[Intrinsic] Add fixed point division intrinsics.

[ebevhan]

Summary

This patch adds intrinsics and ISelDAG nodes for
signed and unsigned fixed-point division:

llvm.sdiv.fix.*
llvm.udiv.fix.*

These intrinsics perform scaled division on two
integers or vectors of integers. They are required
for the implementation of the Embedded-C fixed-point
arithmetic in Clang.

[ebevhan]

I will preface this by saying that I'm not terribly pleased with this solution. There is no (easy) way to construct a wide division during operation legalization if the wider type is not legal, so we have to do a whole bunch of work upfront to avoid having to expand. It feels really hacky, but I'm not sure there's any other way of doing it given ISelDAG's design.

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

7337–7339

I'm unsure if this is acceptable. I wanted to align the default expansions of division, multiplication, and shifts, but correcting for rounding requires extra code since regular signed division rounds toward zero.

[ebevhan]

Updated langref.

[ebevhan]

Ping.

[leonardchan]

Sorry for the holdup. Right now I'm just writing a bunch of tests to validate the IR expansion produces the correct division results. As of now, the patch LGTM, but I'd prefer holding off submitting the revision until I finish validating (unless you think it's better to submit now then fix any potential bugs that may be found later since I don't think anyone will be immediately using this). Here's a few comments I have so far:

• This will need another rebase. I ran into a merge conflict this morning against ToT.
• I think @llvm.sdiv.fix.i4(i4 -2, i4 2, i32 0) produces the incorrect result. This returns -2 instead of -1. I'm looking over the patch to see what could've caused it, but you might be able to find it faster than I can.

[leonardchan]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

7358

Ah, I think you want this to instead be ISD::SETNE and RemNonZero

[leonardchan]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

7362–7364

There's a corner case here where there won't be a subtraction if the true value of a division is between -1 and 0. For -6 / 7, the remainder will be non-zero, but quotient will be zero since sdiv rounds towards zero.

I think you need to instead check if only one of the signs of either operand is negative to check if the quotient is negative:

SDValue LHSNeg = DAG.getSetCC(dl, BoolVT, LHS, Zero, ISD::SETLT);
SDValue RHSNeg = DAG.getSetCC(dl, BoolVT, RHS, Zero, ISD::SETLT);
SDValue QuotNeg = DAG.getNode(ISD::XOR, dl, BoolVT, LHSNeg, RHSNeg);

[ebevhan]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

7358

Oh, you're right. Really surprised I didn't catch this in my testing.

7362–7364

Yep, I see what you mean.

[ebevhan]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

7362–7364

Though, would it also work to check if the Remainder is negative instead of nonzero?

[leonardchan]

Alrighty, it looks like everything in this patch should be ok after those 2 changes. I can send you the script I used to generate tests if you'd like to use it also.

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

7362–7364

Hmm I don't think that would work since srem of 2 negative numbers would return a negative number: srem -7, -3 == -1

[ebevhan]

Fix review comments.

[ebevhan]

I think I'd like someone who has more specialized SelectionDAG knowledge to come with some input as well, but it doesn't seem like anyone else is looking at the patch.

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

7362–7364

Ah, that's true.

[ebevhan]

Maybe @craig.topper has some input on this design.

[craig.topper]

I don't really like the code In SelectionDAGBuilder, but I don't have a better solution either. So LGTM.

[ebevhan]

I've been holding off on this for a while because the design made implementing saturated division a lot more complicated than it ought to be.

I'm very tempted to just scrap the promotion and expansion of the node altogether, determine if the node is legal (or easily promotable) directly in SelectionDAGBuilder and if it is not, expand it straight away. That's also a hack, but saves a ton of headache in the later stages, especially for the saturating operation.

It's either that or a libcall, but I really don't feel like that should be necessary; it should absolutely be possible to do this without one.

[ebevhan]

Rebased.

[uabelho]

I'll submit this for @ebevhan in a little bit.

llvm/docs/LangRef.rst

.. code-block:: llvm
       %res = call i4 @llvm.usub.sat.i4(i4 2, i4 6)  ; %res = 0
 
 
 Fixed Point Arithmetic Intrinsics
 ---------------------------------
 
 A fixed point number represents a real data type for a number that has a fixed
 number of digits after a radix point (equivalent to the decimal point '.').
-The number of digits after the radix point is referred as the ``scale``. These
+The number of digits after the radix point is referred as the `scale`. These
 are useful for representing fractional values to a specific precision. The
 following intrinsics perform fixed point arithmetic operations on 2 operands
 of the same scale, specified as the third argument.
 
-The `llvm.*mul.fix` family of intrinsic functions represents a multiplication
+The ``llvm.*mul.fix`` family of intrinsic functions represents a multiplication
 of fixed point numbers through scaled integers. Therefore, fixed point
-multplication can be represented as
+multiplication can be represented as
+
+.. code-block:: llvm
 
-::
         %result = call i4 @llvm.smul.fix.i4(i4 %a, i4 %b, i32 %scale)
 
         ; Expands to
         %a2 = sext i4 %a to i8
         %b2 = sext i4 %b to i8
         %mul = mul nsw nuw i8 %a, %b
         %scale2 = trunc i32 %scale to i8
         %r = ashr i8 %mul, i8 %scale2  ; this is for a target rounding down towards negative infinity
         %result = trunc i8 %r to i4
 
+The ``llvm.*div.fix`` family of intrinsic functions represents a division of
+fixed point numbers through scaled integers. Fixed point division can be
+represented as:
+
+.. code-block:: llvm
+
+        %result call i4 @llvm.sdiv.fix.i4(i4 %a, i4 %b, i32 %scale)
+
+        ; Expands to
+        %a2 = sext i4 %a to i8
+        %b2 = sext i4 %b to i8
+        %scale2 = trunc i32 %scale to i8
+        %a3 = shl i8 %a2, %scale2
+        %r = sdiv i8 %a3, %b2 ; this is for a target rounding towards zero
+        %result = trunc i8 %r to i4
+
 For each of these functions, if the result cannot be represented exactly with
 the provided scale, the result is rounded. Rounding is unspecified since
 preferred rounding may vary for different targets. Rounding is specified
 through a target hook. Different pipelines should legalize or optimize this
 using the rounding specified by this hook if it is provided. Operations like
 constant folding, instruction combining, KnownBits, and ValueTracking should
 also use this hook, if provided, and not assume the direction of rounding. A
 rounded result must always be within one unit of precision from the true
.. code-block:: llvm
       %res = call i4 @llvm.umul.fix.sat.i4(i4 8, i4 2, i32 0)  ; %res = 15 (8 x 2 -> clamped to 15)
       %res = call i4 @llvm.umul.fix.sat.i4(i4 8, i4 8, i32 2)  ; %res = 15 (2 x 2 -> clamped to 3.75)
 
       ; Scale can affect the saturation result
       %res = call i4 @llvm.umul.fix.sat.i4(i4 2, i4 4, i32 0)  ; %res = 7 (2 x 4 -> clamped to 7)
       %res = call i4 @llvm.umul.fix.sat.i4(i4 2, i4 4, i32 1)  ; %res = 4 (1 x 2 = 2)
 
 
+'``llvm.sdiv.fix.*``' Intrinsics
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Syntax
+"""""""
+
+This is an overloaded intrinsic. You can use ``llvm.sdiv.fix``
+on any integer bit width or vectors of integers.
+
+::
+
+      declare i16 @llvm.sdiv.fix.i16(i16 %a, i16 %b, i32 %scale)
+      declare i32 @llvm.sdiv.fix.i32(i32 %a, i32 %b, i32 %scale)
+      declare i64 @llvm.sdiv.fix.i64(i64 %a, i64 %b, i32 %scale)
+      declare <4 x i32> @llvm.sdiv.fix.v4i32(<4 x i32> %a, <4 x i32> %b, i32 %scale)
+
+Overview
+"""""""""
+
+The '``llvm.sdiv.fix``' family of intrinsic functions perform signed
+fixed point division on 2 arguments of the same scale.
+
+Arguments
+""""""""""
+
+The arguments (%a and %b) and the result may be of integer types of any bit
+width, but they must have the same bit width. The arguments may also work with
+int vectors of the same length and int size. ``%a`` and ``%b`` are the two
+values that will undergo signed fixed point division. The argument
+``%scale`` represents the scale of both operands, and must be a constant
+integer.
+
+Semantics:
+""""""""""
+
+This operation performs fixed point division on the 2 arguments of a
+specified scale. The result will also be returned in the same scale specified
+in the third argument.
+
+If the result value cannot be precisely represented in the given scale, the
+value is rounded up or down to the closest representable value. The rounding
+direction is unspecified.
+
+It is undefined behavior if the result value does not fit within the range of
+the fixed point type, or if the second argument is zero.
+
+
+Examples
+"""""""""
+
+.. code-block:: llvm
+
+      %res = call i4 @llvm.sdiv.fix.i4(i4 6, i4 2, i32 0)  ; %res = 3 (6 / 2 = 3)
+      %res = call i4 @llvm.sdiv.fix.i4(i4 6, i4 4, i32 1)  ; %res = 3 (3 / 2 = 1.5)
+      %res = call i4 @llvm.sdiv.fix.i4(i4 3, i4 -2, i32 1) ; %res = -3 (1.5 / -1 = -1.5)
+
+      ; The result in the following could be rounded up to 1 or down to 0.5
+      %res = call i4 @llvm.sdiv.fix.i4(i4 3, i4 4, i32 1)  ; %res = 2 (or 1) (1.5 / 2 = 0.75)
+
+
+'``llvm.udiv.fix.*``' Intrinsics
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Syntax
+"""""""
+
+This is an overloaded intrinsic. You can use ``llvm.udiv.fix``
+on any integer bit width or vectors of integers.
+
+::
+
+      declare i16 @llvm.udiv.fix.i16(i16 %a, i16 %b, i32 %scale)
+      declare i32 @llvm.udiv.fix.i32(i32 %a, i32 %b, i32 %scale)
+      declare i64 @llvm.udiv.fix.i64(i64 %a, i64 %b, i32 %scale)
+      declare <4 x i32> @llvm.udiv.fix.v4i32(<4 x i32> %a, <4 x i32> %b, i32 %scale)
+
+Overview
+"""""""""
+
+The '``llvm.udiv.fix``' family of intrinsic functions perform unsigned
+fixed point division on 2 arguments of the same scale.
+
+Arguments
+""""""""""
+
+The arguments (%a and %b) and the result may be of integer types of any bit
+width, but they must have the same bit width. The arguments may also work with
+int vectors of the same length and int size. ``%a`` and ``%b`` are the two
+values that will undergo unsigned fixed point division. The argument
+``%scale`` represents the scale of both operands, and must be a constant
+integer.
+
+Semantics:
+""""""""""
+
+This operation performs fixed point division on the 2 arguments of a
+specified scale. The result will also be returned in the same scale specified
+in the third argument.
+
+If the result value cannot be precisely represented in the given scale, the
+value is rounded up or down to the closest representable value. The rounding
+direction is unspecified.
+
+It is undefined behavior if the result value does not fit within the range of
+the fixed point type, or if the second argument is zero.
+
+
+Examples
+"""""""""
+
+.. code-block:: llvm
+
+      %res = call i4 @llvm.udiv.fix.i4(i4 6, i4 2, i32 0)  ; %res = 3 (6 / 2 = 3)
+      %res = call i4 @llvm.udiv.fix.i4(i4 6, i4 4, i32 1)  ; %res = 3 (3 / 2 = 1.5)
+      %res = call i4 @llvm.udiv.fix.i4(i4 1, i4 -8, i32 4) ; %res = 2 (0.0625 / 0.5 = 0.125)
+
+      ; The result in the following could be rounded up to 1 or down to 0.5
+      %res = call i4 @llvm.udiv.fix.i4(i4 3, i4 4, i32 1)  ; %res = 2 (or 1) (1.5 / 2 = 0.75)
+
+
 Specialised Arithmetic Intrinsics
 ---------------------------------
 
 .. _i_intr_llvm_canonicalize:
 
