[Codegen][SelectionDAG] X u% C == 0 fold: non-splat vector improvements

Summary:
Four things here:
1. Generalize the fold to handle non-splat divisors. Reasonably trivial.
2. Unban power-of-two divisors. I don't see any reason why they should
   be illegal.
   * There is no ban in Hacker's Delight
   * I think the ban came from the same bug that caused the miscompile
      in the base patch - in `floor((2^W - 1) / D)` we were dividing by
      `D0` instead of `D`, and we **were** ensuring that `D0` is not `1`,
      which made sense.
3. Unban `1` divisors. I no longer believe Hacker's Delight actually says
   that the fold is invalid for `D = 0`. Further considerations:
   * We know that
     * `(X u% 1) == 0`  can be constant-folded to `1`,
     * `(X u% 1) != 0`  can be constant-folded to `0`,
   *  Also, we know that
     * `X u<= -1` can be constant-folded to `1`,
     * `X u>  -1` can be constant-folded to `0`,
   * https://godbolt.org/z/7jnZJX https://rise4fun.com/Alive/oF6p
   * We know will end up with the following:
       `(setule/setugt (rotr (mul N, P), K), Q)`
   * Therefore, for given new DAG nodes and comparison predicates
     (`ule`/`ugt`), we will still produce the correct answer if:
     `Q` is a all-ones constant; and both `P` and `K` are *anything*
     other than `undef`.
   * The fold will indeed produce `Q = all-ones`.
4. Try to re-splat the `P` and `K` vectors - we don't care about
   their values for the lanes where divisor was `1`.

Reviewers: RKSimon, hermord, craig.topper, spatel, xbolva00

Reviewed By: RKSimon

Subscribers: hiraditya, javed.absar, dexonsmith, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63963

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@366637 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index b260cd91d468057e10eab7c458353ed37be84e82..bab99b2e0293e504e2258cb90c4d9c8de8f1f31f 100644 (file)
--- a/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -4455,6 +4455,34 @@ SDValue TargetLowering::BuildUDIV(SDNode *N, SelectionDAG &DAG,
   return DAG.getSelect(dl, VT, IsOne, N0, Q);
 }
 
+/// If all values in Values that *don't* match the predicate are same 'splat'
+/// value, then replace all values with that splat value.
+/// Else, if AlternativeReplacement was provided, then replace all values that
+/// do match predicate with AlternativeReplacement value.
+static void
+turnVectorIntoSplatVector(MutableArrayRef<SDValue> Values,
+                          std::function<bool(SDValue)> Predicate,
+                          SDValue AlternativeReplacement = SDValue()) {
+  SDValue Replacement;
+  // Is there a value for which the Predicate does *NOT* match? What is it?
+  auto SplatValue = llvm::find_if_not(Values, Predicate);
+  if (SplatValue != Values.end()) {
+    // Does Values consist only of SplatValue's and values matching Predicate?
+    if (llvm::all_of(Values, [Predicate, SplatValue](SDValue Value) {
+          return Value == *SplatValue || Predicate(Value);
+        })) // Then we shall replace values matching predicate with SplatValue.
+      Replacement = *SplatValue;
+  }
+  if (!Replacement) {
+    // Oops, we did not find the "baseline" splat value.
+    if (!AlternativeReplacement)
+      return; // Nothing to do.
+    // Let's replace with provided value then.
+    Replacement = AlternativeReplacement;
+  }
+  std::replace_if(Values.begin(), Values.end(), Predicate, Replacement);
+}
+
 /// Given an ISD::UREM used only by an ISD::SETEQ or ISD::SETNE
 /// where the divisor is constant and the comparison target is zero,
 /// return a DAG expression that will generate the same comparison result
@@ -4482,74 +4510,143 @@ TargetLowering::prepareUREMEqFold(EVT SETCCVT, SDValue REMNode,
                                   DAGCombinerInfo &DCI, const SDLoc &DL,
                                   SmallVectorImpl<SDNode *> &Created) const {
   // fold (seteq/ne (urem N, D), 0) -> (setule/ugt (rotr (mul N, P), K), Q)
-  // - D must be constant with D = D0 * 2^K where D0 is odd and D0 != 1
+  // - D must be constant, with D = D0 * 2^K where D0 is odd
   // - P is the multiplicative inverse of D0 modulo 2^W
   // - Q = floor((2^W - 1) / D0)
   // where W is the width of the common type of N and D.
   assert((Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
          "Only applicable for (in)equality comparisons.");
 
