[llvm] 55b1a14 - [CostModel][X86] getArithmeticInstrCost - use MUL/DIV/REM expansions for all cost kinds

Thu Aug 18 06:23:27 PDT 2022

Author: Simon Pilgrim
Date: 2022-08-18T14:18:54+01:00
New Revision: 55b1a147f263630535ee271dbba0c1b6a863456a

URL: https://github.com/llvm/llvm-project/commit/55b1a147f263630535ee271dbba0c1b6a863456a
DIFF: https://github.com/llvm/llvm-project/commit/55b1a147f263630535ee271dbba0c1b6a863456a.diff

LOG: [CostModel][X86] getArithmeticInstrCost - use MUL/DIV/REM expansions for all cost kinds

The costs tables still assume throughput, but the general expansion patterns should be good for any cost kind

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86TargetTransformInfo.cpp
    llvm/test/Analysis/CostModel/X86/arith-costkinds.ll
    llvm/test/Analysis/CostModel/X86/div-costkinds.ll
    llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
    llvm/test/Analysis/CostModel/X86/rem-costkinds.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index c5bb48edd67fe..b0bf71fe8e454 100644

--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -180,12 +180,6 @@ InstructionCost X86TTIImpl::getArithmeticInstrCost(
     TTI::OperandValueProperties Opd1PropInfo,
     TTI::OperandValueProperties Opd2PropInfo, ArrayRef<const Value *> Args,
     const Instruction *CxtI) {
-  // TODO: Handle more cost kinds.
-  if (CostKind != TTI::TCK_RecipThroughput)
-    return BaseT::getArithmeticInstrCost(Opcode, Ty, CostKind, Op1Info,
-                                         Op2Info, Opd1PropInfo,
-                                         Opd2PropInfo, Args, CxtI);
-
   // vXi8 multiplications are always promoted to vXi16.
   if (Opcode == Instruction::Mul && Ty->isVectorTy() &&
       Ty->getScalarSizeInBits() == 8) {
@@ -291,6 +285,12 @@ InstructionCost X86TTIImpl::getArithmeticInstrCost(
                                   TargetTransformInfo::OP_None);
   }
 
+  // TODO: Handle more cost kinds.
+  if (CostKind != TTI::TCK_RecipThroughput)
+    return BaseT::getArithmeticInstrCost(Opcode, Ty, CostKind, Op1Info, Op2Info,
+                                         Opd1PropInfo, Opd2PropInfo, Args,
+                                         CxtI);
+
   static const CostTblEntry GLMCostTable[] = {
     { ISD::FDIV,  MVT::f32,   18 }, // divss
     { ISD::FDIV,  MVT::v4f32, 35 }, // divps

diff  --git a/llvm/test/Analysis/CostModel/X86/arith-costkinds.ll b/llvm/test/Analysis/CostModel/X86/arith-costkinds.ll
index bfc4e5a1cbe3b..85ea52887ea67 100644
--- a/llvm/test/Analysis/CostModel/X86/arith-costkinds.ll
+++ b/llvm/test/Analysis/CostModel/X86/arith-costkinds.ll
@@ -529,12 +529,12 @@ define i32 @mul(i32 %arg) {
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = mul <16 x i16> undef, undef
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = mul <32 x i16> undef, undef
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, undef
-; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I8 = mul <2 x i8> undef, undef
-; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I8 = mul <4 x i8> undef, undef
-; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I8 = mul <8 x i8> undef, undef
-; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = mul <16 x i8> undef, undef
-; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = mul <32 x i8> undef, undef
-; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64I8 = mul <64 x i8> undef, undef
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I8 = mul <2 x i8> undef, undef
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4I8 = mul <4 x i8> undef, undef
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I8 = mul <8 x i8> undef, undef
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = mul <16 x i8> undef, undef
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = mul <32 x i8> undef, undef
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = mul <64 x i8> undef, undef
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE-LABEL: 'mul'
@@ -551,12 +551,12 @@ define i32 @mul(i32 %arg) {
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = mul <16 x i16> undef, undef
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = mul <32 x i16> undef, undef
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, undef
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I8 = mul <2 x i8> undef, undef
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I8 = mul <4 x i8> undef, undef
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I8 = mul <8 x i8> undef, undef
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = mul <16 x i8> undef, undef
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = mul <32 x i8> undef, undef
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64I8 = mul <64 x i8> undef, undef
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I8 = mul <2 x i8> undef, undef
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4I8 = mul <4 x i8> undef, undef
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I8 = mul <8 x i8> undef, undef
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = mul <16 x i8> undef, undef
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = mul <32 x i8> undef, undef
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = mul <64 x i8> undef, undef
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE_LATE-LABEL: 'mul'
@@ -573,12 +573,12 @@ define i32 @mul(i32 %arg) {
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = mul <16 x i16> undef, undef
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = mul <32 x i16> undef, undef
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = mul i8 undef, undef
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I8 = mul <2 x i8> undef, undef
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I8 = mul <4 x i8> undef, undef
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I8 = mul <8 x i8> undef, undef
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = mul <16 x i8> undef, undef
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = mul <32 x i8> undef, undef
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64I8 = mul <64 x i8> undef, undef
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I8 = mul <2 x i8> undef, undef
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4I8 = mul <4 x i8> undef, undef
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I8 = mul <8 x i8> undef, undef
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = mul <16 x i8> undef, undef
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = mul <32 x i8> undef, undef
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = mul <64 x i8> undef, undef
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = mul i64 undef, undef

