[AArch64][SVE] Fix umin/umax lowering to handle out of range imm.

[huihuiz]

Summary

Immediate must be in an integer range [0,255] for umin/umax instruction.
Extend pattern matching helper SelectSVEArithImm() to take in value type
bitwidth when checking immediate value is in range or not.

[huihuiz]

Current upstream mis-compile, take t.ll , run "llc -mtriple=aarch64-linux-gnu -mattr=+sve < t.ll"

define <vscale x 4 x i32> @test(<vscale x 4 x i32> %a) {
  %pg = shufflevector <vscale x 4 x i1> insertelement (<vscale x 4 x i1> undef, i1 true, i32 0), <vscale x 4 x i1> undef, <vscale x 4 x i32> zeroinitializer
  %elt = insertelement <vscale x 4 x i32> undef, i32 257, i32 0
  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
  %umin = call <vscale x 4 x i32> @llvm.aarch64.sve.umin.nxv4i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat)
  ret <vscale x 4 x i32> %umin
}

declare <vscale x 4 x i32> @llvm.aarch64.sve.umin.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)

Then you see it's mis-compiled into
// %bb.0:

umin    z0.s, z0.s, #1

This patch fixes this error, and generate
// %bb.0:

mov     w8, #257
ptrue   p0.s
mov     z1.s, w8
umin    z0.s, p0/m, z0.s, z1.s

[sdesmalen]

llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

3146

Maybe I'm missing something obvious, but I don't see why any masking is needed.
Is removing ImmVal = ImmVal & 0xFF; not sufficient?

[huihuiz]

llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

3146

We need to apply masking "ImmVal & 0xFF" for i8 at least. Otherwise we got regression.

Take t.ll below, if we don't apply the mask for i8, then we generate "mov z1.b, #-127 ".
With the mask we generate "umax z0.b, z0.b, #129".
There are similar regression for i8 in range [128, 255]

define <vscale x 16 x i8> @umax_i8_large(<vscale x 16 x i8> %a) {
  %elt = insertelement <vscale x 16 x i8> undef, i8 129, i32 0
  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
  %cmp = icmp ugt <vscale x 16 x i8> %a, %splat
  %res = select <vscale x 16 x i1> %cmp, <vscale x 16 x i8> %a, <vscale x 16 x i8> %splat
  ret <vscale x 16 x i8> %res
}

[efriedma]

llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

3146

To add to this, the SelectionDAG node for the operand looks something like the following: nxv16i8 AArch64ISD::DUP(i32 -127).

[sdesmalen]

LGTM, thanks for fixing @huihuiz!

llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

3139

nit: DL can be inlined directly into its single use, i.e.

Imm = CurDAG->getTargetConstant(ImmVal, SDLoc(N), MVT::i32);

3146

Thanks, I see what you mean now!

[huihuiz]

Thanks @sdesmalen for the review!
Fixed in the commit patch.

llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

bool SelectSVEAddSubImm(SDValue N, SDValue &Imm, SDValue &Shift) {
     return SelectSVEAddSubImm(N, VT, Imm, Shift);
   }
 
   template<MVT::SimpleValueType VT>
   bool SelectSVELogicalImm(SDValue N, SDValue &Imm) {
     return SelectSVELogicalImm(N, VT, Imm);
   }
 
+  template <MVT::SimpleValueType VT>
+  bool SelectSVEArithImm(SDValue N, SDValue &Imm) {
+    return SelectSVEArithImm(N, VT, Imm);
+  }
+
   template <unsigned Low, unsigned High, bool AllowSaturation = false>
   bool SelectSVEShiftImm(SDValue N, SDValue &Imm) {
     return SelectSVEShiftImm(N, Low, High, AllowSaturation, Imm);
   }
 
