[CostModel] Express cost(urem) as cost(div+mul+sub) when set to Expand.

[sdesmalen]

Summary

The Legalizer expands the operations of urem/srem into a div+mul+sub or divrem
when those are legal/custom. This patch changes the cost-model to reflect that
cost.

Since there is no 'divrem' Instruction in LLVM IR, the cost of divrem
is assumed to be the same as div+mul+sub since the three operations will
need to be executed at runtime regardless.

Patch co-authored by David Sherwood (@david-arm)

[sdesmalen]

This mimics the implementation in SelectionDAG which uses TargetLowering::expandREM (see e.g. D81511)

[dmgreen]

This sounds OK to me, so long as the X86 numbers are not bonkers.

llvm/include/llvm/CodeGen/BasicTTIImpl.h

788

I don't think you need the brackets here.

[sdesmalen]

In D103799#2818051, @dmgreen wrote:

This sounds OK to me, so long as the X86 numbers are not bonkers.

Thanks!

@RKSimon do the X86 cost numbers look sensible to you?

[sdesmalen]

I had another look at the costs for X86.

These are a factor three bigger because the cost of a scalar REM is now calculated to be expanded to div+sub+mul rather than the default 1 for 'unknown', which is then multiplied in X86TTIImpl::getArithmeticInstrCost by an arbitrary 20 * LT.first * LT.second.getVectorNumElements(). The comment of that code (https://github.com/llvm/llvm-project/blob/6f3f9535fcafcde11d3b3ef72fdc0f357813e9da/llvm/lib/Target/X86/X86TargetTransformInfo.cpp#L1044) says:

// It is not a good idea to vectorize division. We have to scalarize it and
// in the process we will often end up having to spilling regular
// registers. The overhead of division is going to dominate most kernels
// anyways so try hard to prevent vectorization of division - it is
// generally a bad idea. Assume somewhat arbitrarily that we have to be able
// to hide "20 cycles" for each lane.

The options are to:

1. Tweak this number '20' to keep the cost similar to what they are now (that would require special-casing the code in X86TTI for UREM/SREM, since the condition is currently bundled with SDIV/UDIV).
2. Leave it as-is, since the values are already largely arbitrary an arbitrary number and the higher number satisfies the goal of preventing vectorization of division.

Does anyone have a suggestion and/or preference?

[RKSimon]

In D103799#2860104, @sdesmalen wrote:

I had another look at the costs for X86.

....

The options are to:

1. Tweak this number '20' to keep the cost similar to what they are now (that would require special-casing the code in X86TTI for UREM/SREM, since the condition is currently bundled with SDIV/UDIV).
2. Leave it as-is, since the values are already largely arbitrary an arbitrary number and the higher number satisfies the goal of preventing vectorization of division.

Does anyone have a suggestion and/or preference?

This is actually on my backlog - now that we've improved scalarization costs on x86 I'm hoping to add more realistic scalar div/rem costs instead of relying on this kludge. So I'm happy with (2) for now.

[sdesmalen]

In D103799#2860185, @RKSimon wrote:

This is actually on my backlog - now that we've improved scalarization costs on x86 I'm hoping to add more realistic scalar div/rem costs instead of relying on this kludge. So I'm happy with (2) for now.

Thanks for confirming!

Would you or @dmgreen be happy to accept the patch in that case?

[RKSimon]

LGTM cheers

llvm/include/llvm/CodeGen/BasicTTIImpl.h

InstructionCost getArithmeticInstrCost(
     }
 
     if (!TLI->isOperationExpand(ISD, LT.second)) {
       // If the operation is custom lowered, then assume that the code is twice
       // as expensive.
       return LT.first * 2 * OpCost;
     }
 