diff  --git a/llvm/test/Analysis/CostModel/X86/div-costkinds.ll b/llvm/test/Analysis/CostModel/X86/div-costkinds.ll
index 1c7fb5dabf237..129b72ec3753e 100644
--- a/llvm/test/Analysis/CostModel/X86/div-costkinds.ll
+++ b/llvm/test/Analysis/CostModel/X86/div-costkinds.ll
@@ -588,60 +588,60 @@ define i32 @sdiv_constpow2() {
 
 define i32 @udiv_constpow2() {
 ; LATE-LABEL: 'udiv_constpow2'
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = udiv i64 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 8, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = udiv i32 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = udiv i16 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = udiv i8 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 8, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE-LABEL: 'udiv_constpow2'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = udiv i64 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 8, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = udiv i32 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = udiv i16 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = udiv i8 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 8, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE_LATE-LABEL: 'udiv_constpow2'
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = udiv i64 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 8, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = udiv i32 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = udiv i16 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = udiv i8 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 8, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = udiv i64 undef, 16
@@ -750,60 +750,60 @@ define i32 @sdiv_uniformconstpow2() {
 
 define i32 @udiv_uniformconstpow2() {
 ; LATE-LABEL: 'udiv_uniformconstpow2'
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = udiv i64 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = udiv i32 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = udiv i16 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = udiv i8 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE-LABEL: 'udiv_uniformconstpow2'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = udiv i64 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = udiv i32 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = udiv i16 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = udiv i8 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE_LATE-LABEL: 'udiv_uniformconstpow2'
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = udiv i64 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = udiv i32 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = udiv i16 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = udiv i8 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = udiv i64 undef, 16

diff  --git a/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll b/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
index 8d68623e2527c..8f96c850809c0 100644
--- a/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
+++ b/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
@@ -290,17 +290,17 @@ define void @fshl(i32 %a, i32 %b, i32 %c, <16 x i32> %va, <16 x i32> %vb, <16 x
 ;
 ; LATE-LABEL: 'fshl'
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
-; LATE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SIZE-LABEL: 'fshl'
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SIZE_LATE-LABEL: 'fshl'
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)

diff  --git a/llvm/test/Analysis/CostModel/X86/rem-costkinds.ll b/llvm/test/Analysis/CostModel/X86/rem-costkinds.ll
index 876b9bcfe9200..129cfb28f3b30 100644
--- a/llvm/test/Analysis/CostModel/X86/rem-costkinds.ll
+++ b/llvm/test/Analysis/CostModel/X86/rem-costkinds.ll
@@ -597,60 +597,60 @@ define i32 @urem_uniformconst() {
 
 define i32 @srem_constpow2() {
 ; LATE-LABEL: 'srem_constpow2'
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = srem i64 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = srem <2 x i64> undef, <i64 8, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = srem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = srem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = srem i32 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = srem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = srem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = srem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = srem i16 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = srem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = srem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = srem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = srem i8 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = srem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = srem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = srem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I64 = srem i64 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = srem <2 x i64> undef, <i64 8, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = srem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = srem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = srem i32 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = srem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i32 = srem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = srem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I16 = srem i16 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = srem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = srem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = srem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I8 = srem i8 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = srem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i8 = srem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64i8 = srem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE-LABEL: 'srem_constpow2'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = srem i64 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = srem <2 x i64> undef, <i64 8, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = srem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = srem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = srem i32 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = srem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = srem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = srem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = srem i16 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = srem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = srem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = srem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = srem i8 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = srem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = srem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = srem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I64 = srem i64 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = srem <2 x i64> undef, <i64 8, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = srem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = srem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = srem i32 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = srem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i32 = srem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = srem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I16 = srem i16 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = srem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = srem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = srem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I8 = srem i8 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = srem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i8 = srem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64i8 = srem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE_LATE-LABEL: 'srem_constpow2'
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = srem i64 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = srem <2 x i64> undef, <i64 8, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = srem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = srem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = srem i32 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = srem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = srem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = srem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = srem i16 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = srem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = srem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = srem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = srem i8 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = srem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = srem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = srem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I64 = srem i64 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = srem <2 x i64> undef, <i64 8, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = srem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = srem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = srem i32 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = srem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i32 = srem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = srem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I16 = srem i16 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = srem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = srem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = srem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I8 = srem i8 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = srem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i8 = srem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64i8 = srem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; GLM-LABEL: 'srem_constpow2'
@@ -696,60 +696,60 @@ define i32 @srem_constpow2() {
 
