT:    vnsrl.wi v9, v10, 23
+; CHECK-F-NEXT:    vnsrl.wi v8, v10, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e8, mf2, ta, ma
-; CHECK-F-NEXT:    vnsrl.wi v9, v9, 0
+; CHECK-F-NEXT:    vnsrl.wi v8, v8, 0
 ; CHECK-F-NEXT:    li a0, 134
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-F-NEXT:    vrsub.vx v8, v9, a0
-; CHECK-F-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a0
+; CHECK-F-NEXT:    li a0, 8
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv4i8:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
 ; CHECK-D-NEXT:    vzext.vf2 v9, v8
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v10, v9
-; CHECK-D-NEXT:    vnsrl.wi v9, v10, 23
+; CHECK-D-NEXT:    vnsrl.wi v8, v10, 23
 ; CHECK-D-NEXT:    vsetvli zero, zero, e8, mf2, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v9, v9, 0
+; CHECK-D-NEXT:    vnsrl.wi v8, v8, 0
 ; CHECK-D-NEXT:    li a0, 134
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-D-NEXT:    vrsub.vx v8, v9, a0
-; CHECK-D-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
+; CHECK-D-NEXT:    li a0, 8
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 4 x i8> @llvm.ctlz.nxv4i8(<vscale x 4 x i8> %va, i1 false)
   ret <vscale x 4 x i8> %a
 }
 declare <vscale x 4 x i8> @llvm.ctlz.nxv4i8(<vscale x 4 x i8>, i1)
 
 define <vscale x 8 x i8> @ctlz_nxv8i8(<vscale x 8 x i8> %va) {
 ; CHECK-ZVE64X-LABEL: ctlz_nxv8i8:

 ; CHECK-ZVE64X-NEXT:    vand.vi v8, v8, 15
 ; CHECK-ZVE64X-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv8i8:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
 ; CHECK-F-NEXT:    vzext.vf2 v10, v8
 ; CHECK-F-NEXT:    vfwcvt.f.xu.v v12, v10
-; CHECK-F-NEXT:    vnsrl.wi v10, v12, 23
+; CHECK-F-NEXT:    vnsrl.wi v8, v12, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e8, m1, ta, ma
-; CHECK-F-NEXT:    vnsrl.wi v9, v10, 0
+; CHECK-F-NEXT:    vnsrl.wi v10, v8, 0
 ; CHECK-F-NEXT:    li a0, 134
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-F-NEXT:    vrsub.vx v8, v9, a0
-; CHECK-F-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-F-NEXT:    vrsub.vx v8, v10, a0
+; CHECK-F-NEXT:    li a0, 8
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv8i8:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
 ; CHECK-D-NEXT:    vzext.vf2 v10, v8
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v12, v10
-; CHECK-D-NEXT:    vnsrl.wi v10, v12, 23
+; CHECK-D-NEXT:    vnsrl.wi v8, v12, 23
 ; CHECK-D-NEXT:    vsetvli zero, zero, e8, m1, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v9, v10, 0
+; CHECK-D-NEXT:    vnsrl.wi v10, v8, 0
 ; CHECK-D-NEXT:    li a0, 134
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-D-NEXT:    vrsub.vx v8, v9, a0
-; CHECK-D-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-D-NEXT:    vrsub.vx v8, v10, a0
+; CHECK-D-NEXT:    li a0, 8
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 8 x i8> @llvm.ctlz.nxv8i8(<vscale x 8 x i8> %va, i1 false)
   ret <vscale x 8 x i8> %a
 }
 declare <vscale x 8 x i8> @llvm.ctlz.nxv8i8(<vscale x 8 x i8>, i1)
 
 define <vscale x 16 x i8> @ctlz_nxv16i8(<vscale x 16 x i8> %va) {
 ; CHECK-ZVE64X-LABEL: ctlz_nxv16i8:

 ; CHECK-ZVE64X-NEXT:    vand.vi v8, v8, 15
 ; CHECK-ZVE64X-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv16i8:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
 ; CHECK-F-NEXT:    vzext.vf2 v12, v8
 ; CHECK-F-NEXT:    vfwcvt.f.xu.v v16, v12
-; CHECK-F-NEXT:    vnsrl.wi v12, v16, 23
+; CHECK-F-NEXT:    vnsrl.wi v8, v16, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e8, m2, ta, ma
-; CHECK-F-NEXT:    vnsrl.wi v10, v12, 0
+; CHECK-F-NEXT:    vnsrl.wi v12, v8, 0
 ; CHECK-F-NEXT:    li a0, 134
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-F-NEXT:    vrsub.vx v8, v10, a0
-; CHECK-F-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-F-NEXT:    vrsub.vx v8, v12, a0
+; CHECK-F-NEXT:    li a0, 8
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv16i8:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
 ; CHECK-D-NEXT:    vzext.vf2 v12, v8
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v16, v12
-; CHECK-D-NEXT:    vnsrl.wi v12, v16, 23
+; CHECK-D-NEXT:    vnsrl.wi v8, v16, 23
 ; CHECK-D-NEXT:    vsetvli zero, zero, e8, m2, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v10, v12, 0
+; CHECK-D-NEXT:    vnsrl.wi v12, v8, 0
 ; CHECK-D-NEXT:    li a0, 134
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
-; CHECK-D-NEXT:    vrsub.vx v8, v10, a0
-; CHECK-D-NEXT:    vmerge.vim v8, v8, 8, v0
+; CHECK-D-NEXT:    vrsub.vx v8, v12, a0
+; CHECK-D-NEXT:    li a0, 8
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 16 x i8> @llvm.ctlz.nxv16i8(<vscale x 16 x i8> %va, i1 false)
   ret <vscale x 16 x i8> %a
 }
 declare <vscale x 16 x i8> @llvm.ctlz.nxv16i8(<vscale x 16 x i8>, i1)
 
 define <vscale x 32 x i8> @ctlz_nxv32i8(<vscale x 32 x i8> %va) {
 ; CHECK-LABEL: ctlz_nxv32i8:

 ; RV64I-NEXT:    vmul.vx v8, v8, a0
 ; RV64I-NEXT:    vsrl.vi v8, v8, 8
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv1i16:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
 ; CHECK-F-NEXT:    vfwcvt.f.xu.v v9, v8
-; CHECK-F-NEXT:    vnsrl.wi v9, v9, 23
+; CHECK-F-NEXT:    vnsrl.wi v8, v9, 23
 ; CHECK-F-NEXT:    li a0, 142
-; CHECK-F-NEXT:    vrsub.vx v9, v9, a0
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-F-NEXT:    li a0, 16
-; CHECK-F-NEXT:    vmerge.vxm v8, v9, a0, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv1i16:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v9, v8
-; CHECK-D-NEXT:    vnsrl.wi v9, v9, 23
+; CHECK-D-NEXT:    vnsrl.wi v8, v9, 23
 ; CHECK-D-NEXT:    li a0, 142
-; CHECK-D-NEXT:    vrsub.vx v9, v9, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-D-NEXT:    li a0, 16
-; CHECK-D-NEXT:    vmerge.vxm v8, v9, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 1 x i16> @llvm.ctlz.nxv1i16(<vscale x 1 x i16> %va, i1 false)
   ret <vscale x 1 x i16> %a
 }
 declare <vscale x 1 x i16> @llvm.ctlz.nxv1i16(<vscale x 1 x i16>, i1)
 
 define <vscale x 2 x i16> @ctlz_nxv2i16(<vscale x 2 x i16> %va) {
 ; RV32I-LABEL: ctlz_nxv2i16:

 ; RV64I-NEXT:    vmul.vx v8, v8, a0
 ; RV64I-NEXT:    vsrl.vi v8, v8, 8
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv2i16:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
 ; CHECK-F-NEXT:    vfwcvt.f.xu.v v9, v8
-; CHECK-F-NEXT:    vnsrl.wi v9, v9, 23
+; CHECK-F-NEXT:    vnsrl.wi v8, v9, 23
 ; CHECK-F-NEXT:    li a0, 142
-; CHECK-F-NEXT:    vrsub.vx v9, v9, a0
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-F-NEXT:    li a0, 16
-; CHECK-F-NEXT:    vmerge.vxm v8, v9, a0, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv2i16:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v9, v8
-; CHECK-D-NEXT:    vnsrl.wi v9, v9, 23
+; CHECK-D-NEXT:    vnsrl.wi v8, v9, 23
 ; CHECK-D-NEXT:    li a0, 142
-; CHECK-D-NEXT:    vrsub.vx v9, v9, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-D-NEXT:    li a0, 16
-; CHECK-D-NEXT:    vmerge.vxm v8, v9, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 2 x i16> @llvm.ctlz.nxv2i16(<vscale x 2 x i16> %va, i1 false)
   ret <vscale x 2 x i16> %a
 }
 declare <vscale x 2 x i16> @llvm.ctlz.nxv2i16(<vscale x 2 x i16>, i1)
 
 define <vscale x 4 x i16> @ctlz_nxv4i16(<vscale x 4 x i16> %va) {
 ; RV32I-LABEL: ctlz_nxv4i16:

