[SLP] reduce duplicated check lines in tests; NFC

1 files changed, 163 insertions, 313 deletions

diff --git a/llvm/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll b/llvm/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll
index 8b0ae669ae7..40ea60bb5f0 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll
@@ -1,8 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt < %s -mtriple=x86_64-unknown-linux -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SSE
-; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=corei7-avx -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX
-; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=core-avx2 -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=skx -slp-vectorizer -S -slp-threshold=-100 | FileCheck %s --check-prefixes=CHECK,SKX
+; RUN: opt < %s -mtriple=x86_64-unknown-linux -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,DEFAULT,SSE
+; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=corei7-avx -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,DEFAULT,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=core-avx2 -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,DEFAULT,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=skx -slp-vectorizer -S -slp-threshold=-100 | FileCheck %s --check-prefixes=CHECK,THRESH
 
 @arr = local_unnamed_addr global [32 x i32] zeroinitializer, align 16
 @arr1 = local_unnamed_addr global [32 x float] zeroinitializer, align 16
@@ -460,111 +460,61 @@ define float @maxf32(float) {
 }
 
 define i32 @maxi8_mutiple_uses(i32) {
-; SSE-LABEL: @maxi8_mutiple_uses(
-; SSE-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
-; SSE-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
-; SSE-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
-; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
-; SSE-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; SSE-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
-; SSE-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; SSE-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; SSE-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; SSE-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; SSE-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; SSE-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; SSE-NEXT:    [[TMP8:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; SSE-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP8]], [[TMP7]]
-; SSE-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP8]], i32 [[TMP7]]
-; SSE-NEXT:    [[TMP11:%.*]] = icmp sgt i32 [[TMP10]], [[TMP5]]
-; SSE-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP11]], i32 [[TMP10]], i32 [[TMP5]]
-; SSE-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[OP_EXTRA]], [[TMP12]]
-; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[OP_EXTRA]], i32 [[TMP12]]
-; SSE-NEXT:    [[TMP15:%.*]] = select i1 [[TMP4]], i32 3, i32 4
-; SSE-NEXT:    store i32 [[TMP15]], i32* @var, align 8
-; SSE-NEXT:    ret i32 [[TMP14]]
+; DEFAULT-LABEL: @maxi8_mutiple_uses(
+; DEFAULT-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
+; DEFAULT-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
+; DEFAULT-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
+; DEFAULT-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
+; DEFAULT-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
+; DEFAULT-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
+; DEFAULT-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; DEFAULT-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
+; DEFAULT-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
+; DEFAULT-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; DEFAULT-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
+; DEFAULT-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
+; DEFAULT-NEXT:    [[TMP8:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
+; DEFAULT-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP8]], [[TMP7]]
+; DEFAULT-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP8]], i32 [[TMP7]]
+; DEFAULT-NEXT:    [[TMP11:%.*]] = icmp sgt i32 [[TMP10]], [[TMP5]]
+; DEFAULT-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP11]], i32 [[TMP10]], i32 [[TMP5]]
+; DEFAULT-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
+; DEFAULT-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[OP_EXTRA]], [[TMP12]]
+; DEFAULT-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[OP_EXTRA]], i32 [[TMP12]]