   // Returns a suitable CNT/INC/DEC/RDVL multiplier to calculate VSCALE*N.
   template<signed Min, signed Max, signed Scale, bool Shift>
   bool SelectCntImm(SDValue N, SDValue &Imm) {
private:
   bool SelectSVEAddSubImm(SDValue N, MVT VT, SDValue &Imm, SDValue &Shift);
 
   bool SelectSVELogicalImm(SDValue N, MVT VT, SDValue &Imm);
 
   bool SelectSVESignedArithImm(SDValue N, SDValue &Imm);
   bool SelectSVEShiftImm(SDValue N, uint64_t Low, uint64_t High,
                          bool AllowSaturation, SDValue &Imm);
 
-  bool SelectSVEArithImm(SDValue N, SDValue &Imm);
+  bool SelectSVEArithImm(SDValue N, MVT VT, SDValue &Imm);
   bool SelectSVERegRegAddrMode(SDValue N, unsigned Scale, SDValue &Base,
                                SDValue &Offset);
 };
 } // end anonymous namespace
 
 /// isIntImmediate - This method tests to see if the node is a constant
 /// operand. If so Imm will receive the 32-bit value.
 static bool isIntImmediate(const SDNode *N, uint64_t &Imm) {
if (auto CNode = dyn_cast<ConstantSDNode>(N)) {
     if (ImmVal >= -128 && ImmVal < 128) {
       Imm = CurDAG->getTargetConstant(ImmVal, DL, MVT::i32);
       return true;
     }
   }
   return false;
 }
 
-bool AArch64DAGToDAGISel::SelectSVEArithImm(SDValue N, SDValue &Imm) {
+bool AArch64DAGToDAGISel::SelectSVEArithImm(SDValue N, MVT VT, SDValue &Imm) {
   if (auto CNode = dyn_cast<ConstantSDNode>(N)) {
-    uint64_t ImmVal = CNode->getSExtValue();
-    SDLoc DL(N);
-    ImmVal = ImmVal & 0xFF;
+    uint64_t ImmVal = CNode->getZExtValue();
+
+    switch (VT.SimpleTy) {
+    case MVT::i8:
+      ImmVal &= 0xFF;
+      break;
+    case MVT::i16:
+      ImmVal &= 0xFFFF;
+      break;
+    case MVT::i32:
+      ImmVal &= 0xFFFFFFFF;
+      break;
+    case MVT::i64:
+      break;
+    default:
+      llvm_unreachable("Unexpected type");
+    }
+
     if (ImmVal < 256) {
-      Imm = CurDAG->getTargetConstant(ImmVal, DL, MVT::i32);
+      Imm = CurDAG->getTargetConstant(ImmVal, SDLoc(N), MVT::i32);
       return true;
     }
   }
   return false;
 }
 
 bool AArch64DAGToDAGISel::SelectSVELogicalImm(SDValue N, MVT VT, SDValue &Imm) {
   if (auto CNode = dyn_cast<ConstantSDNode>(N)) {

llvm/lib/Target/AArch64/SVEInstrFormats.td

 
 def SVELogicalImm8Pat  : ComplexPattern<i64, 1, "SelectSVELogicalImm<MVT::i8>", []>;
 def SVELogicalImm16Pat : ComplexPattern<i64, 1, "SelectSVELogicalImm<MVT::i16>", []>;
 def SVELogicalImm32Pat : ComplexPattern<i64, 1, "SelectSVELogicalImm<MVT::i32>", []>;
 def SVELogicalImm64Pat : ComplexPattern<i64, 1, "SelectSVELogicalImm<MVT::i64>", []>;
 
 def SVE8BitLslImm : ComplexPattern<i32, 2, "SelectSVE8BitLslImm", [imm]>;
 