 ; RV64I-NEXT:    vmul.vx v8, v8, a0
 ; RV64I-NEXT:    vsrl.vi v8, v8, 8
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv4i16:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
 ; CHECK-F-NEXT:    vfwcvt.f.xu.v v10, v8
-; CHECK-F-NEXT:    vnsrl.wi v9, v10, 23
+; CHECK-F-NEXT:    vnsrl.wi v8, v10, 23
 ; CHECK-F-NEXT:    li a0, 142
-; CHECK-F-NEXT:    vrsub.vx v9, v9, a0
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-F-NEXT:    li a0, 16
-; CHECK-F-NEXT:    vmerge.vxm v8, v9, a0, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv4i16:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v10, v8
-; CHECK-D-NEXT:    vnsrl.wi v9, v10, 23
+; CHECK-D-NEXT:    vnsrl.wi v8, v10, 23
 ; CHECK-D-NEXT:    li a0, 142
-; CHECK-D-NEXT:    vrsub.vx v9, v9, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-D-NEXT:    li a0, 16
-; CHECK-D-NEXT:    vmerge.vxm v8, v9, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 4 x i16> @llvm.ctlz.nxv4i16(<vscale x 4 x i16> %va, i1 false)
   ret <vscale x 4 x i16> %a
 }
 declare <vscale x 4 x i16> @llvm.ctlz.nxv4i16(<vscale x 4 x i16>, i1)
 
 define <vscale x 8 x i16> @ctlz_nxv8i16(<vscale x 8 x i16> %va) {
 ; RV32I-LABEL: ctlz_nxv8i16:

 ; RV64I-NEXT:    vmul.vx v8, v8, a0
 ; RV64I-NEXT:    vsrl.vi v8, v8, 8
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv8i16:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
 ; CHECK-F-NEXT:    vfwcvt.f.xu.v v12, v8
-; CHECK-F-NEXT:    vnsrl.wi v10, v12, 23
+; CHECK-F-NEXT:    vnsrl.wi v8, v12, 23
 ; CHECK-F-NEXT:    li a0, 142
-; CHECK-F-NEXT:    vrsub.vx v10, v10, a0
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-F-NEXT:    li a0, 16
-; CHECK-F-NEXT:    vmerge.vxm v8, v10, a0, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv8i16:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v12, v8
-; CHECK-D-NEXT:    vnsrl.wi v10, v12, 23
+; CHECK-D-NEXT:    vnsrl.wi v8, v12, 23
 ; CHECK-D-NEXT:    li a0, 142
-; CHECK-D-NEXT:    vrsub.vx v10, v10, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-D-NEXT:    li a0, 16
-; CHECK-D-NEXT:    vmerge.vxm v8, v10, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 8 x i16> @llvm.ctlz.nxv8i16(<vscale x 8 x i16> %va, i1 false)
   ret <vscale x 8 x i16> %a
 }
 declare <vscale x 8 x i16> @llvm.ctlz.nxv8i16(<vscale x 8 x i16>, i1)
 
 define <vscale x 16 x i16> @ctlz_nxv16i16(<vscale x 16 x i16> %va) {
 ; RV32I-LABEL: ctlz_nxv16i16:

 ; RV64I-NEXT:    vmul.vx v8, v8, a0
 ; RV64I-NEXT:    vsrl.vi v8, v8, 8
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv16i16:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
 ; CHECK-F-NEXT:    vfwcvt.f.xu.v v16, v8
-; CHECK-F-NEXT:    vnsrl.wi v12, v16, 23
+; CHECK-F-NEXT:    vnsrl.wi v8, v16, 23
 ; CHECK-F-NEXT:    li a0, 142
-; CHECK-F-NEXT:    vrsub.vx v12, v12, a0
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-F-NEXT:    li a0, 16
-; CHECK-F-NEXT:    vmerge.vxm v8, v12, a0, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv16i16:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v16, v8
-; CHECK-D-NEXT:    vnsrl.wi v12, v16, 23
+; CHECK-D-NEXT:    vnsrl.wi v8, v16, 23
 ; CHECK-D-NEXT:    li a0, 142
-; CHECK-D-NEXT:    vrsub.vx v12, v12, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-D-NEXT:    li a0, 16
-; CHECK-D-NEXT:    vmerge.vxm v8, v12, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 16 x i16> @llvm.ctlz.nxv16i16(<vscale x 16 x i16> %va, i1 false)
   ret <vscale x 16 x i16> %a
 }
 declare <vscale x 16 x i16> @llvm.ctlz.nxv16i16(<vscale x 16 x i16>, i1)
 
 define <vscale x 32 x i16> @ctlz_nxv32i16(<vscale x 32 x i16> %va) {
 ; RV32-LABEL: ctlz_nxv32i16:

 ; RV64I-NEXT:    vsrl.vi v8, v8, 24
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv1i32:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, mf2, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
-; CHECK-F-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v9, v9, 23
+; CHECK-F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
+; CHECK-F-NEXT:    vsrl.vi v8, v8, 23
 ; CHECK-F-NEXT:    li a1, 158
-; CHECK-F-NEXT:    vrsub.vx v9, v9, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-F-NEXT:    li a1, 32
-; CHECK-F-NEXT:    vmerge.vxm v8, v9, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv1i32:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e32, mf2, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v9, v8
 ; CHECK-D-NEXT:    li a0, 52
 ; CHECK-D-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
-; CHECK-D-NEXT:    vsrl.vx v9, v9, a0
+; CHECK-D-NEXT:    vsrl.vx v8, v9, a0
 ; CHECK-D-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v9, v9, 0
+; CHECK-D-NEXT:    vnsrl.wi v8, v8, 0
 ; CHECK-D-NEXT:    li a0, 1054
-; CHECK-D-NEXT:    vrsub.vx v9, v9, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a0
 ; CHECK-D-NEXT:    li a0, 32
-; CHECK-D-NEXT:    vmerge.vxm v8, v9, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 1 x i32> @llvm.ctlz.nxv1i32(<vscale x 1 x i32> %va, i1 false)
   ret <vscale x 1 x i32> %a
 }
 declare <vscale x 1 x i32> @llvm.ctlz.nxv1i32(<vscale x 1 x i32>, i1)
 
 define <vscale x 2 x i32> @ctlz_nxv2i32(<vscale x 2 x i32> %va) {
 ; RV32I-LABEL: ctlz_nxv2i32:

 ; RV64I-NEXT:    vsrl.vi v8, v8, 24
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv2i32:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, m1, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
-; CHECK-F-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v9, v9, 23
+; CHECK-F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
+; CHECK-F-NEXT:    vsrl.vi v8, v8, 23
 ; CHECK-F-NEXT:    li a1, 158
-; CHECK-F-NEXT:    vrsub.vx v9, v9, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-F-NEXT:    li a1, 32
-; CHECK-F-NEXT:    vmerge.vxm v8, v9, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv2i32:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e32, m1, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v10, v8
 ; CHECK-D-NEXT:    li a0, 52
 ; CHECK-D-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
-; CHECK-D-NEXT:    vsrl.vx v10, v10, a0
+; CHECK-D-NEXT:    vsrl.vx v8, v10, a0
 ; CHECK-D-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v9, v10, 0
+; CHECK-D-NEXT:    vnsrl.wi v10, v8, 0
 ; CHECK-D-NEXT:    li a0, 1054
-; CHECK-D-NEXT:    vrsub.vx v9, v9, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v10, a0
 ; CHECK-D-NEXT:    li a0, 32
-; CHECK-D-NEXT:    vmerge.vxm v8, v9, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 2 x i32> @llvm.ctlz.nxv2i32(<vscale x 2 x i32> %va, i1 false)
   ret <vscale x 2 x i32> %a
 }
 declare <vscale x 2 x i32> @llvm.ctlz.nxv2i32(<vscale x 2 x i32>, i1)
 
 define <vscale x 4 x i32> @ctlz_nxv4i32(<vscale x 4 x i32> %va) {
 ; RV32I-LABEL: ctlz_nxv4i32:

 ; RV64I-NEXT:    vsrl.vi v8, v8, 24
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv4i32:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
-; CHECK-F-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v10, v10, 23
+; CHECK-F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
+; CHECK-F-NEXT:    vsrl.vi v8, v8, 23
 ; CHECK-F-NEXT:    li a1, 158
-; CHECK-F-NEXT:    vrsub.vx v10, v10, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-F-NEXT:    li a1, 32
-; CHECK-F-NEXT:    vmerge.vxm v8, v10, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv4i32:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v12, v8
 ; CHECK-D-NEXT:    li a0, 52
 ; CHECK-D-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
-; CHECK-D-NEXT:    vsrl.vx v12, v12, a0
+; CHECK-D-NEXT:    vsrl.vx v8, v12, a0
 ; CHECK-D-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v10, v12, 0
+; CHECK-D-NEXT:    vnsrl.wi v12, v8, 0
 ; CHECK-D-NEXT:    li a0, 1054
-; CHECK-D-NEXT:    vrsub.vx v10, v10, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v12, a0
 ; CHECK-D-NEXT:    li a0, 32
-; CHECK-D-NEXT:    vmerge.vxm v8, v10, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 4 x i32> @llvm.ctlz.nxv4i32(<vscale x 4 x i32> %va, i1 false)
   ret <vscale x 4 x i32> %a
 }
 declare <vscale x 4 x i32> @llvm.ctlz.nxv4i32(<vscale x 4 x i32>, i1)
 
 define <vscale x 8 x i32> @ctlz_nxv8i32(<vscale x 8 x i32> %va) {
 ; RV32I-LABEL: ctlz_nxv8i32:

 ; RV64I-NEXT:    vsrl.vi v8, v8, 24
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv8i32:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, m4, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
-; CHECK-F-NEXT:    vfcvt.f.xu.v v12, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v12, v12, 23
+; CHECK-F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
+; CHECK-F-NEXT:    vsrl.vi v8, v8, 23
 ; CHECK-F-NEXT:    li a1, 158
-; CHECK-F-NEXT:    vrsub.vx v12, v12, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-F-NEXT:    li a1, 32
-; CHECK-F-NEXT:    vmerge.vxm v8, v12, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv8i32:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e32, m4, ta, ma
 ; CHECK-D-NEXT:    vfwcvt.f.xu.v v16, v8
 ; CHECK-D-NEXT:    li a0, 52
 ; CHECK-D-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
-; CHECK-D-NEXT:    vsrl.vx v16, v16, a0
+; CHECK-D-NEXT:    vsrl.vx v8, v16, a0
 ; CHECK-D-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
-; CHECK-D-NEXT:    vnsrl.wi v12, v16, 0
+; CHECK-D-NEXT:    vnsrl.wi v16, v8, 0
 ; CHECK-D-NEXT:    li a0, 1054
-; CHECK-D-NEXT:    vrsub.vx v12, v12, a0
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v16, a0
 ; CHECK-D-NEXT:    li a0, 32
-; CHECK-D-NEXT:    vmerge.vxm v8, v12, a0, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 8 x i32> @llvm.ctlz.nxv8i32(<vscale x 8 x i32> %va, i1 false)
   ret <vscale x 8 x i32> %a
 }
 declare <vscale x 8 x i32> @llvm.ctlz.nxv8i32(<vscale x 8 x i32>, i1)
 
 define <vscale x 16 x i32> @ctlz_nxv16i32(<vscale x 16 x i32> %va) {
 ; RV32I-LABEL: ctlz_nxv16i32:

 ; RV64I-NEXT:    vsrl.vi v8, v8, 24
 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv16i32:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, m8, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
-; CHECK-F-NEXT:    vfcvt.f.xu.v v16, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v16, v16, 23
+; CHECK-F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
+; CHECK-F-NEXT:    vsrl.vi v8, v8, 23
 ; CHECK-F-NEXT:    li a1, 158
-; CHECK-F-NEXT:    vrsub.vx v16, v16, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-F-NEXT:    li a1, 32
-; CHECK-F-NEXT:    vmerge.vxm v8, v16, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv16i32:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e32, m8, ta, ma
 ; CHECK-D-NEXT:    vmset.m v0
 ; CHECK-D-NEXT:    fsrmi a0, 1
-; CHECK-D-NEXT:    vfcvt.f.xu.v v16, v8, v0.t
-; CHECK-D-NEXT:    vsrl.vi v16, v16, 23
+; CHECK-D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
+; CHECK-D-NEXT:    vsrl.vi v8, v8, 23
 ; CHECK-D-NEXT:    li a1, 158
-; CHECK-D-NEXT:    vrsub.vx v16, v16, a1
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 32
-; CHECK-D-NEXT:    vmerge.vxm v8, v16, a1, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-D-NEXT:    fsrm a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 16 x i32> @llvm.ctlz.nxv16i32(<vscale x 16 x i32> %va, i1 false)
   ret <vscale x 16 x i32> %a
 }
 declare <vscale x 16 x i32> @llvm.ctlz.nxv16i32(<vscale x 16 x i32>, i1)
 
 define <vscale x 1 x i64> @ctlz_nxv1i64(<vscale x 1 x i64> %va) {

 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv1i64:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, mf2, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
 ; CHECK-F-NEXT:    vfncvt.f.xu.w v9, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v9, v9, 23
+; CHECK-F-NEXT:    vsrl.vi v8, v9, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
-; CHECK-F-NEXT:    vzext.vf2 v10, v9
+; CHECK-F-NEXT:    vzext.vf2 v9, v8
 ; CHECK-F-NEXT:    li a1, 190
-; CHECK-F-NEXT:    vrsub.vx v9, v10, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v9, a1
 ; CHECK-F-NEXT:    li a1, 64
-; CHECK-F-NEXT:    vmerge.vxm v8, v9, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv1i64:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
 ; CHECK-D-NEXT:    vmset.m v0
 ; CHECK-D-NEXT:    fsrmi a0, 1
-; CHECK-D-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
+; CHECK-D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; CHECK-D-NEXT:    li a1, 52
-; CHECK-D-NEXT:    vsrl.vx v9, v9, a1
+; CHECK-D-NEXT:    vsrl.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 1086
-; CHECK-D-NEXT:    vrsub.vx v9, v9, a1
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 64
-; CHECK-D-NEXT:    vmerge.vxm v8, v9, a1, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-D-NEXT:    fsrm a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 1 x i64> @llvm.ctlz.nxv1i64(<vscale x 1 x i64> %va, i1 false)
   ret <vscale x 1 x i64> %a
 }
 declare <vscale x 1 x i64> @llvm.ctlz.nxv1i64(<vscale x 1 x i64>, i1)
 
 define <vscale x 2 x i64> @ctlz_nxv2i64(<vscale x 2 x i64> %va) {

 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv2i64:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, m1, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
 ; CHECK-F-NEXT:    vfncvt.f.xu.w v10, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v10, v10, 23
+; CHECK-F-NEXT:    vsrl.vi v8, v10, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
-; CHECK-F-NEXT:    vzext.vf2 v12, v10
+; CHECK-F-NEXT:    vzext.vf2 v10, v8
 ; CHECK-F-NEXT:    li a1, 190
-; CHECK-F-NEXT:    vrsub.vx v10, v12, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v10, a1
 ; CHECK-F-NEXT:    li a1, 64
-; CHECK-F-NEXT:    vmerge.vxm v8, v10, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv2i64:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
 ; CHECK-D-NEXT:    vmset.m v0
 ; CHECK-D-NEXT:    fsrmi a0, 1
-; CHECK-D-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
+; CHECK-D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; CHECK-D-NEXT:    li a1, 52
-; CHECK-D-NEXT:    vsrl.vx v10, v10, a1
+; CHECK-D-NEXT:    vsrl.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 1086
-; CHECK-D-NEXT:    vrsub.vx v10, v10, a1
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 64
-; CHECK-D-NEXT:    vmerge.vxm v8, v10, a1, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-D-NEXT:    fsrm a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 2 x i64> @llvm.ctlz.nxv2i64(<vscale x 2 x i64> %va, i1 false)
   ret <vscale x 2 x i64> %a
 }
 declare <vscale x 2 x i64> @llvm.ctlz.nxv2i64(<vscale x 2 x i64>, i1)
 