+; DEFAULT-NEXT:    [[TMP15:%.*]] = select i1 [[TMP4]], i32 3, i32 4
+; DEFAULT-NEXT:    store i32 [[TMP15]], i32* @var, align 8
+; DEFAULT-NEXT:    ret i32 [[TMP14]]
 ;
-; AVX-LABEL: @maxi8_mutiple_uses(
-; AVX-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
-; AVX-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
-; AVX-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
-; AVX-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
-; AVX-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; AVX-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
-; AVX-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; AVX-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; AVX-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; AVX-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; AVX-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; AVX-NEXT:    [[TMP8:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; AVX-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP8]], [[TMP7]]
-; AVX-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP8]], i32 [[TMP7]]
-; AVX-NEXT:    [[TMP11:%.*]] = icmp sgt i32 [[TMP10]], [[TMP5]]
-; AVX-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP11]], i32 [[TMP10]], i32 [[TMP5]]
-; AVX-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; AVX-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[OP_EXTRA]], [[TMP12]]
-; AVX-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[OP_EXTRA]], i32 [[TMP12]]
-; AVX-NEXT:    [[TMP15:%.*]] = select i1 [[TMP4]], i32 3, i32 4
-; AVX-NEXT:    store i32 [[TMP15]], i32* @var, align 8
-; AVX-NEXT:    ret i32 [[TMP14]]
-;
-; AVX2-LABEL: @maxi8_mutiple_uses(
-; AVX2-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
-; AVX2-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
-; AVX2-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
-; AVX2-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
-; AVX2-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; AVX2-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
-; AVX2-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; AVX2-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; AVX2-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; AVX2-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; AVX2-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; AVX2-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; AVX2-NEXT:    [[TMP8:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; AVX2-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP8]], [[TMP7]]
-; AVX2-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP8]], i32 [[TMP7]]
-; AVX2-NEXT:    [[TMP11:%.*]] = icmp sgt i32 [[TMP10]], [[TMP5]]
-; AVX2-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP11]], i32 [[TMP10]], i32 [[TMP5]]
-; AVX2-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; AVX2-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[OP_EXTRA]], [[TMP12]]
-; AVX2-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[OP_EXTRA]], i32 [[TMP12]]
-; AVX2-NEXT:    [[TMP15:%.*]] = select i1 [[TMP4]], i32 3, i32 4
-; AVX2-NEXT:    store i32 [[TMP15]], i32* @var, align 8
-; AVX2-NEXT:    ret i32 [[TMP14]]
-;
-; SKX-LABEL: @maxi8_mutiple_uses(
-; SKX-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([32 x i32]* @arr to <2 x i32>*), align 16
-; SKX-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 0
-; SKX-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
-; SKX-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; SKX-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
-; SKX-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; SKX-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP5]], [[RDX_SHUF]]
-; SKX-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP5]], <4 x i32> [[RDX_SHUF]]
-; SKX-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; SKX-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; SKX-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; SKX-NEXT:    [[TMP7:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; SKX-NEXT:    [[TMP8:%.*]] = insertelement <2 x i32> undef, i32 [[TMP7]], i32 0
-; SKX-NEXT:    [[TMP9:%.*]] = insertelement <2 x i32> [[TMP8]], i32 [[TMP3]], i32 1
-; SKX-NEXT:    [[TMP10:%.*]] = insertelement <2 x i32> undef, i32 [[TMP6]], i32 0
-; SKX-NEXT:    [[TMP11:%.*]] = insertelement <2 x i32> [[TMP10]], i32 [[TMP4]], i32 1
-; SKX-NEXT:    [[TMP12:%.*]] = icmp sgt <2 x i32> [[TMP9]], [[TMP11]]
-; SKX-NEXT:    [[TMP13:%.*]] = select <2 x i1> [[TMP12]], <2 x i32> [[TMP9]], <2 x i32> [[TMP11]]