+    // An 'Expand' of URem and SRem is special because it may default
+    // to expanding the operation into a sequence of sub-operations
+    // i.e. X % Y -> X-(X/Y)*Y.
+    if (ISD == ISD::UREM || ISD == ISD::SREM) {
+      bool IsSigned = ISD == ISD::SREM;
+      if (TLI->isOperationLegalOrCustom(IsSigned ? ISD::SDIVREM : ISD::UDIVREM,
+                                        LT.second) ||
+          TLI->isOperationLegalOrCustom(IsSigned ? ISD::SDIV : ISD::UDIV,
+                                        LT.second)) {
+        unsigned DivOpc = (IsSigned ? Instruction::SDiv : Instruction::UDiv);
+        InstructionCost DivCost = thisT()->getArithmeticInstrCost(
+            DivOpc, Ty, CostKind, Opd1Info, Opd2Info, Opd1PropInfo,
+            Opd2PropInfo);
+        InstructionCost MulCost =
+            thisT()->getArithmeticInstrCost(Instruction::Mul, Ty, CostKind);
+        InstructionCost SubCost =
+            thisT()->getArithmeticInstrCost(Instruction::Sub, Ty, CostKind);
+        return DivCost + MulCost + SubCost;
+      }
+    }
+
     // We cannot scalarize scalable vectors, so return Invalid.
     if (isa<ScalableVectorType>(Ty))
       return InstructionCost::getInvalid();
 
     // Else, assume that we need to scalarize this op.
     // TODO: If one of the types get legalized by splitting, handle this
     // similarly to what getCastInstrCost() does.
     if (auto *VTy = dyn_cast<FixedVectorType>(Ty)) {

llvm/test/Analysis/CostModel/AArch64/sve-remainder.ll

 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
 ; RUN: opt -cost-model -analyze -mtriple aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
 
 define void @test_urem_srem_expand() {
 ; CHECK-LABEL: 'test_urem_srem_expand'
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %legal_type_urem_0 = urem <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %legal_type_urem_1 = urem <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %legal_type_urem_2 = urem <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %legal_type_urem_3 = urem <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %legal_type_srem_0 = srem <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %legal_type_srem_1 = srem <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %legal_type_srem_2 = srem <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %legal_type_srem_3 = srem <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %split_type_urem_0 = urem <vscale x 32 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %split_type_urem_1 = urem <vscale x 16 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %split_type_urem_2 = urem <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %split_type_urem_3 = urem <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %split_type_srem_0 = srem <vscale x 32 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %split_type_srem_1 = srem <vscale x 16 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %split_type_srem_2 = srem <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %split_type_srem_3 = srem <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %widen_type_urem_0 = urem <vscale x 31 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %widen_type_urem_1 = urem <vscale x 15 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %widen_type_urem_2 = urem <vscale x 7 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %widen_type_urem_3 = urem <vscale x 3 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %widen_type_srem_0 = srem <vscale x 31 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %widen_type_srem_1 = srem <vscale x 15 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %widen_type_srem_2 = srem <vscale x 7 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %widen_type_srem_3 = srem <vscale x 3 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %legal_type_urem_0 = urem <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %legal_type_urem_1 = urem <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %legal_type_urem_2 = urem <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %legal_type_urem_3 = urem <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %legal_type_srem_0 = srem <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %legal_type_srem_1 = srem <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %legal_type_srem_2 = srem <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %legal_type_srem_3 = srem <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %split_type_urem_0 = urem <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %split_type_urem_1 = urem <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %split_type_urem_2 = urem <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %split_type_urem_3 = urem <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %split_type_srem_0 = srem <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %split_type_srem_1 = srem <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %split_type_srem_2 = srem <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %split_type_srem_3 = srem <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %widen_type_urem_0 = urem <vscale x 31 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %widen_type_urem_1 = urem <vscale x 15 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %widen_type_urem_2 = urem <vscale x 7 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %widen_type_urem_3 = urem <vscale x 3 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %widen_type_srem_0 = srem <vscale x 31 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %widen_type_srem_1 = srem <vscale x 15 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %widen_type_srem_2 = srem <vscale x 7 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %widen_type_srem_3 = srem <vscale x 3 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 entry:
   %legal_type_urem_0 = urem <vscale x 16 x i8> undef, undef
   %legal_type_urem_1 = urem <vscale x 8 x i16> undef, undef
   %legal_type_urem_2 = urem <vscale x 4 x i32> undef, undef
   %legal_type_urem_3 = urem <vscale x 2 x i64> undef, undef
   %legal_type_srem_0 = srem <vscale x 16 x i8> undef, undef