 '``llvm.canonicalize.*``' Intrinsic
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 

llvm/include/llvm/CodeGen/ISDOpcodes.h

enum NodeType {
     /// multiplication on 2 integers.
     SMULFIX, UMULFIX,
 
     /// Same as the corresponding unsaturated fixed point instructions, but the
     /// result is clamped between the min and max values representable by the
     /// bits of the first 2 operands.
     SMULFIXSAT, UMULFIXSAT,
 
+    /// RESULT = [US]DIVFIX(LHS, RHS, SCALE) - Perform fixed point division on
+    /// 2 integers with the same width and scale. SCALE represents the scale
+    /// of both operands as fixed point numbers. This SCALE parameter must be a
+    /// constant integer.
+    SDIVFIX, UDIVFIX,
+
     /// Simple binary floating point operators.
     FADD, FSUB, FMUL, FDIV, FREM,
 
     /// Constrained versions of the binary floating point operators.
     /// These will be lowered to the simple operators before final selection.
     /// They are used to limit optimizations while the DAG is being
     /// optimized.
     STRICT_FADD, STRICT_FSUB, STRICT_FMUL, STRICT_FDIV, STRICT_FREM,

llvm/include/llvm/CodeGen/TargetLowering.h

LegalizeAction getFixedPointOperationAction(unsigned Op, EVT VT,
     bool Supported;
     switch (Op) {
     default:
       llvm_unreachable("Unexpected fixed point operation.");
     case ISD::SMULFIX:
     case ISD::SMULFIXSAT:
     case ISD::UMULFIX:
     case ISD::UMULFIXSAT:
+    case ISD::SDIVFIX:
+    case ISD::UDIVFIX:
       Supported = isSupportedFixedPointOperation(Op, VT, Scale);
       break;
     }
 
     return Supported ? Action : Expand;
   }
 
   // If Op is a strict floating-point operation, return the result

   /// Method for building the DAG expansion of ISD::[US][ADD|SUB]SAT. This
   /// method accepts integers as its arguments.
   SDValue expandAddSubSat(SDNode *Node, SelectionDAG &DAG) const;
 
   /// Method for building the DAG expansion of ISD::[U|S]MULFIX[SAT]. This
   /// method accepts integers as its arguments.
   SDValue expandFixedPointMul(SDNode *Node, SelectionDAG &DAG) const;
 
+  /// Method for building the DAG expansion of ISD::[US]DIVFIX. This
+  /// method accepts integers as its arguments.
+  /// Note: This method may fail if the division could not be performed
+  /// within the type. Clients must retry with a wider type if this happens.
+  SDValue expandFixedPointDiv(unsigned Opcode, const SDLoc &dl,
+                              SDValue LHS, SDValue RHS,
+                              unsigned Scale, SelectionDAG &DAG) const;
+
   /// Method for building the DAG expansion of ISD::U(ADD|SUB)O. Expansion
   /// always suceeds and populates the Result and Overflow arguments.
   void expandUADDSUBO(SDNode *Node, SDValue &Result, SDValue &Overflow,
                       SelectionDAG &DAG) const;
 
   /// Method for building the DAG expansion of ISD::S(ADD|SUB)O. Expansion
   /// always suceeds and populates the Result and Overflow arguments.
   void expandSADDSUBO(SDNode *Node, SDValue &Result, SDValue &Overflow,

llvm/include/llvm/IR/Intrinsics.td

 def int_smul_fix : Intrinsic<[llvm_anyint_ty],
                              [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty],
                              [IntrNoMem, IntrSpeculatable, IntrWillReturn, Commutative, ImmArg<2>]>;
 
 def int_umul_fix : Intrinsic<[llvm_anyint_ty],
                              [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty],
                              [IntrNoMem, IntrSpeculatable, IntrWillReturn, Commutative, ImmArg<2>]>;
 
+def int_sdiv_fix : Intrinsic<[llvm_anyint_ty],
+                             [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty],
+                             [IntrNoMem, ImmArg<2>]>;
+
+def int_udiv_fix : Intrinsic<[llvm_anyint_ty],
+                             [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty],
+                             [IntrNoMem, ImmArg<2>]>;
+
 //===------------------- Fixed Point Saturation Arithmetic Intrinsics ----------------===//
 //
 def int_smul_fix_sat : Intrinsic<[llvm_anyint_ty],
                                  [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty],
                                  [IntrNoMem, IntrSpeculatable, IntrWillReturn, Commutative, ImmArg<2>]>;
 def int_umul_fix_sat : Intrinsic<[llvm_anyint_ty],
                                  [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty],
                                  [IntrNoMem, IntrSpeculatable, IntrWillReturn, Commutative, ImmArg<2>]>;

llvm/include/llvm/Target/TargetSelectionDAG.td

def SDTIntShiftDOp: SDTypeProfile<1, 3, [   // fshl, fshr
   SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>, SDTCisInt<0>, SDTCisInt<3>
 ]>;
 def SDTIntSatNoShOp : SDTypeProfile<1, 2, [   // ssat with no shift
   SDTCisSameAs<0, 1>, SDTCisInt<2>
 ]>;
 def SDTIntBinHiLoOp : SDTypeProfile<2, 2, [ // mulhi, mullo, sdivrem, udivrem
   SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>, SDTCisSameAs<0, 3>,SDTCisInt<0>
 ]>;
-def SDTIntScaledBinOp : SDTypeProfile<1, 3, [  // smulfix, umulfix
+def SDTIntScaledBinOp : SDTypeProfile<1, 3, [  // smulfix, sdivfix, etc
   SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>, SDTCisInt<0>, SDTCisInt<3>
 ]>;
 
 def SDTFPBinOp : SDTypeProfile<1, 2, [      // fadd, fmul, etc.
   SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>, SDTCisFP<0>
 ]>;
 def SDTFPSignOp : SDTypeProfile<1, 2, [     // fcopysign.
   SDTCisSameAs<0, 1>, SDTCisFP<0>, SDTCisFP<2>

 def uaddsat    : SDNode<"ISD::UADDSAT"   , SDTIntBinOp, [SDNPCommutative]>;
 def ssubsat    : SDNode<"ISD::SSUBSAT"   , SDTIntBinOp>;
 def usubsat    : SDNode<"ISD::USUBSAT"   , SDTIntBinOp>;
 
 def smulfix    : SDNode<"ISD::SMULFIX"   , SDTIntScaledBinOp, [SDNPCommutative]>;
 def smulfixsat : SDNode<"ISD::SMULFIXSAT", SDTIntScaledBinOp, [SDNPCommutative]>;
 def umulfix    : SDNode<"ISD::UMULFIX"   , SDTIntScaledBinOp, [SDNPCommutative]>;
 def umulfixsat : SDNode<"ISD::UMULFIXSAT", SDTIntScaledBinOp, [SDNPCommutative]>;
+def sdivfix    : SDNode<"ISD::SDIVFIX"   , SDTIntScaledBinOp>;
+def udivfix    : SDNode<"ISD::UDIVFIX"   , SDTIntScaledBinOp>;
 
 def sext_inreg : SDNode<"ISD::SIGN_EXTEND_INREG", SDTExtInreg>;
 def sext_invec : SDNode<"ISD::SIGN_EXTEND_VECTOR_INREG", SDTExtInvec>;
 def zext_invec : SDNode<"ISD::ZERO_EXTEND_VECTOR_INREG", SDTExtInvec>;
 
 def abs        : SDNode<"ISD::ABS"        , SDTIntUnaryOp>;
 def bitreverse : SDNode<"ISD::BITREVERSE" , SDTIntUnaryOp>;
 def bswap      : SDNode<"ISD::BSWAP"      , SDTIntUnaryOp>;

llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp

#endif
   case ISD::SSUBSAT:
   case ISD::USUBSAT: {
     Action = TLI.getOperationAction(Node->getOpcode(), Node->getValueType(0));
     break;
   }
   case ISD::SMULFIX:
   case ISD::SMULFIXSAT:
   case ISD::UMULFIX:
-  case ISD::UMULFIXSAT: {
+  case ISD::UMULFIXSAT:
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX: {
     unsigned Scale = Node->getConstantOperandVal(2);
     Action = TLI.getFixedPointOperationAction(Node->getOpcode(),
                                               Node->getValueType(0), Scale);
     break;
   }
   case ISD::MSCATTER:
     Action = TLI.getOperationAction(Node->getOpcode(),
                     cast<MaskedScatterSDNode>(Node)->getValue().getValueType());
case ISD::USUBSAT:
     Results.push_back(TLI.expandAddSubSat(Node, DAG));
     break;
   case ISD::SMULFIX:
   case ISD::SMULFIXSAT:
   case ISD::UMULFIX:
   case ISD::UMULFIXSAT:
     Results.push_back(TLI.expandFixedPointMul(Node, DAG));
     break;
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX:
+    if (SDValue V = TLI.expandFixedPointDiv(Node->getOpcode(), SDLoc(Node),
+                                            Node->getOperand(0),
+                                            Node->getOperand(1),
+                                            Node->getConstantOperandVal(2),
+                                            DAG)) {
+      Results.push_back(V);
+      break;
+    }
+    // FIXME: We might want to retry here with a wider type if we fail, if that
+    // type is legal.
+    // FIXME: Technically, so long as we only have sdivfixes where BW+Scale is
+    // <= 128 (which is the case for all of the default Embedded-C types),
+    // we will only get here with types and scales that we could always expand
+    // if we were allowed to generate libcalls to division functions of illegal
+    // type. But we cannot do that.
+    llvm_unreachable("Cannot expand DIVFIX!");
   case ISD::ADDCARRY:
   case ISD::SUBCARRY: {
     SDValue LHS = Node->getOperand(0);
     SDValue RHS = Node->getOperand(1);
     SDValue Carry = Node->getOperand(2);
 
     bool IsAdd = Node->getOpcode() == ISD::ADDCARRY;
 

llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp

#endif
   case ISD::SSUBSAT:
   case ISD::USUBSAT:     Res = PromoteIntRes_ADDSUBSAT(N); break;
 
   case ISD::SMULFIX:
   case ISD::SMULFIXSAT:
   case ISD::UMULFIX:
   case ISD::UMULFIXSAT:  Res = PromoteIntRes_MULFIX(N); break;
 
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX:     Res = PromoteIntRes_DIVFIX(N); break;
+
   case ISD::ABS:         Res = PromoteIntRes_ABS(N); break;
 
   case ISD::ATOMIC_LOAD:
     Res = PromoteIntRes_Atomic0(cast<AtomicSDNode>(N)); break;
 
   case ISD::ATOMIC_LOAD_ADD:
   case ISD::ATOMIC_LOAD_SUB:
   case ISD::ATOMIC_LOAD_AND:
if (Saturating) {
     unsigned ShiftOp = Signed ? ISD::SRA : ISD::SRL;
     return DAG.getNode(ShiftOp, dl, PromotedType, Result,
                        DAG.getConstant(DiffSize, dl, ShiftTy));
   }
   return DAG.getNode(N->getOpcode(), dl, PromotedType, Op1Promoted, Op2Promoted,
                      N->getOperand(2));
 }
 
