[CodeGen] remove instcombine from codegen tests; NFC

The FileCheck lines in these files are auto-generated and complete, so there's very little upside (less CHECK lines) from running -instcombine on them and violating the expected test layering (optimizer developers shouldn't have to be aware of clang tests). Running opt passes like this makes it harder to make changes such as: D93817
2021-06-08 15:12:41 -04:00 · 2021-06-08 15:12:41 -04:00 · d69c4372bf
parent 2a5afb4665
commit d69c4372bf
6 changed files with 901 additions and 155 deletions
--- a/clang/test/CodeGen/aarch64-bf16-dotprod-intrinsics.c
+++ b/clang/test/CodeGen/aarch64-bf16-dotprod-intrinsics.c
@ -1,12 +1,15 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -triple aarch64-arm-none-eabi -target-feature +neon -target-feature +bf16 \
-// RUN: -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg -instcombine | FileCheck %s
+// RUN: -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg | FileCheck %s

 #include <arm_neon.h>

 // CHECK-LABEL: @test_vbfdot_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.bfdot.v2f32.v4bf16(<2 x float> [[R:%.*]], <4 x bfloat> [[A:%.*]], <4 x bfloat> [[B:%.*]]) [[ATTR3:#.*]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[R:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x bfloat> [[A:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x bfloat> [[B:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.bfdot.v2f32.v4bf16(<2 x float> [[R]], <4 x bfloat> [[A]], <4 x bfloat> [[B]]) #[[ATTR3:[0-9]+]]
 // CHECK-NEXT:    ret <2 x float> [[VBFDOT3_I]]
 //
 float32x2_t test_vbfdot_f32(float32x2_t r, bfloat16x4_t a, bfloat16x4_t b) {
@ -15,7 +18,10 @@ float32x2_t test_vbfdot_f32(float32x2_t r, bfloat16x4_t a, bfloat16x4_t b) {

 // CHECK-LABEL: @test_vbfdotq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]]) [[ATTR3]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[B]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <4 x float> [[VBFDOT3_I]]
 //
 float32x4_t test_vbfdotq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b){
@ -24,10 +30,21 @@ float32x4_t test_vbfdotq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b){

 // CHECK-LABEL: @test_vbfdot_lane_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTCAST:%.*]] = bitcast <4 x bfloat> [[B:%.*]] to <2 x float>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[DOTCAST]], <2 x float> undef, <2 x i32> zeroinitializer
-// CHECK-NEXT:    [[DOTCAST1:%.*]] = bitcast <2 x float> [[LANE]] to <4 x bfloat>
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.bfdot.v2f32.v4bf16(<2 x float> [[R:%.*]], <4 x bfloat> [[A:%.*]], <4 x bfloat> [[DOTCAST1]]) [[ATTR3]]
+// CHECK-NEXT:    [[__REINT_144:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-NEXT:    [[__REINT1_144:%.*]] = alloca <2 x float>, align 8
+// CHECK-NEXT:    store <4 x bfloat> [[B:%.*]], <4 x bfloat>* [[__REINT_144]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_144]] to <2 x float>*
+// CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[TMP1]] to <8 x i8>
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <2 x float>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK-NEXT:    store <2 x float> [[LANE]], <2 x float>* [[__REINT1_144]], align 8
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x float>* [[__REINT1_144]] to <4 x bfloat>*
+// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[TMP4]], align 8
+// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x float> [[R:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x bfloat> [[A:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x bfloat> [[TMP5]] to <8 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.bfdot.v2f32.v4bf16(<2 x float> [[R]], <4 x bfloat> [[A]], <4 x bfloat> [[TMP5]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <2 x float> [[VBFDOT3_I]]
 //
 float32x2_t test_vbfdot_lane_f32(float32x2_t r, bfloat16x4_t a, bfloat16x4_t b){
@ -36,10 +53,21 @@ float32x2_t test_vbfdot_lane_f32(float32x2_t r, bfloat16x4_t a, bfloat16x4_t b){

 // CHECK-LABEL: @test_vbfdotq_laneq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTCAST:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <4 x float>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[DOTCAST]], <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-// CHECK-NEXT:    [[DOTCAST1:%.*]] = bitcast <4 x float> [[LANE]] to <8 x bfloat>
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[DOTCAST1]]) [[ATTR3]]
+// CHECK-NEXT:    [[__REINT_146:%.*]] = alloca <8 x bfloat>, align 16
+// CHECK-NEXT:    [[__REINT1_146:%.*]] = alloca <4 x float>, align 16
+// CHECK-NEXT:    store <8 x bfloat> [[B:%.*]], <8 x bfloat>* [[__REINT_146]], align 16
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat>* [[__REINT_146]] to <4 x float>*
+// CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[TMP0]], align 16
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <16 x i8>
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP3]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+// CHECK-NEXT:    store <4 x float> [[LANE]], <4 x float>* [[__REINT1_146]], align 16
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x float>* [[__REINT1_146]] to <8 x bfloat>*
+// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[TMP4]], align 16
+// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x bfloat> [[TMP5]] to <16 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[TMP5]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <4 x float> [[VBFDOT3_I]]
 //
 float32x4_t test_vbfdotq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -48,10 +76,21 @@ float32x4_t test_vbfdotq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b

 // CHECK-LABEL: @test_vbfdot_laneq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTCAST:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <4 x float>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[DOTCAST]], <4 x float> undef, <2 x i32> <i32 3, i32 3>
-// CHECK-NEXT:    [[DOTCAST1:%.*]] = bitcast <2 x float> [[LANE]] to <4 x bfloat>
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.bfdot.v2f32.v4bf16(<2 x float> [[R:%.*]], <4 x bfloat> [[A:%.*]], <4 x bfloat> [[DOTCAST1]]) [[ATTR3]]
+// CHECK-NEXT:    [[__REINT_148:%.*]] = alloca <8 x bfloat>, align 16
+// CHECK-NEXT:    [[__REINT1_148:%.*]] = alloca <2 x float>, align 8
+// CHECK-NEXT:    store <8 x bfloat> [[B:%.*]], <8 x bfloat>* [[__REINT_148]], align 16
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat>* [[__REINT_148]] to <4 x float>*
+// CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[TMP0]], align 16
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <16 x i8>
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP3]], <2 x i32> <i32 3, i32 3>
+// CHECK-NEXT:    store <2 x float> [[LANE]], <2 x float>* [[__REINT1_148]], align 8
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x float>* [[__REINT1_148]] to <4 x bfloat>*
+// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[TMP4]], align 8
+// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x float> [[R:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x bfloat> [[A:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x bfloat> [[TMP5]] to <8 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.bfdot.v2f32.v4bf16(<2 x float> [[R]], <4 x bfloat> [[A]], <4 x bfloat> [[TMP5]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <2 x float> [[VBFDOT3_I]]
 //
 float32x2_t test_vbfdot_laneq_f32(float32x2_t r, bfloat16x4_t a, bfloat16x8_t b) {
@ -60,10 +99,21 @@ float32x2_t test_vbfdot_laneq_f32(float32x2_t r, bfloat16x4_t a, bfloat16x8_t b)

 // CHECK-LABEL: @test_vbfdotq_lane_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTCAST:%.*]] = bitcast <4 x bfloat> [[B:%.*]] to <2 x float>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[DOTCAST]], <2 x float> undef, <4 x i32> zeroinitializer
-// CHECK-NEXT:    [[DOTCAST1:%.*]] = bitcast <4 x float> [[LANE]] to <8 x bfloat>
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[DOTCAST1]]) [[ATTR3]]
+// CHECK-NEXT:    [[__REINT_142:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-NEXT:    [[__REINT1_142:%.*]] = alloca <4 x float>, align 16
+// CHECK-NEXT:    store <4 x bfloat> [[B:%.*]], <4 x bfloat>* [[__REINT_142]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_142]] to <2 x float>*
+// CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[TMP1]] to <8 x i8>
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <2 x float>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK-NEXT:    store <4 x float> [[LANE]], <4 x float>* [[__REINT1_142]], align 16
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x float>* [[__REINT1_142]] to <8 x bfloat>*
+// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[TMP4]], align 16
+// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x bfloat> [[TMP5]] to <16 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfdot.v4f32.v8bf16(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[TMP5]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <4 x float> [[VBFDOT3_I]]
 //
 float32x4_t test_vbfdotq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t b) {
@ -72,7 +122,11 @@ float32x4_t test_vbfdotq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t b)

 // CHECK-LABEL: @test_vbfmmlaq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFMMLAQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmmla(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]]) [[ATTR3]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMMLAQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmmla(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[B]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMMLAQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMMLAQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMMLAQ_V3_I]]
 //
 float32x4_t test_vbfmmlaq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -81,7 +135,11 @@ float32x4_t test_vbfmmlaq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {

 // CHECK-LABEL: @test_vbfmlalbq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalb(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]]) [[ATTR3]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalb(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[B]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALBQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALBQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALBQ_V3_I]]
 //
 float32x4_t test_vbfmlalbq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -90,7 +148,11 @@ float32x4_t test_vbfmlalbq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {

 // CHECK-LABEL: @test_vbfmlaltq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalt(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]]) [[ATTR3]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalt(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[B]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALTQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALTQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALTQ_V3_I]]
 //
 float32x4_t test_vbfmlaltq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -99,8 +161,27 @@ float32x4_t test_vbfmlaltq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {

 // CHECK-LABEL: @test_vbfmlalbq_lane_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VECINIT35:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> undef, <8 x i32> zeroinitializer
-// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalb(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[VECINIT35]]) [[ATTR3]]
+// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[B:%.*]], i32 0
+// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x bfloat> undef, bfloat [[VGET_LANE]], i32 0
+// CHECK-NEXT:    [[VGET_LANE3:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT5:%.*]] = insertelement <8 x bfloat> [[VECINIT]], bfloat [[VGET_LANE3]], i32 1
+// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT10:%.*]] = insertelement <8 x bfloat> [[VECINIT5]], bfloat [[VGET_LANE8]], i32 2
+// CHECK-NEXT:    [[VGET_LANE13:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT15:%.*]] = insertelement <8 x bfloat> [[VECINIT10]], bfloat [[VGET_LANE13]], i32 3
+// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT20:%.*]] = insertelement <8 x bfloat> [[VECINIT15]], bfloat [[VGET_LANE18]], i32 4
+// CHECK-NEXT:    [[VGET_LANE23:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT25:%.*]] = insertelement <8 x bfloat> [[VECINIT20]], bfloat [[VGET_LANE23]], i32 5
+// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT30:%.*]] = insertelement <8 x bfloat> [[VECINIT25]], bfloat [[VGET_LANE28]], i32 6
+// CHECK-NEXT:    [[VGET_LANE33:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT35:%.*]] = insertelement <8 x bfloat> [[VECINIT30]], bfloat [[VGET_LANE33]], i32 7
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[VECINIT35]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalb(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[VECINIT35]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALBQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALBQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALBQ_V3_I]]
 //
 float32x4_t test_vbfmlalbq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t b) {
@ -109,8 +190,27 @@ float32x4_t test_vbfmlalbq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t