llvm/test/Analysis/CostModel/ARM/divrem.ll

 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = udiv i8 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = srem i8 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = urem i8 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; CHECK-V8R-LABEL: 'i8'
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = sdiv i8 undef, undef
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = udiv i8 undef, undef
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = srem i8 undef, undef
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = urem i8 undef, undef
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %3 = srem i8 undef, undef
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %4 = urem i8 undef, undef
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %5 = sdiv i8 undef, 2
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = udiv i8 undef, 2
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = srem i8 undef, 2
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = urem i8 undef, 2
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %7 = srem i8 undef, 2
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %8 = urem i8 undef, 2
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %1 = sdiv i8 undef, undef
   %2 = udiv i8 undef, undef
   %3 = srem i8 undef, undef
   %4 = urem i8 undef, undef
   %5 = sdiv i8 undef, 2
   %6 = udiv i8 undef, 2

 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = udiv i16 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = srem i16 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = urem i16 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; CHECK-V8R-LABEL: 'i16'
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = sdiv i16 undef, undef
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = udiv i16 undef, undef
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = srem i16 undef, undef
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = urem i16 undef, undef
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %3 = srem i16 undef, undef
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %4 = urem i16 undef, undef
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %5 = sdiv i16 undef, 2
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = udiv i16 undef, 2
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = srem i16 undef, 2
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = urem i16 undef, 2
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %7 = srem i16 undef, 2
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %8 = urem i16 undef, 2
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %1 = sdiv i16 undef, undef
   %2 = udiv i16 undef, undef
   %3 = srem i16 undef, undef
   %4 = urem i16 undef, undef
   %5 = sdiv i16 undef, 2
   %6 = udiv i16 undef, 2

 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = udiv i32 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = srem i32 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = urem i32 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; CHECK-V8R-LABEL: 'i32'
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = sdiv i32 undef, undef
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = udiv i32 undef, undef
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = srem i32 undef, undef
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = urem i32 undef, undef
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %3 = srem i32 undef, undef
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %4 = urem i32 undef, undef
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %5 = sdiv i32 undef, 2
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = udiv i32 undef, 2
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = srem i32 undef, 2
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = urem i32 undef, 2
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %7 = srem i32 undef, 2
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %8 = urem i32 undef, 2
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %1 = sdiv i32 undef, undef
   %2 = udiv i32 undef, undef
   %3 = srem i32 undef, undef
   %4 = urem i32 undef, undef
   %5 = sdiv i32 undef, 2
   %6 = udiv i32 undef, 2

 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %6 = udiv i64 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = srem i64 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = urem i64 undef, 2
 ; CHECK-V8M-BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; CHECK-V8R-LABEL: 'i64'
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %1 = sdiv i64 undef, undef
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %2 = udiv i64 undef, undef
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = srem i64 undef, undef
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = urem i64 undef, undef
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %3 = srem i64 undef, undef
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %4 = urem i64 undef, undef
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %5 = sdiv i64 undef, 2
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %6 = udiv i64 undef, 2
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = srem i64 undef, 2
-; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = urem i64 undef, 2
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %7 = srem i64 undef, 2
+; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %8 = urem i64 undef, 2
 ; CHECK-V8R-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %1 = sdiv i64 undef, undef
   %2 = udiv i64 undef, undef
   %3 = srem i64 undef, undef
   %4 = urem i64 undef, undef
   %5 = sdiv i64 undef, 2
   %6 = udiv i64 undef, 2

llvm/test/Analysis/CostModel/X86/rem.ll

 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+ssse3 | FileCheck %s --check-prefixes=CHECK,SSE,SSSE3
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+sse4.2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
 ;
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,GLM
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,BTVER2
 
 define i32 @srem() {
 ; CHECK-LABEL: 'srem'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = srem <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = srem <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = srem <16 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = srem <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = srem <16 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = srem <32 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i32 = srem <4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i32 = srem <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 960 for instruction: %V16i32 = srem <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i16 = srem <8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 960 for instruction: %V16i16 = srem <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1920 for instruction: %V32i16 = srem <32 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 960 for instruction: %V16i8 = srem <16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1920 for instruction: %V32i8 = srem <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3840 for instruction: %V64i8 = srem <64 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %I64 = srem i64 undef, undef
   %V2i64 = srem <2 x i64> undef, undef
   %V4i64 = srem <4 x i64> undef, undef
   %V8i64 = srem <8 x i64> undef, undef
 