-; SKX-NEXT:    [[TMP14:%.*]] = extractelement <2 x i32> [[TMP13]], i32 1
-; SKX-NEXT:    [[TMP15:%.*]] = extractelement <2 x i32> [[TMP13]], i32 0
-; SKX-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP15]], [[TMP14]]
-; SKX-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP16]], i32 [[TMP15]], i32 [[TMP14]]
-; SKX-NEXT:    [[TMP17:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; SKX-NEXT:    [[TMP18:%.*]] = icmp sgt i32 [[OP_EXTRA]], [[TMP17]]
-; SKX-NEXT:    [[TMP19:%.*]] = select i1 [[TMP18]], i32 [[OP_EXTRA]], i32 [[TMP17]]
-; SKX-NEXT:    [[TMP20:%.*]] = extractelement <2 x i1> [[TMP12]], i32 1
-; SKX-NEXT:    [[TMP21:%.*]] = select i1 [[TMP20]], i32 3, i32 4
-; SKX-NEXT:    store i32 [[TMP21]], i32* @var, align 8
-; SKX-NEXT:    ret i32 [[TMP19]]
+; THRESH-LABEL: @maxi8_mutiple_uses(
+; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([32 x i32]* @arr to <2 x i32>*), align 16
+; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 0
+; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
+; THRESH-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
+; THRESH-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
+; THRESH-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; THRESH-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP5]], [[RDX_SHUF]]
+; THRESH-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP5]], <4 x i32> [[RDX_SHUF]]
+; THRESH-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; THRESH-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
+; THRESH-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
+; THRESH-NEXT:    [[TMP7:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
+; THRESH-NEXT:    [[TMP8:%.*]] = insertelement <2 x i32> undef, i32 [[TMP7]], i32 0
+; THRESH-NEXT:    [[TMP9:%.*]] = insertelement <2 x i32> [[TMP8]], i32 [[TMP3]], i32 1
+; THRESH-NEXT:    [[TMP10:%.*]] = insertelement <2 x i32> undef, i32 [[TMP6]], i32 0
+; THRESH-NEXT:    [[TMP11:%.*]] = insertelement <2 x i32> [[TMP10]], i32 [[TMP4]], i32 1
+; THRESH-NEXT:    [[TMP12:%.*]] = icmp sgt <2 x i32> [[TMP9]], [[TMP11]]
+; THRESH-NEXT:    [[TMP13:%.*]] = select <2 x i1> [[TMP12]], <2 x i32> [[TMP9]], <2 x i32> [[TMP11]]
+; THRESH-NEXT:    [[TMP14:%.*]] = extractelement <2 x i32> [[TMP13]], i32 1
+; THRESH-NEXT:    [[TMP15:%.*]] = extractelement <2 x i32> [[TMP13]], i32 0
+; THRESH-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP15]], [[TMP14]]
+; THRESH-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP16]], i32 [[TMP15]], i32 [[TMP14]]
+; THRESH-NEXT:    [[TMP17:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
+; THRESH-NEXT:    [[TMP18:%.*]] = icmp sgt i32 [[OP_EXTRA]], [[TMP17]]
+; THRESH-NEXT:    [[TMP19:%.*]] = select i1 [[TMP18]], i32 [[OP_EXTRA]], i32 [[TMP17]]
+; THRESH-NEXT:    [[TMP20:%.*]] = extractelement <2 x i1> [[TMP12]], i32 1
+; THRESH-NEXT:    [[TMP21:%.*]] = select i1 [[TMP20]], i32 3, i32 4
+; THRESH-NEXT:    store i32 [[TMP21]], i32* @var, align 8
+; THRESH-NEXT:    ret i32 [[TMP19]]
 ;
   %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
   %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
@@ -594,113 +544,63 @@ define i32 @maxi8_mutiple_uses(i32) {
 }
 
 define i32 @maxi8_wrong_parent(i32) {
-; SSE-LABEL: @maxi8_wrong_parent(
-; SSE-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
-; SSE-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
-; SSE-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
-; SSE-NEXT:    br label [[PP:%.*]]
-; SSE:       pp:
-; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
-; SSE-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; SSE-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
-; SSE-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; SSE-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; SSE-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; SSE-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; SSE-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; SSE-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; SSE-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; SSE-NEXT:    [[TMP9:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; SSE-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
-; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
-; SSE-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
-; SSE-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 [[TMP8]]
-; SSE-NEXT:    [[TMP14:%.*]] = icmp sgt i32 [[TMP13]], [[TMP5]]