 define <vscale x 4 x i64> @ctlz_nxv4i64(<vscale x 4 x i64> %va) {

 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv4i64:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
 ; CHECK-F-NEXT:    vfncvt.f.xu.w v12, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v12, v12, 23
+; CHECK-F-NEXT:    vsrl.vi v8, v12, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
-; CHECK-F-NEXT:    vzext.vf2 v16, v12
+; CHECK-F-NEXT:    vzext.vf2 v12, v8
 ; CHECK-F-NEXT:    li a1, 190
-; CHECK-F-NEXT:    vrsub.vx v12, v16, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v12, a1
 ; CHECK-F-NEXT:    li a1, 64
-; CHECK-F-NEXT:    vmerge.vxm v8, v12, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv4i64:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e64, m4, ta, ma
 ; CHECK-D-NEXT:    vmset.m v0
 ; CHECK-D-NEXT:    fsrmi a0, 1
-; CHECK-D-NEXT:    vfcvt.f.xu.v v12, v8, v0.t
+; CHECK-D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; CHECK-D-NEXT:    li a1, 52
-; CHECK-D-NEXT:    vsrl.vx v12, v12, a1
+; CHECK-D-NEXT:    vsrl.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 1086
-; CHECK-D-NEXT:    vrsub.vx v12, v12, a1
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 64
-; CHECK-D-NEXT:    vmerge.vxm v8, v12, a1, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-D-NEXT:    fsrm a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 4 x i64> @llvm.ctlz.nxv4i64(<vscale x 4 x i64> %va, i1 false)
   ret <vscale x 4 x i64> %a
 }
 declare <vscale x 4 x i64> @llvm.ctlz.nxv4i64(<vscale x 4 x i64>, i1)
 
 define <vscale x 8 x i64> @ctlz_nxv8i64(<vscale x 8 x i64> %va) {

 ; RV64I-NEXT:    ret
 ;
 ; CHECK-F-LABEL: ctlz_nxv8i64:
 ; CHECK-F:       # %bb.0:
 ; CHECK-F-NEXT:    vsetvli a0, zero, e32, m4, ta, ma
 ; CHECK-F-NEXT:    vmset.m v0
 ; CHECK-F-NEXT:    fsrmi a0, 1
 ; CHECK-F-NEXT:    vfncvt.f.xu.w v16, v8, v0.t
-; CHECK-F-NEXT:    vsrl.vi v16, v16, 23
+; CHECK-F-NEXT:    vsrl.vi v8, v16, 23
 ; CHECK-F-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
-; CHECK-F-NEXT:    vzext.vf2 v24, v16
+; CHECK-F-NEXT:    vzext.vf2 v16, v8
 ; CHECK-F-NEXT:    li a1, 190
-; CHECK-F-NEXT:    vrsub.vx v16, v24, a1
-; CHECK-F-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-F-NEXT:    vrsub.vx v8, v16, a1
 ; CHECK-F-NEXT:    li a1, 64
-; CHECK-F-NEXT:    vmerge.vxm v8, v16, a1, v0
+; CHECK-F-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-F-NEXT:    fsrm a0
 ; CHECK-F-NEXT:    ret
 ;
 ; CHECK-D-LABEL: ctlz_nxv8i64:
 ; CHECK-D:       # %bb.0:
 ; CHECK-D-NEXT:    vsetvli a0, zero, e64, m8, ta, ma
 ; CHECK-D-NEXT:    vmset.m v0
 ; CHECK-D-NEXT:    fsrmi a0, 1
-; CHECK-D-NEXT:    vfcvt.f.xu.v v16, v8, v0.t
+; CHECK-D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; CHECK-D-NEXT:    li a1, 52
-; CHECK-D-NEXT:    vsrl.vx v16, v16, a1
+; CHECK-D-NEXT:    vsrl.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 1086
-; CHECK-D-NEXT:    vrsub.vx v16, v16, a1
-; CHECK-D-NEXT:    vmseq.vi v0, v8, 0
+; CHECK-D-NEXT:    vrsub.vx v8, v8, a1
 ; CHECK-D-NEXT:    li a1, 64
-; CHECK-D-NEXT:    vmerge.vxm v8, v16, a1, v0
+; CHECK-D-NEXT:    vminu.vx v8, v8, a1
 ; CHECK-D-NEXT:    fsrm a0
 ; CHECK-D-NEXT:    ret
   %a = call <vscale x 8 x i64> @llvm.ctlz.nxv8i64(<vscale x 8 x i64> %va, i1 false)
   ret <vscale x 8 x i64> %a
 }
 declare <vscale x 8 x i64> @llvm.ctlz.nxv8i64(<vscale x 8 x i64>, i1)
 
 define <vscale x 1 x i8> @ctlz_zero_undef_nxv1i8(<vscale x 1 x i8> %va) {

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

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+m,+zve64x -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2,LMULMAX2-RV32,LMULMAX2-RV32I
 ; RUN: llc -mtriple=riscv64 -mattr=+m,+zve64x -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2,LMULMAX2-RV64,LMULMAX2-RV64I
 ; RUN: llc -mtriple=riscv32 -mattr=+m,+zve64x -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=+m,+zve64x -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
 ; RUN: llc -mtriple=riscv32 -mattr=+m,+zve64f,+f -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2,LMULMAX2-RV32,LMULMAX2-RV32F
 ; RUN: llc -mtriple=riscv64 -mattr=+m,+zve64f,+f -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2,LMULMAX2-RV64,LMULMAX2-RV64F
 ; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+d -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2,LMULMAX2-RV32,LMULMAX2-RV32D
 ; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+d -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2,LMULMAX2-RV64,LMULMAX2-RV64D
 ; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+d -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=+m,+v,+d -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
-; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+d -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=8 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=LMULMAX8,LMULMAX8-RV32
-; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+d -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=8 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=LMULMAX8,LMULMAX8-RV64
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+d -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=8 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=LMULMAX8
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+d -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=8 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=LMULMAX8
 
 define void @ctlz_v16i8(ptr %x, ptr %y) nounwind {
 ; CHECK-LABEL: ctlz_v16i8:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
 ; CHECK-NEXT:    vle8.v v8, (a0)
 ; CHECK-NEXT:    vsrl.vi v9, v8, 1
 ; CHECK-NEXT:    vor.vv v8, v8, v9

 ; CHECK-NEXT:    ret
 ;
 ; LMULMAX8-LABEL: ctlz_v16i8:
 ; LMULMAX8:       # %bb.0:
 ; LMULMAX8-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
 ; LMULMAX8-NEXT:    vle8.v v8, (a0)
 ; LMULMAX8-NEXT:    vzext.vf2 v10, v8
 ; LMULMAX8-NEXT:    vfwcvt.f.xu.v v12, v10
-; LMULMAX8-NEXT:    vnsrl.wi v10, v12, 23
+; LMULMAX8-NEXT:    vnsrl.wi v8, v12, 23
 ; LMULMAX8-NEXT:    vsetvli zero, zero, e8, m1, ta, ma
-; LMULMAX8-NEXT:    vnsrl.wi v9, v10, 0
+; LMULMAX8-NEXT:    vnsrl.wi v10, v8, 0
 ; LMULMAX8-NEXT:    li a1, 134
-; LMULMAX8-NEXT:    vmseq.vi v0, v8, 0
-; LMULMAX8-NEXT:    vrsub.vx v8, v9, a1
-; LMULMAX8-NEXT:    vmerge.vim v8, v8, 8, v0
+; LMULMAX8-NEXT:    vrsub.vx v8, v10, a1
+; LMULMAX8-NEXT:    li a1, 8
+; LMULMAX8-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX8-NEXT:    vse8.v v8, (a0)
 ; LMULMAX8-NEXT:    ret
   %a = load <16 x i8>, ptr %x
   %b = load <16 x i8>, ptr %y
   %c = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %a, i1 false)
   store <16 x i8> %c, ptr %x
   ret void
 }