 // CHECK-LABEL: @test_vbfmlalbq_laneq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VECINIT35:%.*]] = shufflevector <8 x bfloat> [[B:%.*]], <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalb(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[VECINIT35]]) [[ATTR3]]
+// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[B:%.*]], i32 3
+// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x bfloat> undef, bfloat [[VGETQ_LANE]], i32 0
+// CHECK-NEXT:    [[VGETQ_LANE3:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT5:%.*]] = insertelement <8 x bfloat> [[VECINIT]], bfloat [[VGETQ_LANE3]], i32 1
+// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT10:%.*]] = insertelement <8 x bfloat> [[VECINIT5]], bfloat [[VGETQ_LANE8]], i32 2
+// CHECK-NEXT:    [[VGETQ_LANE13:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT15:%.*]] = insertelement <8 x bfloat> [[VECINIT10]], bfloat [[VGETQ_LANE13]], i32 3
+// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT20:%.*]] = insertelement <8 x bfloat> [[VECINIT15]], bfloat [[VGETQ_LANE18]], i32 4
+// CHECK-NEXT:    [[VGETQ_LANE23:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT25:%.*]] = insertelement <8 x bfloat> [[VECINIT20]], bfloat [[VGETQ_LANE23]], i32 5
+// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT30:%.*]] = insertelement <8 x bfloat> [[VECINIT25]], bfloat [[VGETQ_LANE28]], i32 6
+// CHECK-NEXT:    [[VGETQ_LANE33:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT35:%.*]] = insertelement <8 x bfloat> [[VECINIT30]], bfloat [[VGETQ_LANE33]], i32 7
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[VECINIT35]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalb(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[VECINIT35]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALBQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALBQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALBQ_V3_I]]
 //
 float32x4_t test_vbfmlalbq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -119,8 +219,27 @@ float32x4_t test_vbfmlalbq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t

 // CHECK-LABEL: @test_vbfmlaltq_lane_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VECINIT35:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> undef, <8 x i32> zeroinitializer
-// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalt(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[VECINIT35]]) [[ATTR3]]
+// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[B:%.*]], i32 0
+// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x bfloat> undef, bfloat [[VGET_LANE]], i32 0
+// CHECK-NEXT:    [[VGET_LANE3:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT5:%.*]] = insertelement <8 x bfloat> [[VECINIT]], bfloat [[VGET_LANE3]], i32 1
+// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT10:%.*]] = insertelement <8 x bfloat> [[VECINIT5]], bfloat [[VGET_LANE8]], i32 2
+// CHECK-NEXT:    [[VGET_LANE13:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT15:%.*]] = insertelement <8 x bfloat> [[VECINIT10]], bfloat [[VGET_LANE13]], i32 3
+// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT20:%.*]] = insertelement <8 x bfloat> [[VECINIT15]], bfloat [[VGET_LANE18]], i32 4
+// CHECK-NEXT:    [[VGET_LANE23:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT25:%.*]] = insertelement <8 x bfloat> [[VECINIT20]], bfloat [[VGET_LANE23]], i32 5
+// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT30:%.*]] = insertelement <8 x bfloat> [[VECINIT25]], bfloat [[VGET_LANE28]], i32 6
+// CHECK-NEXT:    [[VGET_LANE33:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT35:%.*]] = insertelement <8 x bfloat> [[VECINIT30]], bfloat [[VGET_LANE33]], i32 7
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[VECINIT35]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalt(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[VECINIT35]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALTQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALTQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALTQ_V3_I]]
 //
 float32x4_t test_vbfmlaltq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t b) {
@ -129,8 +248,27 @@ float32x4_t test_vbfmlaltq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t

 // CHECK-LABEL: @test_vbfmlaltq_laneq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VECINIT35:%.*]] = shufflevector <8 x bfloat> [[B:%.*]], <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalt(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[VECINIT35]]) [[ATTR3]]
+// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[B:%.*]], i32 3
+// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x bfloat> undef, bfloat [[VGETQ_LANE]], i32 0
+// CHECK-NEXT:    [[VGETQ_LANE3:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT5:%.*]] = insertelement <8 x bfloat> [[VECINIT]], bfloat [[VGETQ_LANE3]], i32 1
+// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT10:%.*]] = insertelement <8 x bfloat> [[VECINIT5]], bfloat [[VGETQ_LANE8]], i32 2
+// CHECK-NEXT:    [[VGETQ_LANE13:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT15:%.*]] = insertelement <8 x bfloat> [[VECINIT10]], bfloat [[VGETQ_LANE13]], i32 3
+// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT20:%.*]] = insertelement <8 x bfloat> [[VECINIT15]], bfloat [[VGETQ_LANE18]], i32 4
+// CHECK-NEXT:    [[VGETQ_LANE23:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT25:%.*]] = insertelement <8 x bfloat> [[VECINIT20]], bfloat [[VGETQ_LANE23]], i32 5
+// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT30:%.*]] = insertelement <8 x bfloat> [[VECINIT25]], bfloat [[VGETQ_LANE28]], i32 6
+// CHECK-NEXT:    [[VGETQ_LANE33:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT35:%.*]] = insertelement <8 x bfloat> [[VECINIT30]], bfloat [[VGETQ_LANE33]], i32 7
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[VECINIT35]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.bfmlalt(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[VECINIT35]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALTQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALTQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALTQ_V3_I]]
 //
 float32x4_t test_vbfmlaltq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
--- a/clang/test/CodeGen/aarch64-bf16-getset-intrinsics.c
+++ b/clang/test/CodeGen/aarch64-bf16-getset-intrinsics.c
@ -1,6 +1,6 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -triple aarch64-arm-none-eabi -target-feature +neon -target-feature +bf16 \
-// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg -instcombine | FileCheck %s
+// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg | FileCheck %s

 #include <arm_neon.h>

@ -16,7 +16,9 @@ bfloat16x4_t test_vcreate_bf16(uint64_t a) {
 // CHECK-LABEL: @test_vdup_n_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x bfloat> undef, bfloat [[V:%.*]], i32 0
-// CHECK-NEXT:    [[VECINIT3_I:%.*]] = shufflevector <4 x bfloat> [[VECINIT_I]], <4 x bfloat> poison, <4 x i32> zeroinitializer
+// CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x bfloat> [[VECINIT_I]], bfloat [[V]], i32 1
+// CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x bfloat> [[VECINIT1_I]], bfloat [[V]], i32 2
+// CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x bfloat> [[VECINIT2_I]], bfloat [[V]], i32 3
 // CHECK-NEXT:    ret <4 x bfloat> [[VECINIT3_I]]
 //
 bfloat16x4_t test_vdup_n_bf16(bfloat16_t v) {
@ -26,7 +28,13 @@ bfloat16x4_t test_vdup_n_bf16(bfloat16_t v) {
 // CHECK-LABEL: @test_vdupq_n_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x bfloat> undef, bfloat [[V:%.*]], i32 0
-// CHECK-NEXT:    [[VECINIT7_I:%.*]] = shufflevector <8 x bfloat> [[VECINIT_I]], <8 x bfloat> poison, <8 x i32> zeroinitializer
+// CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x bfloat> [[VECINIT_I]], bfloat [[V]], i32 1
+// CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x bfloat> [[VECINIT1_I]], bfloat [[V]], i32 2
+// CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x bfloat> [[VECINIT2_I]], bfloat [[V]], i32 3
+// CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x bfloat> [[VECINIT3_I]], bfloat [[V]], i32 4
+// CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x bfloat> [[VECINIT4_I]], bfloat [[V]], i32 5
+// CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x bfloat> [[VECINIT5_I]], bfloat [[V]], i32 6
+// CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x bfloat> [[VECINIT6_I]], bfloat [[V]], i32 7
 // CHECK-NEXT:    ret <8 x bfloat> [[VECINIT7_I]]
 //
 bfloat16x8_t test_vdupq_n_bf16(bfloat16_t v) {
@ -35,7 +43,9 @@ bfloat16x8_t test_vdupq_n_bf16(bfloat16_t v) {

 // CHECK-LABEL: @test_vdup_lane_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x bfloat> [[V:%.*]], <4 x bfloat> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat> [[V:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x bfloat>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x bfloat> [[TMP1]], <4 x bfloat> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    ret <4 x bfloat> [[LANE]]
 //
 bfloat16x4_t test_vdup_lane_bf16(bfloat16x4_t v) {
@ -44,7 +54,9 @@ bfloat16x4_t test_vdup_lane_bf16(bfloat16x4_t v) {

 // CHECK-LABEL: @test_vdupq_lane_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x bfloat> [[V:%.*]], <4 x bfloat> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat> [[V:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x bfloat>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x bfloat> [[TMP1]], <4 x bfloat> [[TMP1]], <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    ret <8 x bfloat> [[LANE]]
 //
 bfloat16x8_t test_vdupq_lane_bf16(bfloat16x4_t v) {
@ -53,7 +65,9 @@ bfloat16x8_t test_vdupq_lane_bf16(bfloat16x4_t v) {

 // CHECK-LABEL: @test_vdup_laneq_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x bfloat> [[V:%.*]], <8 x bfloat> undef, <4 x i32> <i32 7, i32 7, i32 7, i32 7>
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat> [[V:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x bfloat>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x bfloat> [[TMP1]], <8 x bfloat> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    ret <4 x bfloat> [[LANE]]
 //
 bfloat16x4_t test_vdup_laneq_bf16(bfloat16x8_t v) {
@ -62,7 +76,9 @@ bfloat16x4_t test_vdup_laneq_bf16(bfloat16x8_t v) {

 // CHECK-LABEL: @test_vdupq_laneq_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x bfloat> [[V:%.*]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat> [[V:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x bfloat>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x bfloat> [[TMP1]], <8 x bfloat> [[TMP1]], <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    ret <8 x bfloat> [[LANE]]
 //
 bfloat16x8_t test_vdupq_laneq_bf16(bfloat16x8_t v) {
@ -80,7 +96,7 @@ bfloat16x8_t test_vcombine_bf16(bfloat16x4_t low, bfloat16x4_t high) {

 // CHECK-LABEL: @test_vget_high_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    ret <4 x bfloat> [[SHUFFLE_I]]
 //
 bfloat16x4_t test_vget_high_bf16(bfloat16x8_t a) {
@ -89,7 +105,7 @@ bfloat16x4_t test_vget_high_bf16(bfloat16x8_t a) {

 // CHECK-LABEL: @test_vget_low_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 // CHECK-NEXT:    ret <4 x bfloat> [[SHUFFLE_I]]
 //
 bfloat16x4_t test_vget_low_bf16(bfloat16x8_t a) {
--- a/clang/test/CodeGen/aarch64-bf16-lane-intrinsics.c
+++ b/clang/test/CodeGen/aarch64-bf16-lane-intrinsics.c
@ -1,21 +1,24 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -triple aarch64-arm-none-eabi -target-feature +neon -target-feature +bf16 \
-// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg -instcombine | FileCheck --check-prefix=CHECK-LE %s
+// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg | FileCheck --check-prefix=CHECK-LE %s
 // RUN: %clang_cc1 -triple aarch64_be-arm-none-eabi -target-feature +neon -target-feature +bf16 \
-// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg -instcombine | FileCheck --check-prefix=CHECK-BE %s
+// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg | FileCheck --check-prefix=CHECK-BE %s

