diff --git a/lib/Target/X86/X86TargetTransformInfo.cpp b/lib/Target/X86/X86TargetTransformInfo.cpp
index c86790a9326..b5de8f947c9 100644
--- a/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -947,6 +947,24 @@ int X86TTIImpl::getIntrinsicInstrCost(Intrinsic::ID IID, Type *RetTy,
     { ISD::BITREVERSE, MVT::i16,     3 },
     { ISD::BITREVERSE, MVT::i8,      3 }
   };
+  static const CostTblEntry AVX2CostTbl[] = {
+    { ISD::BITREVERSE, MVT::v4i64,   5 },
+    { ISD::BITREVERSE, MVT::v8i32,   5 },
+    { ISD::BITREVERSE, MVT::v16i16,  5 },
+    { ISD::BITREVERSE, MVT::v32i8,   5 }
+  };
+  static const CostTblEntry AVX1CostTbl[] = {
+    { ISD::BITREVERSE, MVT::v4i64,  10 },
+    { ISD::BITREVERSE, MVT::v8i32,  10 },
+    { ISD::BITREVERSE, MVT::v16i16, 10 },
+    { ISD::BITREVERSE, MVT::v32i8,  10 }
+  };
+  static const CostTblEntry SSSE3CostTbl[] = {
+    { ISD::BITREVERSE, MVT::v2i64,   5 },
+    { ISD::BITREVERSE, MVT::v4i32,   5 },
+    { ISD::BITREVERSE, MVT::v8i16,   5 },
+    { ISD::BITREVERSE, MVT::v16i8,   5 }
+  };
 
   unsigned ISD = ISD::DELETED_NODE;
   switch (IID) {
@@ -966,6 +984,18 @@ int X86TTIImpl::getIntrinsicInstrCost(Intrinsic::ID IID, Type *RetTy,
     if (const auto *Entry = CostTableLookup(XOPCostTbl, ISD, MTy))
       return LT.first * Entry->Cost;
 
+  if (ST->hasAVX2())
+    if (const auto *Entry = CostTableLookup(AVX2CostTbl, ISD, MTy))
+      return LT.first * Entry->Cost;
+
+  if (ST->hasAVX())
+    if (const auto *Entry = CostTableLookup(AVX1CostTbl, ISD, MTy))
+      return LT.first * Entry->Cost;
+
+  if (ST->hasSSSE3())
+    if (const auto *Entry = CostTableLookup(SSSE3CostTbl, ISD, MTy))
+      return LT.first * Entry->Cost;
+
   return BaseT::getIntrinsicInstrCost(IID, RetTy, Tys, FMF);
 }
 
diff --git a/test/Analysis/CostModel/X86/bitreverse.ll b/test/Analysis/CostModel/X86/bitreverse.ll
index 4f04f3b0dd9..c9eea20c340 100644
--- a/test/Analysis/CostModel/X86/bitreverse.ll
+++ b/test/Analysis/CostModel/X86/bitreverse.ll
@@ -71,9 +71,9 @@ declare <32 x i8> @llvm.bitreverse.v32i8(<32 x i8>)
 define <2 x i64> @var_bitreverse_v2i64(<2 x i64> %a) {
 ; CHECK: 'Cost Model Analysis' for function 'var_bitreverse_v2i64':
 ; SSE2: Found an estimated cost of 6 for instruction:   %bitreverse
-; SSE42: Found an estimated cost of 6 for instruction:   %bitreverse
-; AVX: Found an estimated cost of 6 for instruction:   %bitreverse
-; AVX2: Found an estimated cost of 6 for instruction:   %bitreverse
+; SSE42: Found an estimated cost of 5 for instruction:   %bitreverse
+; AVX: Found an estimated cost of 5 for instruction:   %bitreverse
+; AVX2: Found an estimated cost of 5 for instruction:   %bitreverse
 ; XOP: Found an estimated cost of 1 for instruction:   %bitreverse
   %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
   ret <2 x i64> %bitreverse
@@ -82,9 +82,9 @@ define <2 x i64> @var_bitreverse_v2i64(<2 x i64> %a) {
 define <4 x i64> @var_bitreverse_v4i64(<4 x i64> %a) {
 ; CHECK: 'Cost Model Analysis' for function 'var_bitreverse_v4i64':
 ; SSE2: Found an estimated cost of 12 for instruction:   %bitreverse
-; SSE42: Found an estimated cost of 12 for instruction:   %bitreverse
-; AVX: Found an estimated cost of 12 for instruction:   %bitreverse
-; AVX2: Found an estimated cost of 12 for instruction:   %bitreverse
+; SSE42: Found an estimated cost of 10 for instruction:   %bitreverse
+; AVX: Found an estimated cost of 10 for instruction:   %bitreverse
+; AVX2: Found an estimated cost of 5 for instruction:   %bitreverse
 ; XOP: Found an estimated cost of 4 for instruction:   %bitreverse
   %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
   ret <4 x i64> %bitreverse
@@ -93,9 +93,9 @@ define <4 x i64> @var_bitreverse_v4i64(<4 x i64> %a) {
 define <4 x i32> @var_bitreverse_v4i32(<4 x i32> %a) {
 ; CHECK: 'Cost Model Analysis' for function 'var_bitreverse_v4i32':
 ; SSE2: Found an estimated cost of 12 for instruction:   %bitreverse
-; SSE42: Found an estimated cost of 12 for instruction:   %bitreverse
-; AVX: Found an estimated cost of 12 for instruction:   %bitreverse
-; AVX2: Found an estimated cost of 12 for instruction:   %bitreverse
+; SSE42: Found an estimated cost of 5 for instruction:   %bitreverse
+; AVX: Found an estimated cost of 5 for instruction:   %bitreverse
+; AVX2: Found an estimated cost of 5 for instruction:   %bitreverse
 ; XOP: Found an estimated cost of 1 for instruction:   %bitreverse
   %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