 ; LMULMAX1-RV64-NEXT:    vse16.v v8, (a0)
 ; LMULMAX1-RV64-NEXT:    ret
 ;
 ; LMULMAX2-RV32F-LABEL: ctlz_v8i16:
 ; LMULMAX2-RV32F:       # %bb.0:
 ; LMULMAX2-RV32F-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
 ; LMULMAX2-RV32F-NEXT:    vle16.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    vfwcvt.f.xu.v v10, v8
-; LMULMAX2-RV32F-NEXT:    vnsrl.wi v9, v10, 23
+; LMULMAX2-RV32F-NEXT:    vnsrl.wi v8, v10, 23
 ; LMULMAX2-RV32F-NEXT:    li a1, 142
-; LMULMAX2-RV32F-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV32F-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV32F-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV32F-NEXT:    li a1, 16
-; LMULMAX2-RV32F-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV32F-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV32F-NEXT:    vse16.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    ret
 ;
 ; LMULMAX2-RV64F-LABEL: ctlz_v8i16:
 ; LMULMAX2-RV64F:       # %bb.0:
 ; LMULMAX2-RV64F-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
 ; LMULMAX2-RV64F-NEXT:    vle16.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    vfwcvt.f.xu.v v10, v8
-; LMULMAX2-RV64F-NEXT:    vnsrl.wi v9, v10, 23
+; LMULMAX2-RV64F-NEXT:    vnsrl.wi v8, v10, 23
 ; LMULMAX2-RV64F-NEXT:    li a1, 142
-; LMULMAX2-RV64F-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV64F-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64F-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV64F-NEXT:    li a1, 16
-; LMULMAX2-RV64F-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV64F-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV64F-NEXT:    vse16.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    ret
 ;
 ; LMULMAX2-RV32D-LABEL: ctlz_v8i16:
 ; LMULMAX2-RV32D:       # %bb.0:
 ; LMULMAX2-RV32D-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
 ; LMULMAX2-RV32D-NEXT:    vle16.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    vfwcvt.f.xu.v v10, v8
-; LMULMAX2-RV32D-NEXT:    vnsrl.wi v9, v10, 23
+; LMULMAX2-RV32D-NEXT:    vnsrl.wi v8, v10, 23
 ; LMULMAX2-RV32D-NEXT:    li a1, 142
-; LMULMAX2-RV32D-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV32D-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV32D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 16
-; LMULMAX2-RV32D-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV32D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    vse16.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    ret
 ;
 ; LMULMAX2-RV64D-LABEL: ctlz_v8i16:
 ; LMULMAX2-RV64D:       # %bb.0:
 ; LMULMAX2-RV64D-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
 ; LMULMAX2-RV64D-NEXT:    vle16.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    vfwcvt.f.xu.v v10, v8
-; LMULMAX2-RV64D-NEXT:    vnsrl.wi v9, v10, 23
+; LMULMAX2-RV64D-NEXT:    vnsrl.wi v8, v10, 23
 ; LMULMAX2-RV64D-NEXT:    li a1, 142
-; LMULMAX2-RV64D-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV64D-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 16
-; LMULMAX2-RV64D-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV64D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    vse16.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    ret
 ;
 ; LMULMAX8-LABEL: ctlz_v8i16:
 ; LMULMAX8:       # %bb.0:
 ; LMULMAX8-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
 ; LMULMAX8-NEXT:    vle16.v v8, (a0)
 ; LMULMAX8-NEXT:    vfwcvt.f.xu.v v10, v8
-; LMULMAX8-NEXT:    vnsrl.wi v9, v10, 23
+; LMULMAX8-NEXT:    vnsrl.wi v8, v10, 23
 ; LMULMAX8-NEXT:    li a1, 142
-; LMULMAX8-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX8-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX8-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX8-NEXT:    li a1, 16
-; LMULMAX8-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX8-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX8-NEXT:    vse16.v v8, (a0)
 ; LMULMAX8-NEXT:    ret
   %a = load <8 x i16>, ptr %x
   %b = load <8 x i16>, ptr %y
   %c = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %a, i1 false)
   store <8 x i16> %c, ptr %x
   ret void
 }

 ; LMULMAX2-RV64I-NEXT:    ret
 ;
 ; LMULMAX2-RV32F-LABEL: ctlz_v4i32:
 ; LMULMAX2-RV32F:       # %bb.0:
 ; LMULMAX2-RV32F-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
 ; LMULMAX2-RV32F-NEXT:    vle32.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    vmset.m v0
 ; LMULMAX2-RV32F-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV32F-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
+; LMULMAX2-RV32F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV32F-NEXT:    fsrm a1
-; LMULMAX2-RV32F-NEXT:    vsrl.vi v9, v9, 23
+; LMULMAX2-RV32F-NEXT:    vsrl.vi v8, v8, 23
 ; LMULMAX2-RV32F-NEXT:    li a1, 158
-; LMULMAX2-RV32F-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV32F-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV32F-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV32F-NEXT:    li a1, 32
-; LMULMAX2-RV32F-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV32F-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV32F-NEXT:    vse32.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    ret
 ;
 ; LMULMAX2-RV64F-LABEL: ctlz_v4i32:
 ; LMULMAX2-RV64F:       # %bb.0:
 ; LMULMAX2-RV64F-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
 ; LMULMAX2-RV64F-NEXT:    vle32.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    vmset.m v0
 ; LMULMAX2-RV64F-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV64F-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
+; LMULMAX2-RV64F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV64F-NEXT:    fsrm a1
-; LMULMAX2-RV64F-NEXT:    vsrl.vi v9, v9, 23
+; LMULMAX2-RV64F-NEXT:    vsrl.vi v8, v8, 23
 ; LMULMAX2-RV64F-NEXT:    li a1, 158
-; LMULMAX2-RV64F-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV64F-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64F-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV64F-NEXT:    li a1, 32
-; LMULMAX2-RV64F-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV64F-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV64F-NEXT:    vse32.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    ret
 ;
 ; LMULMAX2-RV32D-LABEL: ctlz_v4i32:
 ; LMULMAX2-RV32D:       # %bb.0:
 ; LMULMAX2-RV32D-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
 ; LMULMAX2-RV32D-NEXT:    vle32.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    vfwcvt.f.xu.v v10, v8
 ; LMULMAX2-RV32D-NEXT:    li a1, 52
-; LMULMAX2-RV32D-NEXT:    vnsrl.wx v9, v10, a1
+; LMULMAX2-RV32D-NEXT:    vnsrl.wx v8, v10, a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 1054
-; LMULMAX2-RV32D-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV32D-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV32D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 32
-; LMULMAX2-RV32D-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV32D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    vse32.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    ret
 ;
 ; LMULMAX2-RV64D-LABEL: ctlz_v4i32:
 ; LMULMAX2-RV64D:       # %bb.0:
 ; LMULMAX2-RV64D-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
 ; LMULMAX2-RV64D-NEXT:    vle32.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    vfwcvt.f.xu.v v10, v8
 ; LMULMAX2-RV64D-NEXT:    li a1, 52
-; LMULMAX2-RV64D-NEXT:    vnsrl.wx v9, v10, a1
+; LMULMAX2-RV64D-NEXT:    vnsrl.wx v8, v10, a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 1054
-; LMULMAX2-RV64D-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV64D-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 32
-; LMULMAX2-RV64D-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV64D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    vse32.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    ret
 ;
 ; LMULMAX8-LABEL: ctlz_v4i32:
 ; LMULMAX8:       # %bb.0:
 ; LMULMAX8-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
 ; LMULMAX8-NEXT:    vle32.v v8, (a0)
 ; LMULMAX8-NEXT:    vfwcvt.f.xu.v v10, v8
 ; LMULMAX8-NEXT:    li a1, 52
-; LMULMAX8-NEXT:    vnsrl.wx v9, v10, a1
+; LMULMAX8-NEXT:    vnsrl.wx v8, v10, a1
 ; LMULMAX8-NEXT:    li a1, 1054
-; LMULMAX8-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX8-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX8-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX8-NEXT:    li a1, 32
-; LMULMAX8-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX8-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX8-NEXT:    vse32.v v8, (a0)
 ; LMULMAX8-NEXT:    ret
   %a = load <4 x i32>, ptr %x
   %b = load <4 x i32>, ptr %y
   %c = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %a, i1 false)
   store <4 x i32> %c, ptr %x
   ret void
 }