 #include <arm_neon.h>

 // CHECK-LE-LABEL: @test_vcopy_lane_bf16_v1(
 // CHECK-LE-NEXT:  entry:
-// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> [[B:%.*]], <4 x i32> <i32 0, i32 7, i32 2, i32 3>
+// CHECK-LE-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[B:%.*]], i32 3
+// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = insertelement <4 x bfloat> [[A:%.*]], bfloat [[VGET_LANE]], i32 1
 // CHECK-LE-NEXT:    ret <4 x bfloat> [[VSET_LANE]]
 //
 // CHECK-BE-LABEL: @test_vcopy_lane_bf16_v1(
 // CHECK-BE-NEXT:  entry:
-// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> undef, <4 x i32> <i32 3, i32 undef, i32 1, i32 0>
-// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <4 x bfloat> [[SHUFFLE]], <4 x bfloat> [[B:%.*]], <4 x i32> <i32 0, i32 4, i32 2, i32 3>
-// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <4 x bfloat> [[VSET_LANE]], <4 x bfloat> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> [[A]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> [[B]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[SHUFFLE1]], i32 3
+// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = insertelement <4 x bfloat> [[SHUFFLE]], bfloat [[VGET_LANE]], i32 1
+// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <4 x bfloat> [[VSET_LANE]], <4 x bfloat> [[VSET_LANE]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 // CHECK-BE-NEXT:    ret <4 x bfloat> [[SHUFFLE5]]
 //
 bfloat16x4_t test_vcopy_lane_bf16_v1(bfloat16x4_t a, bfloat16x4_t b) {
@ -24,14 +27,17 @@ bfloat16x4_t test_vcopy_lane_bf16_v1(bfloat16x4_t a, bfloat16x4_t b) {

 // CHECK-LE-LABEL: @test_vcopy_lane_bf16_v2(
 // CHECK-LE-NEXT:  entry:
-// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> [[B:%.*]], <4 x i32> <i32 0, i32 1, i32 4, i32 3>
+// CHECK-LE-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[B:%.*]], i32 0
+// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = insertelement <4 x bfloat> [[A:%.*]], bfloat [[VGET_LANE]], i32 2
 // CHECK-LE-NEXT:    ret <4 x bfloat> [[VSET_LANE]]
 //
 // CHECK-BE-LABEL: @test_vcopy_lane_bf16_v2(
 // CHECK-BE-NEXT:  entry:
-// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> undef, <4 x i32> <i32 3, i32 2, i32 undef, i32 0>
-// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <4 x bfloat> [[SHUFFLE]], <4 x bfloat> [[B:%.*]], <4 x i32> <i32 0, i32 1, i32 7, i32 3>
-// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <4 x bfloat> [[VSET_LANE]], <4 x bfloat> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> [[A]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> [[B]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[SHUFFLE1]], i32 0
+// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = insertelement <4 x bfloat> [[SHUFFLE]], bfloat [[VGET_LANE]], i32 2
+// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <4 x bfloat> [[VSET_LANE]], <4 x bfloat> [[VSET_LANE]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 // CHECK-BE-NEXT:    ret <4 x bfloat> [[SHUFFLE5]]
 //
 bfloat16x4_t test_vcopy_lane_bf16_v2(bfloat16x4_t a, bfloat16x4_t b) {
@ -40,16 +46,17 @@ bfloat16x4_t test_vcopy_lane_bf16_v2(bfloat16x4_t a, bfloat16x4_t b) {

 // CHECK-LE-LABEL: @test_vcopyq_lane_bf16_v1(
 // CHECK-LE-NEXT:  entry:
-// CHECK-LE-NEXT:    [[TMP0:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> poison, <8 x i32> <i32 undef, i32 undef, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[TMP0]], <8 x i32> <i32 10, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+// CHECK-LE-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[B:%.*]], i32 2
+// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = insertelement <8 x bfloat> [[A:%.*]], bfloat [[VGET_LANE]], i32 0
 // CHECK-LE-NEXT:    ret <8 x bfloat> [[VSET_LANE]]
 //
 // CHECK-BE-LABEL: @test_vcopyq_lane_bf16_v1(
 // CHECK-BE-NEXT:  entry:
-// CHECK-BE-NEXT:    [[TMP0:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> poison, <8 x i32> <i32 undef, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> undef, <8 x i32> <i32 undef, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <8 x bfloat> [[SHUFFLE]], <8 x bfloat> [[TMP0]], <8 x i32> <i32 9, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <8 x bfloat> [[VSET_LANE]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> [[B]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[SHUFFLE1]], i32 2
+// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = insertelement <8 x bfloat> [[SHUFFLE]], bfloat [[VGET_LANE]], i32 0
+// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <8 x bfloat> [[VSET_LANE]], <8 x bfloat> [[VSET_LANE]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 // CHECK-BE-NEXT:    ret <8 x bfloat> [[SHUFFLE5]]
 //
 bfloat16x8_t test_vcopyq_lane_bf16_v1(bfloat16x8_t a, bfloat16x4_t b) {
@ -58,16 +65,17 @@ bfloat16x8_t test_vcopyq_lane_bf16_v1(bfloat16x8_t a, bfloat16x4_t b) {

 // CHECK-LE-LABEL: @test_vcopyq_lane_bf16_v2(
 // CHECK-LE-NEXT:  entry:
-// CHECK-LE-NEXT:    [[TMP0:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> poison, <8 x i32> <i32 0, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
-// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[TMP0]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 8, i32 7>
+// CHECK-LE-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[B:%.*]], i32 0
+// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = insertelement <8 x bfloat> [[A:%.*]], bfloat [[VGET_LANE]], i32 6
 // CHECK-LE-NEXT:    ret <8 x bfloat> [[VSET_LANE]]
 //
 // CHECK-BE-LABEL: @test_vcopyq_lane_bf16_v2(
 // CHECK-BE-NEXT:  entry:
-// CHECK-BE-NEXT:    [[TMP0:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> poison, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
-// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 undef, i32 0>
-// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <8 x bfloat> [[SHUFFLE]], <8 x bfloat> [[TMP0]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 11, i32 7>
-// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <8 x bfloat> [[VSET_LANE]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> [[B]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[SHUFFLE1]], i32 0
+// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = insertelement <8 x bfloat> [[SHUFFLE]], bfloat [[VGET_LANE]], i32 6
+// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <8 x bfloat> [[VSET_LANE]], <8 x bfloat> [[VSET_LANE]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 // CHECK-BE-NEXT:    ret <8 x bfloat> [[SHUFFLE5]]
 //
 bfloat16x8_t test_vcopyq_lane_bf16_v2(bfloat16x8_t a, bfloat16x4_t b) {
@ -82,10 +90,11 @@ bfloat16x8_t test_vcopyq_lane_bf16_v2(bfloat16x8_t a, bfloat16x4_t b) {
 //
 // CHECK-BE-LABEL: @test_vcopy_laneq_bf16_v1(
 // CHECK-BE-NEXT:  entry:
-// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> undef, <4 x i32> <i32 undef, i32 2, i32 1, i32 0>
-// CHECK-BE-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[B:%.*]], i32 0
+// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> [[A]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <8 x bfloat> [[B:%.*]], <8 x bfloat> [[B]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[SHUFFLE1]], i32 7
 // CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = insertelement <4 x bfloat> [[SHUFFLE]], bfloat [[VGETQ_LANE]], i32 0
-// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <4 x bfloat> [[VSET_LANE]], <4 x bfloat> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <4 x bfloat> [[VSET_LANE]], <4 x bfloat> [[VSET_LANE]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 // CHECK-BE-NEXT:    ret <4 x bfloat> [[SHUFFLE5]]
 //
 bfloat16x4_t test_vcopy_laneq_bf16_v1(bfloat16x4_t a, bfloat16x8_t b) {
@ -100,10 +109,11 @@ bfloat16x4_t test_vcopy_laneq_bf16_v1(bfloat16x4_t a, bfloat16x8_t b) {
 //
 // CHECK-BE-LABEL: @test_vcopy_laneq_bf16_v2(
 // CHECK-BE-NEXT:  entry:
-// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 undef>
-// CHECK-BE-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[B:%.*]], i32 3
+// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x bfloat> [[A:%.*]], <4 x bfloat> [[A]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <8 x bfloat> [[B:%.*]], <8 x bfloat> [[B]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[SHUFFLE1]], i32 4
 // CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = insertelement <4 x bfloat> [[SHUFFLE]], bfloat [[VGETQ_LANE]], i32 3
-// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <4 x bfloat> [[VSET_LANE]], <4 x bfloat> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <4 x bfloat> [[VSET_LANE]], <4 x bfloat> [[VSET_LANE]], <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 // CHECK-BE-NEXT:    ret <4 x bfloat> [[SHUFFLE5]]
 //
 bfloat16x4_t test_vcopy_laneq_bf16_v2(bfloat16x4_t a, bfloat16x8_t b) {
@ -112,14 +122,17 @@ bfloat16x4_t test_vcopy_laneq_bf16_v2(bfloat16x4_t a, bfloat16x8_t b) {

 // CHECK-LE-LABEL: @test_vcopyq_laneq_bf16_v1(
 // CHECK-LE-NEXT:  entry:
-// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 15, i32 4, i32 5, i32 6, i32 7>
+// CHECK-LE-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[B:%.*]], i32 7
+// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = insertelement <8 x bfloat> [[A:%.*]], bfloat [[VGETQ_LANE]], i32 3
 // CHECK-LE-NEXT:    ret <8 x bfloat> [[VSET_LANE]]
 //
 // CHECK-BE-LABEL: @test_vcopyq_laneq_bf16_v1(
 // CHECK-BE-NEXT:  entry:
-// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 undef, i32 3, i32 2, i32 1, i32 0>
-// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <8 x bfloat> [[SHUFFLE]], <8 x bfloat> [[B:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 8, i32 4, i32 5, i32 6, i32 7>
-// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <8 x bfloat> [[VSET_LANE]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <8 x bfloat> [[B:%.*]], <8 x bfloat> [[B]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[SHUFFLE1]], i32 7
+// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = insertelement <8 x bfloat> [[SHUFFLE]], bfloat [[VGETQ_LANE]], i32 3
+// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <8 x bfloat> [[VSET_LANE]], <8 x bfloat> [[VSET_LANE]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 // CHECK-BE-NEXT:    ret <8 x bfloat> [[SHUFFLE5]]
 //
 bfloat16x8_t test_vcopyq_laneq_bf16_v1(bfloat16x8_t a, bfloat16x8_t b) {
@ -129,14 +142,17 @@ bfloat16x8_t test_vcopyq_laneq_bf16_v1(bfloat16x8_t a, bfloat16x8_t b) {