+  SelectionDAG &DAG = DCI.DAG;
+
   EVT VT = REMNode.getValueType();
+  EVT SVT = VT.getScalarType();
+  EVT ShVT = getShiftAmountTy(VT, DAG.getDataLayout());
+  EVT ShSVT = ShVT.getScalarType();
 
   // If MUL is unavailable, we cannot proceed in any case.
   if (!isOperationLegalOrCustom(ISD::MUL, VT))
     return SDValue();
 
-  // TODO: Add non-uniform constant support.
-  ConstantSDNode *Divisor = isConstOrConstSplat(REMNode->getOperand(1));
+  // TODO: Could support comparing with non-zero too.
   ConstantSDNode *CompTarget = isConstOrConstSplat(CompTargetNode);
-  if (!Divisor || !CompTarget || Divisor->isNullValue() ||
-      !CompTarget->isNullValue())
+  if (!CompTarget || !CompTarget->isNullValue())
     return SDValue();
 
-  const APInt &D = Divisor->getAPIntValue();
+  bool HadOneDivisor = false;
+  bool AllDivisorsAreOnes = true;
+  bool HadEvenDivisor = false;
+  bool AllDivisorsArePowerOfTwo = true;
+  SmallVector<SDValue, 16> PAmts, KAmts, QAmts;
+
+  auto BuildUREMPattern = [&](ConstantSDNode *C) {
+    // Division by 0 is UB. Leave it to be constant-folded elsewhere.
+    if (C->isNullValue())
+      return false;
 
-  // Decompose D into D0 * 2^K
-  unsigned K = D.countTrailingZeros();
-  bool DivisorIsEven = (K != 0);
-  APInt D0 = D.lshr(K);
+    const APInt &D = C->getAPIntValue();
+    // If all divisors are ones, we will prefer to avoid the fold.
+    HadOneDivisor |= D.isOneValue();
+    AllDivisorsAreOnes &= D.isOneValue();
+
+    // Decompose D into D0 * 2^K
+    unsigned K = D.countTrailingZeros();
+    assert((!D.isOneValue() || (K == 0)) && "For divisor '1' we won't rotate.");
+    APInt D0 = D.lshr(K);
+
+    // D is even if it has trailing zeros.
+    HadEvenDivisor |= (K != 0);
+    // D is a power-of-two if D0 is one.
+    // If all divisors are power-of-two, we will prefer to avoid the fold.
+    AllDivisorsArePowerOfTwo &= D0.isOneValue();
+
+    // P = inv(D0, 2^W)
+    // 2^W requires W + 1 bits, so we have to extend and then truncate.
+    unsigned W = D.getBitWidth();
+    APInt P = D0.zext(W + 1)
+                  .multiplicativeInverse(APInt::getSignedMinValue(W + 1))
+                  .trunc(W);
+    assert(!P.isNullValue() && "No multiplicative inverse!"); // unreachable
+    assert((D0 * P).isOneValue() && "Multiplicative inverse sanity check.");
+
+    // Q = floor((2^W - 1) / D)
+    APInt Q = APInt::getAllOnesValue(W).udiv(D);
+
+    assert(APInt::getAllOnesValue(ShSVT.getSizeInBits()).ugt(K) &&
+           "We are expecting that K is always less than all-ones for ShSVT");
+
+    // If the divisor is 1 the result can be constant-folded.
+    if (D.isOneValue()) {
+      // Set P and K amount to a bogus values so we can try to splat them.
+      P = 0;
+      K = -1;
+      assert(Q.isAllOnesValue() &&
+             "Expecting all-ones comparison for one divisor");
+    }
 