-; SSE-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP14]], i32 [[TMP13]], i32 [[TMP5]]
-; SSE-NEXT:    ret i32 [[OP_EXTRA]]
-;
-; AVX-LABEL: @maxi8_wrong_parent(
-; AVX-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
-; AVX-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
-; AVX-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
-; AVX-NEXT:    br label [[PP:%.*]]
-; AVX:       pp:
-; AVX-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
-; AVX-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; AVX-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
-; AVX-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; AVX-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; AVX-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; AVX-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; AVX-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; AVX-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; AVX-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; AVX-NEXT:    [[TMP9:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; AVX-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
-; AVX-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
-; AVX-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
-; AVX-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 [[TMP8]]
-; AVX-NEXT:    [[TMP14:%.*]] = icmp sgt i32 [[TMP13]], [[TMP5]]
-; AVX-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP14]], i32 [[TMP13]], i32 [[TMP5]]
-; AVX-NEXT:    ret i32 [[OP_EXTRA]]
+; DEFAULT-LABEL: @maxi8_wrong_parent(
+; DEFAULT-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
+; DEFAULT-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
+; DEFAULT-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
+; DEFAULT-NEXT:    br label [[PP:%.*]]
+; DEFAULT:       pp:
+; DEFAULT-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
+; DEFAULT-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
+; DEFAULT-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
+; DEFAULT-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
+; DEFAULT-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; DEFAULT-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
+; DEFAULT-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
+; DEFAULT-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; DEFAULT-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
+; DEFAULT-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
+; DEFAULT-NEXT:    [[TMP9:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
+; DEFAULT-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
+; DEFAULT-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
+; DEFAULT-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
+; DEFAULT-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 [[TMP8]]
+; DEFAULT-NEXT:    [[TMP14:%.*]] = icmp sgt i32 [[TMP13]], [[TMP5]]
+; DEFAULT-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP14]], i32 [[TMP13]], i32 [[TMP5]]
+; DEFAULT-NEXT:    ret i32 [[OP_EXTRA]]
 ;
-; AVX2-LABEL: @maxi8_wrong_parent(
-; AVX2-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
-; AVX2-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
-; AVX2-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
-; AVX2-NEXT:    br label [[PP:%.*]]
-; AVX2:       pp:
-; AVX2-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
-; AVX2-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; AVX2-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
-; AVX2-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; AVX2-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; AVX2-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; AVX2-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; AVX2-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; AVX2-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; AVX2-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; AVX2-NEXT:    [[TMP9:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; AVX2-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
-; AVX2-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
-; AVX2-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
-; AVX2-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 [[TMP8]]
-; AVX2-NEXT:    [[TMP14:%.*]] = icmp sgt i32 [[TMP13]], [[TMP5]]
-; AVX2-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP14]], i32 [[TMP13]], i32 [[TMP5]]
-; AVX2-NEXT:    ret i32 [[OP_EXTRA]]
-;
-; SKX-LABEL: @maxi8_wrong_parent(