+static SDValue earlyExpandDIVFIX(SDNode *N, SDValue LHS, SDValue RHS,
+                                    unsigned Scale, const TargetLowering &TLI,
+                                    SelectionDAG &DAG) {
+  EVT VT = LHS.getValueType();
+  bool Signed = N->getOpcode() == ISD::SDIVFIX;
+
+  SDLoc dl(N);
+  // See if we can perform the division in this type without widening.
+  if (SDValue V = TLI.expandFixedPointDiv(N->getOpcode(), dl, LHS, RHS, Scale,
+                                          DAG))
+    return V;
+
+  // If that didn't work, double the type width and try again. That must work,
+  // or something is wrong.
+  EVT WideVT = EVT::getIntegerVT(*DAG.getContext(),
+                                 VT.getScalarSizeInBits() * 2);
+  if (Signed) {
+    LHS = DAG.getSExtOrTrunc(LHS, dl, WideVT);
+    RHS = DAG.getSExtOrTrunc(RHS, dl, WideVT);
+  } else {
+    LHS = DAG.getZExtOrTrunc(LHS, dl, WideVT);
+    RHS = DAG.getZExtOrTrunc(RHS, dl, WideVT);
+  }
+
+  // TODO: Saturation.
+
+  SDValue Res = TLI.expandFixedPointDiv(N->getOpcode(), dl, LHS, RHS, Scale,
+                                        DAG);
+  assert(Res && "Expanding DIVFIX with wide type failed?");
+  return DAG.getZExtOrTrunc(Res, dl, VT);
+}
+
+SDValue DAGTypeLegalizer::PromoteIntRes_DIVFIX(SDNode *N) {
+  SDLoc dl(N);
+  SDValue Op1Promoted, Op2Promoted;
+  bool Signed = N->getOpcode() == ISD::SDIVFIX;
+  if (Signed) {
+    Op1Promoted = SExtPromotedInteger(N->getOperand(0));
+    Op2Promoted = SExtPromotedInteger(N->getOperand(1));
+  } else {
+    Op1Promoted = ZExtPromotedInteger(N->getOperand(0));
+    Op2Promoted = ZExtPromotedInteger(N->getOperand(1));
+  }
+  EVT PromotedType = Op1Promoted.getValueType();
+  unsigned Scale = N->getConstantOperandVal(2);
+
+  SDValue Res;
+  // If the type is already legal and the operation is legal in that type, we
+  // should not early expand.
+  if (TLI.isTypeLegal(PromotedType)) {
+    TargetLowering::LegalizeAction Action =
+        TLI.getFixedPointOperationAction(N->getOpcode(), PromotedType, Scale);
+    if (Action == TargetLowering::Legal || Action == TargetLowering::Custom)
+      Res = DAG.getNode(N->getOpcode(), dl, PromotedType, Op1Promoted,
+                        Op2Promoted, N->getOperand(2));
+  }
+
+  if (!Res)
+    Res = earlyExpandDIVFIX(N, Op1Promoted, Op2Promoted, Scale, TLI, DAG);
+
+  // TODO: Saturation.
+
+  return Res;
+}
+
 SDValue DAGTypeLegalizer::PromoteIntRes_SADDSUBO(SDNode *N, unsigned ResNo) {
   if (ResNo == 1)
     return PromoteIntRes_Overflow(N);
 
   // The operation overflowed iff the result in the larger type is not the
   // sign extension of its truncation to the original type.
   SDValue LHS = SExtPromotedInteger(N->getOperand(0));
   SDValue RHS = SExtPromotedInteger(N->getOperand(1));
bool DAGTypeLegalizer::PromoteIntegerOperand(SDNode *N, unsigned OpNo) {
   case ISD::FRAMEADDR:
   case ISD::RETURNADDR: Res = PromoteIntOp_FRAMERETURNADDR(N); break;
 
   case ISD::PREFETCH: Res = PromoteIntOp_PREFETCH(N, OpNo); break;
 
   case ISD::SMULFIX:
   case ISD::SMULFIXSAT:
   case ISD::UMULFIX:
-  case ISD::UMULFIXSAT: Res = PromoteIntOp_MULFIX(N); break;
+  case ISD::UMULFIXSAT:
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX: Res = PromoteIntOp_FIX(N); break;
 
   case ISD::FPOWI: Res = PromoteIntOp_FPOWI(N); break;
 
   case ISD::VECREDUCE_ADD:
   case ISD::VECREDUCE_MUL:
   case ISD::VECREDUCE_AND:
   case ISD::VECREDUCE_OR:
   case ISD::VECREDUCE_XOR:
SDValue DAGTypeLegalizer::PromoteIntOp_ADDSUBCARRY(SDNode *N, unsigned OpNo) {
   SDValue Carry = N->getOperand(2);
   SDLoc DL(N);
 
   Carry = PromoteTargetBoolean(Carry, LHS.getValueType());
 
   return SDValue(DAG.UpdateNodeOperands(N, LHS, RHS, Carry), 0);
 }
 
-SDValue DAGTypeLegalizer::PromoteIntOp_MULFIX(SDNode *N) {
+SDValue DAGTypeLegalizer::PromoteIntOp_FIX(SDNode *N) {
   SDValue Op2 = ZExtPromotedInteger(N->getOperand(2));
   return SDValue(
       DAG.UpdateNodeOperands(N, N->getOperand(0), N->getOperand(1), Op2), 0);
 }
 
 SDValue DAGTypeLegalizer::PromoteIntOp_FRAMERETURNADDR(SDNode *N) {
   // Promote the RETURNADDR/FRAMEADDR argument to a supported integer width.
   SDValue Op = ZExtPromotedInteger(N->getOperand(0));
#endif
   case ISD::SSUBSAT:
   case ISD::USUBSAT: ExpandIntRes_ADDSUBSAT(N, Lo, Hi); break;
 
   case ISD::SMULFIX:
   case ISD::SMULFIXSAT:
   case ISD::UMULFIX:
   case ISD::UMULFIXSAT: ExpandIntRes_MULFIX(N, Lo, Hi); break;
 
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX: ExpandIntRes_DIVFIX(N, Lo, Hi); break;
+
   case ISD::VECREDUCE_ADD:
   case ISD::VECREDUCE_MUL:
   case ISD::VECREDUCE_AND:
   case ISD::VECREDUCE_OR:
   case ISD::VECREDUCE_XOR:
   case ISD::VECREDUCE_SMAX:
   case ISD::VECREDUCE_SMIN:
   case ISD::VECREDUCE_UMAX:
void DAGTypeLegalizer::ExpandIntRes_MULFIX(SDNode *N, SDValue &Lo,
   Hi = DAG.getSelect(dl, NVT, SatMax, DAG.getConstant(MaxHi, dl, NVT), Hi);
   Lo = DAG.getSelect(dl, NVT, SatMax, DAG.getConstant(MaxLo, dl, NVT), Lo);
   // Saturate to signed minimum.
   APInt MinHi = APInt::getSignedMinValue(NVTSize);
   Hi = DAG.getSelect(dl, NVT, SatMin, DAG.getConstant(MinHi, dl, NVT), Hi);
   Lo = DAG.getSelect(dl, NVT, SatMin, NVTZero, Lo);
 }
 
+void DAGTypeLegalizer::ExpandIntRes_DIVFIX(SDNode *N, SDValue &Lo,
+                                           SDValue &Hi) {
+  SDValue Res = earlyExpandDIVFIX(N, N->getOperand(0), N->getOperand(1),
+                                  N->getConstantOperandVal(2), TLI, DAG);
+  SplitInteger(Res, Lo, Hi);
+}
+
 void DAGTypeLegalizer::ExpandIntRes_SADDSUBO(SDNode *Node,
                                              SDValue &Lo, SDValue &Hi) {
   SDValue LHS = Node->getOperand(0);
   SDValue RHS = Node->getOperand(1);
   SDLoc dl(Node);
 
   // Expand the result by simply replacing it with the equivalent
   // non-overflow-checking operation.

llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h

private:
   SDValue PromoteIntRes_TRUNCATE(SDNode *N);
   SDValue PromoteIntRes_UADDSUBO(SDNode *N, unsigned ResNo);
   SDValue PromoteIntRes_ADDSUBCARRY(SDNode *N, unsigned ResNo);
   SDValue PromoteIntRes_UNDEF(SDNode *N);
   SDValue PromoteIntRes_VAARG(SDNode *N);
   SDValue PromoteIntRes_XMULO(SDNode *N, unsigned ResNo);
   SDValue PromoteIntRes_ADDSUBSAT(SDNode *N);
   SDValue PromoteIntRes_MULFIX(SDNode *N);
+  SDValue PromoteIntRes_DIVFIX(SDNode *N);
   SDValue PromoteIntRes_FLT_ROUNDS(SDNode *N);
   SDValue PromoteIntRes_VECREDUCE(SDNode *N);
   SDValue PromoteIntRes_ABS(SDNode *N);
 
   // Integer Operand Promotion.
   bool PromoteIntegerOperand(SDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_ANY_EXTEND(SDNode *N);
   SDValue PromoteIntOp_ATOMIC_STORE(AtomicSDNode *N);
private:
   SDValue PromoteIntOp_ZERO_EXTEND(SDNode *N);
   SDValue PromoteIntOp_MSTORE(MaskedStoreSDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_MLOAD(MaskedLoadSDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_MSCATTER(MaskedScatterSDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_MGATHER(MaskedGatherSDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_ADDSUBCARRY(SDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_FRAMERETURNADDR(SDNode *N);
   SDValue PromoteIntOp_PREFETCH(SDNode *N, unsigned OpNo);
-  SDValue PromoteIntOp_MULFIX(SDNode *N);
+  SDValue PromoteIntOp_FIX(SDNode *N);
   SDValue PromoteIntOp_FPOWI(SDNode *N);
   SDValue PromoteIntOp_VECREDUCE(SDNode *N);
 
   void PromoteSetCCOperands(SDValue &LHS,SDValue &RHS, ISD::CondCode Code);
 
   //===--------------------------------------------------------------------===//
   // Integer Expansion Support: LegalizeIntegerTypes.cpp
   //===--------------------------------------------------------------------===//
private:
 
   void ExpandIntRes_MINMAX            (SDNode *N, SDValue &Lo, SDValue &Hi);
 
   void ExpandIntRes_SADDSUBO          (SDNode *N, SDValue &Lo, SDValue &Hi);
   void ExpandIntRes_UADDSUBO          (SDNode *N, SDValue &Lo, SDValue &Hi);
   void ExpandIntRes_XMULO             (SDNode *N, SDValue &Lo, SDValue &Hi);
   void ExpandIntRes_ADDSUBSAT         (SDNode *N, SDValue &Lo, SDValue &Hi);
   void ExpandIntRes_MULFIX            (SDNode *N, SDValue &Lo, SDValue &Hi);
+  void ExpandIntRes_DIVFIX            (SDNode *N, SDValue &Lo, SDValue &Hi);
 
   void ExpandIntRes_ATOMIC_LOAD       (SDNode *N, SDValue &Lo, SDValue &Hi);
   void ExpandIntRes_VECREDUCE         (SDNode *N, SDValue &Lo, SDValue &Hi);
 
   void ExpandShiftByConstant(SDNode *N, const APInt &Amt,
                              SDValue &Lo, SDValue &Hi);
   bool ExpandShiftWithKnownAmountBit(SDNode *N, SDValue &Lo, SDValue &Hi);
   bool ExpandShiftWithUnknownAmountBit(SDNode *N, SDValue &Lo, SDValue &Hi);
private:
   SDValue ScalarizeVecRes_SCALAR_TO_VECTOR(SDNode *N);
   SDValue ScalarizeVecRes_VSELECT(SDNode *N);
   SDValue ScalarizeVecRes_SELECT(SDNode *N);
   SDValue ScalarizeVecRes_SELECT_CC(SDNode *N);
   SDValue ScalarizeVecRes_SETCC(SDNode *N);
   SDValue ScalarizeVecRes_UNDEF(SDNode *N);
   SDValue ScalarizeVecRes_VECTOR_SHUFFLE(SDNode *N);
 
-  SDValue ScalarizeVecRes_MULFIX(SDNode *N);
+  SDValue ScalarizeVecRes_FIX(SDNode *N);
 
   // Vector Operand Scalarization: <1 x ty> -> ty.
   bool ScalarizeVectorOperand(SDNode *N, unsigned OpNo);
   SDValue ScalarizeVecOp_BITCAST(SDNode *N);
   SDValue ScalarizeVecOp_UnaryOp(SDNode *N);
   SDValue ScalarizeVecOp_UnaryOp_StrictFP(SDNode *N);
   SDValue ScalarizeVecOp_CONCAT_VECTORS(SDNode *N);
   SDValue ScalarizeVecOp_EXTRACT_VECTOR_ELT(SDNode *N);
private:
   void SplitVecRes_UnaryOp(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_ExtendOp(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_InregOp(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_ExtVecInRegOp(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_StrictFPOp(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_OverflowOp(SDNode *N, unsigned ResNo,
                               SDValue &Lo, SDValue &Hi);
 