-  // The fold is invalid when D0 == 1.
-  // This is reachable because visitSetCC happens before visitREM.
-  if (D0.isOneValue())
+    PAmts.push_back(DAG.getConstant(P, DL, SVT));
+    KAmts.push_back(
+        DAG.getConstant(APInt(ShSVT.getSizeInBits(), K), DL, ShSVT));
+    QAmts.push_back(DAG.getConstant(Q, DL, SVT));
+    return true;
+  };
+
+  SDValue N = REMNode.getOperand(0);
+  SDValue D = REMNode.getOperand(1);
+
+  // Collect the values from each element.
+  if (!ISD::matchUnaryPredicate(D, BuildUREMPattern))
     return SDValue();
 
-  // P = inv(D0, 2^W)
-  // 2^W requires W + 1 bits, so we have to extend and then truncate.
-  unsigned W = D.getBitWidth();
-  APInt P = D0.zext(W + 1)
-                .multiplicativeInverse(APInt::getSignedMinValue(W + 1))
-                .trunc(W);
-  assert(!P.isNullValue() && "No multiplicative inverse!"); // unreachable
-  assert((D0 * P).isOneValue() && "Multiplicative inverse sanity check.");
+  // If this is a urem by a one, avoid the fold since it can be constant-folded.
+  if (AllDivisorsAreOnes)
+    return SDValue();
 
-  // Q = floor((2^W - 1) / D)
-  APInt Q = APInt::getAllOnesValue(W).udiv(D);
+  // If this is a urem by a powers-of-two, avoid the fold since it can be
+  // best implemented as a bit test.
+  if (AllDivisorsArePowerOfTwo)
+    return SDValue();
 
-  SelectionDAG &DAG = DCI.DAG;
+  SDValue PVal, KVal, QVal;
+  if (VT.isVector()) {
+    if (HadOneDivisor) {
+      // Try to turn PAmts into a splat, since we don't care about the values
+      // that are currently '0'. If we can't, just keep '0'`s.
+      turnVectorIntoSplatVector(PAmts, isNullConstant);
+      // Try to turn KAmts into a splat, since we don't care about the values
+      // that are currently '-1'. If we can't, change them to '0'`s.
+      turnVectorIntoSplatVector(KAmts, isAllOnesConstant,
+                                DAG.getConstant(0, DL, ShSVT));
+    }
+
+    PVal = DAG.getBuildVector(VT, DL, PAmts);
+    KVal = DAG.getBuildVector(ShVT, DL, KAmts);
+    QVal = DAG.getBuildVector(VT, DL, QAmts);
+  } else {
+    PVal = PAmts[0];
+    KVal = KAmts[0];
+    QVal = QAmts[0];
+  }
 
-  SDValue PVal = DAG.getConstant(P, DL, VT);
-  SDValue QVal = DAG.getConstant(Q, DL, VT);
   // (mul N, P)
-  SDValue Op1 = DAG.getNode(ISD::MUL, DL, VT, REMNode->getOperand(0), PVal);
-  Created.push_back(Op1.getNode());
+  SDValue Op0 = DAG.getNode(ISD::MUL, DL, VT, N, PVal);
+  Created.push_back(Op0.getNode());
 
-  // Rotate right only if D was even.
-  if (DivisorIsEven) {
+  // Rotate right only if any divisor was even. We avoid rotates for all-odd
+  // divisors as a performance improvement, since rotating by 0 is a no-op.
+  if (HadEvenDivisor) {
     // We need ROTR to do this.
     if (!isOperationLegalOrCustom(ISD::ROTR, VT))
       return SDValue();
-    SDValue ShAmt =
-        DAG.getConstant(K, DL, getShiftAmountTy(VT, DAG.getDataLayout()));
     SDNodeFlags Flags;
     Flags.setExact(true);
     // UREM: (rotr (mul N, P), K)
-    Op1 = DAG.getNode(ISD::ROTR, DL, VT, Op1, ShAmt, Flags);
-    Created.push_back(Op1.getNode());
+    Op0 = DAG.getNode(ISD::ROTR, DL, VT, Op0, KVal, Flags);
+    Created.push_back(Op0.getNode());
   }
 