 // CHECK-LE-LABEL: @test_vcopyq_laneq_bf16_v2(
 // CHECK-LE-NEXT:  entry:
-// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 10, i32 7>
+// CHECK-LE-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[B:%.*]], i32 2
+// CHECK-LE-NEXT:    [[VSET_LANE:%.*]] = insertelement <8 x bfloat> [[A:%.*]], bfloat [[VGETQ_LANE]], i32 6
 // CHECK-LE-NEXT:    ret <8 x bfloat> [[VSET_LANE]]
 //
 // CHECK-BE-LABEL: @test_vcopyq_laneq_bf16_v2(
 // CHECK-BE-NEXT:  entry:
-// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 undef, i32 0>
-// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = shufflevector <8 x bfloat> [[SHUFFLE]], <8 x bfloat> [[B:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 13, i32 7>
-// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <8 x bfloat> [[VSET_LANE]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <8 x bfloat> [[B:%.*]], <8 x bfloat> [[B]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+// CHECK-BE-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x bfloat> [[SHUFFLE1]], i32 2
+// CHECK-BE-NEXT:    [[VSET_LANE:%.*]] = insertelement <8 x bfloat> [[SHUFFLE]], bfloat [[VGETQ_LANE]], i32 6
+// CHECK-BE-NEXT:    [[SHUFFLE5:%.*]] = shufflevector <8 x bfloat> [[VSET_LANE]], <8 x bfloat> [[VSET_LANE]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 // CHECK-BE-NEXT:    ret <8 x bfloat> [[SHUFFLE5]]
 //
 bfloat16x8_t test_vcopyq_laneq_bf16_v2(bfloat16x8_t a, bfloat16x8_t b) {
--- a/clang/test/CodeGen/arm-bf16-convert-intrinsics.c
+++ b/clang/test/CodeGen/arm-bf16-convert-intrinsics.c
@ -1,87 +1,509 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 \
 // RUN:   -triple aarch64-arm-none-eabi -target-feature +neon -target-feature +bf16 \
 // RUN:   -disable-O0-optnone -emit-llvm -o - %s \
-// RUN:   | opt -S -mem2reg -instcombine \
+// RUN:   | opt -S -mem2reg \
 // RUN:   | FileCheck --check-prefixes=CHECK,CHECK-A64 %s
 // RUN: %clang_cc1 \
 // RUN:   -triple armv8.6a-arm-none-eabi -target-feature +neon \
 // RUN:   -target-feature +bf16 -mfloat-abi hard \
 // RUN:   -disable-O0-optnone -emit-llvm -o - %s \
-// RUN:   | opt -S -mem2reg -instcombine \
+// RUN:   | opt -S -mem2reg \
 // RUN:   | FileCheck --check-prefixes=CHECK,CHECK-A32-HARDFP %s
 // RUN: %clang_cc1 \
 // RUN:   -triple armv8.6a-arm-none-eabi -target-feature +neon \
 // RUN:   -target-feature +bf16 -mfloat-abi softfp \
 // RUN:   -disable-O0-optnone -emit-llvm -o - %s \
-// RUN:   | opt -S -mem2reg -instcombine \
+// RUN:   | opt -S -mem2reg \
 // RUN:   | FileCheck --check-prefixes=CHECK,CHECK-A32-SOFTFP %s

 #include <arm_neon.h>

-// CHECK-LABEL: test_vcvt_f32_bf16
-// CHECK: %[[EXT:.*]] = zext <4 x i16> %{{.*}} to <4 x i32>
-// CHECK: shl nuw <4 x i32> %[[EXT]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A64-LABEL: @test_vcvt_f32_bf16(
+// CHECK-A64-NEXT:  entry:
+// CHECK-A64-NEXT:    [[__REINT_150_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A64-NEXT:    [[__REINT1_150_I:%.*]] = alloca <4 x i32>, align 16
+// CHECK-A64-NEXT:    store <4 x bfloat> [[A:%.*]], <4 x bfloat>* [[__REINT_150_I]], align 8
+// CHECK-A64-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I]] to <4 x i16>*
+// CHECK-A64-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
+// CHECK-A64-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
+// CHECK-A64-NEXT:    [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
+// CHECK-A64-NEXT:    [[VSHLL_N_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A64-NEXT:    store <4 x i32> [[VSHLL_N_I]], <4 x i32>* [[__REINT1_150_I]], align 16
+// CHECK-A64-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I]] to <4 x float>*
+// CHECK-A64-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 16
+// CHECK-A64-NEXT:    ret <4 x float> [[TMP5]]
+//
+// CHECK-A32-HARDFP-LABEL: @test_vcvt_f32_bf16(
+// CHECK-A32-HARDFP-NEXT:  entry:
+// CHECK-A32-HARDFP-NEXT:    [[__REINT_150_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-HARDFP-NEXT:    [[__REINT1_150_I:%.*]] = alloca <4 x i32>, align 8
+// CHECK-A32-HARDFP-NEXT:    store <4 x bfloat> [[A:%.*]], <4 x bfloat>* [[__REINT_150_I]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I]] to <4 x i16>*
+// CHECK-A32-HARDFP-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
+// CHECK-A32-HARDFP-NEXT:    [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
+// CHECK-A32-HARDFP-NEXT:    [[VSHLL_N_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A32-HARDFP-NEXT:    store <4 x i32> [[VSHLL_N_I]], <4 x i32>* [[__REINT1_150_I]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I]] to <4 x float>*
+// CHECK-A32-HARDFP-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 8
+// CHECK-A32-HARDFP-NEXT:    ret <4 x float> [[TMP5]]
+//
+// CHECK-A32-SOFTFP-LABEL: @test_vcvt_f32_bf16(
+// CHECK-A32-SOFTFP-NEXT:  entry:
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_150_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__REINT_150_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__REINT1_150_I:%.*]] = alloca <4 x i32>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[A:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[A]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[A_COERCE:%.*]], <2 x i32>* [[TMP0]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[A1:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[A]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[A1]], <4 x bfloat>* [[COERCE]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP1:%.*]] = bitcast <4 x bfloat>* [[COERCE]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* [[TMP1]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP3:%.*]] = bitcast <4 x bfloat>* [[__P0_150_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP2]], <2 x i32>* [[TMP3]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_1501_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_150_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[__P0_1501_I]], <4 x bfloat>* [[__REINT_150_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP4:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I]] to <4 x i16>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[TMP4]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK-A32-SOFTFP-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[TMP5]] to <4 x i32>
+// CHECK-A32-SOFTFP-NEXT:    [[VSHLL_N_I:%.*]] = shl <4 x i32> [[TMP7]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[VSHLL_N_I]], <4 x i32>* [[__REINT1_150_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP8:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I]] to <4 x float>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP9:%.*]] = load <4 x float>, <4 x float>* [[TMP8]], align 8
+// CHECK-A32-SOFTFP-NEXT:    ret <4 x float> [[TMP9]]
+//
 float32x4_t test_vcvt_f32_bf16(bfloat16x4_t a) {
  return vcvt_f32_bf16(a);
 }

-// CHECK-LABEL: test_vcvtq_low_f32_bf16
-// CHECK: shufflevector <8 x bfloat> %{{.*}}, <8 x bfloat> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-// CHECK: %[[EXT:.*]] = zext <4 x i16> %{{.*}} to <4 x i32>
-// CHECK: shl nuw <4 x i32> %[[EXT]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A64-LABEL: @test_vcvtq_low_f32_bf16(
+// CHECK-A64-NEXT:  entry:
+// CHECK-A64-NEXT:    [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A64-NEXT:    [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 16
+// CHECK-A64-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+// CHECK-A64-NEXT:    store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
+// CHECK-A64-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
+// CHECK-A64-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
+// CHECK-A64-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
+// CHECK-A64-NEXT:    [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
+// CHECK-A64-NEXT:    [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A64-NEXT:    store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 16
+// CHECK-A64-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
+// CHECK-A64-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 16
+// CHECK-A64-NEXT:    ret <4 x float> [[TMP5]]
+//
+// CHECK-A32-HARDFP-LABEL: @test_vcvtq_low_f32_bf16(
+// CHECK-A32-HARDFP-NEXT:  entry:
+// CHECK-A32-HARDFP-NEXT:    [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-HARDFP-NEXT:    [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 8
+// CHECK-A32-HARDFP-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+// CHECK-A32-HARDFP-NEXT:    store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
+// CHECK-A32-HARDFP-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
+// CHECK-A32-HARDFP-NEXT:    [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
+// CHECK-A32-HARDFP-NEXT:    [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A32-HARDFP-NEXT:    store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
+// CHECK-A32-HARDFP-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 8
+// CHECK-A32-HARDFP-NEXT:    ret <4 x float> [[TMP5]]
+//
+// CHECK-A32-SOFTFP-LABEL: @test_vcvtq_low_f32_bf16(
+// CHECK-A32-SOFTFP-NEXT:  entry:
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_I2:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE2_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE3_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[A:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat>* [[A]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[A_COERCE:%.*]], <4 x i32>* [[TMP0]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[A1:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[A]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[A1]], <8 x bfloat>* [[COERCE]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat>* [[COERCE]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP3:%.*]] = bitcast <8 x bfloat>* [[__P0_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P01_I:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[__P01_I]], <8 x bfloat>* [[COERCE_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP4:%.*]] = bitcast <8 x bfloat>* [[COERCE_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[TMP4]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP6:%.*]] = bitcast <8 x bfloat>* [[__P0_I2]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P01_I5:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I2]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[__P01_I5]], <8 x bfloat> [[__P01_I5]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[RETVAL_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP7:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[TMP7]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP9:%.*]] = bitcast <4 x bfloat>* [[COERCE2_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP8]], <2 x i32>* [[TMP9]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP10:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE2_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP10]], <4 x bfloat>* [[COERCE3_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP11:%.*]] = bitcast <4 x bfloat>* [[COERCE3_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP12:%.*]] = load <2 x i32>, <2 x i32>* [[TMP11]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP13:%.*]] = bitcast <4 x bfloat>* [[__P0_150_I_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP12]], <2 x i32>* [[TMP13]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_1501_I_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_150_I_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[__P0_1501_I_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP14:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP15:%.*]] = load <4 x i16>, <4 x i16>* [[TMP14]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP16:%.*]] = bitcast <4 x i16> [[TMP15]] to <8 x i8>
+// CHECK-A32-SOFTFP-NEXT:    [[TMP17:%.*]] = sext <4 x i16> [[TMP15]] to <4 x i32>
+// CHECK-A32-SOFTFP-NEXT:    [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP17]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP18:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP19:%.*]] = load <4 x float>, <4 x float>* [[TMP18]], align 8
+// CHECK-A32-SOFTFP-NEXT:    ret <4 x float> [[TMP19]]
+//
 float32x4_t test_vcvtq_low_f32_bf16(bfloat16x8_t a) {
  return vcvtq_low_f32_bf16(a);
 }