-; SKX-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([32 x i32]* @arr to <2 x i32>*), align 16
-; SKX-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 0
-; SKX-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
-; SKX-NEXT:    [[TMP5:%.*]] = icmp sgt i32 [[TMP3]], [[TMP4]]
-; SKX-NEXT:    br label [[PP:%.*]]
-; SKX:       pp:
-; SKX-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; SKX-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
-; SKX-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; SKX-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; SKX-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; SKX-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; SKX-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; SKX-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; SKX-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; SKX-NEXT:    [[TMP9:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; SKX-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
-; SKX-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
-; SKX-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
-; SKX-NEXT:    [[TMP13:%.*]] = insertelement <2 x i1> undef, i1 [[TMP12]], i32 0
-; SKX-NEXT:    [[TMP14:%.*]] = insertelement <2 x i1> [[TMP13]], i1 [[TMP5]], i32 1
-; SKX-NEXT:    [[TMP15:%.*]] = insertelement <2 x i32> undef, i32 [[TMP11]], i32 0
-; SKX-NEXT:    [[TMP16:%.*]] = insertelement <2 x i32> [[TMP15]], i32 [[TMP3]], i32 1
-; SKX-NEXT:    [[TMP17:%.*]] = insertelement <2 x i32> undef, i32 [[TMP8]], i32 0
-; SKX-NEXT:    [[TMP18:%.*]] = insertelement <2 x i32> [[TMP17]], i32 [[TMP4]], i32 1
-; SKX-NEXT:    [[TMP19:%.*]] = select <2 x i1> [[TMP14]], <2 x i32> [[TMP16]], <2 x i32> [[TMP18]]
-; SKX-NEXT:    [[TMP20:%.*]] = extractelement <2 x i32> [[TMP19]], i32 1
-; SKX-NEXT:    [[TMP21:%.*]] = extractelement <2 x i32> [[TMP19]], i32 0
-; SKX-NEXT:    [[TMP22:%.*]] = icmp sgt i32 [[TMP21]], [[TMP20]]
-; SKX-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP22]], i32 [[TMP21]], i32 [[TMP20]]
-; SKX-NEXT:    ret i32 [[OP_EXTRA]]
+; THRESH-LABEL: @maxi8_wrong_parent(
+; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([32 x i32]* @arr to <2 x i32>*), align 16
+; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 0
+; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
+; THRESH-NEXT:    [[TMP5:%.*]] = icmp sgt i32 [[TMP3]], [[TMP4]]
+; THRESH-NEXT:    br label [[PP:%.*]]
+; THRESH:       pp:
+; THRESH-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
+; THRESH-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
+; THRESH-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
+; THRESH-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; THRESH-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
+; THRESH-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
+; THRESH-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; THRESH-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
+; THRESH-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
+; THRESH-NEXT:    [[TMP9:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
+; THRESH-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
+; THRESH-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
+; THRESH-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
+; THRESH-NEXT:    [[TMP13:%.*]] = insertelement <2 x i1> undef, i1 [[TMP12]], i32 0
+; THRESH-NEXT:    [[TMP14:%.*]] = insertelement <2 x i1> [[TMP13]], i1 [[TMP5]], i32 1
+; THRESH-NEXT:    [[TMP15:%.*]] = insertelement <2 x i32> undef, i32 [[TMP11]], i32 0
+; THRESH-NEXT:    [[TMP16:%.*]] = insertelement <2 x i32> [[TMP15]], i32 [[TMP3]], i32 1
+; THRESH-NEXT:    [[TMP17:%.*]] = insertelement <2 x i32> undef, i32 [[TMP8]], i32 0
+; THRESH-NEXT:    [[TMP18:%.*]] = insertelement <2 x i32> [[TMP17]], i32 [[TMP4]], i32 1
+; THRESH-NEXT:    [[TMP19:%.*]] = select <2 x i1> [[TMP14]], <2 x i32> [[TMP16]], <2 x i32> [[TMP18]]
+; THRESH-NEXT:    [[TMP20:%.*]] = extractelement <2 x i32> [[TMP19]], i32 1
+; THRESH-NEXT:    [[TMP21:%.*]] = extractelement <2 x i32> [[TMP19]], i32 0
+; THRESH-NEXT:    [[TMP22:%.*]] = icmp sgt i32 [[TMP21]], [[TMP20]]
+; THRESH-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP22]], i32 [[TMP21]], i32 [[TMP20]]
+; THRESH-NEXT:    ret i32 [[OP_EXTRA]]
 ;
   %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
   %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
@@ -732,106 +632,56 @@ pp:
 
 ; PR38191 - We don't handle array-of-pointer reductions.