-def SVEArithUImmPat  : ComplexPattern<i32, 1, "SelectSVEArithImm", []>;
+def SVEArithUImm8Pat  : ComplexPattern<i32, 1, "SelectSVEArithImm<MVT::i8>", []>;
+def SVEArithUImm16Pat  : ComplexPattern<i32, 1, "SelectSVEArithImm<MVT::i16>", []>;
+def SVEArithUImm32Pat  : ComplexPattern<i32, 1, "SelectSVEArithImm<MVT::i32>", []>;
+def SVEArithUImm64Pat  : ComplexPattern<i32, 1, "SelectSVEArithImm<MVT::i64>", []>;
 def SVEArithSImmPat  : ComplexPattern<i32, 1, "SelectSVESignedArithImm", []>;
 
 def SVEShiftImmL8  : ComplexPattern<i32, 1, "SelectSVEShiftImm<0, 7>",  []>;
 def SVEShiftImmL16 : ComplexPattern<i32, 1, "SelectSVEShiftImm<0, 15>", []>;
 def SVEShiftImmL32 : ComplexPattern<i32, 1, "SelectSVEShiftImm<0, 31>", []>;
 def SVEShiftImmL64 : ComplexPattern<i32, 1, "SelectSVEShiftImm<0, 63>", []>;
 def SVEShiftImmR8  : ComplexPattern<i32, 1, "SelectSVEShiftImm<1, 8,  true>", []>;
 def SVEShiftImmR16 : ComplexPattern<i32, 1, "SelectSVEShiftImm<1, 16, true>", []>;

 }
 
 multiclass sve_int_arith_imm1_unsigned<bits<2> opc, string asm, SDPatternOperator op> {
   def _B : sve_int_arith_imm<0b00, { 0b1010, opc }, asm, ZPR8, imm0_255>;
   def _H : sve_int_arith_imm<0b01, { 0b1010, opc }, asm, ZPR16, imm0_255>;
   def _S : sve_int_arith_imm<0b10, { 0b1010, opc }, asm, ZPR32, imm0_255>;
   def _D : sve_int_arith_imm<0b11, { 0b1010, opc }, asm, ZPR64, imm0_255>;
 
-  def : SVE_1_Op_Imm_Arith_Pred_Pat<nxv16i8, nxv16i1, op, ZPR8, i32, SVEArithUImmPat, !cast<Instruction>(NAME # _B)>;
-  def : SVE_1_Op_Imm_Arith_Pred_Pat<nxv8i16, nxv8i1, op, ZPR16, i32, SVEArithUImmPat, !cast<Instruction>(NAME # _H)>;
-  def : SVE_1_Op_Imm_Arith_Pred_Pat<nxv4i32, nxv4i1, op, ZPR32, i32, SVEArithUImmPat, !cast<Instruction>(NAME # _S)>;
-  def : SVE_1_Op_Imm_Arith_Pred_Pat<nxv2i64, nxv2i1, op, ZPR64, i64, SVEArithUImmPat, !cast<Instruction>(NAME # _D)>;
+  def : SVE_1_Op_Imm_Arith_Pred_Pat<nxv16i8, nxv16i1, op, ZPR8, i32, SVEArithUImm8Pat, !cast<Instruction>(NAME # _B)>;
+  def : SVE_1_Op_Imm_Arith_Pred_Pat<nxv8i16, nxv8i1, op, ZPR16, i32, SVEArithUImm16Pat, !cast<Instruction>(NAME # _H)>;
+  def : SVE_1_Op_Imm_Arith_Pred_Pat<nxv4i32, nxv4i1, op, ZPR32, i32, SVEArithUImm32Pat, !cast<Instruction>(NAME # _S)>;
+  def : SVE_1_Op_Imm_Arith_Pred_Pat<nxv2i64, nxv2i1, op, ZPR64, i64, SVEArithUImm64Pat, !cast<Instruction>(NAME # _D)>;
 }
 
 multiclass sve_int_arith_imm2<string asm, SDPatternOperator op> {
   def _B : sve_int_arith_imm<0b00, 0b110000, asm, ZPR8,  simm8>;
   def _H : sve_int_arith_imm<0b01, 0b110000, asm, ZPR16, simm8>;
   def _S : sve_int_arith_imm<0b10, 0b110000, asm, ZPR32, simm8>;
   def _D : sve_int_arith_imm<0b11, 0b110000, asm, ZPR64, simm8>;
 

llvm/test/CodeGen/AArch64/sve-int-arith-imm.ll

 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 -58, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %cmp = icmp sgt <vscale x 8 x i16> %a, %splat
   %res = select <vscale x 8 x i1> %cmp, <vscale x 8 x i16> %a, <vscale x 8 x i16> %splat
   ret <vscale x 8 x i16> %res
 }
 