   // UREM: (setule/setugt (rotr (mul N, P), K), Q)
-  return DAG.getSetCC(DL, SETCCVT, Op1, QVal,
+  return DAG.getSetCC(DL, SETCCVT, Op0, QVal,
                       ((Cond == ISD::SETEQ) ? ISD::SETULE : ISD::SETUGT));
 }
 

diff --git a/test/CodeGen/AArch64/urem-seteq-vec-nonsplat.ll b/test/CodeGen/AArch64/urem-seteq-vec-nonsplat.ll
index 57f6941d5e3c4871cb4988f98a5c084d6370d07b..c5b0148b99a5c062d45b2d97de3030489b8fefff 100644 (file)
--- a/test/CodeGen/AArch64/urem-seteq-vec-nonsplat.ll
+++ b/test/CodeGen/AArch64/urem-seteq-vec-nonsplat.ll
@@ -40,18 +40,11 @@ define <4 x i32> @test_urem_odd_allones_eq(<4 x i32> %X) nounwind {
 ; CHECK-LABEL: test_urem_odd_allones_eq:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    adrp x8, .LCPI1_0
+; CHECK-NEXT:    adrp x9, .LCPI1_1
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI1_0]
-; CHECK-NEXT:    adrp x8, .LCPI1_1
-; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI1_1]
-; CHECK-NEXT:    adrp x8, .LCPI1_2
-; CHECK-NEXT:    ldr q3, [x8, :lo12:.LCPI1_2]
-; CHECK-NEXT:    umull2 v4.2d, v0.4s, v1.4s
-; CHECK-NEXT:    umull v1.2d, v0.2s, v1.2s
-; CHECK-NEXT:    uzp2 v1.4s, v1.4s, v4.4s
-; CHECK-NEXT:    neg v2.4s, v2.4s
-; CHECK-NEXT:    ushl v1.4s, v1.4s, v2.4s
-; CHECK-NEXT:    mls v0.4s, v1.4s, v3.4s
-; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
+; CHECK-NEXT:    ldr q2, [x9, :lo12:.LCPI1_1]
+; CHECK-NEXT:    mul v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    cmhs v0.4s, v2.4s, v0.4s
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -64,19 +57,11 @@ define <4 x i32> @test_urem_odd_allones_ne(<4 x i32> %X) nounwind {
 ; CHECK-LABEL: test_urem_odd_allones_ne:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    adrp x8, .LCPI2_0
+; CHECK-NEXT:    adrp x9, .LCPI2_1
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI2_0]
-; CHECK-NEXT:    adrp x8, .LCPI2_1
-; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI2_1]
-; CHECK-NEXT:    adrp x8, .LCPI2_2
-; CHECK-NEXT:    ldr q3, [x8, :lo12:.LCPI2_2]
-; CHECK-NEXT:    umull2 v4.2d, v0.4s, v1.4s
-; CHECK-NEXT:    umull v1.2d, v0.2s, v1.2s
-; CHECK-NEXT:    uzp2 v1.4s, v1.4s, v4.4s
-; CHECK-NEXT:    neg v2.4s, v2.4s
-; CHECK-NEXT:    ushl v1.4s, v1.4s, v2.4s
-; CHECK-NEXT:    mls v0.4s, v1.4s, v3.4s
-; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
-; CHECK-NEXT:    mvn v0.16b, v0.16b
+; CHECK-NEXT:    ldr q2, [x9, :lo12:.LCPI2_1]
