[TTI] Add OperandValueProperties::OP_NegatedPowerOf2 enum (PR51436)

The mul by constant costmodels handle power-of-2 constants, but not negated-power-of-2, despite the backends handling both.

This patch adds the OperandValueProperties::OP_NegatedPowerOf2 enum and wires it for use for basic mul cost analysis and SLP handling.

Fixes #50778

Differential Revision: https://reviews.llvm.org/D111968

diff --git a/llvm/include/llvm/Analysis/TargetTransformInfo.h b/llvm/include/llvm/Analysis/TargetTransformInfo.h
index 525b5db..9b8f0a9 100644 (file)
--- a/llvm/include/llvm/Analysis/TargetTransformInfo.h
+++ b/llvm/include/llvm/Analysis/TargetTransformInfo.h
@@ -909,7 +909,11 @@ public:
   };
 
   /// Additional properties of an operand's values.
-  enum OperandValueProperties { OP_None = 0, OP_PowerOf2 = 1 };
+  enum OperandValueProperties {
+    OP_None = 0,
+    OP_PowerOf2 = 1,
+    OP_NegatedPowerOf2 = 2,
+  };
 
   // Describe the values an operand can take.  We're in the process
   // of migrating uses of OperandValueKind and OperandValueProperties
@@ -927,6 +931,9 @@ public:
     bool isPowerOf2() const {
       return Properties == OP_PowerOf2;
     }
+    bool isNegatedPowerOf2() const {
+      return Properties == OP_NegatedPowerOf2;
+    }
 
     OperandValueInfo getNoProps() const {
       return {Kind, OP_None};

diff --git a/llvm/lib/Analysis/TargetTransformInfo.cpp b/llvm/lib/Analysis/TargetTransformInfo.cpp
index 1d0b31e..cde0b81 100644 (file)
--- a/llvm/lib/Analysis/TargetTransformInfo.cpp
+++ b/llvm/lib/Analysis/TargetTransformInfo.cpp
@@ -730,6 +730,8 @@ TargetTransformInfo::getOperandInfo(const Value *V) {
     if (const auto *CI = dyn_cast<ConstantInt>(V))
       if (CI->getValue().isPowerOf2())
         OpProps = OP_PowerOf2;
+      else if (CI->getValue().isNegatedPowerOf2())
+        OpProps = OP_NegatedPowerOf2;
     return {OK_UniformConstantValue, OpProps};
   }
 
@@ -751,15 +753,22 @@ TargetTransformInfo::getOperandInfo(const Value *V) {
       if (auto *CI = dyn_cast<ConstantInt>(Splat))
         if (CI->getValue().isPowerOf2())
           OpProps = OP_PowerOf2;
+        else if (CI->getValue().isNegatedPowerOf2())
+          OpProps = OP_NegatedPowerOf2;
     } else if (const auto *CDS = dyn_cast<ConstantDataSequential>(V)) {
-      OpProps = OP_PowerOf2;
+      bool AllPow2 = true, AllNegPow2 = true;
       for (unsigned I = 0, E = CDS->getNumElements(); I != E; ++I) {
-        if (auto *CI = dyn_cast<ConstantInt>(CDS->getElementAsConstant(I)))
-          if (CI->getValue().isPowerOf2())
+        if (auto *CI = dyn_cast<ConstantInt>(CDS->getElementAsConstant(I))) {
+          AllPow2 &= CI->getValue().isPowerOf2();
+          AllNegPow2 &= CI->getValue().isNegatedPowerOf2();
+          if (AllPow2 || AllNegPow2)
             continue;
-        OpProps = OP_None;
+        }
+        AllPow2 = AllNegPow2 = false;
         break;
       }
+      OpProps = AllPow2 ? OP_PowerOf2 : OpProps;
+      OpProps = AllNegPow2 ? OP_NegatedPowerOf2 : OpProps;
     }
   }
 

diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index 5c90e21..1018e9b 100644 (file)
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -288,9 +288,16 @@ InstructionCost X86TTIImpl::getArithmeticInstrCost(
   }
 
   // Vector multiply by pow2 will be simplified to shifts.
-  if (ISD == ISD::MUL && Op2Info.isConstant() && Op2Info.isPowerOf2())
-    return getArithmeticInstrCost(Instruction::Shl, Ty, CostKind,
-                                  Op1Info.getNoProps(), Op2Info.getNoProps());
+  // Vector multiply by -pow2 will be simplified to shifts/negates.
+  if (ISD == ISD::MUL && Op2Info.isConstant() &&
+      (Op2Info.isPowerOf2() || Op2Info.isNegatedPowerOf2())) {
+    InstructionCost Cost =
+        getArithmeticInstrCost(Instruction::Shl, Ty, CostKind,
+                               Op1Info.getNoProps(), Op2Info.getNoProps());
+    if (Op2Info.isNegatedPowerOf2())
+      Cost += getArithmeticInstrCost(Instruction::Sub, Ty, CostKind);
+    return Cost;
+  }
 
   // On X86, vector signed division by constants power-of-two are
   // normally expanded to the sequence SRA + SRL + ADD + SRA.