+define <vscale x 8 x i16> @smax_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: smax_i16_out_of_range:
+; CHECK:    mov w8, #257
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    smax z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %cmp = icmp sgt <vscale x 8 x i16> %a, %splat
+  %res = select <vscale x 8 x i1> %cmp, <vscale x 8 x i16> %a, <vscale x 8 x i16> %splat
+  ret <vscale x 8 x i16> %res
+}
+
 define <vscale x 4 x i32> @smax_i32_pos(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smax_i32_pos
 ; CHECK: smax z0.s, z0.s, #27
 ; CHECK: ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 27, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %cmp = icmp sgt <vscale x 4 x i32> %a, %splat
   %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
   ret <vscale x 4 x i32> %res
 }
 
 define <vscale x 4 x i32> @smax_i32_neg(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smax_i32_neg
 ; CHECK: smax z0.s, z0.s, #-58
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 -58, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %cmp = icmp sgt <vscale x 4 x i32> %a, %splat
   %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
   ret <vscale x 4 x i32> %res
 }
 
+define <vscale x 4 x i32> @smax_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: smax_i32_out_of_range:
+; CHECK:    mov w8, #-129
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    smax z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 4 x i32> undef, i32 -129, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %cmp = icmp sgt <vscale x 4 x i32> %a, %splat
+  %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
+  ret <vscale x 4 x i32> %res
+}
+
 define <vscale x 2 x i64> @smax_i64_pos(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smax_i64_pos
 ; CHECK: smax z0.d, z0.d, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 27, i32 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %cmp = icmp sgt <vscale x 2 x i64> %a, %splat
   %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
   ret <vscale x 2 x i64> %res
 }
 
 define <vscale x 2 x i64> @smax_i64_neg(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smax_i64_neg
 ; CHECK: smax z0.d, z0.d, #-58
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 -58, i32 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %cmp = icmp sgt <vscale x 2 x i64> %a, %splat
   %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
   ret <vscale x 2 x i64> %res
 }
 
+define <vscale x 2 x i64> @smax_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: smax_i64_out_of_range:
+; CHECK:    mov w8, #65535
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    smax z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %cmp = icmp sgt <vscale x 2 x i64> %a, %splat
+  %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
+  ret <vscale x 2 x i64> %res
+}
+
 ;
 ; SMIN
 ;
 define <vscale x 16 x i8> @smin_i8_pos(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: smin_i8_pos
 ; CHECK: smin z0.b, z0.b, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 16 x i8> undef, i8 27, i32 0

 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 -58, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %cmp = icmp slt <vscale x 8 x i16> %a, %splat
   %res = select <vscale x 8 x i1> %cmp, <vscale x 8 x i16> %a, <vscale x 8 x i16> %splat
   ret <vscale x 8 x i16> %res
 }
 
+define <vscale x 8 x i16> @smin_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: smin_i16_out_of_range:
+; CHECK:    mov w8, #257
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    smin z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %cmp = icmp slt <vscale x 8 x i16> %a, %splat
+  %res = select <vscale x 8 x i1> %cmp, <vscale x 8 x i16> %a, <vscale x 8 x i16> %splat
+  ret <vscale x 8 x i16> %res
+}
+
 define <vscale x 4 x i32> @smin_i32_pos(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smin_i32_pos
 ; CHECK: smin z0.s, z0.s, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 27, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %cmp = icmp slt <vscale x 4 x i32> %a, %splat
   %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
   ret <vscale x 4 x i32> %res
 }
 
 define <vscale x 4 x i32> @smin_i32_neg(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smin_i32_neg
 ; CHECK: smin z0.s, z0.s, #-58
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 -58, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %cmp = icmp slt <vscale x 4 x i32> %a, %splat
   %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
   ret <vscale x 4 x i32> %res
 }
 