+; CHECK-NEXT:    mul v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    cmhi v0.4s, v0.4s, v2.4s
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -300,20 +285,11 @@ define <4 x i32> @test_urem_odd_one(<4 x i32> %X) nounwind {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    adrp x8, .LCPI10_0
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI10_0]
-; CHECK-NEXT:    adrp x8, .LCPI10_1
-; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI10_1]
-; CHECK-NEXT:    adrp x8, .LCPI10_2
-; CHECK-NEXT:    ldr q3, [x8, :lo12:.LCPI10_2]
-; CHECK-NEXT:    adrp x8, .LCPI10_3
-; CHECK-NEXT:    umull2 v4.2d, v0.4s, v1.4s
-; CHECK-NEXT:    umull v1.2d, v0.2s, v1.2s
-; CHECK-NEXT:    uzp2 v1.4s, v1.4s, v4.4s
-; CHECK-NEXT:    ldr q4, [x8, :lo12:.LCPI10_3]
-; CHECK-NEXT:    neg v2.4s, v2.4s
-; CHECK-NEXT:    ushl v1.4s, v1.4s, v2.4s
-; CHECK-NEXT:    bsl v3.16b, v0.16b, v1.16b
-; CHECK-NEXT:    mls v0.4s, v3.4s, v4.4s
-; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
+; CHECK-NEXT:    mov w8, #52429
+; CHECK-NEXT:    movk w8, #52428, lsl #16
+; CHECK-NEXT:    dup v2.4s, w8
+; CHECK-NEXT:    mul v0.4s, v0.4s, v2.4s
+; CHECK-NEXT:    cmhs v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -480,21 +456,11 @@ define <4 x i32> @test_urem_odd_allones_and_one(<4 x i32> %X) nounwind {
 ; CHECK-LABEL: test_urem_odd_allones_and_one:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    adrp x8, .LCPI16_0
+; CHECK-NEXT:    adrp x9, .LCPI16_1
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI16_0]
-; CHECK-NEXT:    adrp x8, .LCPI16_1
-; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI16_1]
-; CHECK-NEXT:    adrp x8, .LCPI16_2
-; CHECK-NEXT:    ldr q3, [x8, :lo12:.LCPI16_2]
-; CHECK-NEXT:    adrp x8, .LCPI16_3
-; CHECK-NEXT:    umull2 v4.2d, v0.4s, v1.4s
-; CHECK-NEXT:    umull v1.2d, v0.2s, v1.2s
-; CHECK-NEXT:    uzp2 v1.4s, v1.4s, v4.4s
-; CHECK-NEXT:    ldr q4, [x8, :lo12:.LCPI16_3]
-; CHECK-NEXT:    neg v2.4s, v2.4s
-; CHECK-NEXT:    ushl v1.4s, v1.4s, v2.4s
-; CHECK-NEXT:    bsl v3.16b, v0.16b, v1.16b
-; CHECK-NEXT:    mls v0.4s, v3.4s, v4.4s
-; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
+; CHECK-NEXT:    ldr q2, [x9, :lo12:.LCPI16_1]
+; CHECK-NEXT:    mul v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    cmhs v0.4s, v2.4s, v0.4s
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret

diff --git a/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll b/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll
index 905f45b6985a4ed5bed07e2e902b96ce03a9da54..c1669cf4adb241416d4dfa5d4035459afedc4c58 100644 (file)
--- a/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll
+++ b/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll
@@ -115,18 +115,9 @@ define <4 x i32> @test_urem_odd_even(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_even:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2454267027,1374389535,1374389535]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm3
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm4, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -142,97 +133,33 @@ define <4 x i32> @test_urem_odd_even(<4 x i32> %X) nounwind {
 define <4 x i32> @test_urem_odd_allones_eq(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_allones_eq:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = <3435973837,u,2147483649,u>
-; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    psrld $31, %xmm1
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
-; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
-; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-SSE2-NEXT:    psrld $31, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-SSE2-NEXT:    pxor {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pandn {{.*}}(%rip), %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_allones_eq:
 ; CHECK-SSE41:       # %bb.0:
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq {{.*}}(%rip), %xmm1
-; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = <3435973837,u,2147483649,u>
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-SSE41-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE41-NEXT:    psrld $31, %xmm1
-; CHECK-SSE41-NEXT:    psrld $2, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5],xmm2[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
-; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
-; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [858993459,858993459,1,858993459]
+; CHECK-SSE41-NEXT:    pminud %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
-; CHECK-AVX1-LABEL: test_urem_odd_allones_eq:
-; CHECK-AVX1:       # %bb.0:
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpmuludq {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm2
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT:    vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    vpsrld $31, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    retq
-;
-; CHECK-AVX2-LABEL: test_urem_odd_allones_eq:
-; CHECK-AVX2:       # %bb.0:
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [3435973837,3435973837,3435973837,3435973837]
-; CHECK-AVX2-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm2
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX2-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    vpsrld $31, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    retq
-;
-; CHECK-AVX512VL-LABEL: test_urem_odd_allones_eq:
-; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [3435973837,3435973837,3435973837,3435973837]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm2
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    retq
+; CHECK-AVX-LABEL: test_urem_odd_allones_eq:
+; CHECK-AVX:       # %bb.0:
+; CHECK-AVX-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
+; CHECK-AVX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; CHECK-AVX-NEXT:    vpsrld $31, %xmm0, %xmm0
+; CHECK-AVX-NEXT:    retq
   %urem = urem <4 x i32> %X, <i32 5, i32 5, i32 4294967295, i32 5>
   %cmp = icmp eq <4 x i32> %urem, <i32 0, i32 0, i32 0, i32 0>
   %ret = zext <4 x i1> %cmp to <4 x i32>
@@ -241,98 +168,33 @@ define <4 x i32> @test_urem_odd_allones_eq(<4 x i32> %X) nounwind {
 define <4 x i32> @test_urem_odd_allones_ne(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_allones_ne:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = <3435973837,u,2147483649,u>
-; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    psrld $31, %xmm1
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
-; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
-; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-SSE2-NEXT:    pandn {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-SSE2-NEXT:    pxor {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_allones_ne:
 ; CHECK-SSE41:       # %bb.0:
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq {{.*}}(%rip), %xmm1
-; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = <3435973837,u,2147483649,u>
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-SSE41-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE41-NEXT:    psrld $31, %xmm1
-; CHECK-SSE41-NEXT:    psrld $2, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5],xmm2[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
-; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
-; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [858993460,858993460,2,858993460]
+; CHECK-SSE41-NEXT:    pmaxud %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-SSE41-NEXT:    pandn {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
-; CHECK-AVX1-LABEL: test_urem_odd_allones_ne:
-; CHECK-AVX1:       # %bb.0:
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpmuludq {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm2
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT:    vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    vpandn {{.*}}(%rip), %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    retq
-;
-; CHECK-AVX2-LABEL: test_urem_odd_allones_ne:
-; CHECK-AVX2:       # %bb.0:
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [3435973837,3435973837,3435973837,3435973837]
-; CHECK-AVX2-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm2
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX2-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
-; CHECK-AVX2-NEXT:    vpandn %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    retq
-;
-; CHECK-AVX512VL-LABEL: test_urem_odd_allones_ne:
-; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [3435973837,3435973837,3435973837,3435973837]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm2
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpandnd {{.*}}(%rip){1to4}, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    retq
+; CHECK-AVX-LABEL: test_urem_odd_allones_ne:
+; CHECK-AVX:       # %bb.0:
+; CHECK-AVX-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX-NEXT:    vpmaxud {{.*}}(%rip), %xmm0, %xmm1
+; CHECK-AVX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; CHECK-AVX-NEXT:    vpsrld $31, %xmm0, %xmm0
+; CHECK-AVX-NEXT:    retq
   %urem = urem <4 x i32> %X, <i32 5, i32 5, i32 4294967295, i32 5>
   %cmp = icmp ne <4 x i32> %urem, <i32 0, i32 0, i32 0, i32 0>
   %ret = zext <4 x i1> %cmp to <4 x i32>