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 9aad015..daea8d4 100644 (file)
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -5961,6 +5961,13 @@ TTI::OperandValueInfo BoUpSLP::getOperandInfo(ArrayRef<Value *> VL,
       return CI->getValue().isPowerOf2();
     return false;
   });
+  const bool IsNegatedPowerOfTwo = all_of(VL, [&](Value *V) {
+    // TODO: We should allow undef elements here
+    auto *Op = cast<Instruction>(V)->getOperand(OpIdx);
+    if (auto *CI = dyn_cast<ConstantInt>(Op))
+      return CI->getValue().isNegatedPowerOf2();
+    return false;
+  });
 
   TTI::OperandValueKind VK = TTI::OK_AnyValue;
   if (IsConstant && IsUniform)
@@ -5970,8 +5977,10 @@ TTI::OperandValueInfo BoUpSLP::getOperandInfo(ArrayRef<Value *> VL,
   else if (IsUniform)
     VK = TTI::OK_UniformValue;
 
-  const TTI::OperandValueProperties VP =
-    IsPowerOfTwo ? TTI::OP_PowerOf2 : TTI::OP_None;
+  TTI::OperandValueProperties VP = TTI::OP_None;
+  VP = IsPowerOfTwo ? TTI::OP_PowerOf2 : VP;
+  VP = IsNegatedPowerOfTwo ? TTI::OP_NegatedPowerOf2 : VP;
+
   return {VK, VP};
 }
 

diff --git a/llvm/test/Analysis/CostModel/X86/mul-codesize.ll b/llvm/test/Analysis/CostModel/X86/mul-codesize.ll
index 4291075..1c596ff 100644 (file)
--- a/llvm/test/Analysis/CostModel/X86/mul-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/mul-codesize.ll
@@ -365,155 +365,155 @@ define i32 @mul_uniformconstpow2() {
 
 define i32 @mul_constnegpow2() {
 ; SSE2-LABEL: 'mul_constnegpow2'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'mul_constnegpow2'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'mul_constnegpow2'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'mul_constnegpow2'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'mul_constnegpow2'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'mul_constnegpow2'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'mul_constnegpow2'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'mul_constnegpow2'
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = mul i64 undef, -16
@@ -541,155 +541,155 @@ define i32 @mul_constnegpow2() {
 
 define i32 @mul_uniformconstnegpow2() {
 ; SSE2-LABEL: 'mul_uniformconstnegpow2'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'mul_uniformconstnegpow2'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'mul_uniformconstnegpow2'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'mul_uniformconstnegpow2'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'mul_uniformconstnegpow2'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'mul_uniformconstnegpow2'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'mul_uniformconstnegpow2'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'mul_uniformconstnegpow2'
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = mul i64 undef, -16

diff --git a/llvm/test/Analysis/CostModel/X86/mul-latency.ll b/llvm/test/Analysis/CostModel/X86/mul-latency.ll
index 20c69db..7dfe3be 100644 (file)
--- a/llvm/test/Analysis/CostModel/X86/mul-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/mul-latency.ll
@@ -365,155 +365,155 @@ define i32 @mul_uniformconstpow2() {
 
 define i32 @mul_constnegpow2() {
 ; SSE2-LABEL: 'mul_constnegpow2'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'mul_constnegpow2'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'mul_constnegpow2'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'mul_constnegpow2'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'mul_constnegpow2'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'mul_constnegpow2'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'mul_constnegpow2'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'mul_constnegpow2'
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 45 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = mul i64 undef, -16
@@ -541,155 +541,155 @@ define i32 @mul_constnegpow2() {
 
 define i32 @mul_uniformconstnegpow2() {
 ; SSE2-LABEL: 'mul_uniformconstnegpow2'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'mul_uniformconstnegpow2'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'mul_uniformconstnegpow2'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'mul_uniformconstnegpow2'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'mul_uniformconstnegpow2'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'mul_uniformconstnegpow2'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'mul_uniformconstnegpow2'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'mul_uniformconstnegpow2'
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = mul i64 undef, -16

diff --git a/llvm/test/Analysis/CostModel/X86/mul-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/mul-sizelatency.ll
index ac8ba56..7fc7d5f 100644 (file)
--- a/llvm/test/Analysis/CostModel/X86/mul-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/mul-sizelatency.ll
@@ -365,155 +365,155 @@ define i32 @mul_uniformconstpow2() {
 
