clang/test/CodeGen/aarch64-neon-2velem.c - llvm-project - Git at Google

 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon -disable-O0-optnone -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s

 // REQUIRES: aarch64-registered-target || arm-registered-target

 #include <arm_neon.h>

 // CHECK-LABEL: @test_vmla_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD]]
 //
 int16x4_t test_vmla_lane_s16(int16x4_t a, int16x4_t b, int16x4_t v) {
   return vmla_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlaq_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD]]
 //
 int16x8_t test_vmlaq_lane_s16(int16x8_t a, int16x8_t b, int16x4_t v) {
   return vmlaq_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmla_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD]]
 //
 int32x2_t test_vmla_lane_s32(int32x2_t a, int32x2_t b, int32x2_t v) {
   return vmla_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlaq_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlaq_lane_s32(int32x4_t a, int32x4_t b, int32x2_t v) {
   return vmlaq_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmla_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD]]
 //
 int16x4_t test_vmla_laneq_s16(int16x4_t a, int16x4_t b, int16x8_t v) {
   return vmla_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlaq_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD]]
 //
 int16x8_t test_vmlaq_laneq_s16(int16x8_t a, int16x8_t b, int16x8_t v) {
   return vmlaq_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmla_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD]]
 //
 int32x2_t test_vmla_laneq_s32(int32x2_t a, int32x2_t b, int32x4_t v) {
   return vmla_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlaq_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlaq_laneq_s32(int32x4_t a, int32x4_t b, int32x4_t v) {
   return vmlaq_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmls_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB]]
 //
 int16x4_t test_vmls_lane_s16(int16x4_t a, int16x4_t b, int16x4_t v) {
   return vmls_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsq_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB]]
 //
 int16x8_t test_vmlsq_lane_s16(int16x8_t a, int16x8_t b, int16x4_t v) {
   return vmlsq_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmls_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB]]
 //
 int32x2_t test_vmls_lane_s32(int32x2_t a, int32x2_t b, int32x2_t v) {
   return vmls_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlsq_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsq_lane_s32(int32x4_t a, int32x4_t b, int32x2_t v) {
   return vmlsq_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmls_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB]]
 //
 int16x4_t test_vmls_laneq_s16(int16x4_t a, int16x4_t b, int16x8_t v) {
   return vmls_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlsq_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB]]
 //
 int16x8_t test_vmlsq_laneq_s16(int16x8_t a, int16x8_t b, int16x8_t v) {
   return vmlsq_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmls_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB]]
 //
 int32x2_t test_vmls_laneq_s32(int32x2_t a, int32x2_t b, int32x4_t v) {
   return vmls_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsq_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsq_laneq_s32(int32x4_t a, int32x4_t b, int32x4_t v) {
   return vmlsq_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmul_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL]]
 //
 int16x4_t test_vmul_lane_s16(int16x4_t a, int16x4_t v) {
   return vmul_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vmulq_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL]]
 //
 int16x8_t test_vmulq_lane_s16(int16x8_t a, int16x4_t v) {
   return vmulq_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vmul_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL]]
 //
 int32x2_t test_vmul_lane_s32(int32x2_t a, int32x2_t v) {
   return vmul_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmulq_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL]]
 //
 int32x4_t test_vmulq_lane_s32(int32x4_t a, int32x2_t v) {
   return vmulq_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmul_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL]]
 //
 uint16x4_t test_vmul_lane_u16(uint16x4_t a, uint16x4_t v) {
   return vmul_lane_u16(a, v, 3);
 }

 // CHECK-LABEL: @test_vmulq_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL]]
 //
 uint16x8_t test_vmulq_lane_u16(uint16x8_t a, uint16x4_t v) {
   return vmulq_lane_u16(a, v, 3);
 }