+define <vscale x 4 x i32> @smin_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: smin_i32_out_of_range:
+; CHECK:    mov w8, #-129
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    smin z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 4 x i32> undef, i32 -129, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %cmp = icmp slt <vscale x 4 x i32> %a, %splat
+  %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
+  ret <vscale x 4 x i32> %res
+}
+
 define <vscale x 2 x i64> @smin_i64_pos(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smin_i64_pos
 ; CHECK: smin z0.d, z0.d, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 27, i32 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %cmp = icmp slt <vscale x 2 x i64> %a, %splat
   %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
   ret <vscale x 2 x i64> %res
 }
 
 define <vscale x 2 x i64> @smin_i64_neg(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smin_i64_neg
 ; CHECK: smin z0.d, z0.d, #-58
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 -58, i32 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %cmp = icmp slt <vscale x 2 x i64> %a, %splat
   %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
   ret <vscale x 2 x i64> %res
 }
 
+define <vscale x 2 x i64> @smin_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: smin_i64_out_of_range:
+; CHECK:    mov w8, #65535
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    smin z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %cmp = icmp slt <vscale x 2 x i64> %a, %splat
+  %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
+  ret <vscale x 2 x i64> %res
+}
+
 ;
 ; UMAX
 ;
 define <vscale x 16 x i8> @umax_i8_pos(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: umax_i8_pos
 ; CHECK: umax z0.b, z0.b, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 16 x i8> undef, i8 27, i32 0

 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 27, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %cmp = icmp ugt <vscale x 8 x i16> %a, %splat
   %res = select <vscale x 8 x i1> %cmp, <vscale x 8 x i16> %a, <vscale x 8 x i16> %splat
   ret <vscale x 8 x i16> %res
 }
 
-define <vscale x 8 x i16> @umax_i16_large(<vscale x 8 x i16> %a) {
-; CHECK-LABEL: umax_i16_large
-; CHECK: umax z0.h, z0.h, #129
-; CHECK-NEXT: ret
-  %elt = insertelement <vscale x 8 x i16> undef, i16 129, i32 0
+define <vscale x 8 x i16> @umax_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: umax_i16_out_of_range:
+; CHECK:    mov w8, #257
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %cmp = icmp ugt <vscale x 8 x i16> %a, %splat
   %res = select <vscale x 8 x i1> %cmp, <vscale x 8 x i16> %a, <vscale x 8 x i16> %splat
   ret <vscale x 8 x i16> %res
 }
 
 define <vscale x 4 x i32> @umax_i32_pos(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: umax_i32_pos
 ; CHECK: umax z0.s, z0.s, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 27, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %cmp = icmp ugt <vscale x 4 x i32> %a, %splat
   %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
   ret <vscale x 4 x i32> %res
 }
 
-define <vscale x 4 x i32> @umax_i32_large(<vscale x 4 x i32> %a) {
-; CHECK-LABEL: umax_i32_large
-; CHECK: umax z0.s, z0.s, #129
-; CHECK-NEXT: ret
-  %elt = insertelement <vscale x 4 x i32> undef, i32 129, i32 0
+define <vscale x 4 x i32> @umax_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: umax_i32_out_of_range:
+; CHECK:    mov w8, #257
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 4 x i32> undef, i32 257, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %cmp = icmp ugt <vscale x 4 x i32> %a, %splat
   %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
   ret <vscale x 4 x i32> %res
 }
 
 define <vscale x 2 x i64> @umax_i64_pos(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: umax_i64_pos
 ; CHECK: umax z0.d, z0.d, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 27, i32 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %cmp = icmp ugt <vscale x 2 x i64> %a, %splat
   %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
   ret <vscale x 2 x i64> %res
 }
 