 define i32 @mul_constnegpow2() {
 ; SSE2-LABEL: 'mul_constnegpow2'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'mul_constnegpow2'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'mul_constnegpow2'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'mul_constnegpow2'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'mul_constnegpow2'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'mul_constnegpow2'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'mul_constnegpow2'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'mul_constnegpow2'
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = mul i64 undef, -16
@@ -541,155 +541,155 @@ define i32 @mul_constnegpow2() {
 
 define i32 @mul_uniformconstnegpow2() {
 ; SSE2-LABEL: 'mul_uniformconstnegpow2'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'mul_uniformconstnegpow2'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'mul_uniformconstnegpow2'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'mul_uniformconstnegpow2'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'mul_uniformconstnegpow2'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'mul_uniformconstnegpow2'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'mul_uniformconstnegpow2'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'mul_uniformconstnegpow2'
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = mul i64 undef, -16

diff --git a/llvm/test/Analysis/CostModel/X86/mul.ll b/llvm/test/Analysis/CostModel/X86/mul.ll
index ce9b5e0..6581c0f 100644 (file)
--- a/llvm/test/Analysis/CostModel/X86/mul.ll
+++ b/llvm/test/Analysis/CostModel/X86/mul.ll
@@ -366,154 +366,154 @@ define i32 @mul_uniformconstpow2() {
 define i32 @mul_constnegpow2() {
 ; SSE2-LABEL: 'mul_constnegpow2'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'mul_constnegpow2'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'mul_constnegpow2'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'mul_constnegpow2'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'mul_constnegpow2'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'mul_constnegpow2'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'mul_constnegpow2'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'mul_constnegpow2'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %I64 = mul i64 undef, -16
@@ -542,151 +542,151 @@ define i32 @mul_constnegpow2() {
 define i32 @mul_uniformconstnegpow2() {
 ; SSE2-LABEL: 'mul_uniformconstnegpow2'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'mul_uniformconstnegpow2'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'mul_uniformconstnegpow2'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'mul_uniformconstnegpow2'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'mul_uniformconstnegpow2'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'mul_uniformconstnegpow2'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'mul_uniformconstnegpow2'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = mul i64 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'mul_uniformconstnegpow2'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I64 = mul i64 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = mul i32 undef, -16
-; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = mul i16 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, -16
+; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>