-  void SplitVecRes_MULFIX(SDNode *N, SDValue &Lo, SDValue &Hi);
+  void SplitVecRes_FIX(SDNode *N, SDValue &Lo, SDValue &Hi);
 
   void SplitVecRes_BITCAST(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_BUILD_VECTOR(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_CONCAT_VECTORS(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_EXTRACT_SUBVECTOR(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_INSERT_SUBVECTOR(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_FPOWI(SDNode *N, SDValue &Lo, SDValue &Hi);
   void SplitVecRes_FCOPYSIGN(SDNode *N, SDValue &Lo, SDValue &Hi);

llvm/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp

class VectorLegalizer {
   SDValue ExpandFunnelShift(SDValue Op);
   SDValue ExpandROT(SDValue Op);
   SDValue ExpandFMINNUM_FMAXNUM(SDValue Op);
   SDValue ExpandUADDSUBO(SDValue Op);
   SDValue ExpandSADDSUBO(SDValue Op);
   SDValue ExpandMULO(SDValue Op);
   SDValue ExpandAddSubSat(SDValue Op);
   SDValue ExpandFixedPointMul(SDValue Op);
+  SDValue ExpandFixedPointDiv(SDValue Op);
   SDValue ExpandStrictFPOp(SDValue Op);
 
   SDValue UnrollStrictFPOp(SDValue Op);
 
   /// Implements vector promotion.
   ///
   /// This is essentially just bitcasting the operands to a different type and
   /// bitcasting the result back to the original type.
#include "llvm/IR/ConstrainedOps.def"
   case ISD::UADDSAT:
   case ISD::SSUBSAT:
   case ISD::USUBSAT:
     Action = TLI.getOperationAction(Node->getOpcode(), Node->getValueType(0));
     break;
   case ISD::SMULFIX:
   case ISD::SMULFIXSAT:
   case ISD::UMULFIX:
-  case ISD::UMULFIXSAT: {
+  case ISD::UMULFIXSAT:
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX: {
     unsigned Scale = Node->getConstantOperandVal(2);
     Action = TLI.getFixedPointOperationAction(Node->getOpcode(),
                                               Node->getValueType(0), Scale);
     break;
   }
   case ISD::SINT_TO_FP:
   case ISD::UINT_TO_FP:
   case ISD::VECREDUCE_ADD:
case ISD::UMULFIX:
     return ExpandFixedPointMul(Op);
   case ISD::SMULFIXSAT:
   case ISD::UMULFIXSAT:
     // FIXME: We do not expand SMULFIXSAT/UMULFIXSAT here yet, not sure exactly
     // why. Maybe it results in worse codegen compared to the unroll for some
     // targets? This should probably be investigated. And if we still prefer to
     // unroll an explanation could be helpful.
     return DAG.UnrollVectorOp(Op.getNode());
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX:
+    return ExpandFixedPointDiv(Op);
 #define INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC, DAGN)                   \
   case ISD::STRICT_##DAGN:
 #include "llvm/IR/ConstrainedOps.def"
     return ExpandStrictFPOp(Op);
   case ISD::VECREDUCE_ADD:
   case ISD::VECREDUCE_MUL:
   case ISD::VECREDUCE_AND:
   case ISD::VECREDUCE_OR:

 }
 
 SDValue VectorLegalizer::ExpandFixedPointMul(SDValue Op) {
   if (SDValue Expanded = TLI.expandFixedPointMul(Op.getNode(), DAG))
     return Expanded;
   return DAG.UnrollVectorOp(Op.getNode());
 }
 
+SDValue VectorLegalizer::ExpandFixedPointDiv(SDValue Op) {
+  SDNode *N = Op.getNode();
+  if (SDValue Expanded = TLI.expandFixedPointDiv(N->getOpcode(), SDLoc(N),
+          N->getOperand(0), N->getOperand(1), N->getConstantOperandVal(2), DAG))
+    return Expanded;
+  return DAG.UnrollVectorOp(N);
+}
+
 SDValue VectorLegalizer::ExpandStrictFPOp(SDValue Op) {
   if (Op.getOpcode() == ISD::STRICT_UINT_TO_FP)
     return ExpandUINT_TO_FLOAT(Op);
   if (Op.getOpcode() == ISD::STRICT_FP_TO_UINT)
     return ExpandFP_TO_UINT(Op);
 
   return UnrollStrictFPOp(Op);
 }

llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp

#include "llvm/IR/ConstrainedOps.def"
   case ISD::UMULO:
   case ISD::SMULO:
     R = ScalarizeVecRes_OverflowOp(N, ResNo);
     break;
   case ISD::SMULFIX:
   case ISD::SMULFIXSAT:
   case ISD::UMULFIX:
   case ISD::UMULFIXSAT:
-    R = ScalarizeVecRes_MULFIX(N);
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX:
+    R = ScalarizeVecRes_FIX(N);
     break;
   }
 
   // If R is null, the sub-method took care of registering the result.
   if (R.getNode())
     SetScalarizedVector(SDValue(N, ResNo), R);
 }
 
 SDValue DAGTypeLegalizer::ScalarizeVecRes_BinOp(SDNode *N) {
   SDValue LHS = GetScalarizedVector(N->getOperand(0));
   SDValue RHS = GetScalarizedVector(N->getOperand(1));
   return DAG.getNode(N->getOpcode(), SDLoc(N),
                      LHS.getValueType(), LHS, RHS, N->getFlags());
 }
 
 SDValue DAGTypeLegalizer::ScalarizeVecRes_TernaryOp(SDNode *N) {
   SDValue Op0 = GetScalarizedVector(N->getOperand(0));
   SDValue Op1 = GetScalarizedVector(N->getOperand(1));
   SDValue Op2 = GetScalarizedVector(N->getOperand(2));
   return DAG.getNode(N->getOpcode(), SDLoc(N),
                      Op0.getValueType(), Op0, Op1, Op2);
 }
 
-SDValue DAGTypeLegalizer::ScalarizeVecRes_MULFIX(SDNode *N) {
+SDValue DAGTypeLegalizer::ScalarizeVecRes_FIX(SDNode *N) {
   SDValue Op0 = GetScalarizedVector(N->getOperand(0));
   SDValue Op1 = GetScalarizedVector(N->getOperand(1));
   SDValue Op2 = N->getOperand(2);
   return DAG.getNode(N->getOpcode(), SDLoc(N), Op0.getValueType(), Op0, Op1,
                      Op2);
 }
 
 SDValue DAGTypeLegalizer::ScalarizeVecRes_StrictFPOp(SDNode *N) {
#include "llvm/IR/ConstrainedOps.def"
   case ISD::UMULO:
   case ISD::SMULO:
     SplitVecRes_OverflowOp(N, ResNo, Lo, Hi);
     break;
   case ISD::SMULFIX:
   case ISD::SMULFIXSAT:
   case ISD::UMULFIX:
   case ISD::UMULFIXSAT:
-    SplitVecRes_MULFIX(N, Lo, Hi);
+  case ISD::SDIVFIX:
+  case ISD::UDIVFIX:
+    SplitVecRes_FIX(N, Lo, Hi);
     break;
   }
 
   // If Lo/Hi is null, the sub-method took care of registering results etc.
   if (Lo.getNode())
     SetSplitVector(SDValue(N, ResNo), Lo, Hi);
 }
 
void DAGTypeLegalizer::SplitVecRes_TernaryOp(SDNode *N, SDValue &Lo,
   SDLoc dl(N);
 
   Lo = DAG.getNode(N->getOpcode(), dl, Op0Lo.getValueType(),
                    Op0Lo, Op1Lo, Op2Lo);
   Hi = DAG.getNode(N->getOpcode(), dl, Op0Hi.getValueType(),
                    Op0Hi, Op1Hi, Op2Hi);
 }
 
-void DAGTypeLegalizer::SplitVecRes_MULFIX(SDNode *N, SDValue &Lo, SDValue &Hi) {
+void DAGTypeLegalizer::SplitVecRes_FIX(SDNode *N, SDValue &Lo, SDValue &Hi) {
   SDValue LHSLo, LHSHi;
   GetSplitVector(N->getOperand(0), LHSLo, LHSHi);
   SDValue RHSLo, RHSHi;
   GetSplitVector(N->getOperand(1), RHSLo, RHSHi);
   SDLoc dl(N);
   SDValue Op2 = N->getOperand(2);
 
   unsigned Opcode = N->getOpcode();

llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

if (!OptForSize ||
       return Res;
     }
   }
 
   // Otherwise, expand to a libcall.
   return DAG.getNode(ISD::FPOWI, DL, LHS.getValueType(), LHS, RHS);
 }
 
+static SDValue expandDivFix(unsigned Opcode, const SDLoc &DL,
+                            SDValue LHS, SDValue RHS, SDValue Scale,
+                            SelectionDAG &DAG, const TargetLowering &TLI) {
+  EVT VT = LHS.getValueType();
+  bool Signed = Opcode == ISD::SDIVFIX;
+  LLVMContext &Ctx = *DAG.getContext();
+
+  // If the type is legal but the operation isn't, this node might survive all
+  // the way to operation legalization. If we end up there and we do not have
+  // the ability to widen the type (if VT*2 is not legal), we cannot expand the
+  // node.
+
+  // Coax the legalizer into expanding the node during type legalization instead
+  // by bumping the size by one bit. This will force it to Promote, enabling the
+  // early expansion and avoiding the need to expand later.
+
+  // We don't have to do this if Scale is 0; that can always be expanded.
+
+  // FIXME: We wouldn't have to do this (or any of the early
+  // expansion/promotion) if it was possible to expand a libcall of an
+  // illegal type during operation legalization. But it's not, so things
+  // get a bit hacky.
+  unsigned ScaleInt = cast<ConstantSDNode>(Scale)->getZExtValue();
+  if (ScaleInt > 0 &&
+      (TLI.isTypeLegal(VT) ||
+       (VT.isVector() && TLI.isTypeLegal(VT.getVectorElementType())))) {
+    TargetLowering::LegalizeAction Action = TLI.getFixedPointOperationAction(
+        Opcode, VT, ScaleInt);
+    if (Action != TargetLowering::Legal && Action != TargetLowering::Custom) {
+      EVT PromVT;
+      if (VT.isScalarInteger())
+        PromVT = EVT::getIntegerVT(Ctx, VT.getSizeInBits() + 1);
+      else if (VT.isVector()) {
+        PromVT = VT.getVectorElementType();
+        PromVT = EVT::getIntegerVT(Ctx, PromVT.getSizeInBits() + 1);
+        PromVT = EVT::getVectorVT(Ctx, PromVT, VT.getVectorElementCount());
+      } else
+        llvm_unreachable("Wrong VT for DIVFIX?");
+      if (Signed) {
+        LHS = DAG.getSExtOrTrunc(LHS, DL, PromVT);
+        RHS = DAG.getSExtOrTrunc(RHS, DL, PromVT);
+      } else {
+        LHS = DAG.getZExtOrTrunc(LHS, DL, PromVT);
+        RHS = DAG.getZExtOrTrunc(RHS, DL, PromVT);
+      }
+      // TODO: Saturation.
+      SDValue Res = DAG.getNode(Opcode, DL, PromVT, LHS, RHS, Scale);
+      return DAG.getZExtOrTrunc(Res, DL, VT);
+    }
+  }
+
+  return DAG.getNode(Opcode, DL, VT, LHS, RHS, Scale);
+}
+
 // getUnderlyingArgRegs - Find underlying registers used for a truncated,
 // bitcasted, or split argument. Returns a list of <Register, size in bits>
 static void
 getUnderlyingArgRegs(SmallVectorImpl<std::pair<unsigned, unsigned>> &Regs,
                      const SDValue &N) {
   switch (N.getOpcode()) {
   case ISD::CopyFromReg: {
     SDValue Op = N.getOperand(1);