 // CHECK-LABEL: @test_vmul_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL]]
 //
 uint32x2_t test_vmul_lane_u32(uint32x2_t a, uint32x2_t v) {
   return vmul_lane_u32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmulq_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL]]
 //
 uint32x4_t test_vmulq_lane_u32(uint32x4_t a, uint32x2_t v) {
   return vmulq_lane_u32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmul_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL]]
 //
 int16x4_t test_vmul_laneq_s16(int16x4_t a, int16x8_t v) {
   return vmul_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vmulq_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL]]
 //
 int16x8_t test_vmulq_laneq_s16(int16x8_t a, int16x8_t v) {
   return vmulq_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vmul_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL]]
 //
 int32x2_t test_vmul_laneq_s32(int32x2_t a, int32x4_t v) {
   return vmul_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmulq_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL]]
 //
 int32x4_t test_vmulq_laneq_s32(int32x4_t a, int32x4_t v) {
   return vmulq_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmul_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL]]
 //
 uint16x4_t test_vmul_laneq_u16(uint16x4_t a, uint16x8_t v) {
   return vmul_laneq_u16(a, v, 7);
 }

 // CHECK-LABEL: @test_vmulq_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL]]
 //
 uint16x8_t test_vmulq_laneq_u16(uint16x8_t a, uint16x8_t v) {
   return vmulq_laneq_u16(a, v, 7);
 }

 // CHECK-LABEL: @test_vmul_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL]]
 //
 uint32x2_t test_vmul_laneq_u32(uint32x2_t a, uint32x4_t v) {
   return vmul_laneq_u32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmulq_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL]]
 //
 uint32x4_t test_vmulq_laneq_u32(uint32x4_t a, uint32x4_t v) {
   return vmulq_laneq_u32(a, v, 3);
 }