 define i32* @maxp8(i32) {
-; SSE-LABEL: @maxp8(
-; SSE-NEXT:    [[TMP2:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 0), align 16
-; SSE-NEXT:    [[TMP3:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 1), align 4
-; SSE-NEXT:    [[TMP4:%.*]] = icmp ugt i32* [[TMP2]], [[TMP3]]
-; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32* [[TMP2]], i32* [[TMP3]]
-; SSE-NEXT:    [[TMP6:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
-; SSE-NEXT:    [[TMP7:%.*]] = icmp ugt i32* [[TMP5]], [[TMP6]]
-; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32* [[TMP5]], i32* [[TMP6]]
-; SSE-NEXT:    [[TMP9:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
-; SSE-NEXT:    [[TMP10:%.*]] = icmp ugt i32* [[TMP8]], [[TMP9]]
-; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32* [[TMP8]], i32* [[TMP9]]
-; SSE-NEXT:    [[TMP12:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
-; SSE-NEXT:    [[TMP13:%.*]] = icmp ugt i32* [[TMP11]], [[TMP12]]
-; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32* [[TMP11]], i32* [[TMP12]]
-; SSE-NEXT:    [[TMP15:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
-; SSE-NEXT:    [[TMP16:%.*]] = icmp ugt i32* [[TMP14]], [[TMP15]]
-; SSE-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32* [[TMP14]], i32* [[TMP15]]
-; SSE-NEXT:    [[TMP18:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
-; SSE-NEXT:    [[TMP19:%.*]] = icmp ugt i32* [[TMP17]], [[TMP18]]
-; SSE-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32* [[TMP17]], i32* [[TMP18]]
-; SSE-NEXT:    [[TMP21:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
-; SSE-NEXT:    [[TMP22:%.*]] = icmp ugt i32* [[TMP20]], [[TMP21]]
-; SSE-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32* [[TMP20]], i32* [[TMP21]]
-; SSE-NEXT:    ret i32* [[TMP23]]
-;
-; AVX-LABEL: @maxp8(
-; AVX-NEXT:    [[TMP2:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 0), align 16
-; AVX-NEXT:    [[TMP3:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 1), align 4
-; AVX-NEXT:    [[TMP4:%.*]] = icmp ugt i32* [[TMP2]], [[TMP3]]
-; AVX-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32* [[TMP2]], i32* [[TMP3]]
-; AVX-NEXT:    [[TMP6:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
-; AVX-NEXT:    [[TMP7:%.*]] = icmp ugt i32* [[TMP5]], [[TMP6]]
-; AVX-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32* [[TMP5]], i32* [[TMP6]]
-; AVX-NEXT:    [[TMP9:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
-; AVX-NEXT:    [[TMP10:%.*]] = icmp ugt i32* [[TMP8]], [[TMP9]]
-; AVX-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32* [[TMP8]], i32* [[TMP9]]
-; AVX-NEXT:    [[TMP12:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
-; AVX-NEXT:    [[TMP13:%.*]] = icmp ugt i32* [[TMP11]], [[TMP12]]
-; AVX-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32* [[TMP11]], i32* [[TMP12]]
-; AVX-NEXT:    [[TMP15:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
-; AVX-NEXT:    [[TMP16:%.*]] = icmp ugt i32* [[TMP14]], [[TMP15]]
-; AVX-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32* [[TMP14]], i32* [[TMP15]]
-; AVX-NEXT:    [[TMP18:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
-; AVX-NEXT:    [[TMP19:%.*]] = icmp ugt i32* [[TMP17]], [[TMP18]]
-; AVX-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32* [[TMP17]], i32* [[TMP18]]
-; AVX-NEXT:    [[TMP21:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
-; AVX-NEXT:    [[TMP22:%.*]] = icmp ugt i32* [[TMP20]], [[TMP21]]
-; AVX-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32* [[TMP20]], i32* [[TMP21]]
-; AVX-NEXT:    ret i32* [[TMP23]]
-;
-; AVX2-LABEL: @maxp8(
-; AVX2-NEXT:    [[TMP2:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 0), align 16
-; AVX2-NEXT:    [[TMP3:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 1), align 4
