[CostModel][X86] Add SLM/GLM/BtVer2 compare + division/remainder cost tests

llvm-svn: 330435

3 files changed, 194 insertions, 0 deletions

diff --git a/llvm/test/Analysis/CostModel/X86/cmp.ll b/llvm/test/Analysis/CostModel/X86/cmp.ll
index 0f033e0814d..044de0c4f13 100644
--- a/llvm/test/Analysis/CostModel/X86/cmp.ll
+++ b/llvm/test/Analysis/CostModel/X86/cmp.ll
@@ -8,6 +8,10 @@
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+;
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=btver2 | FileCheck %s --check-prefixes=BTVER2
 
 define i32 @cmp_float(i32 %arg) {
 ; SSE2-LABEL: 'cmp_float'
@@ -87,6 +91,17 @@ define i32 @cmp_float(i32 %arg) {
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F64 = fcmp olt <16 x double> undef, undef
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'cmp_float'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F32 = fcmp olt <2 x float> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fcmp olt <4 x float> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fcmp olt <8 x float> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fcmp olt <16 x float> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fcmp olt <2 x double> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fcmp olt <4 x double> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fcmp olt <8 x double> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F64 = fcmp olt <16 x double> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %V2F32 = fcmp olt <2 x float> undef, undef
   %V4F32 = fcmp olt <4 x float> undef, undef
   %V8F32 = fcmp olt <8 x float> undef, undef
@@ -272,6 +287,25 @@ define i32 @cmp_int(i32 %arg) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I64 = icmp eq <16 x i64> undef, undef
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'cmp_int'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = icmp eq <16 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = icmp eq <32 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = icmp eq <64 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V128I8 = icmp eq <128 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = icmp eq <8 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = icmp eq <16 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = icmp eq <32 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I16 = icmp eq <64 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = icmp eq <4 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I32 = icmp eq <8 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I32 = icmp eq <16 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I32 = icmp eq <32 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = icmp eq <2 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4I64 = icmp eq <4 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I64 = icmp eq <8 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16I64 = icmp eq <16 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %V16I8 = icmp eq <16 x i8> undef, undef
   %V32I8 = icmp eq <32 x i8> undef, undef
   %V64I8 = icmp eq <64 x i8> undef, undef
diff --git a/llvm/test/Analysis/CostModel/X86/div.ll b/llvm/test/Analysis/CostModel/X86/div.ll
index 483da9a765f..f5d9822abfc 100644
--- a/llvm/test/Analysis/CostModel/X86/div.ll
+++ b/llvm/test/Analysis/CostModel/X86/div.ll
@@ -6,6 +6,10 @@
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+;
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=btver2 | FileCheck %s --check-prefixes=BTVER2
 
 define i32 @sdiv() {
 ; CHECK-LABEL: 'sdiv'
@@ -27,6 +31,25 @@ define i32 @sdiv() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'sdiv'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = sdiv <4 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = sdiv <8 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = sdiv <16 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = sdiv <8 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = sdiv <16 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = sdiv <32 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %I64 = sdiv i64 undef, undef
   %V2i64 = sdiv <2 x i64> undef, undef
   %V4i64 = sdiv <4 x i64> undef, undef
@@ -70,6 +93,25 @@ define i32 @udiv() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'udiv'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = udiv <4 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = udiv <8 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = udiv <16 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = udiv <8 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = udiv <16 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = udiv <32 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %I64 = udiv i64 undef, undef
   %V2i64 = udiv <2 x i64> undef, undef
   %V4i64 = udiv <4 x i64> undef, undef
@@ -227,6 +269,25 @@ define i32 @sdiv_uniformconst() {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'sdiv_uniformconst'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %I64 = sdiv i64 undef, 7
   %V2i64 = sdiv <2 x i64> undef, <i64 7, i64 7>
   %V4i64 = sdiv <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
@@ -346,6 +407,25 @@ define i32 @udiv_uniformconst() {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'udiv_uniformconst'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 7, i32 7, i32 7, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %I64 = udiv i64 undef, 7
   %V2i64 = udiv <2 x i64> undef, <i64 7, i64 7>
   %V4i64 = udiv <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
@@ -503,6 +583,25 @@ define i32 @sdiv_uniformconstpow2() {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'sdiv_uniformconstpow2'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 16
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 16, i64 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 16
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16i32 = sdiv <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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 16
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i16 = sdiv <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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32i16 = sdiv <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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 16
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %I64 = sdiv i64 undef, 16
   %V2i64 = sdiv <2 x i64> undef, <i64 16, i64 16>
   %V4i64 = sdiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
@@ -622,6 +721,25 @@ define i32 @udiv_uniformconstpow2() {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1280 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>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'udiv_uniformconstpow2'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 16
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 16
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 64 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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 16
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 16
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1280 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>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %I64 = udiv i64 undef, 16
   %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
   %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
diff --git a/llvm/test/Analysis/CostModel/X86/rem.ll b/llvm/test/Analysis/CostModel/X86/rem.ll
index c4db00e50ee..b68b55f1e76 100644
--- a/llvm/test/Analysis/CostModel/X86/rem.ll
+++ b/llvm/test/Analysis/CostModel/X86/rem.ll
@@ -6,6 +6,10 @@
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
 ; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+;
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -cost-model -analyze -mcpu=btver2 | FileCheck %s --check-prefixes=BTVER2
 
 define i32 @srem() {
 ; CHECK-LABEL: 'srem'
@@ -27,6 +31,25 @@ define i32 @srem() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 192 for instruction: %V64i8 = srem <64 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'srem'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = srem <2 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = srem <4 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = srem <8 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i32 = srem <4 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i32 = srem <8 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V16i32 = srem <16 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i16 = srem <8 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V16i16 = srem <16 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %V32i16 = srem <32 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V16i8 = srem <16 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %V32i8 = srem <32 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 192 for instruction: %V64i8 = srem <64 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %I64 = srem i64 undef, undef
   %V2i64 = srem <2 x i64> undef, undef
   %V4i64 = srem <4 x i64> undef, undef
@@ -70,6 +93,25 @@ define i32 @urem() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 192 for instruction: %V64i8 = urem <64 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
+; BTVER2-LABEL: 'urem'
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2i64 = urem <2 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i64 = urem <4 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = urem <8 x i64> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4i32 = urem <4 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i32 = urem <8 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V16i32 = urem <16 x i32> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i16 = urem <8 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V16i16 = urem <16 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %V32i16 = urem <32 x i16> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V16i8 = urem <16 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %V32i8 = urem <32 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 192 for instruction: %V64i8 = urem <64 x i8> undef, undef
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
   %I64 = urem i64 undef, undef
   %V2i64 = urem <2 x i64> undef, undef
   %V4i64 = urem <4 x i64> undef, undef