 // CHECK-LABEL: @test_vfma_lane_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[V:%.*]] 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> <i32 1, i32 1>
 // CHECK-NEXT:    [[FMLA:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[FMLA]], <2 x float> [[LANE]], <2 x float> [[FMLA1]])
 // CHECK-NEXT:    ret <2 x float> [[FMLA2]]
 //
 float32x2_t test_vfma_lane_f32(float32x2_t a, float32x2_t b, float32x2_t v) {
   return vfma_lane_f32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vfmaq_lane_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[V:%.*]] 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> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[FMLA:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[FMLA]], <4 x float> [[LANE]], <4 x float> [[FMLA1]])
 // CHECK-NEXT:    ret <4 x float> [[FMLA2]]
 //
 float32x4_t test_vfmaq_lane_f32(float32x4_t a, float32x4_t b, float32x2_t v) {
   return vfmaq_lane_f32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vfma_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP5]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP6:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[LANE]], <2 x float> [[TMP4]], <2 x float> [[TMP3]])
 // CHECK-NEXT:    ret <2 x float> [[TMP6]]
 //
 float32x2_t test_vfma_laneq_f32(float32x2_t a, float32x2_t b, float32x4_t v) {
   return vfma_laneq_f32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vfmaq_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP5]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP6:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[LANE]], <4 x float> [[TMP4]], <4 x float> [[TMP3]])
 // CHECK-NEXT:    ret <4 x float> [[TMP6]]
 //
 float32x4_t test_vfmaq_laneq_f32(float32x4_t a, float32x4_t b, float32x4_t v) {
   return vfmaq_laneq_f32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vfms_lane_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <2 x float> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[FNEG]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[V:%.*]] 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> <i32 1, i32 1>
 // CHECK-NEXT:    [[FMLA:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[FMLA]], <2 x float> [[LANE]], <2 x float> [[FMLA1]])
 // CHECK-NEXT:    ret <2 x float> [[FMLA2]]
 //
 float32x2_t test_vfms_lane_f32(float32x2_t a, float32x2_t b, float32x2_t v) {
   return vfms_lane_f32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vfmsq_lane_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <4 x float> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[FNEG]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[V:%.*]] 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> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[FMLA:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[FMLA]], <4 x float> [[LANE]], <4 x float> [[FMLA1]])
 // CHECK-NEXT:    ret <4 x float> [[FMLA2]]
 //
 float32x4_t test_vfmsq_lane_f32(float32x4_t a, float32x4_t b, float32x2_t v) {
   return vfmsq_lane_f32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vfms_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <2 x float> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[FNEG]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP5]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP6:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[LANE]], <2 x float> [[TMP4]], <2 x float> [[TMP3]])
 // CHECK-NEXT:    ret <2 x float> [[TMP6]]
 //
 float32x2_t test_vfms_laneq_f32(float32x2_t a, float32x2_t b, float32x4_t v) {
   return vfms_laneq_f32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vfmsq_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <4 x float> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[FNEG]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP5]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP6:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[LANE]], <4 x float> [[TMP4]], <4 x float> [[TMP3]])
 // CHECK-NEXT:    ret <4 x float> [[TMP6]]
 //
 float32x4_t test_vfmsq_laneq_f32(float32x4_t a, float32x4_t b, float32x4_t v) {
   return vfmsq_laneq_f32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vfmaq_lane_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <1 x double> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <1 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x double> [[TMP3]], <1 x double> [[TMP3]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[FMLA:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <2 x double> @llvm.fma.v2f64(<2 x double> [[FMLA]], <2 x double> [[LANE]], <2 x double> [[FMLA1]])
 // CHECK-NEXT:    ret <2 x double> [[FMLA2]]
 //
 float64x2_t test_vfmaq_lane_f64(float64x2_t a, float64x2_t b, float64x1_t v) {
   return vfmaq_lane_f64(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmaq_laneq_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP5]], <2 x double> [[TMP5]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP6:%.*]] = call <2 x double> @llvm.fma.v2f64(<2 x double> [[LANE]], <2 x double> [[TMP4]], <2 x double> [[TMP3]])
 // CHECK-NEXT:    ret <2 x double> [[TMP6]]
 //
 float64x2_t test_vfmaq_laneq_f64(float64x2_t a, float64x2_t b, float64x2_t v) {
   return vfmaq_laneq_f64(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vfmsq_lane_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <2 x double> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[FNEG]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <1 x double> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <1 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x double> [[TMP3]], <1 x double> [[TMP3]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[FMLA:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <2 x double> @llvm.fma.v2f64(<2 x double> [[FMLA]], <2 x double> [[LANE]], <2 x double> [[FMLA1]])
 // CHECK-NEXT:    ret <2 x double> [[FMLA2]]
 //
 float64x2_t test_vfmsq_lane_f64(float64x2_t a, float64x2_t b, float64x1_t v) {
   return vfmsq_lane_f64(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmsq_laneq_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <2 x double> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[FNEG]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP5]], <2 x double> [[TMP5]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP6:%.*]] = call <2 x double> @llvm.fma.v2f64(<2 x double> [[LANE]], <2 x double> [[TMP4]], <2 x double> [[TMP3]])
 // CHECK-NEXT:    ret <2 x double> [[TMP6]]
 //
 float64x2_t test_vfmsq_laneq_f64(float64x2_t a, float64x2_t b, float64x2_t v) {
   return vfmsq_laneq_f64(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vfmas_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <4 x float> [[V:%.*]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = call float @llvm.fma.f32(float [[B:%.*]], float [[EXTRACT]], float [[A:%.*]])
 // CHECK-NEXT:    ret float [[TMP0]]
 //
 float32_t test_vfmas_laneq_f32(float32_t a, float32_t b, float32x4_t v) {
   return vfmas_laneq_f32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vfmsd_lane_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg double [[B:%.*]]
 // CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <1 x double> [[V:%.*]], i32 0
 // CHECK-NEXT:    [[TMP0:%.*]] = call double @llvm.fma.f64(double [[FNEG]], double [[EXTRACT]], double [[A:%.*]])
 // CHECK-NEXT:    ret double [[TMP0]]
 //
 float64_t test_vfmsd_lane_f64(float64_t a, float64_t b, float64x1_t v) {
   return vfmsd_lane_f64(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmss_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg float [[B:%.*]]
 // CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <4 x float> [[V:%.*]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = call float @llvm.fma.f32(float [[FNEG]], float [[EXTRACT]], float [[A:%.*]])
 // CHECK-NEXT:    ret float [[TMP0]]
 //
 float32_t test_vfmss_laneq_f32(float32_t a, float32_t b, float32x4_t v) {
   return vfmss_laneq_f32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vfmsd_laneq_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg double [[B:%.*]]
 // CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <2 x double> [[V:%.*]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = call double @llvm.fma.f64(double [[FNEG]], double [[EXTRACT]], double [[A:%.*]])
 // CHECK-NEXT:    ret double [[TMP0]]
 //
 float64_t test_vfmsd_laneq_f64(float64_t a, float64_t b, float64x2_t v) {
   return vfmsd_laneq_f64(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlal_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_lane_s16(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vmlal_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlal_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_lane_s32(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vmlal_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlal_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_laneq_s16(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vmlal_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlal_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_laneq_s32(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vmlal_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlal_high_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_high_lane_s16(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vmlal_high_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlal_high_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_high_lane_s32(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vmlal_high_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlal_high_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_high_laneq_s16(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vmlal_high_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlal_high_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_high_laneq_s32(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vmlal_high_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsl_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_lane_s16(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vmlsl_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsl_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_lane_s32(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vmlsl_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlsl_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_laneq_s16(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vmlsl_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlsl_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_laneq_s32(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vmlsl_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsl_high_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_high_lane_s16(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vmlsl_high_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsl_high_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_high_lane_s32(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vmlsl_high_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlsl_high_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_high_laneq_s16(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vmlsl_high_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlsl_high_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_high_laneq_s32(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vmlsl_high_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlal_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_lane_u16(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vmlal_lane_u16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlal_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_lane_u32(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vmlal_lane_u32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlal_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_laneq_u16(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vmlal_laneq_u16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlal_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_laneq_u32(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vmlal_laneq_u32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlal_high_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_high_lane_u16(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vmlal_high_lane_u16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlal_high_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_high_lane_u32(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vmlal_high_lane_u32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlal_high_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_high_laneq_u16(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vmlal_high_laneq_u16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlal_high_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_high_laneq_u32(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vmlal_high_laneq_u32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsl_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_lane_u16(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vmlsl_lane_u16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsl_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_lane_u32(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vmlsl_lane_u32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlsl_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_laneq_u16(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vmlsl_laneq_u16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlsl_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_laneq_u32(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vmlsl_laneq_u32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsl_high_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_high_lane_u16(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vmlsl_high_lane_u16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsl_high_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_high_lane_u32(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vmlsl_high_lane_u32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlsl_high_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_high_laneq_u16(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vmlsl_high_laneq_u16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlsl_high_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_high_laneq_u32(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vmlsl_high_laneq_u32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmull_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 int32x4_t test_vmull_lane_s16(int16x4_t a, int16x4_t v) {
   return vmull_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vmull_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 int64x2_t test_vmull_lane_s32(int32x2_t a, int32x2_t v) {
   return vmull_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmull_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 uint32x4_t test_vmull_lane_u16(uint16x4_t a, uint16x4_t v) {
   return vmull_lane_u16(a, v, 3);
 }