-; AVX2-NEXT:    [[TMP4:%.*]] = icmp ugt i32* [[TMP2]], [[TMP3]]
-; AVX2-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32* [[TMP2]], i32* [[TMP3]]
-; AVX2-NEXT:    [[TMP6:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
-; AVX2-NEXT:    [[TMP7:%.*]] = icmp ugt i32* [[TMP5]], [[TMP6]]
-; AVX2-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32* [[TMP5]], i32* [[TMP6]]
-; AVX2-NEXT:    [[TMP9:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
-; AVX2-NEXT:    [[TMP10:%.*]] = icmp ugt i32* [[TMP8]], [[TMP9]]
-; AVX2-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32* [[TMP8]], i32* [[TMP9]]
-; AVX2-NEXT:    [[TMP12:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
-; AVX2-NEXT:    [[TMP13:%.*]] = icmp ugt i32* [[TMP11]], [[TMP12]]
-; AVX2-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32* [[TMP11]], i32* [[TMP12]]
-; AVX2-NEXT:    [[TMP15:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
-; AVX2-NEXT:    [[TMP16:%.*]] = icmp ugt i32* [[TMP14]], [[TMP15]]
-; AVX2-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32* [[TMP14]], i32* [[TMP15]]
-; AVX2-NEXT:    [[TMP18:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
-; AVX2-NEXT:    [[TMP19:%.*]] = icmp ugt i32* [[TMP17]], [[TMP18]]
-; AVX2-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32* [[TMP17]], i32* [[TMP18]]
-; AVX2-NEXT:    [[TMP21:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
-; AVX2-NEXT:    [[TMP22:%.*]] = icmp ugt i32* [[TMP20]], [[TMP21]]
-; AVX2-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32* [[TMP20]], i32* [[TMP21]]
-; AVX2-NEXT:    ret i32* [[TMP23]]
+; DEFAULT-LABEL: @maxp8(
+; DEFAULT-NEXT:    [[TMP2:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 0), align 16
+; DEFAULT-NEXT:    [[TMP3:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 1), align 4
+; DEFAULT-NEXT:    [[TMP4:%.*]] = icmp ugt i32* [[TMP2]], [[TMP3]]
+; DEFAULT-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32* [[TMP2]], i32* [[TMP3]]
+; DEFAULT-NEXT:    [[TMP6:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
+; DEFAULT-NEXT:    [[TMP7:%.*]] = icmp ugt i32* [[TMP5]], [[TMP6]]
+; DEFAULT-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32* [[TMP5]], i32* [[TMP6]]
+; DEFAULT-NEXT:    [[TMP9:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
+; DEFAULT-NEXT:    [[TMP10:%.*]] = icmp ugt i32* [[TMP8]], [[TMP9]]
+; DEFAULT-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32* [[TMP8]], i32* [[TMP9]]
+; DEFAULT-NEXT:    [[TMP12:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
+; DEFAULT-NEXT:    [[TMP13:%.*]] = icmp ugt i32* [[TMP11]], [[TMP12]]
+; DEFAULT-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32* [[TMP11]], i32* [[TMP12]]
+; DEFAULT-NEXT:    [[TMP15:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
+; DEFAULT-NEXT:    [[TMP16:%.*]] = icmp ugt i32* [[TMP14]], [[TMP15]]
+; DEFAULT-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32* [[TMP14]], i32* [[TMP15]]
+; DEFAULT-NEXT:    [[TMP18:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
+; DEFAULT-NEXT:    [[TMP19:%.*]] = icmp ugt i32* [[TMP17]], [[TMP18]]
+; DEFAULT-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32* [[TMP17]], i32* [[TMP18]]
+; DEFAULT-NEXT:    [[TMP21:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
+; DEFAULT-NEXT:    [[TMP22:%.*]] = icmp ugt i32* [[TMP20]], [[TMP21]]
+; DEFAULT-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32* [[TMP20]], i32* [[TMP21]]
+; DEFAULT-NEXT:    ret i32* [[TMP23]]
 ;
-; SKX-LABEL: @maxp8(
-; SKX-NEXT:    [[TMP2:%.*]] = load <2 x i32*>, <2 x i32*>* bitcast ([32 x i32*]* @arrp to <2 x i32*>*), align 16
-; SKX-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32*> [[TMP2]], i32 0