-// CHECK-LABEL: test_vcvtq_high_f32_bf16
-// CHECK: shufflevector <8 x bfloat> %{{.*}}, <8 x bfloat> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-// CHECK: %[[EXT:.*]] = zext <4 x i16> %{{.*}} to <4 x i32>
-// CHECK: shl nuw <4 x i32> %[[EXT]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A64-LABEL: @test_vcvtq_high_f32_bf16(
+// CHECK-A64-NEXT:  entry:
+// CHECK-A64-NEXT:    [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A64-NEXT:    [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 16
+// CHECK-A64-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+// CHECK-A64-NEXT:    store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
+// CHECK-A64-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
+// CHECK-A64-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
+// CHECK-A64-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
+// CHECK-A64-NEXT:    [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
+// CHECK-A64-NEXT:    [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A64-NEXT:    store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 16
+// CHECK-A64-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
+// CHECK-A64-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 16
+// CHECK-A64-NEXT:    ret <4 x float> [[TMP5]]
+//
+// CHECK-A32-HARDFP-LABEL: @test_vcvtq_high_f32_bf16(
+// CHECK-A32-HARDFP-NEXT:  entry:
+// CHECK-A32-HARDFP-NEXT:    [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-HARDFP-NEXT:    [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 8
+// CHECK-A32-HARDFP-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+// CHECK-A32-HARDFP-NEXT:    store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
+// CHECK-A32-HARDFP-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
+// CHECK-A32-HARDFP-NEXT:    [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
+// CHECK-A32-HARDFP-NEXT:    [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A32-HARDFP-NEXT:    store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 8
+// CHECK-A32-HARDFP-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
+// CHECK-A32-HARDFP-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 8
+// CHECK-A32-HARDFP-NEXT:    ret <4 x float> [[TMP5]]
+//
+// CHECK-A32-SOFTFP-LABEL: @test_vcvtq_high_f32_bf16(
+// CHECK-A32-SOFTFP-NEXT:  entry:
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_I2:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE2_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE3_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[A:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat>* [[A]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[A_COERCE:%.*]], <4 x i32>* [[TMP0]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[A1:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[A]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[A1]], <8 x bfloat>* [[COERCE]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat>* [[COERCE]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP3:%.*]] = bitcast <8 x bfloat>* [[__P0_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P01_I:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[__P01_I]], <8 x bfloat>* [[COERCE_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP4:%.*]] = bitcast <8 x bfloat>* [[COERCE_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[TMP4]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP6:%.*]] = bitcast <8 x bfloat>* [[__P0_I2]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P01_I5:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I2]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[__P01_I5]], <8 x bfloat> [[__P01_I5]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[RETVAL_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP7:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[TMP7]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP9:%.*]] = bitcast <4 x bfloat>* [[COERCE2_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP8]], <2 x i32>* [[TMP9]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP10:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE2_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP10]], <4 x bfloat>* [[COERCE3_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP11:%.*]] = bitcast <4 x bfloat>* [[COERCE3_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP12:%.*]] = load <2 x i32>, <2 x i32>* [[TMP11]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP13:%.*]] = bitcast <4 x bfloat>* [[__P0_150_I_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP12]], <2 x i32>* [[TMP13]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_1501_I_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_150_I_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[__P0_1501_I_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP14:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP15:%.*]] = load <4 x i16>, <4 x i16>* [[TMP14]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP16:%.*]] = bitcast <4 x i16> [[TMP15]] to <8 x i8>
+// CHECK-A32-SOFTFP-NEXT:    [[TMP17:%.*]] = sext <4 x i16> [[TMP15]] to <4 x i32>
+// CHECK-A32-SOFTFP-NEXT:    [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP17]], <i32 16, i32 16, i32 16, i32 16>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP18:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP19:%.*]] = load <4 x float>, <4 x float>* [[TMP18]], align 8
+// CHECK-A32-SOFTFP-NEXT:    ret <4 x float> [[TMP19]]
+//
 float32x4_t test_vcvtq_high_f32_bf16(bfloat16x8_t a) {
  return vcvtq_high_f32_bf16(a);
 }

-// CHECK-LABEL: test_vcvt_bf16_f32
-// CHECK-A64: %[[CVT:.*]] = call <8 x bfloat> @llvm.aarch64.neon.bfcvtn(<4 x float> %a)
-// CHECK-A64: shufflevector <8 x bfloat> %[[CVT]], <8 x bfloat> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-// CHECK-A32-HARDFP: call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16(<4 x float> %a)
-// CHECK-A32-SOFTFP: call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16(<4 x float> %a)
+// CHECK-A64-LABEL: @test_vcvt_bf16_f32(
+// CHECK-A64-NEXT:  entry:
+// CHECK-A64-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A64-NEXT:    [[__A64_VCVTQ_LOW_BF16_V1_I:%.*]] = call <8 x bfloat> @llvm.aarch64.neon.bfcvtn(<4 x float> [[A]]) #[[ATTR3:[0-9]+]]
+// CHECK-A64-NEXT:    [[__A64_VCVTQ_LOW_BF16_V2_I:%.*]] = bitcast <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]] to <16 x i8>
+// CHECK-A64-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]], <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+// CHECK-A64-NEXT:    ret <4 x bfloat> [[SHUFFLE_I]]
+//
+// CHECK-A32-HARDFP-LABEL: @test_vcvt_bf16_f32(
+// CHECK-A32-HARDFP-NEXT:  entry:
+// CHECK-A32-HARDFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A32-HARDFP-NEXT:    [[VCVTFP2BF1_I:%.*]] = call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16(<4 x float> [[A]]) #[[ATTR3:[0-9]+]]
+// CHECK-A32-HARDFP-NEXT:    ret <4 x bfloat> [[VCVTFP2BF1_I]]
+//
+// CHECK-A32-SOFTFP-LABEL: @test_vcvt_bf16_f32(
+// CHECK-A32-SOFTFP-NEXT:  entry:
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I1:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A32-SOFTFP-NEXT:    [[VCVTFP2BF1_I:%.*]] = call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16(<4 x float> [[A]]) #[[ATTR3:[0-9]+]]
+// CHECK-A32-SOFTFP-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VCVTFP2BF1_I]] to <4 x bfloat>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP1]], <4 x bfloat>* [[RETVAL_I1]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP2:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I1]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[TMP2]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP4:%.*]] = bitcast <4 x bfloat>* [[COERCE_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP3]], <2 x i32>* [[TMP4]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP5:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP5]], <4 x bfloat>* [[RETVAL_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP6:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[TMP6]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP8:%.*]] = bitcast <4 x bfloat>* [[COERCE]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP7]], <2 x i32>* [[TMP8]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP9:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP9]], <4 x bfloat>* [[RETVAL]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP10:%.*]] = bitcast <4 x bfloat>* [[RETVAL]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP11:%.*]] = load <2 x i32>, <2 x i32>* [[TMP10]], align 8
+// CHECK-A32-SOFTFP-NEXT:    ret <2 x i32> [[TMP11]]
+//
 bfloat16x4_t test_vcvt_bf16_f32(float32x4_t a) {
  return vcvt_bf16_f32(a);
 }

-// CHECK-LABEL: test_vcvtq_low_bf16_f32
-// CHECK-A64: call <8 x bfloat> @llvm.aarch64.neon.bfcvtn(<4 x float> %a)
-// CHECK-A32-HARDFP: %[[CVT:.*]] = call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16
-// CHECK-A32-HARDFP: shufflevector <4 x bfloat> zeroinitializer, <4 x bfloat> %[[CVT]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-// CHECK-A32-SOFTFP: call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16
-// CHECK-A32-SOFTFP: shufflevector <4 x bfloat> zeroinitializer, <4 x bfloat> %{{.*}}, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+// CHECK-A64-LABEL: @test_vcvtq_low_bf16_f32(
+// CHECK-A64-NEXT:  entry:
+// CHECK-A64-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A64-NEXT:    [[__A64_VCVTQ_LOW_BF16_V1_I:%.*]] = call <8 x bfloat> @llvm.aarch64.neon.bfcvtn(<4 x float> [[A]]) #[[ATTR3]]
+// CHECK-A64-NEXT:    [[__A64_VCVTQ_LOW_BF16_V2_I:%.*]] = bitcast <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]] to <16 x i8>
+// CHECK-A64-NEXT:    ret <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]]
+//
+// CHECK-A32-HARDFP-LABEL: @test_vcvtq_low_bf16_f32(
+// CHECK-A32-HARDFP-NEXT:  entry:
+// CHECK-A32-HARDFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A32-HARDFP-NEXT:    [[VCVTFP2BF1_I:%.*]] = call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16(<4 x float> [[A]]) #[[ATTR3]]
+// CHECK-A32-HARDFP-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x bfloat> zeroinitializer, <4 x bfloat> [[VCVTFP2BF1_I]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+// CHECK-A32-HARDFP-NEXT:    ret <8 x bfloat> [[SHUFFLE_I]]
+//
+// CHECK-A32-SOFTFP-LABEL: @test_vcvtq_low_bf16_f32(
+// CHECK-A32-SOFTFP-NEXT:  entry:
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I4:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P1_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I1:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE1_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE2_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE4_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A32-SOFTFP-NEXT:    [[VCVTFP2BF1_I:%.*]] = call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16(<4 x float> [[A]]) #[[ATTR3]]
+// CHECK-A32-SOFTFP-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VCVTFP2BF1_I]] to <4 x bfloat>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP1]], <4 x bfloat>* [[RETVAL_I1]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP2:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I1]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[TMP2]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP4:%.*]] = bitcast <4 x bfloat>* [[COERCE_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP3]], <2 x i32>* [[TMP4]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP5:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> zeroinitializer, <4 x bfloat>* [[COERCE1_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP6:%.*]] = bitcast <4 x bfloat>* [[COERCE1_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[TMP6]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP5]], <4 x bfloat>* [[COERCE2_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP8:%.*]] = bitcast <4 x bfloat>* [[COERCE2_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP9:%.*]] = load <2 x i32>, <2 x i32>* [[TMP8]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP10:%.*]] = bitcast <4 x bfloat>* [[__P0_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP7]], <2 x i32>* [[TMP10]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P01_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP11:%.*]] = bitcast <4 x bfloat>* [[__P1_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP9]], <2 x i32>* [[TMP11]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P12_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P1_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x bfloat> [[__P01_I]], <4 x bfloat> [[__P12_I]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[SHUFFLE_I]], <8 x bfloat>* [[RETVAL_I4]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP12:%.*]] = bitcast <8 x bfloat>* [[RETVAL_I4]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP13:%.*]] = load <4 x i32>, <4 x i32>* [[TMP12]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP14:%.*]] = bitcast <8 x bfloat>* [[COERCE4_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP13]], <4 x i32>* [[TMP14]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP15:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[COERCE4_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[TMP15]], <8 x bfloat>* [[RETVAL_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP16:%.*]] = bitcast <8 x bfloat>* [[RETVAL_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP17:%.*]] = load <4 x i32>, <4 x i32>* [[TMP16]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP18:%.*]] = bitcast <8 x bfloat>* [[COERCE]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP17]], <4 x i32>* [[TMP18]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP19:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[COERCE]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[TMP19]], <8 x bfloat>* [[RETVAL]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP20:%.*]] = bitcast <8 x bfloat>* [[RETVAL]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP21:%.*]] = load <4 x i32>, <4 x i32>* [[TMP20]], align 8
+// CHECK-A32-SOFTFP-NEXT:    ret <4 x i32> [[TMP21]]
+//
 bfloat16x8_t test_vcvtq_low_bf16_f32(float32x4_t a) {
  return vcvtq_low_bf16_f32(a);
 }