@@ -430,17 +292,9 @@ define <4 x i32> @test_urem_even_allones_eq(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_even_allones_eq:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [2454267027,2454267027,2454267027,2454267027]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm3, %xmm2, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -540,19 +394,11 @@ define <4 x i32> @test_urem_even_allones_ne(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_even_allones_ne:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [2454267027,2454267027,2454267027,2454267027]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm3, %xmm2, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpmaxud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpandnd {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
   %urem = urem <4 x i32> %X, <i32 14, i32 14, i32 4294967295, i32 14>
   %cmp = icmp ne <4 x i32> %urem, <i32 0, i32 0, i32 0, i32 0>
@@ -668,18 +514,9 @@ define <4 x i32> @test_urem_odd_even_allones_eq(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_even_allones_eq:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2454267027,2147483649,1374389535]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm3
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm4, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -796,20 +633,11 @@ define <4 x i32> @test_urem_odd_even_allones_ne(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_even_allones_ne:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2454267027,2147483649,1374389535]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm3
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm4, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpmaxud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpandnd {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
   %urem = urem <4 x i32> %X, <i32 5, i32 14, i32 4294967295, i32 100>
   %cmp = icmp ne <4 x i32> %urem, <i32 0, i32 0, i32 0, i32 0>
@@ -897,16 +725,9 @@ define <4 x i32> @test_urem_odd_poweroftwo(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_poweroftwo:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [3435973837,3435973837,3435973837,3435973837]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm2
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1003,17 +824,9 @@ define <4 x i32> @test_urem_even_poweroftwo(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_even_poweroftwo:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [2454267027,2454267027,2454267027,2454267027]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm3, %xmm2, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1128,18 +941,9 @@ define <4 x i32> @test_urem_odd_even_poweroftwo(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_even_poweroftwo:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2454267027,268435456,1374389535]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm3
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm4, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1155,100 +959,48 @@ define <4 x i32> @test_urem_odd_even_poweroftwo(<4 x i32> %X) nounwind {
 define <4 x i32> @test_urem_odd_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_one:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movl $-858993459, %eax # imm = 0xCCCCCCCD
-; CHECK-SSE2-NEXT:    movd %eax, %xmm1
-; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,3435973837,3435973837,3435973837]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
-; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
-; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-SSE2-NEXT:    psrld $31, %xmm0
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-SSE2-NEXT:    pxor {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pandn {{.*}}(%rip), %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_one:
 ; CHECK-SSE41:       # %bb.0:
-; CHECK-SSE41-NEXT:    movl $-858993459, %eax # imm = 0xCCCCCCCD
-; CHECK-SSE41-NEXT:    movd %eax, %xmm1
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm1
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
-; CHECK-SSE41-NEXT:    psrld $2, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm0[4,5],xmm2[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
-; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
-; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [858993459,858993459,4294967295,858993459]
+; CHECK-SSE41-NEXT:    pminud %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
 ; CHECK-AVX1-LABEL: test_urem_odd_one:
 ; CHECK-AVX1:       # %bb.0:
-; CHECK-AVX1-NEXT:    movl $-858993459, %eax # imm = 0xCCCCCCCD
-; CHECK-AVX1-NEXT:    vmovd %eax, %xmm1
-; CHECK-AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpmuludq {{.*}}(%rip), %xmm2, %xmm2
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX1-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    retq
 ;
 ; CHECK-AVX2-LABEL: test_urem_odd_one:
 ; CHECK-AVX2:       # %bb.0:
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [3435973837,3435973837,3435973837,3435973837]
-; CHECK-AVX2-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    movl $-858993459, %eax # imm = 0xCCCCCCCD
-; CHECK-AVX2-NEXT:    vmovd %eax, %xmm2
-; CHECK-AVX2-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX2-NEXT:    vpsrld $2, %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX2-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [3435973837,3435973837,3435973837,3435973837]
+; CHECK-AVX2-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX2-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX2-NEXT:    retq
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [3435973837,3435973837,3435973837,3435973837]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    movl $-858993459, %eax # imm = 0xCCCCCCCD
-; CHECK-AVX512VL-NEXT:    vmovd %eax, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpsrld $2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1354,20 +1106,9 @@ define <4 x i32> @test_urem_even_one(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_even_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [2454267027,2454267027,2454267027,2454267027]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm3, %xmm2, %xmm2
-; CHECK-AVX512VL-NEXT:    movl $-1840700269, %eax # imm = 0x92492493
-; CHECK-AVX512VL-NEXT:    vmovd %eax, %xmm3
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrld $2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprord $1, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1482,19 +1223,9 @@ define <4 x i32> @test_urem_odd_even_one(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_even_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2454267027,0,1374389535]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm3