   ret <4 x i32> %bitreverse
@@ -104,9 +104,9 @@ define <4 x i32> @var_bitreverse_v4i32(<4 x i32> %a) {
 define <8 x i32> @var_bitreverse_v8i32(<8 x i32> %a) {
 ; CHECK: 'Cost Model Analysis' for function 'var_bitreverse_v8i32':
 ; SSE2: Found an estimated cost of 24 for instruction:   %bitreverse
-; SSE42: Found an estimated cost of 24 for instruction:   %bitreverse
-; AVX: Found an estimated cost of 24 for instruction:   %bitreverse
-; AVX2: Found an estimated cost of 24 for instruction:   %bitreverse
+; SSE42: Found an estimated cost of 10 for instruction:   %bitreverse
+; AVX: Found an estimated cost of 10 for instruction:   %bitreverse
+; AVX2: Found an estimated cost of 5 for instruction:   %bitreverse
 ; XOP: Found an estimated cost of 4 for instruction:   %bitreverse
   %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
   ret <8 x i32> %bitreverse
@@ -115,9 +115,9 @@ define <8 x i32> @var_bitreverse_v8i32(<8 x i32> %a) {
 define <8 x i16> @var_bitreverse_v8i16(<8 x i16> %a) {
 ; CHECK: 'Cost Model Analysis' for function 'var_bitreverse_v8i16':
 ; SSE2: Found an estimated cost of 24 for instruction:   %bitreverse
-; SSE42: Found an estimated cost of 24 for instruction:   %bitreverse
-; AVX: Found an estimated cost of 24 for instruction:   %bitreverse
-; AVX2: Found an estimated cost of 24 for instruction:   %bitreverse
+; SSE42: Found an estimated cost of 5 for instruction:   %bitreverse
+; AVX: Found an estimated cost of 5 for instruction:   %bitreverse
+; AVX2: Found an estimated cost of 5 for instruction:   %bitreverse
 ; XOP: Found an estimated cost of 1 for instruction:   %bitreverse
   %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
   ret <8 x i16> %bitreverse
@@ -126,9 +126,9 @@ define <8 x i16> @var_bitreverse_v8i16(<8 x i16> %a) {
 define <16 x i16> @var_bitreverse_v16i16(<16 x i16> %a) {
 ; CHECK: 'Cost Model Analysis' for function 'var_bitreverse_v16i16':
 ; SSE2: Found an estimated cost of 48 for instruction:   %bitreverse
-; SSE42: Found an estimated cost of 48 for instruction:   %bitreverse
-; AVX: Found an estimated cost of 48 for instruction:   %bitreverse
-; AVX2: Found an estimated cost of 48 for instruction:   %bitreverse
+; SSE42: Found an estimated cost of 10 for instruction:   %bitreverse
+; AVX: Found an estimated cost of 10 for instruction:   %bitreverse
+; AVX2: Found an estimated cost of 5 for instruction:   %bitreverse
 ; XOP: Found an estimated cost of 4 for instruction:   %bitreverse
   %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
   ret <16 x i16> %bitreverse
@@ -137,9 +137,9 @@ define <16 x i16> @var_bitreverse_v16i16(<16 x i16> %a) {
 define <16 x i8> @var_bitreverse_v16i8(<16 x i8> %a) {
 ; CHECK: 'Cost Model Analysis' for function 'var_bitreverse_v16i8':
 ; SSE2: Found an estimated cost of 48 for instruction:   %bitreverse
-; SSE42: Found an estimated cost of 48 for instruction:   %bitreverse
-; AVX: Found an estimated cost of 48 for instruction:   %bitreverse
-; AVX2: Found an estimated cost of 48 for instruction:   %bitreverse
+; SSE42: Found an estimated cost of 5 for instruction:   %bitreverse
+; AVX: Found an estimated cost of 5 for instruction:   %bitreverse
+; AVX2: Found an estimated cost of 5 for instruction:   %bitreverse
 ; XOP: Found an estimated cost of 1 for instruction:   %bitreverse
   %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
   ret <16 x i8> %bitreverse
@@ -148,9 +148,9 @@ define <16 x i8> @var_bitreverse_v16i8(<16 x i8> %a) {
 define <32 x i8> @var_bitreverse_v32i8(<32 x i8> %a) {
 ; CHECK: 'Cost Model Analysis' for function 'var_bitreverse_v32i8':
 ; SSE2: Found an estimated cost of 96 for instruction:   %bitreverse
-; SSE42: Found an estimated cost of 96 for instruction:   %bitreverse
-; AVX: Found an estimated cost of 96 for instruction:   %bitreverse
-; AVX2: Found an estimated cost of 96 for instruction:   %bitreverse
+; SSE42: Found an estimated cost of 10 for instruction:   %bitreverse
+; AVX: Found an estimated cost of 10 for instruction:   %bitreverse
+; AVX2: Found an estimated cost of 5 for instruction:   %bitreverse
 ; XOP: Found an estimated cost of 4 for instruction:   %bitreverse
   %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
   ret <32 x i8> %bitreverse
diff --git a/test/Transforms/SLPVectorizer/X86/bitreverse.ll b/test/Transforms/SLPVectorizer/X86/bitreverse.ll
index d794bcab545..c6d65bbe684 100644
--- a/test/Transforms/SLPVectorizer/X86/bitreverse.ll