 // CHECK-LABEL: @test_vmull_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 uint64x2_t test_vmull_lane_u32(uint32x2_t a, uint32x2_t v) {
   return vmull_lane_u32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmull_high_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 int32x4_t test_vmull_high_lane_s16(int16x8_t a, int16x4_t v) {
   return vmull_high_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vmull_high_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 int64x2_t test_vmull_high_lane_s32(int32x4_t a, int32x2_t v) {
   return vmull_high_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmull_high_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 uint32x4_t test_vmull_high_lane_u16(uint16x8_t a, uint16x4_t v) {
   return vmull_high_lane_u16(a, v, 3);
 }

 // CHECK-LABEL: @test_vmull_high_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 uint64x2_t test_vmull_high_lane_u32(uint32x4_t a, uint32x2_t v) {
   return vmull_high_lane_u32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmull_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 int32x4_t test_vmull_laneq_s16(int16x4_t a, int16x8_t v) {
   return vmull_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vmull_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 int64x2_t test_vmull_laneq_s32(int32x2_t a, int32x4_t v) {
   return vmull_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmull_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 uint32x4_t test_vmull_laneq_u16(uint16x4_t a, uint16x8_t v) {
   return vmull_laneq_u16(a, v, 7);
 }

 // CHECK-LABEL: @test_vmull_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 uint64x2_t test_vmull_laneq_u32(uint32x2_t a, uint32x4_t v) {
   return vmull_laneq_u32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmull_high_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 int32x4_t test_vmull_high_laneq_s16(int16x8_t a, int16x8_t v) {
   return vmull_high_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vmull_high_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 int64x2_t test_vmull_high_laneq_s32(int32x4_t a, int32x4_t v) {
   return vmull_high_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmull_high_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 uint32x4_t test_vmull_high_laneq_u16(uint16x8_t a, uint16x8_t v) {
   return vmull_high_laneq_u16(a, v, 7);
 }