 ; LMULMAX2-RV32F-LABEL: ctlz_v2i64:
 ; LMULMAX2-RV32F:       # %bb.0:
 ; LMULMAX2-RV32F-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
 ; LMULMAX2-RV32F-NEXT:    vle64.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    vmset.m v0
 ; LMULMAX2-RV32F-NEXT:    fsrmi a1, 1
 ; LMULMAX2-RV32F-NEXT:    vfncvt.f.xu.w v9, v8, v0.t
 ; LMULMAX2-RV32F-NEXT:    fsrm a1
-; LMULMAX2-RV32F-NEXT:    vsrl.vi v9, v9, 23
+; LMULMAX2-RV32F-NEXT:    vsrl.vi v8, v9, 23
 ; LMULMAX2-RV32F-NEXT:    li a1, 190
 ; LMULMAX2-RV32F-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
-; LMULMAX2-RV32F-NEXT:    vmv.v.x v10, a1
+; LMULMAX2-RV32F-NEXT:    vmv.v.x v9, a1
 ; LMULMAX2-RV32F-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
-; LMULMAX2-RV32F-NEXT:    vwsubu.wv v10, v10, v9
-; LMULMAX2-RV32F-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
-; LMULMAX2-RV32F-NEXT:    vmv.v.i v9, 0
-; LMULMAX2-RV32F-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
-; LMULMAX2-RV32F-NEXT:    vmseq.vv v0, v8, v9
+; LMULMAX2-RV32F-NEXT:    vwsubu.wv v9, v9, v8
 ; LMULMAX2-RV32F-NEXT:    li a1, 64
-; LMULMAX2-RV32F-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX2-RV32F-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; LMULMAX2-RV32F-NEXT:    vminu.vx v8, v9, a1
 ; LMULMAX2-RV32F-NEXT:    vse64.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    ret
 ;
 ; LMULMAX2-RV64F-LABEL: ctlz_v2i64:
 ; LMULMAX2-RV64F:       # %bb.0:
 ; LMULMAX2-RV64F-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
 ; LMULMAX2-RV64F-NEXT:    vle64.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    vmset.m v0
 ; LMULMAX2-RV64F-NEXT:    fsrmi a1, 1
 ; LMULMAX2-RV64F-NEXT:    vfncvt.f.xu.w v9, v8, v0.t
 ; LMULMAX2-RV64F-NEXT:    fsrm a1
-; LMULMAX2-RV64F-NEXT:    vsrl.vi v9, v9, 23
+; LMULMAX2-RV64F-NEXT:    vsrl.vi v8, v9, 23
 ; LMULMAX2-RV64F-NEXT:    li a1, 190
-; LMULMAX2-RV64F-NEXT:    vmv.v.x v10, a1
-; LMULMAX2-RV64F-NEXT:    vwsubu.vv v11, v10, v9
-; LMULMAX2-RV64F-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
-; LMULMAX2-RV64F-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64F-NEXT:    vmv.v.x v9, a1
+; LMULMAX2-RV64F-NEXT:    vwsubu.vv v10, v9, v8
 ; LMULMAX2-RV64F-NEXT:    li a1, 64
-; LMULMAX2-RV64F-NEXT:    vmerge.vxm v8, v11, a1, v0
+; LMULMAX2-RV64F-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; LMULMAX2-RV64F-NEXT:    vminu.vx v8, v10, a1
 ; LMULMAX2-RV64F-NEXT:    vse64.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    ret
 ;
 ; LMULMAX2-RV32D-LABEL: ctlz_v2i64:
 ; LMULMAX2-RV32D:       # %bb.0:
 ; LMULMAX2-RV32D-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
 ; LMULMAX2-RV32D-NEXT:    vle64.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    vmset.m v0
 ; LMULMAX2-RV32D-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV32D-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
+; LMULMAX2-RV32D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV32D-NEXT:    fsrm a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 52
-; LMULMAX2-RV32D-NEXT:    vsrl.vx v9, v9, a1
+; LMULMAX2-RV32D-NEXT:    vsrl.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 1086
-; LMULMAX2-RV32D-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV32D-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
-; LMULMAX2-RV32D-NEXT:    vmv.v.i v10, 0
-; LMULMAX2-RV32D-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
-; LMULMAX2-RV32D-NEXT:    vmseq.vv v0, v8, v10
+; LMULMAX2-RV32D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 64
-; LMULMAX2-RV32D-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV32D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    vse64.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    ret
 ;
 ; LMULMAX2-RV64D-LABEL: ctlz_v2i64:
 ; LMULMAX2-RV64D:       # %bb.0:
 ; LMULMAX2-RV64D-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
 ; LMULMAX2-RV64D-NEXT:    vle64.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    vmset.m v0
 ; LMULMAX2-RV64D-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV64D-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
+; LMULMAX2-RV64D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV64D-NEXT:    fsrm a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 52
-; LMULMAX2-RV64D-NEXT:    vsrl.vx v9, v9, a1
+; LMULMAX2-RV64D-NEXT:    vsrl.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 1086
-; LMULMAX2-RV64D-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX2-RV64D-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 64
-; LMULMAX2-RV64D-NEXT:    vmerge.vxm v8, v9, a1, v0
+; LMULMAX2-RV64D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    vse64.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    ret
 ;
-; LMULMAX8-RV32-LABEL: ctlz_v2i64:
-; LMULMAX8-RV32:       # %bb.0:
-; LMULMAX8-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
-; LMULMAX8-RV32-NEXT:    vle64.v v8, (a0)
-; LMULMAX8-RV32-NEXT:    vmset.m v0
-; LMULMAX8-RV32-NEXT:    fsrmi a1, 1
-; LMULMAX8-RV32-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
-; LMULMAX8-RV32-NEXT:    fsrm a1
-; LMULMAX8-RV32-NEXT:    li a1, 52
-; LMULMAX8-RV32-NEXT:    vsrl.vx v9, v9, a1
-; LMULMAX8-RV32-NEXT:    li a1, 1086
-; LMULMAX8-RV32-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX8-RV32-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
-; LMULMAX8-RV32-NEXT:    vmv.v.i v10, 0
-; LMULMAX8-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
-; LMULMAX8-RV32-NEXT:    vmseq.vv v0, v8, v10
-; LMULMAX8-RV32-NEXT:    li a1, 64
-; LMULMAX8-RV32-NEXT:    vmerge.vxm v8, v9, a1, v0
-; LMULMAX8-RV32-NEXT:    vse64.v v8, (a0)
-; LMULMAX8-RV32-NEXT:    ret
-;
-; LMULMAX8-RV64-LABEL: ctlz_v2i64:
-; LMULMAX8-RV64:       # %bb.0:
-; LMULMAX8-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
-; LMULMAX8-RV64-NEXT:    vle64.v v8, (a0)
-; LMULMAX8-RV64-NEXT:    vmset.m v0
-; LMULMAX8-RV64-NEXT:    fsrmi a1, 1
-; LMULMAX8-RV64-NEXT:    vfcvt.f.xu.v v9, v8, v0.t
-; LMULMAX8-RV64-NEXT:    fsrm a1
-; LMULMAX8-RV64-NEXT:    li a1, 52
-; LMULMAX8-RV64-NEXT:    vsrl.vx v9, v9, a1
-; LMULMAX8-RV64-NEXT:    li a1, 1086
-; LMULMAX8-RV64-NEXT:    vrsub.vx v9, v9, a1
-; LMULMAX8-RV64-NEXT:    vmseq.vi v0, v8, 0
-; LMULMAX8-RV64-NEXT:    li a1, 64
-; LMULMAX8-RV64-NEXT:    vmerge.vxm v8, v9, a1, v0
-; LMULMAX8-RV64-NEXT:    vse64.v v8, (a0)
-; LMULMAX8-RV64-NEXT:    ret
+; LMULMAX8-LABEL: ctlz_v2i64:
+; LMULMAX8:       # %bb.0:
+; LMULMAX8-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
+; LMULMAX8-NEXT:    vle64.v v8, (a0)
+; LMULMAX8-NEXT:    vmset.m v0
+; LMULMAX8-NEXT:    fsrmi a1, 1
+; LMULMAX8-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
+; LMULMAX8-NEXT:    fsrm a1
+; LMULMAX8-NEXT:    li a1, 52
+; LMULMAX8-NEXT:    vsrl.vx v8, v8, a1
+; LMULMAX8-NEXT:    li a1, 1086
+; LMULMAX8-NEXT:    vrsub.vx v8, v8, a1
+; LMULMAX8-NEXT:    li a1, 64
+; LMULMAX8-NEXT:    vminu.vx v8, v8, a1
+; LMULMAX8-NEXT:    vse64.v v8, (a0)
+; LMULMAX8-NEXT:    ret
   %a = load <2 x i64>, ptr %x
   %b = load <2 x i64>, ptr %y
   %c = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %a, i1 false)
   store <2 x i64> %c, ptr %x
   ret void
 }
 declare <2 x i64> @llvm.ctlz.v2i64(<2 x i64>, i1)
 