+++ b/test/Transforms/SLPVectorizer/X86/bitreverse.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
-; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
-; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=XOP
 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver4 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=XOP
 
@@ -71,20 +71,26 @@ define void @bitreverse_4i64() #0 {
 ; SSE-NEXT:    store i64 [[BITREVERSE3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
 ; SSE-NEXT:    ret void
 ;
-; AVX-LABEL: @bitreverse_4i64(
-; AVX-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
-; AVX-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
-; AVX-NEXT:    [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
-; AVX-NEXT:    [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
-; AVX-NEXT:    [[BITREVERSE0:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD0]])
-; AVX-NEXT:    [[BITREVERSE1:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD1]])
-; AVX-NEXT:    [[BITREVERSE2:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD2]])
-; AVX-NEXT:    [[BITREVERSE3:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD3]])
-; AVX-NEXT:    store i64 [[BITREVERSE0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
-; AVX-NEXT:    store i64 [[BITREVERSE1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
-; AVX-NEXT:    store i64 [[BITREVERSE2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
-; AVX-NEXT:    store i64 [[BITREVERSE3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
-; AVX-NEXT:    ret void
+; AVX1-LABEL: @bitreverse_4i64(
+; AVX1-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+; AVX1-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+; AVX1-NEXT:    [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+; AVX1-NEXT:    [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+; AVX1-NEXT:    [[BITREVERSE0:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD0]])
+; AVX1-NEXT:    [[BITREVERSE1:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD1]])
+; AVX1-NEXT:    [[BITREVERSE2:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD2]])
+; AVX1-NEXT:    [[BITREVERSE3:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD3]])
+; AVX1-NEXT:    store i64 [[BITREVERSE0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+; AVX1-NEXT:    store i64 [[BITREVERSE1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+; AVX1-NEXT:    store i64 [[BITREVERSE2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+; AVX1-NEXT:    store i64 [[BITREVERSE3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+; AVX1-NEXT:    ret void
+;
+; AVX2-LABEL: @bitreverse_4i64(
+; AVX2-NEXT:    [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([4 x i64]* @src64 to <4 x i64>*), align 4
+; AVX2-NEXT:    [[TMP2:%.*]] = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> [[TMP1]])
+; AVX2-NEXT:    store <4 x i64> [[TMP2]], <4 x i64>* bitcast ([4 x i64]* @dst64 to <4 x i64>*), align 4
+; AVX2-NEXT:    ret void
 ;
 ; XOP-LABEL: @bitreverse_4i64(
 ; XOP-NEXT:    [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([4 x i64]* @src64 to <4 x i64>*), align 4
@@ -124,18 +130,9 @@ define void @bitreverse_4i32() #0 {
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @bitreverse_4i32(
-; AVX-NEXT:    [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
-; AVX-NEXT:    [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
-; AVX-NEXT:    [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
-; AVX-NEXT:    [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
-; AVX-NEXT:    [[BITREVERSE0:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD0]])
-; AVX-NEXT:    [[BITREVERSE1:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD1]])
-; AVX-NEXT:    [[BITREVERSE2:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD2]])
-; AVX-NEXT:    [[BITREVERSE3:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD3]])
-; AVX-NEXT:    store i32 [[BITREVERSE0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
-; AVX-NEXT:    store i32 [[BITREVERSE1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
-; AVX-NEXT:    store i32 [[BITREVERSE2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
-; AVX-NEXT:    store i32 [[BITREVERSE3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+; AVX-NEXT:    [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 4
+; AVX-NEXT:    [[TMP2:%.*]] = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> [[TMP1]])
+; AVX-NEXT:    store <4 x i32> [[TMP2]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 4