 }
 
 static unsigned FixedPointIntrinsicToOpcode(unsigned Intrinsic) {
   switch (Intrinsic) {
   case Intrinsic::smul_fix:
     return ISD::SMULFIX;
   case Intrinsic::umul_fix:
     return ISD::UMULFIX;
+  case Intrinsic::smul_fix_sat:
+    return ISD::SMULFIXSAT;
+  case Intrinsic::umul_fix_sat:
+    return ISD::UMULFIXSAT;
+  case Intrinsic::sdiv_fix:
+    return ISD::SDIVFIX;
+  case Intrinsic::udiv_fix:
+    return ISD::UDIVFIX;
   default:
     llvm_unreachable("Unhandled fixed point intrinsic");
   }
 }
 
 void SelectionDAGBuilder::lowerCallToExternalSymbol(const CallInst &I,
                                            const char *FunctionName) {
   assert(FunctionName && "FunctionName must not be nullptr");
#include "llvm/IR/ConstrainedOps.def"
   }
   case Intrinsic::usub_sat: {
     SDValue Op1 = getValue(I.getArgOperand(0));
     SDValue Op2 = getValue(I.getArgOperand(1));
     setValue(&I, DAG.getNode(ISD::USUBSAT, sdl, Op1.getValueType(), Op1, Op2));
     return;
   }
   case Intrinsic::smul_fix:
-  case Intrinsic::umul_fix: {
+  case Intrinsic::umul_fix:
+  case Intrinsic::smul_fix_sat:
+  case Intrinsic::umul_fix_sat: {
     SDValue Op1 = getValue(I.getArgOperand(0));
     SDValue Op2 = getValue(I.getArgOperand(1));
     SDValue Op3 = getValue(I.getArgOperand(2));
     setValue(&I, DAG.getNode(FixedPointIntrinsicToOpcode(Intrinsic), sdl,
                              Op1.getValueType(), Op1, Op2, Op3));
     return;
   }
-  case Intrinsic::smul_fix_sat: {
-    SDValue Op1 = getValue(I.getArgOperand(0));
-    SDValue Op2 = getValue(I.getArgOperand(1));
-    SDValue Op3 = getValue(I.getArgOperand(2));
-    setValue(&I, DAG.getNode(ISD::SMULFIXSAT, sdl, Op1.getValueType(), Op1, Op2,
-                             Op3));
-    return;
-  }
-  case Intrinsic::umul_fix_sat: {
+  case Intrinsic::sdiv_fix:
+  case Intrinsic::udiv_fix: {
     SDValue Op1 = getValue(I.getArgOperand(0));
     SDValue Op2 = getValue(I.getArgOperand(1));
     SDValue Op3 = getValue(I.getArgOperand(2));
-    setValue(&I, DAG.getNode(ISD::UMULFIXSAT, sdl, Op1.getValueType(), Op1, Op2,
-                             Op3));
+    setValue(&I, expandDivFix(FixedPointIntrinsicToOpcode(Intrinsic), sdl,
+                              Op1, Op2, Op3, DAG, TLI));
     return;
   }
   case Intrinsic::stacksave: {
     SDValue Op = getRoot();
     Res = DAG.getNode(
         ISD::STACKSAVE, sdl,
         DAG.getVTList(TLI.getPointerTy(DAG.getDataLayout()), MVT::Other), Op);
     setValue(&I, Res);

llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp

#endif
   case ISD::SSUBSAT:                    return "ssubsat";
   case ISD::USUBSAT:                    return "usubsat";
 
   case ISD::SMULFIX:                    return "smulfix";
   case ISD::SMULFIXSAT:                 return "smulfixsat";
   case ISD::UMULFIX:                    return "umulfix";
   case ISD::UMULFIXSAT:                 return "umulfixsat";
 
+  case ISD::SDIVFIX:                    return "sdivfix";
+  case ISD::UDIVFIX:                    return "udivfix";
+
   // Conversion operators.
   case ISD::SIGN_EXTEND:                return "sign_extend";
   case ISD::ZERO_EXTEND:                return "zero_extend";
   case ISD::ANY_EXTEND:                 return "any_extend";
   case ISD::SIGN_EXTEND_INREG:          return "sign_extend_inreg";
   case ISD::ANY_EXTEND_VECTOR_INREG:    return "any_extend_vector_inreg";
   case ISD::SIGN_EXTEND_VECTOR_INREG:   return "sign_extend_vector_inreg";
   case ISD::ZERO_EXTEND_VECTOR_INREG:   return "zero_extend_vector_inreg";

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

TargetLowering::expandFixedPointMul(SDNode *Node, SelectionDAG &DAG) const {
   // which is the same as if (HI < (-1 << (Scale - 1))
   SDValue HighMask =
       DAG.getConstant(APInt::getHighBitsSet(VTSize, VTSize - Scale + 1),
                       dl, VT);
   Result = DAG.getSelectCC(dl, Hi, HighMask, SatMin, Result, ISD::SETLT);
   return Result;
 }
 
+SDValue
+TargetLowering::expandFixedPointDiv(unsigned Opcode, const SDLoc &dl,
+                                    SDValue LHS, SDValue RHS,
+                                    unsigned Scale, SelectionDAG &DAG) const {
+  assert((Opcode == ISD::SDIVFIX ||
+          Opcode == ISD::UDIVFIX) &&
+         "Expected a fixed point division opcode");
+
+  EVT VT = LHS.getValueType();
+  bool Signed = Opcode == ISD::SDIVFIX;
+  EVT BoolVT = getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), VT);
+
+  // If there is enough room in the type to upscale the LHS or downscale the
+  // RHS before the division, we can perform it in this type without having to
+  // resize. For signed operations, the LHS headroom is the number of
+  // redundant sign bits, and for unsigned ones it is the number of zeroes.
+  // The headroom for the RHS is the number of trailing zeroes.
+  unsigned LHSLead = Signed ? DAG.ComputeNumSignBits(LHS) - 1
+                            : DAG.computeKnownBits(LHS).countMinLeadingZeros();
+  unsigned RHSTrail = DAG.computeKnownBits(RHS).countMinTrailingZeros();
+
+  if (LHSLead + RHSTrail < Scale)
+    return SDValue();
+
+  unsigned LHSShift = std::min(LHSLead, Scale);
+  unsigned RHSShift = Scale - LHSShift;
+
+  // At this point, we know that if we shift the LHS up by LHSShift and the
+  // RHS down by RHSShift, we can emit a regular division with a final scaling
+  // factor of Scale.
+
+  EVT ShiftTy = getShiftAmountTy(VT, DAG.getDataLayout());
+  if (LHSShift)
+    LHS = DAG.getNode(ISD::SHL, dl, VT, LHS,
+                      DAG.getConstant(LHSShift, dl, ShiftTy));
+  if (RHSShift)
+    RHS = DAG.getNode(Signed ? ISD::SRA : ISD::SRL, dl, VT, RHS,
+                      DAG.getConstant(RHSShift, dl, ShiftTy));
+
+  SDValue Quot;
+  if (Signed) {
+    // For signed operations, if the resulting quotient is negative and the
+    // remainder is nonzero, subtract 1 from the quotient to round towards
+    // negative infinity.
+    SDValue Rem;
+    // FIXME: Ideally we would always produce an SDIVREM here, but if the
+    // type isn't legal, SDIVREM cannot be expanded. There is no reason why
+    // we couldn't just form a libcall, but the type legalizer doesn't do it.
+    if (isTypeLegal(VT) &&
+        isOperationLegalOrCustom(ISD::SDIVREM, VT)) {
+      Quot = DAG.getNode(ISD::SDIVREM, dl,
+                         DAG.getVTList(VT, VT),
+                         LHS, RHS);
+      Rem = Quot.getValue(1);
+      Quot = Quot.getValue(0);
+    } else {
+      Quot = DAG.getNode(ISD::SDIV, dl, VT,
+                         LHS, RHS);
+      Rem = DAG.getNode(ISD::SREM, dl, VT,
+                        LHS, RHS);
+    }
+    SDValue Zero = DAG.getConstant(0, dl, VT);
+    SDValue RemNonZero = DAG.getSetCC(dl, BoolVT, Rem, Zero, ISD::SETNE);
+    SDValue LHSNeg = DAG.getSetCC(dl, BoolVT, LHS, Zero, ISD::SETLT);
+    SDValue RHSNeg = DAG.getSetCC(dl, BoolVT, RHS, Zero, ISD::SETLT);
+    SDValue QuotNeg = DAG.getNode(ISD::XOR, dl, BoolVT, LHSNeg, RHSNeg);
+    SDValue Sub1 = DAG.getNode(ISD::SUB, dl, VT, Quot,
+                               DAG.getConstant(1, dl, VT));
+    Quot = DAG.getSelect(dl, VT,
+                         DAG.getNode(ISD::AND, dl, BoolVT, RemNonZero, QuotNeg),
+                         Sub1, Quot);
+  } else
+    Quot = DAG.getNode(ISD::UDIV, dl, VT,
+                       LHS, RHS);
+
+  // TODO: Saturation.
+
+  return Quot;
+}
+
 void TargetLowering::expandUADDSUBO(
     SDNode *Node, SDValue &Result, SDValue &Overflow, SelectionDAG &DAG) const {
   SDLoc dl(Node);
   SDValue LHS = Node->getOperand(0);
   SDValue RHS = Node->getOperand(1);
   bool IsAdd = Node->getOpcode() == ISD::UADDO;
 
   // If ADD/SUBCARRY is legal, use that instead.

llvm/lib/CodeGen/TargetLoweringBase.cpp

for (MVT VT : MVT::all_valuetypes()) {
     setOperationAction(ISD::SADDSAT, VT, Expand);
     setOperationAction(ISD::UADDSAT, VT, Expand);
     setOperationAction(ISD::SSUBSAT, VT, Expand);
     setOperationAction(ISD::USUBSAT, VT, Expand);
     setOperationAction(ISD::SMULFIX, VT, Expand);
     setOperationAction(ISD::SMULFIXSAT, VT, Expand);
     setOperationAction(ISD::UMULFIX, VT, Expand);
     setOperationAction(ISD::UMULFIXSAT, VT, Expand);
+    setOperationAction(ISD::SDIVFIX, VT, Expand);
+    setOperationAction(ISD::UDIVFIX, VT, Expand);
 
     // Overflow operations default to expand
     setOperationAction(ISD::SADDO, VT, Expand);
     setOperationAction(ISD::SSUBO, VT, Expand);
     setOperationAction(ISD::UADDO, VT, Expand);
     setOperationAction(ISD::USUBO, VT, Expand);
     setOperationAction(ISD::SMULO, VT, Expand);
     setOperationAction(ISD::UMULO, VT, Expand);

llvm/lib/IR/Verifier.cpp

case Intrinsic::usub_sat: {
     Assert(Op2->getType()->isIntOrIntVectorTy(),
            "second operand of [us][add|sub]_sat must be an int type or vector "
            "of ints");
     break;
   }
   case Intrinsic::smul_fix:
   case Intrinsic::smul_fix_sat:
   case Intrinsic::umul_fix:
-  case Intrinsic::umul_fix_sat: {
+  case Intrinsic::umul_fix_sat:
+  case Intrinsic::sdiv_fix:
+  case Intrinsic::udiv_fix: {
     Value *Op1 = Call.getArgOperand(0);
     Value *Op2 = Call.getArgOperand(1);
     Assert(Op1->getType()->isIntOrIntVectorTy(),
-           "first operand of [us]mul_fix[_sat] must be an int type or vector "
-           "of ints");
+           "first operand of [us][mul|div]_fix[_sat] must be an int type or "
+           "vector of ints");
     Assert(Op2->getType()->isIntOrIntVectorTy(),
-           "second operand of [us]mul_fix_[sat] must be an int type or vector "
-           "of ints");
+           "second operand of [us][mul|div]_fix[_sat] must be an int type or "
+           "vector of ints");
 
     auto *Op3 = cast<ConstantInt>(Call.getArgOperand(2));
     Assert(Op3->getType()->getBitWidth() <= 32,
-           "third argument of [us]mul_fix[_sat] must fit within 32 bits");
+           "third argument of [us][mul|div]_fix[_sat] must fit within 32 bits");
 