 ;
 ; LMULMAX8-LABEL: ctlz_v32i8:
 ; LMULMAX8:       # %bb.0:
 ; LMULMAX8-NEXT:    li a1, 32
 ; LMULMAX8-NEXT:    vsetvli zero, a1, e16, m4, ta, ma
 ; LMULMAX8-NEXT:    vle8.v v8, (a0)
 ; LMULMAX8-NEXT:    vzext.vf2 v12, v8
 ; LMULMAX8-NEXT:    vfwcvt.f.xu.v v16, v12
-; LMULMAX8-NEXT:    vnsrl.wi v12, v16, 23
+; LMULMAX8-NEXT:    vnsrl.wi v8, v16, 23
 ; LMULMAX8-NEXT:    vsetvli zero, zero, e8, m2, ta, ma
-; LMULMAX8-NEXT:    vnsrl.wi v10, v12, 0
+; LMULMAX8-NEXT:    vnsrl.wi v12, v8, 0
 ; LMULMAX8-NEXT:    li a1, 134
-; LMULMAX8-NEXT:    vmseq.vi v0, v8, 0
-; LMULMAX8-NEXT:    vrsub.vx v8, v10, a1
-; LMULMAX8-NEXT:    vmerge.vim v8, v8, 8, v0
+; LMULMAX8-NEXT:    vrsub.vx v8, v12, a1
+; LMULMAX8-NEXT:    li a1, 8
+; LMULMAX8-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX8-NEXT:    vse8.v v8, (a0)
 ; LMULMAX8-NEXT:    ret
   %a = load <32 x i8>, ptr %x
   %b = load <32 x i8>, ptr %y
   %c = call <32 x i8> @llvm.ctlz.v32i8(<32 x i8> %a, i1 false)
   store <32 x i8> %c, ptr %x
   ret void
 }

 ; LMULMAX1-RV64-NEXT:    vse16.v v8, (a1)
 ; LMULMAX1-RV64-NEXT:    ret
 ;
 ; LMULMAX8-LABEL: ctlz_v16i16:
 ; LMULMAX8:       # %bb.0:
 ; LMULMAX8-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
 ; LMULMAX8-NEXT:    vle16.v v8, (a0)
 ; LMULMAX8-NEXT:    vfwcvt.f.xu.v v12, v8
-; LMULMAX8-NEXT:    vnsrl.wi v10, v12, 23
+; LMULMAX8-NEXT:    vnsrl.wi v8, v12, 23
 ; LMULMAX8-NEXT:    li a1, 142
-; LMULMAX8-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX8-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX8-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX8-NEXT:    li a1, 16
-; LMULMAX8-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX8-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX8-NEXT:    vse16.v v8, (a0)
 ; LMULMAX8-NEXT:    ret
   %a = load <16 x i16>, ptr %x
   %b = load <16 x i16>, ptr %y
   %c = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> %a, i1 false)
   store <16 x i16> %c, ptr %x
   ret void
 }

 ; LMULMAX2-RV64I-NEXT:    ret
 ;
 ; LMULMAX2-RV32F-LABEL: ctlz_v8i32:
 ; LMULMAX2-RV32F:       # %bb.0:
 ; LMULMAX2-RV32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; LMULMAX2-RV32F-NEXT:    vle32.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    vmset.m v0
 ; LMULMAX2-RV32F-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV32F-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
+; LMULMAX2-RV32F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV32F-NEXT:    fsrm a1
-; LMULMAX2-RV32F-NEXT:    vsrl.vi v10, v10, 23
+; LMULMAX2-RV32F-NEXT:    vsrl.vi v8, v8, 23
 ; LMULMAX2-RV32F-NEXT:    li a1, 158
-; LMULMAX2-RV32F-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX2-RV32F-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV32F-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV32F-NEXT:    li a1, 32
-; LMULMAX2-RV32F-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX2-RV32F-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV32F-NEXT:    vse32.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    ret
 ;
 ; LMULMAX2-RV64F-LABEL: ctlz_v8i32:
 ; LMULMAX2-RV64F:       # %bb.0:
 ; LMULMAX2-RV64F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; LMULMAX2-RV64F-NEXT:    vle32.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    vmset.m v0
 ; LMULMAX2-RV64F-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV64F-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
+; LMULMAX2-RV64F-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV64F-NEXT:    fsrm a1
-; LMULMAX2-RV64F-NEXT:    vsrl.vi v10, v10, 23
+; LMULMAX2-RV64F-NEXT:    vsrl.vi v8, v8, 23
 ; LMULMAX2-RV64F-NEXT:    li a1, 158
-; LMULMAX2-RV64F-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX2-RV64F-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64F-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV64F-NEXT:    li a1, 32
-; LMULMAX2-RV64F-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX2-RV64F-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV64F-NEXT:    vse32.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    ret
 ;
 ; LMULMAX2-RV32D-LABEL: ctlz_v8i32:
 ; LMULMAX2-RV32D:       # %bb.0:
 ; LMULMAX2-RV32D-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; LMULMAX2-RV32D-NEXT:    vle32.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    vmset.m v0
 ; LMULMAX2-RV32D-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV32D-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
+; LMULMAX2-RV32D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV32D-NEXT:    fsrm a1
-; LMULMAX2-RV32D-NEXT:    vsrl.vi v10, v10, 23
+; LMULMAX2-RV32D-NEXT:    vsrl.vi v8, v8, 23
 ; LMULMAX2-RV32D-NEXT:    li a1, 158
-; LMULMAX2-RV32D-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX2-RV32D-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV32D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 32
-; LMULMAX2-RV32D-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX2-RV32D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    vse32.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    ret
 ;
 ; LMULMAX2-RV64D-LABEL: ctlz_v8i32:
 ; LMULMAX2-RV64D:       # %bb.0:
 ; LMULMAX2-RV64D-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; LMULMAX2-RV64D-NEXT:    vle32.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    vmset.m v0
 ; LMULMAX2-RV64D-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV64D-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
+; LMULMAX2-RV64D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV64D-NEXT:    fsrm a1
-; LMULMAX2-RV64D-NEXT:    vsrl.vi v10, v10, 23
+; LMULMAX2-RV64D-NEXT:    vsrl.vi v8, v8, 23
 ; LMULMAX2-RV64D-NEXT:    li a1, 158
-; LMULMAX2-RV64D-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX2-RV64D-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 32
-; LMULMAX2-RV64D-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX2-RV64D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    vse32.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    ret
 ;
 ; LMULMAX8-LABEL: ctlz_v8i32:
 ; LMULMAX8:       # %bb.0:
 ; LMULMAX8-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
 ; LMULMAX8-NEXT:    vle32.v v8, (a0)
 ; LMULMAX8-NEXT:    vfwcvt.f.xu.v v12, v8
 ; LMULMAX8-NEXT:    li a1, 52
-; LMULMAX8-NEXT:    vnsrl.wx v10, v12, a1
+; LMULMAX8-NEXT:    vnsrl.wx v8, v12, a1
 ; LMULMAX8-NEXT:    li a1, 1054
-; LMULMAX8-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX8-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX8-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX8-NEXT:    li a1, 32
-; LMULMAX8-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX8-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX8-NEXT:    vse32.v v8, (a0)
 ; LMULMAX8-NEXT:    ret
   %a = load <8 x i32>, ptr %x
   %b = load <8 x i32>, ptr %y
   %c = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> %a, i1 false)
   store <8 x i32> %c, ptr %x
   ret void
 }