 ; AVX-NEXT:    ret void
 ;
 ; XOP-LABEL: @bitreverse_4i32(
@@ -188,30 +185,9 @@ define void @bitreverse_8i32() #0 {
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @bitreverse_8i32(
-; AVX-NEXT:    [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
-; AVX-NEXT:    [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
-; AVX-NEXT:    [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
-; AVX-NEXT:    [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
-; AVX-NEXT:    [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
-; AVX-NEXT:    [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
-; AVX-NEXT:    [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
-; AVX-NEXT:    [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
-; AVX-NEXT:    [[BITREVERSE0:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD0]])
-; AVX-NEXT:    [[BITREVERSE1:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD1]])
-; AVX-NEXT:    [[BITREVERSE2:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD2]])
-; AVX-NEXT:    [[BITREVERSE3:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD3]])
-; AVX-NEXT:    [[BITREVERSE4:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD4]])
-; AVX-NEXT:    [[BITREVERSE5:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD5]])
-; AVX-NEXT:    [[BITREVERSE6:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD6]])
-; AVX-NEXT:    [[BITREVERSE7:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD7]])
-; AVX-NEXT:    store i32 [[BITREVERSE0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
-; AVX-NEXT:    store i32 [[BITREVERSE1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
-; AVX-NEXT:    store i32 [[BITREVERSE2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
-; AVX-NEXT:    store i32 [[BITREVERSE3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
-; AVX-NEXT:    store i32 [[BITREVERSE4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
-; AVX-NEXT:    store i32 [[BITREVERSE5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
-; AVX-NEXT:    store i32 [[BITREVERSE6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
-; AVX-NEXT:    store i32 [[BITREVERSE7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; AVX-NEXT:    [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; AVX-NEXT:    [[TMP2:%.*]] = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> [[TMP1]])
+; AVX-NEXT:    store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
 ; AVX-NEXT:    ret void
 ;
 ; XOP-LABEL: @bitreverse_8i32(
@@ -276,30 +252,9 @@ define void @bitreverse_8i16() #0 {
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @bitreverse_8i16(
-; AVX-NEXT:    [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
-; AVX-NEXT:    [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
-; AVX-NEXT:    [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
-; AVX-NEXT:    [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
-; AVX-NEXT:    [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
-; AVX-NEXT:    [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
-; AVX-NEXT:    [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
-; AVX-NEXT:    [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
-; AVX-NEXT:    [[BITREVERSE0:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD0]])
-; AVX-NEXT:    [[BITREVERSE1:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD1]])
-; AVX-NEXT:    [[BITREVERSE2:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD2]])
-; AVX-NEXT:    [[BITREVERSE3:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD3]])
-; AVX-NEXT:    [[BITREVERSE4:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD4]])
-; AVX-NEXT:    [[BITREVERSE5:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD5]])
-; AVX-NEXT:    [[BITREVERSE6:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD6]])
-; AVX-NEXT:    [[BITREVERSE7:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD7]])
-; AVX-NEXT:    store i16 [[BITREVERSE0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+; AVX-NEXT:    [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; AVX-NEXT:    [[TMP2:%.*]] = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> [[TMP1]])
+; AVX-NEXT:    store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
 ; AVX-NEXT:    ret void
 ;
 ; XOP-LABEL: @bitreverse_8i16(
@@ -388,54 +343,9 @@ define void @bitreverse_16i16() #0 {
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @bitreverse_16i16(
-; AVX-NEXT:    [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