-    if (ID == Intrinsic::smul_fix || ID == Intrinsic::smul_fix_sat) {
+    if (ID == Intrinsic::smul_fix || ID == Intrinsic::smul_fix_sat ||
+        ID == Intrinsic::sdiv_fix) {
       Assert(
           Op3->getZExtValue() < Op1->getType()->getScalarSizeInBits(),
-          "the scale of smul_fix[_sat] must be less than the width of the operands");
+          "the scale of s[mul|div]_fix[_sat] must be less than the width of "
+          "the operands");
     } else {
       Assert(Op3->getZExtValue() <= Op1->getType()->getScalarSizeInBits(),
-             "the scale of umul_fix[_sat] must be less than or equal to the width of "
-             "the operands");
+             "the scale of u[mul|div]_fix[_sat] must be less than or equal "
+             "to the width of the operands");
     }
     break;
   }
   case Intrinsic::lround:
   case Intrinsic::llround:
   case Intrinsic::lrint:
   case Intrinsic::llrint: {
     Type *ValTy = Call.getArgOperand(0)->getType();

llvm/test/CodeGen/X86/sdiv_fix.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-linux | FileCheck %s --check-prefix=X64
+; RUN: llc < %s -mtriple=i686 -mattr=cmov | FileCheck %s --check-prefix=X86
+
+declare  i4  @llvm.sdiv.fix.i4   (i4,  i4,  i32)
+declare  i15 @llvm.sdiv.fix.i15  (i15, i15, i32)
+declare  i16 @llvm.sdiv.fix.i16  (i16, i16, i32)
+declare  i18 @llvm.sdiv.fix.i18  (i18, i18, i32)
+declare  i64 @llvm.sdiv.fix.i64  (i64, i64, i32)
+declare  <4 x i32> @llvm.sdiv.fix.v4i32(<4 x i32>, <4 x i32>, i32)
+
+define i16 @func(i16 %x, i16 %y) nounwind {
+; X64-LABEL: func:
+; X64:       # %bb.0:
+; X64-NEXT:    movswl %si, %esi
+; X64-NEXT:    movswl %di, %ecx
+; X64-NEXT:    shll $7, %ecx
+; X64-NEXT:    movl %ecx, %eax
+; X64-NEXT:    cltd
+; X64-NEXT:    idivl %esi
+; X64-NEXT:    # kill: def $eax killed $eax def $rax
+; X64-NEXT:    leal -1(%rax), %edi
+; X64-NEXT:    testl %esi, %esi
+; X64-NEXT:    sets %sil
+; X64-NEXT:    testl %ecx, %ecx
+; X64-NEXT:    sets %cl
+; X64-NEXT:    xorb %sil, %cl
+; X64-NEXT:    testl %edx, %edx
+; X64-NEXT:    setne %dl
+; X64-NEXT:    testb %cl, %dl
+; X64-NEXT:    cmovnel %edi, %eax
+; X64-NEXT:    # kill: def $ax killed $ax killed $rax
+; X64-NEXT:    retq
+;
+; X86-LABEL: func:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movswl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movswl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    shll $7, %ecx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    cltd
+; X86-NEXT:    idivl %esi
+; X86-NEXT:    leal -1(%eax), %edi
+; X86-NEXT:    testl %esi, %esi
+; X86-NEXT:    sets %bl
+; X86-NEXT:    testl %ecx, %ecx
+; X86-NEXT:    sets %cl
+; X86-NEXT:    xorb %bl, %cl
+; X86-NEXT:    testl %edx, %edx
+; X86-NEXT:    setne %dl
+; X86-NEXT:    testb %cl, %dl
+; X86-NEXT:    cmovnel %edi, %eax
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    retl
+  %tmp = call i16 @llvm.sdiv.fix.i16(i16 %x, i16 %y, i32 7)
+  ret i16 %tmp
+}
+
+define i16 @func2(i8 %x, i8 %y) nounwind {
+; X64-LABEL: func2:
+; X64:       # %bb.0:
+; X64-NEXT:    movsbl %dil, %eax
+; X64-NEXT:    movsbl %sil, %ecx
+; X64-NEXT:    movswl %cx, %esi
+; X64-NEXT:    movswl %ax, %ecx
+; X64-NEXT:    shll $14, %ecx
+; X64-NEXT:    movl %ecx, %eax
+; X64-NEXT:    cltd
+; X64-NEXT:    idivl %esi
+; X64-NEXT:    # kill: def $eax killed $eax def $rax
+; X64-NEXT:    leal -1(%rax), %edi
+; X64-NEXT:    testl %esi, %esi
+; X64-NEXT:    sets %sil
+; X64-NEXT:    testl %ecx, %ecx
+; X64-NEXT:    sets %cl
+; X64-NEXT:    xorb %sil, %cl
+; X64-NEXT:    testl %edx, %edx
+; X64-NEXT:    setne %dl
+; X64-NEXT:    testb %cl, %dl
+; X64-NEXT:    cmovel %eax, %edi
+; X64-NEXT:    addl %edi, %edi
+; X64-NEXT:    movswl %di, %eax
+; X64-NEXT:    shrl %eax
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    retq
+;
+; X86-LABEL: func2:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movsbl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movsbl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    shll $14, %ecx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    cltd
+; X86-NEXT:    idivl %esi
+; X86-NEXT:    leal -1(%eax), %edi
+; X86-NEXT:    testl %esi, %esi
+; X86-NEXT:    sets %bl
+; X86-NEXT:    testl %ecx, %ecx
+; X86-NEXT:    sets %cl
+; X86-NEXT:    xorb %bl, %cl
+; X86-NEXT:    testl %edx, %edx
+; X86-NEXT:    setne %dl
+; X86-NEXT:    testb %cl, %dl
+; X86-NEXT:    cmovel %eax, %edi
+; X86-NEXT:    addl %edi, %edi
+; X86-NEXT:    movswl %di, %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    retl
+  %x2 = sext i8 %x to i15
+  %y2 = sext i8 %y to i15
+  %tmp = call i15 @llvm.sdiv.fix.i15(i15 %x2, i15 %y2, i32 14)
+  %tmp2 = sext i15 %tmp to i16
+  ret i16 %tmp2
+}
+
+define i16 @func3(i15 %x, i8 %y) nounwind {
+; X64-LABEL: func3:
+; X64:       # %bb.0:
+; X64-NEXT:    shll $8, %esi
+; X64-NEXT:    movswl %si, %ecx
+; X64-NEXT:    addl %edi, %edi
+; X64-NEXT:    shrl $4, %ecx
+; X64-NEXT:    movl %edi, %eax
+; X64-NEXT:    cwtd
+; X64-NEXT:    idivw %cx
+; X64-NEXT:    # kill: def $ax killed $ax def $rax
+; X64-NEXT:    leal -1(%rax), %esi
+; X64-NEXT:    testw %di, %di
+; X64-NEXT:    sets %dil
+; X64-NEXT:    testw %cx, %cx
+; X64-NEXT:    sets %cl
+; X64-NEXT:    xorb %dil, %cl
+; X64-NEXT:    testw %dx, %dx
+; X64-NEXT:    setne %dl
+; X64-NEXT:    testb %cl, %dl
+; X64-NEXT:    cmovel %eax, %esi
+; X64-NEXT:    addl %esi, %esi
+; X64-NEXT:    movswl %si, %eax
+; X64-NEXT:    shrl %eax
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    retq
+;
+; X86-LABEL: func3:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shll $8, %eax
+; X86-NEXT:    movswl %ax, %esi
+; X86-NEXT:    addl %ecx, %ecx
+; X86-NEXT:    shrl $4, %esi
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    cwtd
+; X86-NEXT:    idivw %si
+; X86-NEXT:    # kill: def $ax killed $ax def $eax
+; X86-NEXT:    leal -1(%eax), %edi
+; X86-NEXT:    testw %cx, %cx
+; X86-NEXT:    sets %cl
+; X86-NEXT:    testw %si, %si
+; X86-NEXT:    sets %ch
+; X86-NEXT:    xorb %cl, %ch
+; X86-NEXT:    testw %dx, %dx
+; X86-NEXT:    setne %cl
+; X86-NEXT:    testb %ch, %cl
+; X86-NEXT:    cmovel %eax, %edi
+; X86-NEXT:    addl %edi, %edi
+; X86-NEXT:    movswl %di, %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    retl
+  %y2 = sext i8 %y to i15
+  %y3 = shl i15 %y2, 7
+  %tmp = call i15 @llvm.sdiv.fix.i15(i15 %x, i15 %y3, i32 4)
+  %tmp2 = sext i15 %tmp to i16
+  ret i16 %tmp2
+}
+
+define i4 @func4(i4 %x, i4 %y) nounwind {
+; X64-LABEL: func4:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    shlb $4, %sil
+; X64-NEXT:    sarb $4, %sil
+; X64-NEXT:    shlb $4, %dil
+; X64-NEXT:    sarb $4, %dil
+; X64-NEXT:    shlb $2, %dil
+; X64-NEXT:    movsbl %dil, %ecx
+; X64-NEXT:    movl %ecx, %eax
+; X64-NEXT:    idivb %sil
+; X64-NEXT:    movsbl %ah, %ebx
+; X64-NEXT:    movzbl %al, %edi
+; X64-NEXT:    leal -1(%rdi), %eax
+; X64-NEXT:    movzbl %al, %eax
+; X64-NEXT:    testb %sil, %sil
+; X64-NEXT:    sets %dl
+; X64-NEXT:    testb %cl, %cl
+; X64-NEXT:    sets %cl
+; X64-NEXT:    xorb %dl, %cl
+; X64-NEXT:    testb %bl, %bl
+; X64-NEXT:    setne %dl
+; X64-NEXT:    testb %cl, %dl
+; X64-NEXT:    cmovel %edi, %eax
+; X64-NEXT:    # kill: def $al killed $al killed $eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    retq
+;
+; X86-LABEL: func4:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %dl
+; X86-NEXT:    shlb $4, %dl
+; X86-NEXT:    sarb $4, %dl
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %dh
+; X86-NEXT:    shlb $4, %dh
+; X86-NEXT:    sarb $4, %dh
+; X86-NEXT:    shlb $2, %dh
+; X86-NEXT:    movsbl %dh, %eax
+; X86-NEXT:    idivb %dl
+; X86-NEXT:    movsbl %ah, %ecx
+; X86-NEXT:    movzbl %al, %esi
+; X86-NEXT:    decb %al
+; X86-NEXT:    movzbl %al, %eax
+; X86-NEXT:    testb %dl, %dl
+; X86-NEXT:    sets %dl
+; X86-NEXT:    testb %dh, %dh
+; X86-NEXT:    sets %dh
+; X86-NEXT:    xorb %dl, %dh
+; X86-NEXT:    testb %cl, %cl
+; X86-NEXT:    setne %cl
+; X86-NEXT:    testb %dh, %cl
+; X86-NEXT:    cmovel %esi, %eax
+; X86-NEXT:    # kill: def $al killed $al killed $eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    retl
+  %tmp = call i4 @llvm.sdiv.fix.i4(i4 %x, i4 %y, i32 2)
+  ret i4 %tmp
+}
+
+define i64 @func5(i64 %x, i64 %y) nounwind {
+; X64-LABEL: func5:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbp
+; X64-NEXT:    pushq %r15
+; X64-NEXT:    pushq %r14
+; X64-NEXT:    pushq %r13
+; X64-NEXT:    pushq %r12
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    subq $24, %rsp
+; X64-NEXT:    movq %rsi, %r14
+; X64-NEXT:    movq %rdi, %r15
+; X64-NEXT:    movq %rdi, %rax
+; X64-NEXT:    shrq $33, %rax
+; X64-NEXT:    movq %rdi, %rbx
+; X64-NEXT:    sarq $63, %rbx
+; X64-NEXT:    shlq $31, %rbx
+; X64-NEXT:    orq %rax, %rbx
+; X64-NEXT:    sets {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Folded Spill
+; X64-NEXT:    shlq $31, %r15
+; X64-NEXT:    movq %rsi, %r12
+; X64-NEXT:    sarq $63, %r12
+; X64-NEXT:    movq %r15, %rdi
+; X64-NEXT:    movq %rbx, %rsi
+; X64-NEXT:    movq %r14, %rdx
+; X64-NEXT:    movq %r12, %rcx
+; X64-NEXT:    callq __divti3
+; X64-NEXT:    movq %rax, %r13
+; X64-NEXT:    decq %rax
+; X64-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT:    testq %r12, %r12
+; X64-NEXT:    sets %bpl
+; X64-NEXT:    xorb {{[-0-9]+}}(%r{{[sb]}}p), %bpl # 1-byte Folded Reload
+; X64-NEXT:    movq %r15, %rdi
+; X64-NEXT:    movq %rbx, %rsi
+; X64-NEXT:    movq %r14, %rdx
+; X64-NEXT:    movq %r12, %rcx
+; X64-NEXT:    callq __modti3
+; X64-NEXT:    orq %rax, %rdx
+; X64-NEXT:    setne %al
+; X64-NEXT:    testb %bpl, %al
+; X64-NEXT:    cmovneq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Folded Reload
+; X64-NEXT:    movq %r13, %rax
+; X64-NEXT:    addq $24, %rsp
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    popq %r12
+; X64-NEXT:    popq %r13
+; X64-NEXT:    popq %r14
+; X64-NEXT:    popq %r15
+; X64-NEXT:    popq %rbp
+; X64-NEXT:    retq
+;
+; X86-LABEL: func5:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    movl %esp, %ebp
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    andl $-8, %esp
+; X86-NEXT:    subl $72, %esp
+; X86-NEXT:    movl 8(%ebp), %ecx
+; X86-NEXT:    movl 12(%ebp), %edx
+; X86-NEXT:    movl 20(%ebp), %ebx
+; X86-NEXT:    sarl $31, %ebx
+; X86-NEXT:    movl %edx, %eax
+; X86-NEXT:    shldl $31, %ecx, %eax
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    shll $31, %ecx
+; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %edx, %esi
+; X86-NEXT:    sarl $31, %esi
+; X86-NEXT:    movl %esi, %edi
+; X86-NEXT:    shldl $31, %edx, %esi
+; X86-NEXT:    leal {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    rorl %edi
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl 20(%ebp)
+; X86-NEXT:    pushl 16(%ebp)
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %edx
+; X86-NEXT:    calll __divti3
+; X86-NEXT:    addl $32, %esp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    subl $1, %ecx
+; X86-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    sbbl $0, %eax