-// CHECK-LABEL: test_vcvtq_high_bf16_f32
-// CHECK-A64: call <8 x bfloat> @llvm.aarch64.neon.bfcvtn2(<8 x bfloat> %inactive, <4 x float> %a)
-// CHECK-A32-HARDFP: %[[CVT:.*]] = call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16(<4 x float> %a)
-// CHECK-A32-HARDFP: %[[INACT:.*]] = shufflevector <8 x bfloat> %inactive, <8 x bfloat> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-// CHECK-A32-HARDFP: shufflevector <4 x bfloat> %[[CVT]], <4 x bfloat> %[[INACT]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-// CHECK-A32-SOFTFP: call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16(<4 x float> %a)
-// CHECK-A32-SOFTFP: shufflevector <8 x bfloat> %{{.*}}, <8 x bfloat> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-// CHECK-A32-SOFTFP: shufflevector <4 x bfloat> %{{.*}}, <4 x bfloat> %{{.*}}, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+// CHECK-A64-LABEL: @test_vcvtq_high_bf16_f32(
+// CHECK-A64-NEXT:  entry:
+// CHECK-A64-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat> [[INACTIVE:%.*]] to <16 x i8>
+// CHECK-A64-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A64-NEXT:    [[VCVTQ_HIGH_BF16_V2_I:%.*]] = call <8 x bfloat> @llvm.aarch64.neon.bfcvtn2(<8 x bfloat> [[INACTIVE]], <4 x float> [[A]]) #[[ATTR3]]
+// CHECK-A64-NEXT:    [[VCVTQ_HIGH_BF16_V3_I:%.*]] = bitcast <8 x bfloat> [[VCVTQ_HIGH_BF16_V2_I]] to <16 x i8>
+// CHECK-A64-NEXT:    ret <8 x bfloat> [[VCVTQ_HIGH_BF16_V2_I]]
+//
+// CHECK-A32-HARDFP-LABEL: @test_vcvtq_high_bf16_f32(
+// CHECK-A32-HARDFP-NEXT:  entry:
+// CHECK-A32-HARDFP-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A32-HARDFP-NEXT:    [[VCVTFP2BF1_I:%.*]] = call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16(<4 x float> [[A]]) #[[ATTR3]]
+// CHECK-A32-HARDFP-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[INACTIVE:%.*]], <8 x bfloat> [[INACTIVE]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+// CHECK-A32-HARDFP-NEXT:    [[SHUFFLE_I8:%.*]] = shufflevector <4 x bfloat> [[VCVTFP2BF1_I]], <4 x bfloat> [[SHUFFLE_I]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+// CHECK-A32-HARDFP-NEXT:    ret <8 x bfloat> [[SHUFFLE_I8]]
+//
+// CHECK-A32-SOFTFP-LABEL: @test_vcvtq_high_bf16_f32(
+// CHECK-A32-SOFTFP-NEXT:  entry:
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I11:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_I12:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P1_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I8:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I3:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_I4:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P0_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE2_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE4_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE5_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE6_I:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE8_I:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[RETVAL:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[INACTIVE:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[COERCE2:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat>* [[INACTIVE]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[INACTIVE_COERCE:%.*]], <4 x i32>* [[TMP0]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[INACTIVE1:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[INACTIVE]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[INACTIVE1]], <8 x bfloat>* [[COERCE]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat>* [[COERCE]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP3:%.*]] = bitcast <8 x bfloat>* [[__P0_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P01_I:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP4:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
+// CHECK-A32-SOFTFP-NEXT:    [[VCVTFP2BF1_I:%.*]] = call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16(<4 x float> [[A]]) #[[ATTR3]]
+// CHECK-A32-SOFTFP-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[VCVTFP2BF1_I]] to <4 x bfloat>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP5]], <4 x bfloat>* [[RETVAL_I8]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP6:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I8]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[TMP6]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP8:%.*]] = bitcast <4 x bfloat>* [[COERCE_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP7]], <2 x i32>* [[TMP8]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP9:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[__P01_I]], <8 x bfloat>* [[COERCE2_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP10:%.*]] = bitcast <8 x bfloat>* [[COERCE2_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP11:%.*]] = load <4 x i32>, <4 x i32>* [[TMP10]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP12:%.*]] = bitcast <8 x bfloat>* [[__P0_I4]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP11]], <4 x i32>* [[TMP12]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P01_I7:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I4]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[__P01_I7]], <8 x bfloat> [[__P01_I7]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[RETVAL_I3]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP13:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I3]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP14:%.*]] = load <2 x i32>, <2 x i32>* [[TMP13]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP15:%.*]] = bitcast <4 x bfloat>* [[COERCE4_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP14]], <2 x i32>* [[TMP15]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP16:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE4_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP9]], <4 x bfloat>* [[COERCE5_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP17:%.*]] = bitcast <4 x bfloat>* [[COERCE5_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP18:%.*]] = load <2 x i32>, <2 x i32>* [[TMP17]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <4 x bfloat> [[TMP16]], <4 x bfloat>* [[COERCE6_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP19:%.*]] = bitcast <4 x bfloat>* [[COERCE6_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP20:%.*]] = load <2 x i32>, <2 x i32>* [[TMP19]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP21:%.*]] = bitcast <4 x bfloat>* [[__P0_I12]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP18]], <2 x i32>* [[TMP21]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P01_I16:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_I12]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP22:%.*]] = bitcast <4 x bfloat>* [[__P1_I]] to <2 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <2 x i32> [[TMP20]], <2 x i32>* [[TMP22]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[__P12_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P1_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[SHUFFLE_I17:%.*]] = shufflevector <4 x bfloat> [[__P01_I16]], <4 x bfloat> [[__P12_I]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[SHUFFLE_I17]], <8 x bfloat>* [[RETVAL_I11]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP23:%.*]] = bitcast <8 x bfloat>* [[RETVAL_I11]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP24:%.*]] = load <4 x i32>, <4 x i32>* [[TMP23]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP25:%.*]] = bitcast <8 x bfloat>* [[COERCE8_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP24]], <4 x i32>* [[TMP25]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP26:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[COERCE8_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[TMP26]], <8 x bfloat>* [[RETVAL_I]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP27:%.*]] = bitcast <8 x bfloat>* [[RETVAL_I]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP28:%.*]] = load <4 x i32>, <4 x i32>* [[TMP27]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP29:%.*]] = bitcast <8 x bfloat>* [[COERCE2]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    store <4 x i32> [[TMP28]], <4 x i32>* [[TMP29]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP30:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[COERCE2]], align 8
+// CHECK-A32-SOFTFP-NEXT:    store <8 x bfloat> [[TMP30]], <8 x bfloat>* [[RETVAL]], align 8
+// CHECK-A32-SOFTFP-NEXT:    [[TMP31:%.*]] = bitcast <8 x bfloat>* [[RETVAL]] to <4 x i32>*
+// CHECK-A32-SOFTFP-NEXT:    [[TMP32:%.*]] = load <4 x i32>, <4 x i32>* [[TMP31]], align 8
+// CHECK-A32-SOFTFP-NEXT:    ret <4 x i32> [[TMP32]]
+//
 bfloat16x8_t test_vcvtq_high_bf16_f32(bfloat16x8_t inactive, float32x4_t a) {
  return vcvtq_high_bf16_f32(inactive, a);
 }

-// CHECK-LABEL: test_vcvth_bf16_f32
-// CHECK-A64: call bfloat @llvm.aarch64.neon.bfcvt(float %a)
-// CHECK-A32-HARDFP: call bfloat @llvm.arm.neon.vcvtbfp2bf(float %a)
-// CHECK-A32-SOFTFP: call bfloat @llvm.arm.neon.vcvtbfp2bf(float %a)
+// CHECK-A64-LABEL: @test_vcvth_bf16_f32(
+// CHECK-A64-NEXT:  entry:
+// CHECK-A64-NEXT:    [[VCVTH_BF16_F32_I:%.*]] = call bfloat @llvm.aarch64.neon.bfcvt(float [[A:%.*]]) #[[ATTR3]]
+// CHECK-A64-NEXT:    ret bfloat [[VCVTH_BF16_F32_I]]
+//
+// CHECK-A32-HARDFP-LABEL: @test_vcvth_bf16_f32(
+// CHECK-A32-HARDFP-NEXT:  entry:
+// CHECK-A32-HARDFP-NEXT:    [[VCVTBFP2BF_I:%.*]] = call bfloat @llvm.arm.neon.vcvtbfp2bf(float [[A:%.*]]) #[[ATTR3]]
+// CHECK-A32-HARDFP-NEXT:    ret bfloat [[VCVTBFP2BF_I]]
+//
+// CHECK-A32-SOFTFP-LABEL: @test_vcvth_bf16_f32(
+// CHECK-A32-SOFTFP-NEXT:  entry:
+// CHECK-A32-SOFTFP-NEXT:    [[VCVTBFP2BF_I:%.*]] = call bfloat @llvm.arm.neon.vcvtbfp2bf(float [[A:%.*]]) #[[ATTR3]]
+// CHECK-A32-SOFTFP-NEXT:    ret bfloat [[VCVTBFP2BF_I]]
+//
 bfloat16_t test_vcvth_bf16_f32(float32_t a) {
  return vcvth_bf16_f32(a);
 }

-// CHECK-LABEL: test_vcvtah_f32_bf16
-// CHECK: shl i32 %{{.*}}, 16
+// CHECK-LABEL: @test_vcvtah_f32_bf16(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[__REINT_I:%.*]] = alloca bfloat, align 2
+// CHECK-NEXT:    [[__REINT1_I:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    store bfloat [[A:%.*]], bfloat* [[__REINT_I]], align 2
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast bfloat* [[__REINT_I]] to i32*
+// CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 2
+// CHECK-NEXT:    [[SHL_I:%.*]] = shl i32 [[TMP1]], 16
+// CHECK-NEXT:    store i32 [[SHL_I]], i32* [[__REINT1_I]], align 4
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[__REINT1_I]] to float*
+// CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP2]], align 4
+// CHECK-NEXT:    ret float [[TMP3]]
+//
 float32_t test_vcvtah_f32_bf16(bfloat16_t a) {
  return vcvtah_f32_bf16(a);
 }
--- a/clang/test/CodeGen/arm-bf16-dotprod-intrinsics.c
+++ b/clang/test/CodeGen/arm-bf16-dotprod-intrinsics.c
@ -2,17 +2,20 @@
 // RUN: %clang_cc1 -triple armv8-arm-none-eabi \
 // RUN:   -target-feature +neon -target-feature +bf16 -mfloat-abi soft \
 // RUN:   -disable-O0-optnone -S -emit-llvm -o - %s \
-// RUN: | opt -S -mem2reg -instcombine | FileCheck %s
+// RUN: | opt -S -mem2reg | FileCheck %s
 // RUN: %clang_cc1 -triple armv8-arm-none-eabi \
 // RUN:   -target-feature +neon -target-feature +bf16 -mfloat-abi hard \
 // RUN:   -disable-O0-optnone -S -emit-llvm -o - %s \
-// RUN: | opt -S -mem2reg -instcombine | FileCheck %s
+// RUN: | opt -S -mem2reg | FileCheck %s