-; AVX-NEXT:    [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
-; AVX-NEXT:    [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
-; AVX-NEXT:    [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
-; AVX-NEXT:    [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
-; AVX-NEXT:    [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
-; AVX-NEXT:    [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
-; AVX-NEXT:    [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
-; AVX-NEXT:    [[LD8:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2
-; AVX-NEXT:    [[LD9:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2
-; AVX-NEXT:    [[LD10:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
-; AVX-NEXT:    [[LD11:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
-; AVX-NEXT:    [[LD12:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
-; AVX-NEXT:    [[LD13:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
-; AVX-NEXT:    [[LD14:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
-; AVX-NEXT:    [[LD15:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
-; AVX-NEXT:    [[BITREVERSE0:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD0]])
-; AVX-NEXT:    [[BITREVERSE1:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD1]])
-; AVX-NEXT:    [[BITREVERSE2:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD2]])
-; AVX-NEXT:    [[BITREVERSE3:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD3]])
-; AVX-NEXT:    [[BITREVERSE4:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD4]])
-; AVX-NEXT:    [[BITREVERSE5:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD5]])
-; AVX-NEXT:    [[BITREVERSE6:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD6]])
-; AVX-NEXT:    [[BITREVERSE7:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD7]])
-; AVX-NEXT:    [[BITREVERSE8:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD8]])
-; AVX-NEXT:    [[BITREVERSE9:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD9]])
-; AVX-NEXT:    [[BITREVERSE10:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD10]])
-; AVX-NEXT:    [[BITREVERSE11:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD11]])
-; AVX-NEXT:    [[BITREVERSE12:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD12]])
-; AVX-NEXT:    [[BITREVERSE13:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD13]])
-; AVX-NEXT:    [[BITREVERSE14:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD14]])
-; AVX-NEXT:    [[BITREVERSE15:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD15]])
-; AVX-NEXT:    store i16 [[BITREVERSE0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE8]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE9]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE10]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE11]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE12]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE13]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE14]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
-; AVX-NEXT:    store i16 [[BITREVERSE15]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
+; AVX-NEXT:    [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; AVX-NEXT:    [[TMP2:%.*]] = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> [[TMP1]])
+; AVX-NEXT:    store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
 ; AVX-NEXT:    ret void
 ;
 ; XOP-LABEL: @bitreverse_16i16(
@@ -548,54 +458,9 @@ define void @bitreverse_16i8() #0 {
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @bitreverse_16i8(
-; AVX-NEXT:    [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
-; AVX-NEXT:    [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
-; AVX-NEXT:    [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
-; AVX-NEXT:    [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
-; AVX-NEXT:    [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
-; AVX-NEXT:    [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
-; AVX-NEXT:    [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
-; AVX-NEXT:    [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
-; AVX-NEXT:    [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
-; AVX-NEXT:    [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
-; AVX-NEXT:    [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
-; AVX-NEXT:    [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
-; AVX-NEXT:    [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