 // CHECK-LABEL: @test_vmull_high_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 uint64x2_t test_vmull_high_laneq_u32(uint32x4_t a, uint32x4_t v) {
   return vmull_high_laneq_u32(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlal_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I]]
 //
 int32x4_t test_vqdmlal_lane_s16(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vqdmlal_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlal_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I]]
 //
 int64x2_t test_vqdmlal_lane_s32(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vqdmlal_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vqdmlal_high_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I]]
 //
 int32x4_t test_vqdmlal_high_lane_s16(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vqdmlal_high_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlal_high_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I]]
 //
 int64x2_t test_vqdmlal_high_lane_s32(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vqdmlal_high_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vqdmlsl_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I]]
 //
 int32x4_t test_vqdmlsl_lane_s16(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vqdmlsl_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlsl_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I]]
 //
 int64x2_t test_vqdmlsl_lane_s32(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vqdmlsl_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I]]
 //
 int32x4_t test_vqdmlsl_high_lane_s16(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vqdmlsl_high_lane_s16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I]]
 //
 int64x2_t test_vqdmlsl_high_lane_s32(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vqdmlsl_high_lane_s32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vqdmull_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I]]
 //
 int32x4_t test_vqdmull_lane_s16(int16x4_t a, int16x4_t v) {
   return vqdmull_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmull_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I]]
 //
 int64x2_t test_vqdmull_lane_s32(int32x2_t a, int32x2_t v) {
   return vqdmull_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vqdmull_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I]]
 //
 int32x4_t test_vqdmull_laneq_s16(int16x4_t a, int16x8_t v) {
   return vqdmull_laneq_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmull_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I]]
 //
 int64x2_t test_vqdmull_laneq_s32(int32x2_t a, int32x4_t v) {
   return vqdmull_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmull_high_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I]]
 //
 int32x4_t test_vqdmull_high_lane_s16(int16x8_t a, int16x4_t v) {
   return vqdmull_high_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmull_high_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I]]
 //
 int64x2_t test_vqdmull_high_lane_s32(int32x4_t a, int32x2_t v) {
   return vqdmull_high_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vqdmull_high_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I]]
 //
 int32x4_t test_vqdmull_high_laneq_s16(int16x8_t a, int16x8_t v) {
   return vqdmull_high_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vqdmull_high_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I]]
 //
 int64x2_t test_vqdmull_high_laneq_s32(int32x4_t a, int32x4_t v) {
   return vqdmull_high_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmulh_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULH_LANE_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[VQDMULH_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
 // CHECK-NEXT:    [[VQDMULH_LANE_V2:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.lane.v4i16.v4i16(<4 x i16> [[VQDMULH_LANE_V]], <4 x i16> [[VQDMULH_LANE_V1]], i32 3)
 // CHECK-NEXT:    ret <4 x i16> [[VQDMULH_LANE_V2]]
 //
 int16x4_t test_vqdmulh_lane_s16(int16x4_t a, int16x4_t v) {
   return vqdmulh_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmulhq_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V2:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqdmulh.lane.v8i16.v4i16(<8 x i16> [[VQDMULHQ_LANE_V]], <4 x i16> [[VQDMULHQ_LANE_V1]], i32 3)
 // CHECK-NEXT:    ret <8 x i16> [[VQDMULHQ_LANE_V2]]
 //
 int16x8_t test_vqdmulhq_lane_s16(int16x8_t a, int16x4_t v) {
   return vqdmulhq_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmulh_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULH_LANE_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[VQDMULH_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
 // CHECK-NEXT:    [[VQDMULH_LANE_V2:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqdmulh.lane.v2i32.v2i32(<2 x i32> [[VQDMULH_LANE_V]], <2 x i32> [[VQDMULH_LANE_V1]], i32 1)
 // CHECK-NEXT:    ret <2 x i32> [[VQDMULH_LANE_V2]]
 //
 int32x2_t test_vqdmulh_lane_s32(int32x2_t a, int32x2_t v) {
   return vqdmulh_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vqdmulhq_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V2:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmulh.lane.v4i32.v2i32(<4 x i32> [[VQDMULHQ_LANE_V]], <2 x i32> [[VQDMULHQ_LANE_V1]], i32 1)
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULHQ_LANE_V2]]
 //
 int32x4_t test_vqdmulhq_lane_s32(int32x4_t a, int32x2_t v) {
   return vqdmulhq_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vqrdmulh_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V2:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.lane.v4i16.v4i16(<4 x i16> [[VQRDMULH_LANE_V]], <4 x i16> [[VQRDMULH_LANE_V1]], i32 3)
 // CHECK-NEXT:    ret <4 x i16> [[VQRDMULH_LANE_V2]]
 //
 int16x4_t test_vqrdmulh_lane_s16(int16x4_t a, int16x4_t v) {
   return vqrdmulh_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vqrdmulhq_lane_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V2:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqrdmulh.lane.v8i16.v4i16(<8 x i16> [[VQRDMULHQ_LANE_V]], <4 x i16> [[VQRDMULHQ_LANE_V1]], i32 3)
 // CHECK-NEXT:    ret <8 x i16> [[VQRDMULHQ_LANE_V2]]
 //
 int16x8_t test_vqrdmulhq_lane_s16(int16x8_t a, int16x4_t v) {
   return vqrdmulhq_lane_s16(a, v, 3);
 }