-define <vscale x 2 x i64> @umax_i64_large(<vscale x 2 x i64> %a) {
-; CHECK-LABEL: umax_i64_large
-; CHECK: umax z0.d, z0.d, #129
-; CHECK-NEXT: ret
-  %elt = insertelement <vscale x 2 x i64> undef, i64 129, i32 0
+define <vscale x 2 x i64> @umax_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: umax_i64_out_of_range:
+; CHECK:    mov w8, #65535
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    umax z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i32 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %cmp = icmp ugt <vscale x 2 x i64> %a, %splat
   %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
   ret <vscale x 2 x i64> %res
 }
 
 ;
 ; UMIN

 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 27, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %cmp = icmp ult <vscale x 8 x i16> %a, %splat
   %res = select <vscale x 8 x i1> %cmp, <vscale x 8 x i16> %a, <vscale x 8 x i16> %splat
   ret <vscale x 8 x i16> %res
 }
 
-define <vscale x 8 x i16> @umin_i16_large(<vscale x 8 x i16> %a) {
-; CHECK-LABEL: umin_i16_large
-; CHECK: umin z0.h, z0.h, #129
-; CHECK-NEXT: ret
-  %elt = insertelement <vscale x 8 x i16> undef, i16 129, i32 0
+define <vscale x 8 x i16> @umin_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: umin_i16_out_of_range:
+; CHECK:    mov w8, #257
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    umin z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %cmp = icmp ult <vscale x 8 x i16> %a, %splat
   %res = select <vscale x 8 x i1> %cmp, <vscale x 8 x i16> %a, <vscale x 8 x i16> %splat
   ret <vscale x 8 x i16> %res
 }
 
 define <vscale x 4 x i32> @umin_i32_pos(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: umin_i32_pos
 ; CHECK: umin z0.s, z0.s, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 27, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %cmp = icmp ult <vscale x 4 x i32> %a, %splat
   %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
   ret <vscale x 4 x i32> %res
 }
 
-define <vscale x 4 x i32> @umin_i32_large(<vscale x 4 x i32> %a) {
-; CHECK-LABEL: umin_i32_large
-; CHECK: umin z0.s, z0.s, #129
-; CHECK-NEXT: ret
-  %elt = insertelement <vscale x 4 x i32> undef, i32 129, i32 0
+define <vscale x 4 x i32> @umin_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: umin_i32_out_of_range:
+; CHECK:    mov w8, #257
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    umin z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 4 x i32> undef, i32 257, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %cmp = icmp ult <vscale x 4 x i32> %a, %splat
   %res = select <vscale x 4 x i1> %cmp, <vscale x 4 x i32> %a, <vscale x 4 x i32> %splat
   ret <vscale x 4 x i32> %res
 }
 
 define <vscale x 2 x i64> @umin_i64_pos(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: umin_i64_pos
 ; CHECK: umin z0.d, z0.d, #27
 ; CHECK-NEXT: ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 27, i32 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %cmp = icmp ult <vscale x 2 x i64> %a, %splat
   %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
   ret <vscale x 2 x i64> %res
 }
 
-define <vscale x 2 x i64> @umin_i64_large(<vscale x 2 x i64> %a) {
-; CHECK-LABEL: umin_i64_large
-; CHECK: umin z0.d, z0.d, #129
-; CHECK-NEXT: ret
-  %elt = insertelement <vscale x 2 x i64> undef, i64 129, i32 0
+define <vscale x 2 x i64> @umin_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: umin_i64_out_of_range:
+; CHECK:    mov w8, #65535
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    umin z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i32 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %cmp = icmp ult <vscale x 2 x i64> %a, %splat
   %res = select <vscale x 2 x i1> %cmp, <vscale x 2 x i64> %a, <vscale x 2 x i64> %splat
   ret <vscale x 2 x i64> %res
 }
 
 ;
 ; MUL

llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll

; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 127, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %out = call <vscale x 8 x i16> @llvm.aarch64.sve.smax.nxv8i16(<vscale x 8 x i1> %pg,
                                                                 <vscale x 8 x i16> %a,
                                                                 <vscale x 8 x i16> %splat)
   ret <vscale x 8 x i16> %out
 }
 