 ; LMULMAX2-RV32F-LABEL: ctlz_v4i64:
 ; LMULMAX2-RV32F:       # %bb.0:
 ; LMULMAX2-RV32F-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
 ; LMULMAX2-RV32F-NEXT:    vle64.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    vmset.m v0
 ; LMULMAX2-RV32F-NEXT:    fsrmi a1, 1
 ; LMULMAX2-RV32F-NEXT:    vfncvt.f.xu.w v10, v8, v0.t
 ; LMULMAX2-RV32F-NEXT:    fsrm a1
-; LMULMAX2-RV32F-NEXT:    vsrl.vi v10, v10, 23
+; LMULMAX2-RV32F-NEXT:    vsrl.vi v8, v10, 23
 ; LMULMAX2-RV32F-NEXT:    li a1, 190
 ; LMULMAX2-RV32F-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
-; LMULMAX2-RV32F-NEXT:    vmv.v.x v12, a1
+; LMULMAX2-RV32F-NEXT:    vmv.v.x v10, a1
 ; LMULMAX2-RV32F-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
-; LMULMAX2-RV32F-NEXT:    vwsubu.wv v12, v12, v10
-; LMULMAX2-RV32F-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
-; LMULMAX2-RV32F-NEXT:    vmv.v.i v10, 0
-; LMULMAX2-RV32F-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
-; LMULMAX2-RV32F-NEXT:    vmseq.vv v0, v8, v10
+; LMULMAX2-RV32F-NEXT:    vwsubu.wv v10, v10, v8
 ; LMULMAX2-RV32F-NEXT:    li a1, 64
-; LMULMAX2-RV32F-NEXT:    vmerge.vxm v8, v12, a1, v0
+; LMULMAX2-RV32F-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; LMULMAX2-RV32F-NEXT:    vminu.vx v8, v10, a1
 ; LMULMAX2-RV32F-NEXT:    vse64.v v8, (a0)
 ; LMULMAX2-RV32F-NEXT:    ret
 ;
 ; LMULMAX2-RV64F-LABEL: ctlz_v4i64:
 ; LMULMAX2-RV64F:       # %bb.0:
 ; LMULMAX2-RV64F-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
 ; LMULMAX2-RV64F-NEXT:    vle64.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    vmset.m v0
 ; LMULMAX2-RV64F-NEXT:    fsrmi a1, 1
 ; LMULMAX2-RV64F-NEXT:    vfncvt.f.xu.w v10, v8, v0.t
 ; LMULMAX2-RV64F-NEXT:    fsrm a1
-; LMULMAX2-RV64F-NEXT:    vsrl.vi v10, v10, 23
+; LMULMAX2-RV64F-NEXT:    vsrl.vi v8, v10, 23
 ; LMULMAX2-RV64F-NEXT:    li a1, 190
-; LMULMAX2-RV64F-NEXT:    vmv.v.x v11, a1
-; LMULMAX2-RV64F-NEXT:    vwsubu.vv v12, v11, v10
-; LMULMAX2-RV64F-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
-; LMULMAX2-RV64F-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64F-NEXT:    vmv.v.x v9, a1
+; LMULMAX2-RV64F-NEXT:    vwsubu.vv v10, v9, v8
 ; LMULMAX2-RV64F-NEXT:    li a1, 64
-; LMULMAX2-RV64F-NEXT:    vmerge.vxm v8, v12, a1, v0
+; LMULMAX2-RV64F-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; LMULMAX2-RV64F-NEXT:    vminu.vx v8, v10, a1
 ; LMULMAX2-RV64F-NEXT:    vse64.v v8, (a0)
 ; LMULMAX2-RV64F-NEXT:    ret
 ;
 ; LMULMAX2-RV32D-LABEL: ctlz_v4i64:
 ; LMULMAX2-RV32D:       # %bb.0:
 ; LMULMAX2-RV32D-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
 ; LMULMAX2-RV32D-NEXT:    vle64.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    vmset.m v0
 ; LMULMAX2-RV32D-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV32D-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
+; LMULMAX2-RV32D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV32D-NEXT:    fsrm a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 52
-; LMULMAX2-RV32D-NEXT:    vsrl.vx v10, v10, a1
+; LMULMAX2-RV32D-NEXT:    vsrl.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 1086
-; LMULMAX2-RV32D-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX2-RV32D-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
-; LMULMAX2-RV32D-NEXT:    vmv.v.i v12, 0
-; LMULMAX2-RV32D-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
-; LMULMAX2-RV32D-NEXT:    vmseq.vv v0, v8, v12
+; LMULMAX2-RV32D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    li a1, 64
-; LMULMAX2-RV32D-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX2-RV32D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV32D-NEXT:    vse64.v v8, (a0)
 ; LMULMAX2-RV32D-NEXT:    ret
 ;
 ; LMULMAX2-RV64D-LABEL: ctlz_v4i64:
 ; LMULMAX2-RV64D:       # %bb.0:
 ; LMULMAX2-RV64D-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
 ; LMULMAX2-RV64D-NEXT:    vle64.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    vmset.m v0
 ; LMULMAX2-RV64D-NEXT:    fsrmi a1, 1
-; LMULMAX2-RV64D-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
+; LMULMAX2-RV64D-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
 ; LMULMAX2-RV64D-NEXT:    fsrm a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 52
-; LMULMAX2-RV64D-NEXT:    vsrl.vx v10, v10, a1
+; LMULMAX2-RV64D-NEXT:    vsrl.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 1086
-; LMULMAX2-RV64D-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX2-RV64D-NEXT:    vmseq.vi v0, v8, 0
+; LMULMAX2-RV64D-NEXT:    vrsub.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    li a1, 64
-; LMULMAX2-RV64D-NEXT:    vmerge.vxm v8, v10, a1, v0
+; LMULMAX2-RV64D-NEXT:    vminu.vx v8, v8, a1
 ; LMULMAX2-RV64D-NEXT:    vse64.v v8, (a0)
 ; LMULMAX2-RV64D-NEXT:    ret
 ;
-; LMULMAX8-RV32-LABEL: ctlz_v4i64:
-; LMULMAX8-RV32:       # %bb.0:
-; LMULMAX8-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
-; LMULMAX8-RV32-NEXT:    vle64.v v8, (a0)
-; LMULMAX8-RV32-NEXT:    vmset.m v0
-; LMULMAX8-RV32-NEXT:    fsrmi a1, 1
-; LMULMAX8-RV32-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
-; LMULMAX8-RV32-NEXT:    fsrm a1
-; LMULMAX8-RV32-NEXT:    li a1, 52
-; LMULMAX8-RV32-NEXT:    vsrl.vx v10, v10, a1
-; LMULMAX8-RV32-NEXT:    li a1, 1086
-; LMULMAX8-RV32-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX8-RV32-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
-; LMULMAX8-RV32-NEXT:    vmv.v.i v12, 0
-; LMULMAX8-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
-; LMULMAX8-RV32-NEXT:    vmseq.vv v0, v8, v12
-; LMULMAX8-RV32-NEXT:    li a1, 64
-; LMULMAX8-RV32-NEXT:    vmerge.vxm v8, v10, a1, v0
-; LMULMAX8-RV32-NEXT:    vse64.v v8, (a0)
-; LMULMAX8-RV32-NEXT:    ret
-;
-; LMULMAX8-RV64-LABEL: ctlz_v4i64:
-; LMULMAX8-RV64:       # %bb.0:
-; LMULMAX8-RV64-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
-; LMULMAX8-RV64-NEXT:    vle64.v v8, (a0)
-; LMULMAX8-RV64-NEXT:    vmset.m v0
-; LMULMAX8-RV64-NEXT:    fsrmi a1, 1
-; LMULMAX8-RV64-NEXT:    vfcvt.f.xu.v v10, v8, v0.t
-; LMULMAX8-RV64-NEXT:    fsrm a1
-; LMULMAX8-RV64-NEXT:    li a1, 52
-; LMULMAX8-RV64-NEXT:    vsrl.vx v10, v10, a1
-; LMULMAX8-RV64-NEXT:    li a1, 1086
-; LMULMAX8-RV64-NEXT:    vrsub.vx v10, v10, a1
-; LMULMAX8-RV64-NEXT:    vmseq.vi v0, v8, 0
-; LMULMAX8-RV64-NEXT:    li a1, 64
-; LMULMAX8-RV64-NEXT:    vmerge.vxm v8, v10, a1, v0
-; LMULMAX8-RV64-NEXT:    vse64.v v8, (a0)
-; LMULMAX8-RV64-NEXT:    ret
+; LMULMAX8-LABEL: ctlz_v4i64:
+; LMULMAX8:       # %bb.0:
+; LMULMAX8-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
+; LMULMAX8-NEXT:    vle64.v v8, (a0)
+; LMULMAX8-NEXT:    vmset.m v0
+; LMULMAX8-NEXT:    fsrmi a1, 1
+; LMULMAX8-NEXT:    vfcvt.f.xu.v v8, v8, v0.t
+; LMULMAX8-NEXT:    fsrm a1
+; LMULMAX8-NEXT:    li a1, 52
+; LMULMAX8-NEXT:    vsrl.vx v8, v8, a1
+; LMULMAX8-NEXT:    li a1, 1086
+; LMULMAX8-NEXT:    vrsub.vx v8, v8, a1
+; LMULMAX8-NEXT:    li a1, 64
+; LMULMAX8-NEXT:    vminu.vx v8, v8, a1
+; LMULMAX8-NEXT:    vse64.v v8, (a0)
+; LMULMAX8-NEXT:    ret
   %a = load <4 x i64>, ptr %x
   %b = load <4 x i64>, ptr %y
   %c = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> %a, i1 false)
   store <4 x i64> %c, ptr %x
   ret void
 }
 declare <4 x i64> @llvm.ctlz.v4i64(<4 x i64>, i1)