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    testl %ebx, %ebx
+; X86-NEXT:    sets %al
+; X86-NEXT:    testl %edi, %edi
+; X86-NEXT:    sets %cl
+; X86-NEXT:    xorb %al, %cl
+; X86-NEXT:    movb %cl, {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Spill
+; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl 20(%ebp)
+; X86-NEXT:    pushl 16(%ebp)
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    calll __modti3
+; X86-NEXT:    addl $32, %esp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    orl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    orl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    orl %eax, %ecx
+; X86-NEXT:    setne %al
+; X86-NEXT:    testb %al, {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Reload
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
+; X86-NEXT:    cmovel {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Folded Reload
+; X86-NEXT:    leal -12(%ebp), %esp
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    popl %ebp
+; X86-NEXT:    retl
+  %tmp = call i64 @llvm.sdiv.fix.i64(i64 %x, i64 %y, i32 31)
+  ret i64 %tmp
+}
+
+define i18 @func6(i16 %x, i16 %y) nounwind {
+; X64-LABEL: func6:
+; X64:       # %bb.0:
+; X64-NEXT:    movswl %di, %ecx
+; X64-NEXT:    movswl %si, %esi
+; X64-NEXT:    shll $7, %ecx
+; X64-NEXT:    movl %ecx, %eax
+; X64-NEXT:    cltd
+; X64-NEXT:    idivl %esi
+; X64-NEXT:    # kill: def $eax killed $eax def $rax
+; X64-NEXT:    leal -1(%rax), %edi
+; X64-NEXT:    testl %esi, %esi
+; X64-NEXT:    sets %sil
+; X64-NEXT:    testl %ecx, %ecx
+; X64-NEXT:    sets %cl
+; X64-NEXT:    xorb %sil, %cl
+; X64-NEXT:    testl %edx, %edx
+; X64-NEXT:    setne %dl
+; X64-NEXT:    testb %cl, %dl
+; X64-NEXT:    cmovnel %edi, %eax
+; X64-NEXT:    # kill: def $eax killed $eax killed $rax
+; X64-NEXT:    retq
+;
+; X86-LABEL: func6:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movswl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movswl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    shll $7, %ecx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    cltd
+; X86-NEXT:    idivl %esi
+; X86-NEXT:    leal -1(%eax), %edi
+; X86-NEXT:    testl %esi, %esi
+; X86-NEXT:    sets %bl
+; X86-NEXT:    testl %ecx, %ecx
+; X86-NEXT:    sets %cl
+; X86-NEXT:    xorb %bl, %cl
+; X86-NEXT:    testl %edx, %edx
+; X86-NEXT:    setne %dl
+; X86-NEXT:    testb %cl, %dl
+; X86-NEXT:    cmovnel %edi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    retl
+  %x2 = sext i16 %x to i18
+  %y2 = sext i16 %y to i18
+  %tmp = call i18 @llvm.sdiv.fix.i18(i18 %x2, i18 %y2, i32 7)
+  ret i18 %tmp
+}
+
+define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
+; X64-LABEL: vec:
+; X64:       # %bb.0:
+; X64-NEXT:    pxor %xmm2, %xmm2
+; X64-NEXT:    pcmpgtd %xmm1, %xmm2
+; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,3,0,1]
+; X64-NEXT:    movdqa %xmm1, %xmm4
+; X64-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; X64-NEXT:    movq %xmm4, %rcx
+; X64-NEXT:    pxor %xmm2, %xmm2
+; X64-NEXT:    pcmpgtd %xmm0, %xmm2
+; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; X64-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X64-NEXT:    psllq $31, %xmm0
+; X64-NEXT:    movq %xmm0, %rax
+; X64-NEXT:    cqto
+; X64-NEXT:    idivq %rcx
+; X64-NEXT:    movq %rax, %r8
+; X64-NEXT:    movq %rdx, %r11
+; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[2,3,0,1]
+; X64-NEXT:    movq %xmm2, %rcx
+; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; X64-NEXT:    movq %xmm2, %rax
+; X64-NEXT:    cqto
+; X64-NEXT:    idivq %rcx
+; X64-NEXT:    movq %rax, %r10
+; X64-NEXT:    movq %rdx, %rcx
+; X64-NEXT:    pxor %xmm2, %xmm2
+; X64-NEXT:    pcmpgtd %xmm3, %xmm2
+; X64-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; X64-NEXT:    movq %xmm3, %rdi
+; X64-NEXT:    pxor %xmm2, %xmm2
+; X64-NEXT:    pcmpgtd %xmm1, %xmm2
+; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X64-NEXT:    psllq $31, %xmm1
+; X64-NEXT:    movq %xmm1, %rax
+; X64-NEXT:    cqto
+; X64-NEXT:    idivq %rdi
+; X64-NEXT:    movq %rax, %r9
+; X64-NEXT:    movq %rdx, %rdi
+; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,0,1]
+; X64-NEXT:    movq %xmm2, %rsi
+; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,3,0,1]
+; X64-NEXT:    movq %xmm2, %rax
+; X64-NEXT:    cqto
+; X64-NEXT:    idivq %rsi
+; X64-NEXT:    movq %r11, %xmm2
+; X64-NEXT:    movq %rcx, %xmm5
+; X64-NEXT:    pxor %xmm6, %xmm6
+; X64-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm5[0]
+; X64-NEXT:    pcmpeqd %xmm6, %xmm2
+; X64-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[1,0,3,2]
+; X64-NEXT:    pand %xmm2, %xmm5
+; X64-NEXT:    pxor %xmm2, %xmm2
+; X64-NEXT:    pcmpgtd %xmm4, %xmm2
+; X64-NEXT:    pxor %xmm4, %xmm4
+; X64-NEXT:    pcmpgtd %xmm0, %xmm4
+; X64-NEXT:    movq %r8, %xmm0
+; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; X64-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; X64-NEXT:    pxor %xmm2, %xmm4
+; X64-NEXT:    movq %r10, %xmm2
+; X64-NEXT:    pandn %xmm4, %xmm5
+; X64-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; X64-NEXT:    movdqa %xmm5, %xmm2
+; X64-NEXT:    pandn %xmm0, %xmm2
+; X64-NEXT:    pcmpeqd %xmm4, %xmm4
+; X64-NEXT:    paddq %xmm4, %xmm0
+; X64-NEXT:    pand %xmm5, %xmm0
+; X64-NEXT:    por %xmm2, %xmm0
+; X64-NEXT:    movq %rdi, %xmm2
+; X64-NEXT:    movq %rdx, %xmm5
+; X64-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm5[0]
+; X64-NEXT:    pcmpeqd %xmm6, %xmm2
+; X64-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[1,0,3,2]
+; X64-NEXT:    pand %xmm2, %xmm5
+; X64-NEXT:    pxor %xmm2, %xmm2
+; X64-NEXT:    pcmpgtd %xmm3, %xmm2
+; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; X64-NEXT:    pcmpgtd %xmm1, %xmm6
+; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,3,3]
+; X64-NEXT:    pxor %xmm2, %xmm1
+; X64-NEXT:    pandn %xmm1, %xmm5
+; X64-NEXT:    movq %r9, %xmm1
+; X64-NEXT:    movq %rax, %xmm2
+; X64-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; X64-NEXT:    movdqa %xmm5, %xmm2
+; X64-NEXT:    pandn %xmm1, %xmm2
+; X64-NEXT:    paddq %xmm4, %xmm1
+; X64-NEXT:    pand %xmm5, %xmm1
+; X64-NEXT:    por %xmm2, %xmm1
+; X64-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
+; X64-NEXT:    retq
+;
+; X86-LABEL: vec:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    subl $64, %esp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %ecx, %edx
+; X86-NEXT:    sarl $31, %edx
+; X86-NEXT:    movl %edi, %esi
+; X86-NEXT:    shll $31, %esi
+; X86-NEXT:    movl %ebx, %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    andl $-2147483648, %ebx # imm = 0x80000000
+; X86-NEXT:    orl %eax, %ebx
+; X86-NEXT:    sets {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
+; X86-NEXT:    movl %ebp, %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    andl $-2147483648, %ebp # imm = 0x80000000
+; X86-NEXT:    orl %eax, %ebp
+; X86-NEXT:    movl %ebp, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    sets {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; X86-NEXT:    andl $-2147483648, %ebp # imm = 0x80000000
+; X86-NEXT:    orl %eax, %ebp
+; X86-NEXT:    movl %ebp, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    sets (%esp) # 1-byte Folded Spill
+; X86-NEXT:    movl %edi, %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    andl $-2147483648, %edi # imm = 0x80000000
+; X86-NEXT:    orl %eax, %edi
+; X86-NEXT:    sets {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
+; X86-NEXT:    pushl %edx
+; X86-NEXT:    movl %edx, %ebp
+; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    calll __moddi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    calll __divdi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    shll $31, %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    movl %edx, %eax
+; X86-NEXT:    sarl $31, %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    movl %eax, %esi
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    pushl %edx
+; X86-NEXT:    movl %edx, %ebp
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    movl %ecx, %edi
+; X86-NEXT:    calll __moddi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    calll __divdi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shll $31, %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %ecx, %edx
+; X86-NEXT:    sarl $31, %edx
+; X86-NEXT:    pushl %edx
+; X86-NEXT:    movl %edx, %ebp
+; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    movl %ecx, %edi
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    movl %eax, %esi
+; X86-NEXT:    calll __moddi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    calll __divdi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shll $31, %eax
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %ecx, %ebp
+; X86-NEXT:    sarl $31, %ebp
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    calll __moddi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT:    movl %edx, %edi
+; X86-NEXT:    testl %ebp, %ebp
+; X86-NEXT:    sets %bl
+; X86-NEXT:    xorb (%esp), %bl # 1-byte Folded Reload
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    calll __divdi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %edi # 4-byte Folded Reload
+; X86-NEXT:    setne %cl
+; X86-NEXT:    testb %bl, %cl
+; X86-NEXT:    leal -1(%eax), %ecx
+; X86-NEXT:    cmovel %eax, %ecx
+; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    sets %al
+; X86-NEXT:    xorb {{[-0-9]+}}(%e{{[sb]}}p), %al # 1-byte Folded Reload
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
+; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Folded Reload
+; X86-NEXT:    setne %dl
+; X86-NEXT:    testb %al, %dl
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X86-NEXT:    leal -1(%eax), %edi
+; X86-NEXT:    cmovel %eax, %edi
+; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    sets %dl
+; X86-NEXT:    xorb {{[-0-9]+}}(%e{{[sb]}}p), %dl # 1-byte Folded Reload
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
+; X86-NEXT:    setne %dh
+; X86-NEXT:    testb %dl, %dh
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X86-NEXT:    leal -1(%eax), %edx
+; X86-NEXT:    cmovel %eax, %edx
+; X86-NEXT:    cmpl $0, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
+; X86-NEXT:    sets %bl
+; X86-NEXT:    xorb {{[-0-9]+}}(%e{{[sb]}}p), %bl # 1-byte Folded Reload
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X86-NEXT:    orl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
+; X86-NEXT:    setne %bh
+; X86-NEXT:    testb %bl, %bh
+; X86-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X86-NEXT:    leal -1(%eax), %esi
+; X86-NEXT:    cmovel %eax, %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl %esi, 12(%eax)
+; X86-NEXT:    movl %edx, 8(%eax)
+; X86-NEXT:    movl %edi, 4(%eax)
+; X86-NEXT:    movl %ecx, (%eax)
+; X86-NEXT:    addl $64, %esp
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    popl %ebp
+; X86-NEXT:    retl $4
+  %tmp = call <4 x i32> @llvm.sdiv.fix.v4i32(<4 x i32> %x, <4 x i32> %y, i32 31)
+  ret <4 x i32> %tmp
+}