+define <vscale x 8 x i16> @smax_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: smax_i16_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #129
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    smax z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 129, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.smax.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                <vscale x 8 x i16> %a,
+                                                                <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
 define <vscale x 4 x i32> @smax_i32(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smax_i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    smax z0.s, z0.s, #-128
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
   %elt = insertelement <vscale x 4 x i32> undef, i32 -128, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %out = call <vscale x 4 x i32> @llvm.aarch64.sve.smax.nxv4i32(<vscale x 4 x i1> %pg,
                                                                 <vscale x 4 x i32> %a,
                                                                 <vscale x 4 x i32> %splat)
   ret <vscale x 4 x i32> %out
 }
 
+define <vscale x 4 x i32> @smax_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: smax_i32_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #-129
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    smax z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 -129, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.smax.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                <vscale x 4 x i32> %a,
+                                                                <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
 define <vscale x 2 x i64> @smax_i64(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smax_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    smax z0.d, z0.d, #127
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
   %elt = insertelement <vscale x 2 x i64> undef, i64 127, i64 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %out = call <vscale x 2 x i64> @llvm.aarch64.sve.smax.nxv2i64(<vscale x 2 x i1> %pg,
                                                                 <vscale x 2 x i64> %a,
                                                                 <vscale x 2 x i64> %splat)
   ret <vscale x 2 x i64> %out
 }
 
+define <vscale x 2 x i64> @smax_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: smax_i64_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #65535
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    smax z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i64 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.smax.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                <vscale x 2 x i64> %a,
+                                                                <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+
 ; SMIN
 
 define <vscale x 16 x i8> @smin_i8(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: smin_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    smin z0.b, z0.b, #127
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 -128, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %out = call <vscale x 8 x i16> @llvm.aarch64.sve.smin.nxv8i16(<vscale x 8 x i1> %pg,
                                                                 <vscale x 8 x i16> %a,
                                                                 <vscale x 8 x i16> %splat)
   ret <vscale x 8 x i16> %out
 }
 
+define <vscale x 8 x i16> @smin_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: smin_i16_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #-129
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    smin z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 -129, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.smin.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                <vscale x 8 x i16> %a,
+                                                                <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
 define <vscale x 4 x i32> @smin_i32(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smin_i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    smin z0.s, z0.s, #127
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
   %elt = insertelement <vscale x 4 x i32> undef, i32 127, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %out = call <vscale x 4 x i32> @llvm.aarch64.sve.smin.nxv4i32(<vscale x 4 x i1> %pg,
                                                                 <vscale x 4 x i32> %a,
                                                                 <vscale x 4 x i32> %splat)
   ret <vscale x 4 x i32> %out
 }
 
+define <vscale x 4 x i32> @smin_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: smin_i32_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    smin z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 257, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.smin.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                <vscale x 4 x i32> %a,
+                                                                <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+
 define <vscale x 2 x i64> @smin_i64(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smin_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    smin z0.d, z0.d, #-128
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
   %elt = insertelement <vscale x 2 x i64> undef, i64 -128, i64 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %out = call <vscale x 2 x i64> @llvm.aarch64.sve.smin.nxv2i64(<vscale x 2 x i1> %pg,
                                                                 <vscale x 2 x i64> %a,
                                                                 <vscale x 2 x i64> %splat)
   ret <vscale x 2 x i64> %out
 }
 
+define <vscale x 2 x i64> @smin_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: smin_i64_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z1.d, #-256 // =0xffffffffffffff00
+; CHECK-NEXT:    smin z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 -256, i64 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.smin.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                <vscale x 2 x i64> %a,
+                                                                <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
 ; UMAX
 
 define <vscale x 16 x i8> @umax_i8(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: umax_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    umax z0.b, z0.b, #0
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 255, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %out = call <vscale x 8 x i16> @llvm.aarch64.sve.umax.nxv8i16(<vscale x 8 x i1> %pg,
                                                                 <vscale x 8 x i16> %a,
                                                                 <vscale x 8 x i16> %splat)
   ret <vscale x 8 x i16> %out
 }
 