 // CHECK-LABEL: @test_vqrdmulh_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V2:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqrdmulh.lane.v2i32.v2i32(<2 x i32> [[VQRDMULH_LANE_V]], <2 x i32> [[VQRDMULH_LANE_V1]], i32 1)
 // CHECK-NEXT:    ret <2 x i32> [[VQRDMULH_LANE_V2]]
 //
 int32x2_t test_vqrdmulh_lane_s32(int32x2_t a, int32x2_t v) {
   return vqrdmulh_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vqrdmulhq_lane_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V2:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqrdmulh.lane.v4i32.v2i32(<4 x i32> [[VQRDMULHQ_LANE_V]], <2 x i32> [[VQRDMULHQ_LANE_V1]], i32 1)
 // CHECK-NEXT:    ret <4 x i32> [[VQRDMULHQ_LANE_V2]]
 //
 int32x4_t test_vqrdmulhq_lane_s32(int32x4_t a, int32x2_t v) {
   return vqrdmulhq_lane_s32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmul_lane_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <2 x float> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x float> [[MUL]]
 //
 float32x2_t test_vmul_lane_f32(float32x2_t a, float32x2_t v) {
   return vmul_lane_f32(a, v, 1);
 }


 // CHECK-LABEL: @test_vmul_lane_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to double
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
 // CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <1 x double> [[TMP3]], i32 0
 // CHECK-NEXT:    [[TMP4:%.*]] = fmul double [[TMP2]], [[EXTRACT]]
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast double [[TMP4]] to <1 x double>
 // CHECK-NEXT:    ret <1 x double> [[TMP5]]
 //
 float64x1_t test_vmul_lane_f64(float64x1_t a, float64x1_t v) {
   return vmul_lane_f64(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_lane_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <4 x float> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x float> [[MUL]]
 //
 float32x4_t test_vmulq_lane_f32(float32x4_t a, float32x2_t v) {
   return vmulq_lane_f32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmulq_lane_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x double> [[TMP1]], <1 x double> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <2 x double> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x double> [[MUL]]
 //
 float64x2_t test_vmulq_lane_f64(float64x2_t a, float64x1_t v) {
   return vmulq_lane_f64(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <2 x float> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x float> [[MUL]]
 //
 float32x2_t test_vmul_laneq_f32(float32x2_t a, float32x4_t v) {
   return vmul_laneq_f32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmul_laneq_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to double
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
 // CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <2 x double> [[TMP3]], i32 1
 // CHECK-NEXT:    [[TMP4:%.*]] = fmul double [[TMP2]], [[EXTRACT]]
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast double [[TMP4]] to <1 x double>
 // CHECK-NEXT:    ret <1 x double> [[TMP5]]
 //
 float64x1_t test_vmul_laneq_f64(float64x1_t a, float64x2_t v) {
   return vmul_laneq_f64(a, v, 1);
 }

 // CHECK-LABEL: @test_vmulq_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <4 x float> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x float> [[MUL]]
 //
 float32x4_t test_vmulq_laneq_f32(float32x4_t a, float32x4_t v) {
   return vmulq_laneq_f32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmulq_laneq_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <2 x double> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x double> [[MUL]]
 //
 float64x2_t test_vmulq_laneq_f64(float64x2_t a, float64x2_t v) {
   return vmulq_laneq_f64(a, v, 1);
 }