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm4, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1601,17 +1332,9 @@ define <4 x i32> @test_urem_odd_allones_and_poweroftwo(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_allones_and_poweroftwo:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,3435973837]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1726,18 +1449,9 @@ define <4 x i32> @test_urem_even_allones_and_poweroftwo(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_even_allones_and_poweroftwo:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [2454267027,2147483649,268435456,2454267027]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm3
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm4, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1845,17 +1559,9 @@ define <4 x i32> @test_urem_odd_even_allones_and_poweroftwo(<4 x i32> %X) nounwi
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_even_allones_and_poweroftwo:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,1374389535]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -1871,113 +1577,35 @@ define <4 x i32> @test_urem_odd_even_allones_and_poweroftwo(<4 x i32> %X) nounwi
 define <4 x i32> @test_urem_odd_allones_and_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_allones_and_one:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2147483649,0,3435973837]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,4294967295,0,3435973837]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    psrld $31, %xmm2
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [5,4294967295,1,5]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm5
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm1[0,0]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm5
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm5[0,2]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
-; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
-; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-SSE2-NEXT:    psrld $31, %xmm0
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-SSE2-NEXT:    pxor {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT:    pandn {{.*}}(%rip), %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_allones_and_one:
 ; CHECK-SSE41:       # %bb.0:
-; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2147483649,0,3435973837]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm3
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm1
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
-; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE41-NEXT:    psrld $31, %xmm2
-; CHECK-SSE41-NEXT:    psrld $2, %xmm1
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5],xmm1[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm1
-; CHECK-SSE41-NEXT:    psubd %xmm1, %xmm0
-; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [858993459,1,4294967295,858993459]
+; CHECK-SSE41-NEXT:    pminud %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
-; CHECK-AVX1-LABEL: test_urem_odd_allones_and_one:
-; CHECK-AVX1:       # %bb.0:
-; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,0,3435973837]
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
-; CHECK-AVX1-NEXT:    vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    vpsrld $31, %xmm0, %xmm0
-; CHECK-AVX1-NEXT:    retq
-;
-; CHECK-AVX2-LABEL: test_urem_odd_allones_and_one:
-; CHECK-AVX2:       # %bb.0:
-; CHECK-AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,0,3435973837]
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX2-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX2-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    vpsrld $31, %xmm0, %xmm0
-; CHECK-AVX2-NEXT:    retq
-;
-; CHECK-AVX512VL-LABEL: test_urem_odd_allones_and_one:
-; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,0,3435973837]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    retq
+; CHECK-AVX-LABEL: test_urem_odd_allones_and_one:
+; CHECK-AVX:       # %bb.0:
+; CHECK-AVX-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
+; CHECK-AVX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; CHECK-AVX-NEXT:    vpsrld $31, %xmm0, %xmm0
+; CHECK-AVX-NEXT:    retq
   %urem = urem <4 x i32> %X, <i32 5, i32 4294967295, i32 1, i32 5>
   %cmp = icmp eq <4 x i32> %urem, <i32 0, i32 0, i32 0, i32 0>
   %ret = zext <4 x i1> %cmp to <4 x i32>
@@ -2090,19 +1718,9 @@ define <4 x i32> @test_urem_even_allones_and_one(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_even_allones_and_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [2454267027,2147483649,0,2454267027]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm3
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm4, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -2213,18 +1831,9 @@ define <4 x i32> @test_urem_odd_even_allones_and_one(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_even_allones_and_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,0,1374389535]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -2328,18 +1937,9 @@ define <4 x i32> @test_urem_odd_poweroftwo_and_one(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_poweroftwo_and_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,268435456,0,3435973837]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -2451,19 +2051,9 @@ define <4 x i32> @test_urem_even_poweroftwo_and_one(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_even_poweroftwo_and_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [2454267027,268435456,0,2454267027]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm3
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm4, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -2570,18 +2160,9 @@ define <4 x i32> @test_urem_odd_even_poweroftwo_and_one(<4 x i32> %X) nounwind {
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_even_poweroftwo_and_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,268435456,0,1374389535]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -2688,18 +2269,9 @@ define <4 x i32> @test_urem_odd_allones_and_poweroftwo_and_one(<4 x i32> %X) nou
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_odd_allones_and_poweroftwo_and_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,0]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm0[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq
@@ -2815,21 +2387,9 @@ define <4 x i32> @test_urem_even_allones_and_poweroftwo_and_one(<4 x i32> %X) no
 ;
 ; CHECK-AVX512VL-LABEL: test_urem_even_allones_and_poweroftwo_and_one:
 ; CHECK-AVX512VL:       # %bb.0:
-; CHECK-AVX512VL-NEXT:    movl $1, %eax
-; CHECK-AVX512VL-NEXT:    vmovd %eax, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm3 = [2454267027,2147483649,268435456,0]
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm4, %xmm2, %xmm2
-; CHECK-AVX512VL-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm0[3]
-; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vprorvd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm1
 ; CHECK-AVX512VL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX512VL-NEXT:    retq