llvm/test/CodeGen/X86/udiv_fix.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-linux | FileCheck %s --check-prefix=X64
+; RUN: llc < %s -mtriple=i686 -mattr=cmov | FileCheck %s --check-prefix=X86
+
+declare  i4  @llvm.udiv.fix.i4   (i4,  i4,  i32)
+declare  i15 @llvm.udiv.fix.i15  (i15, i15, i32)
+declare  i16 @llvm.udiv.fix.i16  (i16, i16, i32)
+declare  i18 @llvm.udiv.fix.i18  (i18, i18, i32)
+declare  i64 @llvm.udiv.fix.i64  (i64, i64, i32)
+declare  <4 x i32> @llvm.udiv.fix.v4i32(<4 x i32>, <4 x i32>, i32)
+
+define i16 @func(i16 %x, i16 %y) nounwind {
+; X64-LABEL: func:
+; X64:       # %bb.0:
+; X64-NEXT:    movzwl %si, %ecx
+; X64-NEXT:    movzwl %di, %eax
+; X64-NEXT:    shll $7, %eax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divl %ecx
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    retq
+;
+; X86-LABEL: func:
+; X86:       # %bb.0:
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shll $7, %eax
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    divl %ecx
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    retl
+  %tmp = call i16 @llvm.udiv.fix.i16(i16 %x, i16 %y, i32 7)
+  ret i16 %tmp
+}
+
+define i16 @func2(i8 %x, i8 %y) nounwind {
+; X64-LABEL: func2:
+; X64:       # %bb.0:
+; X64-NEXT:    movsbl %dil, %eax
+; X64-NEXT:    andl $32767, %eax # imm = 0x7FFF
+; X64-NEXT:    movsbl %sil, %ecx
+; X64-NEXT:    andl $32767, %ecx # imm = 0x7FFF
+; X64-NEXT:    shll $14, %eax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divl %ecx
+; X64-NEXT:    addl %eax, %eax
+; X64-NEXT:    cwtl
+; X64-NEXT:    shrl %eax
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    retq
+;
+; X86-LABEL: func2:
+; X86:       # %bb.0:
+; X86-NEXT:    movsbl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    andl $32767, %ecx # imm = 0x7FFF
+; X86-NEXT:    movsbl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl $32767, %eax # imm = 0x7FFF
+; X86-NEXT:    shll $14, %eax
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    divl %ecx
+; X86-NEXT:    addl %eax, %eax
+; X86-NEXT:    cwtl
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    retl
+  %x2 = sext i8 %x to i15
+  %y2 = sext i8 %y to i15
+  %tmp = call i15 @llvm.udiv.fix.i15(i15 %x2, i15 %y2, i32 14)
+  %tmp2 = sext i15 %tmp to i16
+  ret i16 %tmp2
+}
+
+define i16 @func3(i15 %x, i8 %y) nounwind {
+; X64-LABEL: func3:
+; X64:       # %bb.0:
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal (%rdi,%rdi), %eax
+; X64-NEXT:    movzbl %sil, %ecx
+; X64-NEXT:    shll $4, %ecx
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divw %cx
+; X64-NEXT:    # kill: def $ax killed $ax def $eax
+; X64-NEXT:    addl %eax, %eax
+; X64-NEXT:    cwtl
+; X64-NEXT:    shrl %eax
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    retq
+;
+; X86-LABEL: func3:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    addl %eax, %eax
+; X86-NEXT:    movzbl %cl, %ecx
+; X86-NEXT:    shll $4, %ecx
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    divw %cx
+; X86-NEXT:    # kill: def $ax killed $ax def $eax
+; X86-NEXT:    addl %eax, %eax
+; X86-NEXT:    cwtl
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    retl
+  %y2 = sext i8 %y to i15
+  %y3 = shl i15 %y2, 7
+  %tmp = call i15 @llvm.udiv.fix.i15(i15 %x, i15 %y3, i32 4)
+  %tmp2 = sext i15 %tmp to i16
+  ret i16 %tmp2
+}
+
+define i4 @func4(i4 %x, i4 %y) nounwind {
+; X64-LABEL: func4:
+; X64:       # %bb.0:
+; X64-NEXT:    andb $15, %sil
+; X64-NEXT:    andb $15, %dil
+; X64-NEXT:    shlb $2, %dil
+; X64-NEXT:    movzbl %dil, %eax
+; X64-NEXT:    divb %sil
+; X64-NEXT:    retq
+;
+; X86-LABEL: func4:
+; X86:       # %bb.0:
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %cl
+; X86-NEXT:    andb $15, %cl
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-NEXT:    andb $15, %al
+; X86-NEXT:    shlb $2, %al
+; X86-NEXT:    movzbl %al, %eax
+; X86-NEXT:    divb %cl
+; X86-NEXT:    retl
+  %tmp = call i4 @llvm.udiv.fix.i4(i4 %x, i4 %y, i32 2)
+  ret i4 %tmp
+}
+
+define i64 @func5(i64 %x, i64 %y) nounwind {
+; X64-LABEL: func5:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rax
+; X64-NEXT:    movq %rsi, %rdx
+; X64-NEXT:    movq %rdi, %rsi
+; X64-NEXT:    shlq $31, %rdi
+; X64-NEXT:    shrq $33, %rsi
+; X64-NEXT:    xorl %ecx, %ecx
+; X64-NEXT:    callq __udivti3
+; X64-NEXT:    popq %rcx
+; X64-NEXT:    retq
+;
+; X86-LABEL: func5:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    movl %esp, %ebp
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    andl $-8, %esp
+; X86-NEXT:    subl $24, %esp
+; X86-NEXT:    movl 8(%ebp), %eax
+; X86-NEXT:    movl 12(%ebp), %ecx
+; X86-NEXT:    movl %ecx, %edx
+; X86-NEXT:    shrl %edx
+; X86-NEXT:    shldl $31, %eax, %ecx
+; X86-NEXT:    shll $31, %eax
+; X86-NEXT:    movl %esp, %esi
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl 20(%ebp)
+; X86-NEXT:    pushl 16(%ebp)
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl %edx
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    calll __udivti3
+; X86-NEXT:    addl $32, %esp
+; X86-NEXT:    movl (%esp), %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    leal -4(%ebp), %esp
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %ebp
+; X86-NEXT:    retl
+  %tmp = call i64 @llvm.udiv.fix.i64(i64 %x, i64 %y, i32 31)
+  ret i64 %tmp
+}
+
+define i18 @func6(i16 %x, i16 %y) nounwind {
+; X64-LABEL: func6:
+; X64:       # %bb.0:
+; X64-NEXT:    movswl %di, %eax
+; X64-NEXT:    andl $262143, %eax # imm = 0x3FFFF
+; X64-NEXT:    movswl %si, %ecx
+; X64-NEXT:    andl $262143, %ecx # imm = 0x3FFFF
+; X64-NEXT:    shll $7, %eax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divl %ecx
+; X64-NEXT:    retq
+;
+; X86-LABEL: func6:
+; X86:       # %bb.0:
+; X86-NEXT:    movswl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    andl $262143, %ecx # imm = 0x3FFFF
+; X86-NEXT:    movswl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl $262143, %eax # imm = 0x3FFFF
+; X86-NEXT:    shll $7, %eax
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    divl %ecx
+; X86-NEXT:    retl
+  %x2 = sext i16 %x to i18
+  %y2 = sext i16 %y to i18
+  %tmp = call i18 @llvm.udiv.fix.i18(i18 %x2, i18 %y2, i32 7)
+  ret i18 %tmp
+}
+
+define i16 @func7(i16 %x, i16 %y) nounwind {
+; X64-LABEL: func7:
+; X64:       # %bb.0:
+; X64-NEXT:    movl %edi, %eax
+; X64-NEXT:    shll $16, %eax
+; X64-NEXT:    movzwl %si, %ecx
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divl %ecx
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    retq
+;
+; X86-LABEL: func7:
+; X86:       # %bb.0:
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shll $16, %eax
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    divl %ecx
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    retl
+  %tmp = call i16 @llvm.udiv.fix.i16(i16 %x, i16 %y, i32 16)
+  ret i16 %tmp
+}
+
+define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
+; X64-LABEL: vec:
+; X64:       # %bb.0:
+; X64-NEXT:    pxor %xmm2, %xmm2
+; X64-NEXT:    movdqa %xmm1, %xmm4
+; X64-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; X64-NEXT:    movq %xmm4, %rcx
+; X64-NEXT:    movdqa %xmm0, %xmm5
+; X64-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; X64-NEXT:    psllq $31, %xmm5
+; X64-NEXT:    movq %xmm5, %rax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divq %rcx
+; X64-NEXT:    movq %rax, %xmm3
+; X64-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,0,1]
+; X64-NEXT:    movq %xmm4, %rcx
+; X64-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[2,3,0,1]
+; X64-NEXT:    movq %xmm4, %rax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divq %rcx
+; X64-NEXT:    movq %rax, %xmm4
+; X64-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm4[0]
+; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X64-NEXT:    movq %xmm1, %rcx
+; X64-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X64-NEXT:    psllq $31, %xmm0
+; X64-NEXT:    movq %xmm0, %rax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divq %rcx
+; X64-NEXT:    movq %rax, %xmm2
+; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; X64-NEXT:    movq %xmm1, %rcx
+; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X64-NEXT:    movq %xmm0, %rax
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divq %rcx
+; X64-NEXT:    movq %rax, %xmm0
+; X64-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; X64-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
+; X64-NEXT:    movaps %xmm2, %xmm0
+; X64-NEXT:    retq
+;
+; X86-LABEL: vec:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl %eax, %ecx
+; X86-NEXT:    shrl %ecx
+; X86-NEXT:    shll $31, %eax
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl %ecx
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    calll __udivdi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, (%esp) # 4-byte Spill
+; X86-NEXT:    movl %ebx, %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    shll $31, %ebx
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    calll __udivdi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, %ebx
+; X86-NEXT:    movl %ebp, %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    shll $31, %ebp
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    calll __udivdi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, %ebp
+; X86-NEXT:    movl %edi, %eax
+; X86-NEXT:    shrl %eax
+; X86-NEXT:    shll $31, %edi
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    calll __udivdi3
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    movl %eax, 12(%esi)
+; X86-NEXT:    movl %ebp, 8(%esi)
+; X86-NEXT:    movl %ebx, 4(%esi)
+; X86-NEXT:    movl (%esp), %eax # 4-byte Reload
+; X86-NEXT:    movl %eax, (%esi)
+; X86-NEXT:    movl %esi, %eax
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    popl %ebp
+; X86-NEXT:    retl $4
+  %tmp = call <4 x i32> @llvm.udiv.fix.v4i32(<4 x i32> %x, <4 x i32> %y, i32 31)
+  ret <4 x i32> %tmp
+}