-; AVX-NEXT:    [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
-; AVX-NEXT:    [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
-; AVX-NEXT:    [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
-; AVX-NEXT:    [[BITREVERSE0:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD0]])
-; AVX-NEXT:    [[BITREVERSE1:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD1]])
-; AVX-NEXT:    [[BITREVERSE2:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD2]])
-; AVX-NEXT:    [[BITREVERSE3:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD3]])
-; AVX-NEXT:    [[BITREVERSE4:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD4]])
-; AVX-NEXT:    [[BITREVERSE5:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD5]])
-; AVX-NEXT:    [[BITREVERSE6:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD6]])
-; AVX-NEXT:    [[BITREVERSE7:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD7]])
-; AVX-NEXT:    [[BITREVERSE8:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD8]])
-; AVX-NEXT:    [[BITREVERSE9:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD9]])
-; AVX-NEXT:    [[BITREVERSE10:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD10]])
-; AVX-NEXT:    [[BITREVERSE11:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD11]])
-; AVX-NEXT:    [[BITREVERSE12:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD12]])
-; AVX-NEXT:    [[BITREVERSE13:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD13]])
-; AVX-NEXT:    [[BITREVERSE14:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD14]])
-; AVX-NEXT:    [[BITREVERSE15:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD15]])
-; AVX-NEXT:    store i8 [[BITREVERSE0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+; AVX-NEXT:    [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; AVX-NEXT:    [[TMP2:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP1]])
+; AVX-NEXT:    store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
 ; AVX-NEXT:    ret void
 ;
 ; XOP-LABEL: @bitreverse_16i8(
@@ -756,102 +621,12 @@ define void @bitreverse_32i8() #0 {
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @bitreverse_32i8(
-; AVX-NEXT:    [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
-; AVX-NEXT:    [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
-; AVX-NEXT:    [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
-; AVX-NEXT:    [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
-; AVX-NEXT:    [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
-; AVX-NEXT:    [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
-; AVX-NEXT:    [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
-; AVX-NEXT:    [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
-; AVX-NEXT:    [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
-; AVX-NEXT:    [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
-; AVX-NEXT:    [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
-; AVX-NEXT:    [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
-; AVX-NEXT:    [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
-; AVX-NEXT:    [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
-; AVX-NEXT:    [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
-; AVX-NEXT:    [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
-; AVX-NEXT:    [[LD16:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
-; AVX-NEXT:    [[LD17:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
-; AVX-NEXT:    [[LD18:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
-; AVX-NEXT:    [[LD19:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
-; AVX-NEXT:    [[LD20:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
-; AVX-NEXT:    [[LD21:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
-; AVX-NEXT:    [[LD22:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
-; AVX-NEXT:    [[LD23:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
-; AVX-NEXT:    [[LD24:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
-; AVX-NEXT:    [[LD25:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
-; AVX-NEXT:    [[LD26:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
-; AVX-NEXT:    [[LD27:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
-; AVX-NEXT:    [[LD28:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
-; AVX-NEXT:    [[LD29:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
-; AVX-NEXT:    [[LD30:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