+define <vscale x 8 x i16> @umax_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: umax_i16_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.umax.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                <vscale x 8 x i16> %a,
+                                                                <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
 define <vscale x 4 x i32> @umax_i32(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: umax_i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    umax z0.s, z0.s, #0
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
   %elt = insertelement <vscale x 4 x i32> undef, i32 0, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %out = call <vscale x 4 x i32> @llvm.aarch64.sve.umax.nxv4i32(<vscale x 4 x i1> %pg,
                                                                 <vscale x 4 x i32> %a,
                                                                 <vscale x 4 x i32> %splat)
   ret <vscale x 4 x i32> %out
 }
 
+define <vscale x 4 x i32> @umax_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: umax_i32_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 257, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.umax.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                <vscale x 4 x i32> %a,
+                                                                <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
 define <vscale x 2 x i64> @umax_i64(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: umax_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    umax z0.d, z0.d, #255
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
   %elt = insertelement <vscale x 2 x i64> undef, i64 255, i64 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %out = call <vscale x 2 x i64> @llvm.aarch64.sve.umax.nxv2i64(<vscale x 2 x i1> %pg,
                                                                 <vscale x 2 x i64> %a,
                                                                 <vscale x 2 x i64> %splat)
   ret <vscale x 2 x i64> %out
 }
 
+define <vscale x 2 x i64> @umax_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: umax_i64_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #65535
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    umax z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i64 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.umax.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                <vscale x 2 x i64> %a,
+                                                                <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
 ; UMIN
 
 define <vscale x 16 x i8> @umin_i8(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: umin_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    umin z0.b, z0.b, #255
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 0, i32 0
   %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
   %out = call <vscale x 8 x i16> @llvm.aarch64.sve.umin.nxv8i16(<vscale x 8 x i1> %pg,
                                                                 <vscale x 8 x i16> %a,
                                                                 <vscale x 8 x i16> %splat)
   ret <vscale x 8 x i16> %out
 }
 
+define <vscale x 8 x i16> @umin_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: umin_i16_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    umin z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.umin.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                <vscale x 8 x i16> %a,
+                                                                <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
 define <vscale x 4 x i32> @umin_i32(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: umin_i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    umin z0.s, z0.s, #255
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
   %elt = insertelement <vscale x 4 x i32> undef, i32 255, i32 0
   %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
   %out = call <vscale x 4 x i32> @llvm.aarch64.sve.umin.nxv4i32(<vscale x 4 x i1> %pg,
                                                                 <vscale x 4 x i32> %a,
                                                                 <vscale x 4 x i32> %splat)
   ret <vscale x 4 x i32> %out
 }
 
+define <vscale x 4 x i32> @umin_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: umin_i32_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    umin z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 257, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.umin.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                <vscale x 4 x i32> %a,
+                                                                <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
 define <vscale x 2 x i64> @umin_i64(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: umin_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    umin z0.d, z0.d, #0
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
   %elt = insertelement <vscale x 2 x i64> undef, i64 0, i64 0
   %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
   %out = call <vscale x 2 x i64> @llvm.aarch64.sve.umin.nxv2i64(<vscale x 2 x i1> %pg,
                                                                 <vscale x 2 x i64> %a,
                                                                 <vscale x 2 x i64> %splat)
   ret <vscale x 2 x i64> %out
 }
 
+define <vscale x 2 x i64> @umin_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: umin_i64_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #65535
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    umin z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i64 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.umin.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                <vscale x 2 x i64> %a,
+                                                                <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
 ; SQADD
 
 define <vscale x 16 x i8> @sqadd_b_lowimm(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: sqadd_b_lowimm:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sqadd z0.b, z0.b, #27 // =0x1b
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 16 x i8> undef, i8 27, i32 0