-; SKX-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32*> [[TMP2]], i32 1
-; SKX-NEXT:    [[TMP5:%.*]] = icmp ugt i32* [[TMP3]], [[TMP4]]
-; SKX-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], i32* [[TMP3]], i32* [[TMP4]]
-; SKX-NEXT:    [[TMP7:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
-; SKX-NEXT:    [[TMP8:%.*]] = icmp ugt i32* [[TMP6]], [[TMP7]]
-; SKX-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i32* [[TMP6]], i32* [[TMP7]]
-; SKX-NEXT:    [[TMP10:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
-; SKX-NEXT:    [[TMP11:%.*]] = icmp ugt i32* [[TMP9]], [[TMP10]]
-; SKX-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i32* [[TMP9]], i32* [[TMP10]]
-; SKX-NEXT:    [[TMP13:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
-; SKX-NEXT:    [[TMP14:%.*]] = icmp ugt i32* [[TMP12]], [[TMP13]]
-; SKX-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i32* [[TMP12]], i32* [[TMP13]]
-; SKX-NEXT:    [[TMP16:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
-; SKX-NEXT:    [[TMP17:%.*]] = icmp ugt i32* [[TMP15]], [[TMP16]]
-; SKX-NEXT:    [[TMP18:%.*]] = select i1 [[TMP17]], i32* [[TMP15]], i32* [[TMP16]]
-; SKX-NEXT:    [[TMP19:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
-; SKX-NEXT:    [[TMP20:%.*]] = icmp ugt i32* [[TMP18]], [[TMP19]]
-; SKX-NEXT:    [[TMP21:%.*]] = select i1 [[TMP20]], i32* [[TMP18]], i32* [[TMP19]]
-; SKX-NEXT:    [[TMP22:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
-; SKX-NEXT:    [[TMP23:%.*]] = icmp ugt i32* [[TMP21]], [[TMP22]]
-; SKX-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], i32* [[TMP21]], i32* [[TMP22]]
-; SKX-NEXT:    ret i32* [[TMP24]]
+; THRESH-LABEL: @maxp8(
+; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x i32*>, <2 x i32*>* bitcast ([32 x i32*]* @arrp to <2 x i32*>*), align 16
+; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32*> [[TMP2]], i32 0
+; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32*> [[TMP2]], i32 1
+; THRESH-NEXT:    [[TMP5:%.*]] = icmp ugt i32* [[TMP3]], [[TMP4]]
+; THRESH-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], i32* [[TMP3]], i32* [[TMP4]]
+; THRESH-NEXT:    [[TMP7:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
+; THRESH-NEXT:    [[TMP8:%.*]] = icmp ugt i32* [[TMP6]], [[TMP7]]
+; THRESH-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i32* [[TMP6]], i32* [[TMP7]]
+; THRESH-NEXT:    [[TMP10:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
+; THRESH-NEXT:    [[TMP11:%.*]] = icmp ugt i32* [[TMP9]], [[TMP10]]
+; THRESH-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i32* [[TMP9]], i32* [[TMP10]]
+; THRESH-NEXT:    [[TMP13:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
+; THRESH-NEXT:    [[TMP14:%.*]] = icmp ugt i32* [[TMP12]], [[TMP13]]
+; THRESH-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i32* [[TMP12]], i32* [[TMP13]]
+; THRESH-NEXT:    [[TMP16:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
+; THRESH-NEXT:    [[TMP17:%.*]] = icmp ugt i32* [[TMP15]], [[TMP16]]
+; THRESH-NEXT:    [[TMP18:%.*]] = select i1 [[TMP17]], i32* [[TMP15]], i32* [[TMP16]]
+; THRESH-NEXT:    [[TMP19:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
+; THRESH-NEXT:    [[TMP20:%.*]] = icmp ugt i32* [[TMP18]], [[TMP19]]
+; THRESH-NEXT:    [[TMP21:%.*]] = select i1 [[TMP20]], i32* [[TMP18]], i32* [[TMP19]]
+; THRESH-NEXT:    [[TMP22:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
+; THRESH-NEXT:    [[TMP23:%.*]] = icmp ugt i32* [[TMP21]], [[TMP22]]
+; THRESH-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], i32* [[TMP21]], i32* [[TMP22]]
+; THRESH-NEXT:    ret i32* [[TMP24]]
 ;
   %2 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 0), align 16
   %3 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 1), align 4