-; AVX-NEXT:    [[LD31:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
-; AVX-NEXT:    [[BITREVERSE0:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD0]])
-; AVX-NEXT:    [[BITREVERSE1:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD1]])
-; AVX-NEXT:    [[BITREVERSE2:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD2]])
-; AVX-NEXT:    [[BITREVERSE3:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD3]])
-; AVX-NEXT:    [[BITREVERSE4:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD4]])
-; AVX-NEXT:    [[BITREVERSE5:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD5]])
-; AVX-NEXT:    [[BITREVERSE6:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD6]])
-; AVX-NEXT:    [[BITREVERSE7:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD7]])
-; AVX-NEXT:    [[BITREVERSE8:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD8]])
-; AVX-NEXT:    [[BITREVERSE9:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD9]])
-; AVX-NEXT:    [[BITREVERSE10:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD10]])
-; AVX-NEXT:    [[BITREVERSE11:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD11]])
-; AVX-NEXT:    [[BITREVERSE12:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD12]])
-; AVX-NEXT:    [[BITREVERSE13:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD13]])
-; AVX-NEXT:    [[BITREVERSE14:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD14]])
-; AVX-NEXT:    [[BITREVERSE15:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD15]])
-; AVX-NEXT:    [[BITREVERSE16:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD16]])
-; AVX-NEXT:    [[BITREVERSE17:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD17]])
-; AVX-NEXT:    [[BITREVERSE18:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD18]])
-; AVX-NEXT:    [[BITREVERSE19:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD19]])
-; AVX-NEXT:    [[BITREVERSE20:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD20]])
-; AVX-NEXT:    [[BITREVERSE21:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD21]])
-; AVX-NEXT:    [[BITREVERSE22:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD22]])
-; AVX-NEXT:    [[BITREVERSE23:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD23]])
-; AVX-NEXT:    [[BITREVERSE24:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD24]])
-; AVX-NEXT:    [[BITREVERSE25:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD25]])
-; AVX-NEXT:    [[BITREVERSE26:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD26]])
-; AVX-NEXT:    [[BITREVERSE27:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD27]])
-; AVX-NEXT:    [[BITREVERSE28:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD28]])
-; AVX-NEXT:    [[BITREVERSE29:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD29]])
-; AVX-NEXT:    [[BITREVERSE30:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD30]])
-; AVX-NEXT:    [[BITREVERSE31:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD31]])
-; AVX-NEXT:    store i8 [[BITREVERSE0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE16]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE17]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE18]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE19]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE20]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE21]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE22]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE23]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE24]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE25]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE26]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE27]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE28]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE29]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE30]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
-; AVX-NEXT:    store i8 [[BITREVERSE31]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
+; AVX-NEXT:    [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; AVX-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
+; AVX-NEXT:    [[TMP3:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP1]])
+; AVX-NEXT:    [[TMP4:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP2]])
+; AVX-NEXT:    store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; AVX-NEXT:    store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
 ; AVX-NEXT:    ret void
 ;
 ; XOP-LABEL: @bitreverse_32i8(