diff --git a/llvm/test/Transforms/SLPVectorizer/X86/powof2mul.ll b/llvm/test/Transforms/SLPVectorizer/X86/powof2mul.ll
index 44c4e3b..b70df33 100644 (file)
--- a/llvm/test/Transforms/SLPVectorizer/X86/powof2mul.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/powof2mul.ll
@@ -190,45 +190,33 @@ entry:
 define void @PR51436(i64* nocapture %a) {
 ; SSE-LABEL: @PR51436(
 ; SSE-NEXT:  entry:
-; SSE-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 1
-; SSE-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 2
-; SSE-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 3
+; SSE-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 2
 ; SSE-NEXT:    [[GEP4:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 4
-; SSE-NEXT:    [[GEP5:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 5
 ; SSE-NEXT:    [[GEP6:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 6
-; SSE-NEXT:    [[GEP7:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 7
-; SSE-NEXT:    [[LOAD0:%.*]] = load i64, i64* [[A]], align 8
-; SSE-NEXT:    [[LOAD1:%.*]] = load i64, i64* [[GEP1]], align 8
-; SSE-NEXT:    [[LOAD2:%.*]] = load i64, i64* [[GEP2]], align 8
-; SSE-NEXT:    [[LOAD3:%.*]] = load i64, i64* [[GEP3]], align 8
-; SSE-NEXT:    [[LOAD4:%.*]] = load i64, i64* [[GEP4]], align 8
-; SSE-NEXT:    [[LOAD5:%.*]] = load i64, i64* [[GEP5]], align 8
-; SSE-NEXT:    [[LOAD6:%.*]] = load i64, i64* [[GEP6]], align 8
-; SSE-NEXT:    [[LOAD7:%.*]] = load i64, i64* [[GEP7]], align 8
-; SSE-NEXT:    [[MUL0:%.*]] = mul i64 [[LOAD0]], -17592186044416
-; SSE-NEXT:    [[MUL1:%.*]] = mul i64 [[LOAD1]], -17592186044416
-; SSE-NEXT:    [[MUL2:%.*]] = mul i64 [[LOAD2]], -17592186044416
-; SSE-NEXT:    [[MUL3:%.*]] = mul i64 [[LOAD3]], -17592186044416
-; SSE-NEXT:    [[MUL4:%.*]] = mul i64 [[LOAD4]], -17592186044416
-; SSE-NEXT:    [[MUL5:%.*]] = mul i64 [[LOAD5]], -17592186044416
-; SSE-NEXT:    [[MUL6:%.*]] = mul i64 [[LOAD6]], -17592186044416
-; SSE-NEXT:    [[MUL7:%.*]] = mul i64 [[LOAD7]], -17592186044416
-; SSE-NEXT:    [[ADD0:%.*]] = add i64 [[MUL0]], -17592186044416
-; SSE-NEXT:    [[ADD1:%.*]] = add i64 [[MUL1]], -17592186044416
-; SSE-NEXT:    [[ADD2:%.*]] = add i64 [[MUL2]], -17592186044416
-; SSE-NEXT:    [[ADD3:%.*]] = add i64 [[MUL3]], -17592186044416
-; SSE-NEXT:    [[ADD4:%.*]] = add i64 [[MUL4]], -17592186044416
-; SSE-NEXT:    [[ADD5:%.*]] = add i64 [[MUL5]], -17592186044416
-; SSE-NEXT:    [[ADD6:%.*]] = add i64 [[MUL6]], -17592186044416
-; SSE-NEXT:    [[ADD7:%.*]] = add i64 [[MUL7]], -17592186044416
-; SSE-NEXT:    store i64 [[ADD0]], i64* [[A]], align 8
-; SSE-NEXT:    store i64 [[ADD1]], i64* [[GEP1]], align 8
-; SSE-NEXT:    store i64 [[ADD2]], i64* [[GEP2]], align 8
-; SSE-NEXT:    store i64 [[ADD3]], i64* [[GEP3]], align 8
-; SSE-NEXT:    store i64 [[ADD4]], i64* [[GEP4]], align 8
-; SSE-NEXT:    store i64 [[ADD5]], i64* [[GEP5]], align 8
-; SSE-NEXT:    store i64 [[ADD6]], i64* [[GEP6]], align 8
-; SSE-NEXT:    store i64 [[ADD7]], i64* [[GEP7]], align 8
+; SSE-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A]] to <2 x i64>*
+; SSE-NEXT:    [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* [[TMP0]], align 8
+; SSE-NEXT:    [[TMP2:%.*]] = mul <2 x i64> [[TMP1]], <i64 -17592186044416, i64 -17592186044416>
+; SSE-NEXT:    [[TMP3:%.*]] = add <2 x i64> [[TMP2]], <i64 -17592186044416, i64 -17592186044416>
+; SSE-NEXT:    [[TMP4:%.*]] = bitcast i64* [[A]] to <2 x i64>*
+; SSE-NEXT:    store <2 x i64> [[TMP3]], <2 x i64>* [[TMP4]], align 8
+; SSE-NEXT:    [[TMP5:%.*]] = bitcast i64* [[GEP2]] to <2 x i64>*
+; SSE-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[TMP5]], align 8
+; SSE-NEXT:    [[TMP7:%.*]] = mul <2 x i64> [[TMP6]], <i64 -17592186044416, i64 -17592186044416>
+; SSE-NEXT:    [[TMP8:%.*]] = add <2 x i64> [[TMP7]], <i64 -17592186044416, i64 -17592186044416>
+; SSE-NEXT:    [[TMP9:%.*]] = bitcast i64* [[GEP2]] to <2 x i64>*
+; SSE-NEXT:    store <2 x i64> [[TMP8]], <2 x i64>* [[TMP9]], align 8
+; SSE-NEXT:    [[TMP10:%.*]] = bitcast i64* [[GEP4]] to <2 x i64>*
+; SSE-NEXT:    [[TMP11:%.*]] = load <2 x i64>, <2 x i64>* [[TMP10]], align 8
+; SSE-NEXT:    [[TMP12:%.*]] = mul <2 x i64> [[TMP11]], <i64 -17592186044416, i64 -17592186044416>
+; SSE-NEXT:    [[TMP13:%.*]] = add <2 x i64> [[TMP12]], <i64 -17592186044416, i64 -17592186044416>
+; SSE-NEXT:    [[TMP14:%.*]] = bitcast i64* [[GEP4]] to <2 x i64>*
+; SSE-NEXT:    store <2 x i64> [[TMP13]], <2 x i64>* [[TMP14]], align 8
+; SSE-NEXT:    [[TMP15:%.*]] = bitcast i64* [[GEP6]] to <2 x i64>*
+; SSE-NEXT:    [[TMP16:%.*]] = load <2 x i64>, <2 x i64>* [[TMP15]], align 8
+; SSE-NEXT:    [[TMP17:%.*]] = mul <2 x i64> [[TMP16]], <i64 -17592186044416, i64 -17592186044416>
+; SSE-NEXT:    [[TMP18:%.*]] = add <2 x i64> [[TMP17]], <i64 -17592186044416, i64 -17592186044416>
+; SSE-NEXT:    [[TMP19:%.*]] = bitcast i64* [[GEP6]] to <2 x i64>*
+; SSE-NEXT:    store <2 x i64> [[TMP18]], <2 x i64>* [[TMP19]], align 8
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @PR51436(