   %I32 = srem i32 undef, undef
;
   %V32i8 = srem <32 x i8> undef, undef
   %V64i8 = srem <64 x i8> undef, undef
 
   ret i32 undef
 }
 
 define i32 @urem() {
 ; CHECK-LABEL: 'urem'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = urem <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = urem <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = urem <16 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = urem <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = urem <16 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = urem <32 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = urem <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = urem <32 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = urem <64 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i32 = urem <4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i32 = urem <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 960 for instruction: %V16i32 = urem <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i16 = urem <8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 960 for instruction: %V16i16 = urem <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1920 for instruction: %V32i16 = urem <32 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 960 for instruction: %V16i8 = urem <16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1920 for instruction: %V32i8 = urem <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3840 for instruction: %V64i8 = urem <64 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %I64 = urem i64 undef, undef
   %V2i64 = urem <2 x i64> undef, undef
   %V4i64 = urem <4 x i64> undef, undef
   %V8i64 = urem <8 x i64> undef, undef
 
   %I32 = urem i32 undef, undef
;
   %V32i8 = urem <32 x i8> undef, undef
   %V64i8 = urem <64 x i8> undef, undef
 
   ret i32 undef
 }
 
 define i32 @srem_const() {
 ; SSE2-LABEL: 'srem_const'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'srem_const'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'srem_const'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'srem_const'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'srem_const'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'srem_const'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'srem_const'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SLM-LABEL: 'srem_const'
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; GLM-LABEL: 'srem_const'
-; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; GLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; GLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; GLM-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; GLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; GLM-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; GLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; GLM-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'srem_const'
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %I64 = srem i64 undef, 7
   %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
   %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
;
   %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
   %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 
   ret i32 undef
 }
 
 define i32 @urem_const() {
 ; SSE-LABEL: 'urem_const'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; SSE-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'urem_const'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'urem_const'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'urem_const'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'urem_const'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'urem_const'
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %I64 = urem i64 undef, 7
   %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
   %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
;
   %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
   %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 
   ret i32 undef
 }
 
 define i32 @srem_uniformconst() {
 ; SSE-LABEL: 'srem_uniformconst'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; SSE-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = srem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = srem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16i32 = srem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i8 = srem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V64i8 = srem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'srem_uniformconst'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = srem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8i32 = srem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16i32 = srem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = srem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = srem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32i8 = srem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64i8 = srem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'srem_uniformconst'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = srem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i32 = srem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i32 = srem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = srem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = srem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = srem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'srem_uniformconst'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = srem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i32 = srem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = srem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = srem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = srem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = srem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'srem_uniformconst'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = srem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i32 = srem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i32 = srem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = srem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = srem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64i8 = srem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'srem_uniformconst'
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = srem i64 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = srem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = srem i32 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = srem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8i32 = srem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V16i32 = srem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = srem i16 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = srem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = srem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = srem i8 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = srem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32i8 = srem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64i8 = srem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %I64 = srem i64 undef, 7
   %V2i64 = srem <2 x i64> undef, <i64 7, i64 7>
   %V4i64 = srem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
;
   %V32i8 = srem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
   %V64i8 = srem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 
   ret i32 undef
 }
 
 define i32 @urem_uniformconst() {
 ; SSE-LABEL: 'urem_uniformconst'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; SSE-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = urem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8i32 = urem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16i32 = urem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = urem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = urem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i8 = urem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V64i8 = urem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'urem_uniformconst'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = urem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = urem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16i32 = urem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = urem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = urem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32i8 = urem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64i8 = urem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'urem_uniformconst'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = urem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i32 = urem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i32 = urem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = urem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = urem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = urem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'urem_uniformconst'
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = urem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i32 = urem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i32 = urem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = urem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = urem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64i8 = urem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'urem_uniformconst'
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = urem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i32 = urem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i32 = urem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = urem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i8 = urem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64i8 = urem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'urem_uniformconst'
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = urem i64 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = urem <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = urem i32 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4i32 = urem <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = urem <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16i32 = urem <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = urem i16 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = urem <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = urem <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = urem i8 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i8 = urem <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V32i8 = urem <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64i8 = urem <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %I64 = urem i64 undef, 7
   %V2i64 = urem <2 x i64> undef, <i64 7, i64 7>
   %V4i64 = urem <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>