 // CHECK-LABEL: @test_vmulx_lane_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x float> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmulx.v2f32(<2 x float> [[A]], <2 x float> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x float> [[VMULX2_I]]
 //
 float32x2_t test_vmulx_lane_f32(float32x2_t a, float32x2_t v) {
   return vmulx_lane_f32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmulxq_lane_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x float> [[LANE]] to <16 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmulx.v4f32(<4 x float> [[A]], <4 x float> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x float> [[VMULX2_I]]
 //
 float32x4_t test_vmulxq_lane_f32(float32x4_t a, float32x2_t v) {
   return vmulxq_lane_f32(a, v, 1);
 }

 // CHECK-LABEL: @test_vmulxq_lane_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x double> [[TMP1]], <1 x double> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x double> [[LANE]] to <16 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <2 x double> @llvm.aarch64.neon.fmulx.v2f64(<2 x double> [[A]], <2 x double> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x double> [[VMULX2_I]]
 //
 float64x2_t test_vmulxq_lane_f64(float64x2_t a, float64x1_t v) {
   return vmulxq_lane_f64(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulx_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x float> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmulx.v2f32(<2 x float> [[A]], <2 x float> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x float> [[VMULX2_I]]
 //
 float32x2_t test_vmulx_laneq_f32(float32x2_t a, float32x4_t v) {
   return vmulx_laneq_f32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmulxq_laneq_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x float> [[LANE]] to <16 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmulx.v4f32(<4 x float> [[A]], <4 x float> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x float> [[VMULX2_I]]
 //
 float32x4_t test_vmulxq_laneq_f32(float32x4_t a, float32x4_t v) {
   return vmulxq_laneq_f32(a, v, 3);
 }

 // CHECK-LABEL: @test_vmulxq_laneq_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x double> [[LANE]] to <16 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <2 x double> @llvm.aarch64.neon.fmulx.v2f64(<2 x double> [[A]], <2 x double> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x double> [[VMULX2_I]]
 //
 float64x2_t test_vmulxq_laneq_f64(float64x2_t a, float64x2_t v) {
   return vmulxq_laneq_f64(a, v, 1);
 }