 define i32 @urem_constpow2() {
 ; LATE-LABEL: 'urem_constpow2'
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = urem i64 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = urem <2 x i64> undef, <i64 8, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = urem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = urem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = urem i32 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = urem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = urem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = urem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = urem i16 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = urem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = urem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = urem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = urem i8 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = urem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = urem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = urem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = urem <2 x i64> undef, <i64 8, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = urem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = urem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = urem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = urem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = urem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = urem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = urem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = urem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = urem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = urem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = urem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE-LABEL: 'urem_constpow2'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = urem i64 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = urem <2 x i64> undef, <i64 8, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = urem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = urem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = urem i32 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = urem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = urem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = urem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = urem i16 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = urem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = urem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = urem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = urem i8 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = urem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = urem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = urem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = urem <2 x i64> undef, <i64 8, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = urem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = urem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = urem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = urem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = urem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = urem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = urem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = urem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = urem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = urem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = urem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE_LATE-LABEL: 'urem_constpow2'
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = urem i64 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = urem <2 x i64> undef, <i64 8, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = urem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = urem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = urem i32 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = urem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = urem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = urem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = urem i16 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = urem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = urem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = urem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = urem i8 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = urem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = urem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = urem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = urem <2 x i64> undef, <i64 8, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = urem <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = urem <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = urem <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = urem <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = urem <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = urem <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = urem <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = urem <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = urem <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = urem <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = urem <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX-LABEL: 'urem_constpow2'
@@ -813,60 +813,60 @@ define i32 @urem_constpow2() {
 
 define i32 @srem_uniformconstpow2() {
 ; LATE-LABEL: 'srem_uniformconstpow2'
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = srem i64 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = srem <2 x i64> undef, <i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = srem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = srem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = srem i32 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = srem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = srem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = srem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = srem i16 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = srem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = srem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = srem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = srem i8 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = srem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = srem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = srem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I64 = srem i64 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = srem <2 x i64> undef, <i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = srem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = srem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = srem i32 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = srem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i32 = srem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = srem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I16 = srem i16 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = srem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = srem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = srem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I8 = srem i8 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = srem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i8 = srem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64i8 = srem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE-LABEL: 'srem_uniformconstpow2'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = srem i64 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = srem <2 x i64> undef, <i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = srem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = srem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = srem i32 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = srem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = srem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = srem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = srem i16 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = srem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = srem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = srem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = srem i8 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = srem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = srem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = srem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I64 = srem i64 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = srem <2 x i64> undef, <i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = srem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = srem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = srem i32 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = srem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i32 = srem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = srem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I16 = srem i16 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = srem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = srem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = srem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I8 = srem i8 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = srem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i8 = srem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64i8 = srem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE_LATE-LABEL: 'srem_uniformconstpow2'
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = srem i64 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = srem <2 x i64> undef, <i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = srem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = srem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = srem i32 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = srem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = srem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = srem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = srem i16 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = srem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = srem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = srem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = srem i8 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = srem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = srem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = srem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I64 = srem i64 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = srem <2 x i64> undef, <i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = srem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i64 = srem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = srem i32 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i32 = srem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i32 = srem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = srem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I16 = srem i16 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = srem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = srem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = srem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I8 = srem i8 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i8 = srem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i8 = srem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64i8 = srem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; GLM-LABEL: 'srem_uniformconstpow2'
@@ -912,60 +912,60 @@ define i32 @srem_uniformconstpow2() {
 
 define i32 @urem_uniformconstpow2() {
 ; LATE-LABEL: 'urem_uniformconstpow2'
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = urem i64 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = urem <2 x i64> undef, <i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = urem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = urem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = urem i32 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = urem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = urem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = urem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = urem i16 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = urem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = urem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = urem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = urem i8 undef, 16
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = urem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = urem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = urem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = urem <2 x i64> undef, <i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = urem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = urem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = urem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = urem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = urem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = urem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = urem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = urem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 16
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = urem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = urem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = urem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE-LABEL: 'urem_uniformconstpow2'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = urem i64 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = urem <2 x i64> undef, <i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = urem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = urem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = urem i32 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = urem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = urem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = urem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = urem i16 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = urem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = urem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = urem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = urem i8 undef, 16
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = urem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = urem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = urem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = urem <2 x i64> undef, <i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = urem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = urem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = urem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = urem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = urem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = urem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = urem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = urem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 16
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = urem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = urem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = urem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; SIZE_LATE-LABEL: 'urem_uniformconstpow2'
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = urem i64 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2i64 = urem <2 x i64> undef, <i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = urem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = urem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = urem i32 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i32 = urem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = urem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = urem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = urem i16 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i16 = urem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = urem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i16 = urem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = urem i8 undef, 16
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i8 = urem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32i8 = urem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64i8 = urem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = urem <2 x i64> undef, <i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = urem <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = urem <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = urem <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = urem <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = urem <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = urem <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = urem <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = urem <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 16
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i8 = urem <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i8 = urem <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i8 = urem <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
 ; AVX-LABEL: 'urem_uniformconstpow2'


        