 #include <arm_neon.h>

 // CHECK-LABEL: @test_vbfdot_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.arm.neon.bfdot.v2f32.v4bf16(<2 x float> [[R:%.*]], <4 x bfloat> [[A:%.*]], <4 x bfloat> [[B:%.*]]) [[ATTR3:#.*]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[R:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x bfloat> [[A:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x bfloat> [[B:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.arm.neon.bfdot.v2f32.v4bf16(<2 x float> [[R]], <4 x bfloat> [[A]], <4 x bfloat> [[B]]) #[[ATTR3:[0-9]+]]
 // CHECK-NEXT:    ret <2 x float> [[VBFDOT3_I]]
 //
 float32x2_t test_vbfdot_f32(float32x2_t r, bfloat16x4_t a, bfloat16x4_t b) {
@ -21,7 +24,10 @@ float32x2_t test_vbfdot_f32(float32x2_t r, bfloat16x4_t a, bfloat16x4_t b) {

 // CHECK-LABEL: @test_vbfdotq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]]) [[ATTR3]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[B]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <4 x float> [[VBFDOT3_I]]
 //
 float32x4_t test_vbfdotq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b){
@ -30,10 +36,21 @@ float32x4_t test_vbfdotq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b){

 // CHECK-LABEL: @test_vbfdot_lane_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTCAST:%.*]] = bitcast <4 x bfloat> [[B:%.*]] to <2 x float>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[DOTCAST]], <2 x float> undef, <2 x i32> zeroinitializer
-// CHECK-NEXT:    [[DOTCAST1:%.*]] = bitcast <2 x float> [[LANE]] to <4 x bfloat>
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.arm.neon.bfdot.v2f32.v4bf16(<2 x float> [[R:%.*]], <4 x bfloat> [[A:%.*]], <4 x bfloat> [[DOTCAST1]]) [[ATTR3]]
+// CHECK-NEXT:    [[__REINT_144:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-NEXT:    [[__REINT1_144:%.*]] = alloca <2 x float>, align 8
+// CHECK-NEXT:    store <4 x bfloat> [[B:%.*]], <4 x bfloat>* [[__REINT_144]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_144]] to <2 x float>*
+// CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[TMP1]] to <8 x i8>
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <2 x float>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK-NEXT:    store <2 x float> [[LANE]], <2 x float>* [[__REINT1_144]], align 8
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x float>* [[__REINT1_144]] to <4 x bfloat>*
+// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[TMP4]], align 8
+// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x float> [[R:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x bfloat> [[A:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x bfloat> [[TMP5]] to <8 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.arm.neon.bfdot.v2f32.v4bf16(<2 x float> [[R]], <4 x bfloat> [[A]], <4 x bfloat> [[TMP5]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <2 x float> [[VBFDOT3_I]]
 //
 float32x2_t test_vbfdot_lane_f32(float32x2_t r, bfloat16x4_t a, bfloat16x4_t b){
@ -42,10 +59,21 @@ float32x2_t test_vbfdot_lane_f32(float32x2_t r, bfloat16x4_t a, bfloat16x4_t b){

 // CHECK-LABEL: @test_vbfdotq_laneq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTCAST:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <4 x float>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[DOTCAST]], <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-// CHECK-NEXT:    [[DOTCAST1:%.*]] = bitcast <4 x float> [[LANE]] to <8 x bfloat>
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[DOTCAST1]]) [[ATTR3]]
+// CHECK-NEXT:    [[__REINT_146:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-NEXT:    [[__REINT1_146:%.*]] = alloca <4 x float>, align 8
+// CHECK-NEXT:    store <8 x bfloat> [[B:%.*]], <8 x bfloat>* [[__REINT_146]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat>* [[__REINT_146]] to <4 x float>*
+// CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[TMP0]], align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <16 x i8>
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP3]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+// CHECK-NEXT:    store <4 x float> [[LANE]], <4 x float>* [[__REINT1_146]], align 8
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x float>* [[__REINT1_146]] to <8 x bfloat>*
+// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[TMP4]], align 8
+// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x bfloat> [[TMP5]] to <16 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[TMP5]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <4 x float> [[VBFDOT3_I]]
 //
 float32x4_t test_vbfdotq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -54,10 +82,21 @@ float32x4_t test_vbfdotq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b

 // CHECK-LABEL: @test_vbfdot_laneq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTCAST:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <4 x float>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[DOTCAST]], <4 x float> undef, <2 x i32> <i32 3, i32 3>
-// CHECK-NEXT:    [[DOTCAST1:%.*]] = bitcast <2 x float> [[LANE]] to <4 x bfloat>
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.arm.neon.bfdot.v2f32.v4bf16(<2 x float> [[R:%.*]], <4 x bfloat> [[A:%.*]], <4 x bfloat> [[DOTCAST1]]) [[ATTR3]]
+// CHECK-NEXT:    [[__REINT_148:%.*]] = alloca <8 x bfloat>, align 8
+// CHECK-NEXT:    [[__REINT1_148:%.*]] = alloca <2 x float>, align 8
+// CHECK-NEXT:    store <8 x bfloat> [[B:%.*]], <8 x bfloat>* [[__REINT_148]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat>* [[__REINT_148]] to <4 x float>*
+// CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[TMP0]], align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <16 x i8>
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP3]], <2 x i32> <i32 3, i32 3>
+// CHECK-NEXT:    store <2 x float> [[LANE]], <2 x float>* [[__REINT1_148]], align 8
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x float>* [[__REINT1_148]] to <4 x bfloat>*
+// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[TMP4]], align 8
+// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x float> [[R:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x bfloat> [[A:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x bfloat> [[TMP5]] to <8 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <2 x float> @llvm.arm.neon.bfdot.v2f32.v4bf16(<2 x float> [[R]], <4 x bfloat> [[A]], <4 x bfloat> [[TMP5]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <2 x float> [[VBFDOT3_I]]
 //
 float32x2_t test_vbfdot_laneq_f32(float32x2_t r, bfloat16x4_t a, bfloat16x8_t b) {
@ -66,10 +105,21 @@ float32x2_t test_vbfdot_laneq_f32(float32x2_t r, bfloat16x4_t a, bfloat16x8_t b)

 // CHECK-LABEL: @test_vbfdotq_lane_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTCAST:%.*]] = bitcast <4 x bfloat> [[B:%.*]] to <2 x float>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[DOTCAST]], <2 x float> undef, <4 x i32> zeroinitializer
-// CHECK-NEXT:    [[DOTCAST1:%.*]] = bitcast <4 x float> [[LANE]] to <8 x bfloat>
-// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[DOTCAST1]]) [[ATTR3]]
+// CHECK-NEXT:    [[__REINT_142:%.*]] = alloca <4 x bfloat>, align 8
+// CHECK-NEXT:    [[__REINT1_142:%.*]] = alloca <4 x float>, align 8
+// CHECK-NEXT:    store <4 x bfloat> [[B:%.*]], <4 x bfloat>* [[__REINT_142]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_142]] to <2 x float>*
+// CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[TMP1]] to <8 x i8>
+// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <2 x float>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK-NEXT:    store <4 x float> [[LANE]], <4 x float>* [[__REINT1_142]], align 8
+// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x float>* [[__REINT1_142]] to <8 x bfloat>*
+// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[TMP4]], align 8
+// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x bfloat> [[TMP5]] to <16 x i8>
+// CHECK-NEXT:    [[VBFDOT3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfdot.v4f32.v8bf16(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[TMP5]]) #[[ATTR3]]
 // CHECK-NEXT:    ret <4 x float> [[VBFDOT3_I]]
 //
 float32x4_t test_vbfdotq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t b) {
@ -78,7 +128,11 @@ float32x4_t test_vbfdotq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t b)

 // CHECK-LABEL: @test_vbfmmlaq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFMMLAQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmmla(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]]) [[ATTR3]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMMLAQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmmla(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[B]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMMLAQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMMLAQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMMLAQ_V3_I]]
 //
 float32x4_t test_vbfmmlaq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -87,7 +141,11 @@ float32x4_t test_vbfmmlaq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {

 // CHECK-LABEL: @test_vbfmlalbq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalb(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]]) [[ATTR3]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalb(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[B]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALBQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALBQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALBQ_V3_I]]
 //
 float32x4_t test_vbfmlalbq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -96,7 +154,11 @@ float32x4_t test_vbfmlalbq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {

 // CHECK-LABEL: @test_vbfmlaltq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalt(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[B:%.*]]) [[ATTR3]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[B:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalt(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[B]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALTQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALTQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALTQ_V3_I]]
 //
 float32x4_t test_vbfmlaltq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -105,8 +167,27 @@ float32x4_t test_vbfmlaltq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {

 // CHECK-LABEL: @test_vbfmlalbq_lane_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VECINIT35:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> undef, <8 x i32> zeroinitializer
-// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalb(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[VECINIT35]]) [[ATTR3]]
+// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[B:%.*]], i32 0
+// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x bfloat> undef, bfloat [[VGET_LANE]], i32 0
+// CHECK-NEXT:    [[VGET_LANE3:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT5:%.*]] = insertelement <8 x bfloat> [[VECINIT]], bfloat [[VGET_LANE3]], i32 1
+// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT10:%.*]] = insertelement <8 x bfloat> [[VECINIT5]], bfloat [[VGET_LANE8]], i32 2
+// CHECK-NEXT:    [[VGET_LANE13:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT15:%.*]] = insertelement <8 x bfloat> [[VECINIT10]], bfloat [[VGET_LANE13]], i32 3
+// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT20:%.*]] = insertelement <8 x bfloat> [[VECINIT15]], bfloat [[VGET_LANE18]], i32 4
+// CHECK-NEXT:    [[VGET_LANE23:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT25:%.*]] = insertelement <8 x bfloat> [[VECINIT20]], bfloat [[VGET_LANE23]], i32 5
+// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT30:%.*]] = insertelement <8 x bfloat> [[VECINIT25]], bfloat [[VGET_LANE28]], i32 6
+// CHECK-NEXT:    [[VGET_LANE33:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT35:%.*]] = insertelement <8 x bfloat> [[VECINIT30]], bfloat [[VGET_LANE33]], i32 7
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[VECINIT35]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalb(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[VECINIT35]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALBQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALBQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALBQ_V3_I]]
 //
 float32x4_t test_vbfmlalbq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t b) {
@ -115,8 +196,27 @@ float32x4_t test_vbfmlalbq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t

 // CHECK-LABEL: @test_vbfmlalbq_laneq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VECINIT35:%.*]] = shufflevector <8 x bfloat> [[B:%.*]], <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalb(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[VECINIT35]]) [[ATTR3]]
+// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <8 x bfloat> [[B:%.*]], i32 3
+// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x bfloat> undef, bfloat [[VGET_LANE]], i32 0
+// CHECK-NEXT:    [[VGET_LANE3:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT5:%.*]] = insertelement <8 x bfloat> [[VECINIT]], bfloat [[VGET_LANE3]], i32 1
+// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT10:%.*]] = insertelement <8 x bfloat> [[VECINIT5]], bfloat [[VGET_LANE8]], i32 2
+// CHECK-NEXT:    [[VGET_LANE13:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT15:%.*]] = insertelement <8 x bfloat> [[VECINIT10]], bfloat [[VGET_LANE13]], i32 3
+// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT20:%.*]] = insertelement <8 x bfloat> [[VECINIT15]], bfloat [[VGET_LANE18]], i32 4
+// CHECK-NEXT:    [[VGET_LANE23:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT25:%.*]] = insertelement <8 x bfloat> [[VECINIT20]], bfloat [[VGET_LANE23]], i32 5
+// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT30:%.*]] = insertelement <8 x bfloat> [[VECINIT25]], bfloat [[VGET_LANE28]], i32 6
+// CHECK-NEXT:    [[VGET_LANE33:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT35:%.*]] = insertelement <8 x bfloat> [[VECINIT30]], bfloat [[VGET_LANE33]], i32 7
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[VECINIT35]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALBQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalb(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[VECINIT35]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALBQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALBQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALBQ_V3_I]]
 //
 float32x4_t test_vbfmlalbq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
@ -125,8 +225,27 @@ float32x4_t test_vbfmlalbq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t

 // CHECK-LABEL: @test_vbfmlaltq_lane_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VECINIT35:%.*]] = shufflevector <4 x bfloat> [[B:%.*]], <4 x bfloat> undef, <8 x i32> zeroinitializer
-// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalt(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[VECINIT35]]) [[ATTR3]]
+// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x bfloat> [[B:%.*]], i32 0
+// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x bfloat> undef, bfloat [[VGET_LANE]], i32 0
+// CHECK-NEXT:    [[VGET_LANE3:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT5:%.*]] = insertelement <8 x bfloat> [[VECINIT]], bfloat [[VGET_LANE3]], i32 1
+// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT10:%.*]] = insertelement <8 x bfloat> [[VECINIT5]], bfloat [[VGET_LANE8]], i32 2
+// CHECK-NEXT:    [[VGET_LANE13:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT15:%.*]] = insertelement <8 x bfloat> [[VECINIT10]], bfloat [[VGET_LANE13]], i32 3
+// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT20:%.*]] = insertelement <8 x bfloat> [[VECINIT15]], bfloat [[VGET_LANE18]], i32 4
+// CHECK-NEXT:    [[VGET_LANE23:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT25:%.*]] = insertelement <8 x bfloat> [[VECINIT20]], bfloat [[VGET_LANE23]], i32 5
+// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT30:%.*]] = insertelement <8 x bfloat> [[VECINIT25]], bfloat [[VGET_LANE28]], i32 6
+// CHECK-NEXT:    [[VGET_LANE33:%.*]] = extractelement <4 x bfloat> [[B]], i32 0
+// CHECK-NEXT:    [[VECINIT35:%.*]] = insertelement <8 x bfloat> [[VECINIT30]], bfloat [[VGET_LANE33]], i32 7
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[VECINIT35]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalt(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[VECINIT35]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALTQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALTQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALTQ_V3_I]]
 //
 float32x4_t test_vbfmlaltq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t b) {
@ -135,8 +254,27 @@ float32x4_t test_vbfmlaltq_lane_f32(float32x4_t r, bfloat16x8_t a, bfloat16x4_t

 // CHECK-LABEL: @test_vbfmlaltq_laneq_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VECINIT35:%.*]] = shufflevector <8 x bfloat> [[B:%.*]], <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalt(<4 x float> [[R:%.*]], <8 x bfloat> [[A:%.*]], <8 x bfloat> [[VECINIT35]]) [[ATTR3]]
+// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <8 x bfloat> [[B:%.*]], i32 3
+// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x bfloat> undef, bfloat [[VGET_LANE]], i32 0
+// CHECK-NEXT:    [[VGET_LANE3:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT5:%.*]] = insertelement <8 x bfloat> [[VECINIT]], bfloat [[VGET_LANE3]], i32 1
+// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT10:%.*]] = insertelement <8 x bfloat> [[VECINIT5]], bfloat [[VGET_LANE8]], i32 2
+// CHECK-NEXT:    [[VGET_LANE13:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT15:%.*]] = insertelement <8 x bfloat> [[VECINIT10]], bfloat [[VGET_LANE13]], i32 3
+// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT20:%.*]] = insertelement <8 x bfloat> [[VECINIT15]], bfloat [[VGET_LANE18]], i32 4
+// CHECK-NEXT:    [[VGET_LANE23:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT25:%.*]] = insertelement <8 x bfloat> [[VECINIT20]], bfloat [[VGET_LANE23]], i32 5
+// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT30:%.*]] = insertelement <8 x bfloat> [[VECINIT25]], bfloat [[VGET_LANE28]], i32 6
+// CHECK-NEXT:    [[VGET_LANE33:%.*]] = extractelement <8 x bfloat> [[B]], i32 3
+// CHECK-NEXT:    [[VECINIT35:%.*]] = insertelement <8 x bfloat> [[VECINIT30]], bfloat [[VGET_LANE33]], i32 7
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[R:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x bfloat> [[A:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x bfloat> [[VECINIT35]] to <16 x i8>
+// CHECK-NEXT:    [[VBFMLALTQ_V3_I:%.*]] = call <4 x float> @llvm.arm.neon.bfmlalt(<4 x float> [[R]], <8 x bfloat> [[A]], <8 x bfloat> [[VECINIT35]]) #[[ATTR3]]
+// CHECK-NEXT:    [[VBFMLALTQ_V4_I:%.*]] = bitcast <4 x float> [[VBFMLALTQ_V3_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x float> [[VBFMLALTQ_V3_I]]
 //
 float32x4_t test_vbfmlaltq_laneq_f32(float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) {
--- a/clang/test/CodeGen/arm-bf16-getset-intrinsics.c
+++ b/clang/test/CodeGen/arm-bf16-getset-intrinsics.c
@ -1,8 +1,8 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -triple armv8.6a-arm-none-eabi -target-feature +neon -target-feature +bf16 -mfloat-abi hard \
-// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg -instcombine | FileCheck %s
+// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg | FileCheck %s
 // RUN: %clang_cc1 -triple armv8.6a-arm-none-eabi -target-feature +neon -target-feature +bf16 -mfloat-abi soft \
-// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg -instcombine | FileCheck %s
+// RUN:  -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg | FileCheck %s

 #include <arm_neon.h>

@ -18,7 +18,9 @@ bfloat16x4_t test_vcreate_bf16(uint64_t a) {
 // CHECK-LABEL: @test_vdup_n_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x bfloat> undef, bfloat [[V:%.*]], i32 0
-// CHECK-NEXT:    [[VECINIT3_I:%.*]] = shufflevector <4 x bfloat> [[VECINIT_I]], <4 x bfloat> poison, <4 x i32> zeroinitializer
+// CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x bfloat> [[VECINIT_I]], bfloat [[V]], i32 1
+// CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x bfloat> [[VECINIT1_I]], bfloat [[V]], i32 2
+// CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x bfloat> [[VECINIT2_I]], bfloat [[V]], i32 3
 // CHECK-NEXT:    ret <4 x bfloat> [[VECINIT3_I]]
 //
 bfloat16x4_t test_vdup_n_bf16(bfloat16_t v) {
@ -28,7 +30,13 @@ bfloat16x4_t test_vdup_n_bf16(bfloat16_t v) {
 // CHECK-LABEL: @test_vdupq_n_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x bfloat> undef, bfloat [[V:%.*]], i32 0
-// CHECK-NEXT:    [[VECINIT7_I:%.*]] = shufflevector <8 x bfloat> [[VECINIT_I]], <8 x bfloat> poison, <8 x i32> zeroinitializer
+// CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x bfloat> [[VECINIT_I]], bfloat [[V]], i32 1
+// CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x bfloat> [[VECINIT1_I]], bfloat [[V]], i32 2
+// CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x bfloat> [[VECINIT2_I]], bfloat [[V]], i32 3
+// CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x bfloat> [[VECINIT3_I]], bfloat [[V]], i32 4
+// CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x bfloat> [[VECINIT4_I]], bfloat [[V]], i32 5
+// CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x bfloat> [[VECINIT5_I]], bfloat [[V]], i32 6
+// CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x bfloat> [[VECINIT6_I]], bfloat [[V]], i32 7
 // CHECK-NEXT:    ret <8 x bfloat> [[VECINIT7_I]]
 //
 bfloat16x8_t test_vdupq_n_bf16(bfloat16_t v) {
@ -37,7 +45,9 @@ bfloat16x8_t test_vdupq_n_bf16(bfloat16_t v) {

 // CHECK-LABEL: @test_vdup_lane_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x bfloat> [[V:%.*]], <4 x bfloat> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat> [[V:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x bfloat>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x bfloat> [[TMP1]], <4 x bfloat> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    ret <4 x bfloat> [[LANE]]
 //
 bfloat16x4_t test_vdup_lane_bf16(bfloat16x4_t v) {
@ -46,7 +56,9 @@ bfloat16x4_t test_vdup_lane_bf16(bfloat16x4_t v) {

 // CHECK-LABEL: @test_vdupq_lane_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x bfloat> [[V:%.*]], <4 x bfloat> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x bfloat> [[V:%.*]] to <8 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x bfloat>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x bfloat> [[TMP1]], <4 x bfloat> [[TMP1]], <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    ret <8 x bfloat> [[LANE]]
 //
 bfloat16x8_t test_vdupq_lane_bf16(bfloat16x4_t v) {
@ -55,7 +67,9 @@ bfloat16x8_t test_vdupq_lane_bf16(bfloat16x4_t v) {

 // CHECK-LABEL: @test_vdup_laneq_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x bfloat> [[V:%.*]], <8 x bfloat> undef, <4 x i32> <i32 7, i32 7, i32 7, i32 7>
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat> [[V:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x bfloat>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x bfloat> [[TMP1]], <8 x bfloat> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    ret <4 x bfloat> [[LANE]]
 //
 bfloat16x4_t test_vdup_laneq_bf16(bfloat16x8_t v) {
@ -64,7 +78,9 @@ bfloat16x4_t test_vdup_laneq_bf16(bfloat16x8_t v) {

 // CHECK-LABEL: @test_vdupq_laneq_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x bfloat> [[V:%.*]], <8 x bfloat> undef, <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x bfloat> [[V:%.*]] to <16 x i8>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x bfloat>
+// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x bfloat> [[TMP1]], <8 x bfloat> [[TMP1]], <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    ret <8 x bfloat> [[LANE]]
 //
 bfloat16x8_t test_vdupq_laneq_bf16(bfloat16x8_t v) {
@ -82,7 +98,7 @@ bfloat16x8_t test_vcombine_bf16(bfloat16x4_t low, bfloat16x4_t high) {

 // CHECK-LABEL: @test_vget_high_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    ret <4 x bfloat> [[SHUFFLE_I]]
 //
 bfloat16x4_t test_vget_high_bf16(bfloat16x8_t a) {
@ -91,7 +107,7 @@ bfloat16x4_t test_vget_high_bf16(bfloat16x8_t a) {

 // CHECK-LABEL: @test_vget_low_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 // CHECK-NEXT:    ret <4 x bfloat> [[SHUFFLE_I]]
 //
 bfloat16x4_t test_vget_low_bf16(bfloat16x8_t a) {