 // CHECK-LABEL: @test_vmla_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD]]
 //
 int16x4_t test_vmla_lane_s16_0(int16x4_t a, int16x4_t b, int16x4_t v) {
   return vmla_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlaq_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD]]
 //
 int16x8_t test_vmlaq_lane_s16_0(int16x8_t a, int16x8_t b, int16x4_t v) {
   return vmlaq_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmla_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD]]
 //
 int32x2_t test_vmla_lane_s32_0(int32x2_t a, int32x2_t b, int32x2_t v) {
   return vmla_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlaq_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlaq_lane_s32_0(int32x4_t a, int32x4_t b, int32x2_t v) {
   return vmlaq_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmla_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD]]
 //
 int16x4_t test_vmla_laneq_s16_0(int16x4_t a, int16x4_t b, int16x8_t v) {
   return vmla_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlaq_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD]]
 //
 int16x8_t test_vmlaq_laneq_s16_0(int16x8_t a, int16x8_t b, int16x8_t v) {
   return vmlaq_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmla_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD]]
 //
 int32x2_t test_vmla_laneq_s32_0(int32x2_t a, int32x2_t b, int32x4_t v) {
   return vmla_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlaq_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlaq_laneq_s32_0(int32x4_t a, int32x4_t b, int32x4_t v) {
   return vmlaq_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmls_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB]]
 //
 int16x4_t test_vmls_lane_s16_0(int16x4_t a, int16x4_t b, int16x4_t v) {
   return vmls_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsq_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB]]
 //
 int16x8_t test_vmlsq_lane_s16_0(int16x8_t a, int16x8_t b, int16x4_t v) {
   return vmlsq_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmls_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB]]
 //
 int32x2_t test_vmls_lane_s32_0(int32x2_t a, int32x2_t b, int32x2_t v) {
   return vmls_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsq_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsq_lane_s32_0(int32x4_t a, int32x4_t b, int32x2_t v) {
   return vmlsq_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmls_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB]]
 //
 int16x4_t test_vmls_laneq_s16_0(int16x4_t a, int16x4_t b, int16x8_t v) {
   return vmls_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsq_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB]]
 //
 int16x8_t test_vmlsq_laneq_s16_0(int16x8_t a, int16x8_t b, int16x8_t v) {
   return vmlsq_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmls_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB]]
 //
 int32x2_t test_vmls_laneq_s32_0(int32x2_t a, int32x2_t b, int32x4_t v) {
   return vmls_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsq_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsq_laneq_s32_0(int32x4_t a, int32x4_t b, int32x4_t v) {
   return vmlsq_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmul_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL]]
 //
 int16x4_t test_vmul_lane_s16_0(int16x4_t a, int16x4_t v) {
   return vmul_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL]]
 //
 int16x8_t test_vmulq_lane_s16_0(int16x8_t a, int16x4_t v) {
   return vmulq_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL]]
 //
 int32x2_t test_vmul_lane_s32_0(int32x2_t a, int32x2_t v) {
   return vmul_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL]]
 //
 int32x4_t test_vmulq_lane_s32_0(int32x4_t a, int32x2_t v) {
   return vmulq_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL]]
 //
 uint16x4_t test_vmul_lane_u16_0(uint16x4_t a, uint16x4_t v) {
   return vmul_lane_u16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL]]
 //
 uint16x8_t test_vmulq_lane_u16_0(uint16x8_t a, uint16x4_t v) {
   return vmulq_lane_u16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL]]
 //
 uint32x2_t test_vmul_lane_u32_0(uint32x2_t a, uint32x2_t v) {
   return vmul_lane_u32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL]]
 //
 uint32x4_t test_vmulq_lane_u32_0(uint32x4_t a, uint32x2_t v) {
   return vmulq_lane_u32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL]]
 //
 int16x4_t test_vmul_laneq_s16_0(int16x4_t a, int16x8_t v) {
   return vmul_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL]]
 //
 int16x8_t test_vmulq_laneq_s16_0(int16x8_t a, int16x8_t v) {
   return vmulq_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL]]
 //
 int32x2_t test_vmul_laneq_s32_0(int32x2_t a, int32x4_t v) {
   return vmul_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL]]
 //
 int32x4_t test_vmulq_laneq_s32_0(int32x4_t a, int32x4_t v) {
   return vmulq_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL]]
 //
 uint16x4_t test_vmul_laneq_u16_0(uint16x4_t a, uint16x8_t v) {
   return vmul_laneq_u16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL]]
 //
 uint16x8_t test_vmulq_laneq_u16_0(uint16x8_t a, uint16x8_t v) {
   return vmulq_laneq_u16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL]]
 //
 uint32x2_t test_vmul_laneq_u32_0(uint32x2_t a, uint32x4_t v) {
   return vmul_laneq_u32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL]]
 //
 uint32x4_t test_vmulq_laneq_u32_0(uint32x4_t a, uint32x4_t v) {
   return vmulq_laneq_u32(a, v, 0);
 }

 // CHECK-LABEL: @test_vfma_lane_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[V:%.*]] 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:    [[FMLA:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[FMLA]], <2 x float> [[LANE]], <2 x float> [[FMLA1]])
 // CHECK-NEXT:    ret <2 x float> [[FMLA2]]
 //
 float32x2_t test_vfma_lane_f32_0(float32x2_t a, float32x2_t b, float32x2_t v) {
   return vfma_lane_f32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmaq_lane_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[V:%.*]] 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:    [[FMLA:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[FMLA]], <4 x float> [[LANE]], <4 x float> [[FMLA1]])
 // CHECK-NEXT:    ret <4 x float> [[FMLA2]]
 //
 float32x4_t test_vfmaq_lane_f32_0(float32x4_t a, float32x4_t b, float32x2_t v) {
   return vfmaq_lane_f32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfma_laneq_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP5]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP6:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[LANE]], <2 x float> [[TMP4]], <2 x float> [[TMP3]])
 // CHECK-NEXT:    ret <2 x float> [[TMP6]]
 //
 float32x2_t test_vfma_laneq_f32_0(float32x2_t a, float32x2_t b, float32x4_t v) {
   return vfma_laneq_f32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmaq_laneq_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP5]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP6:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[LANE]], <4 x float> [[TMP4]], <4 x float> [[TMP3]])
 // CHECK-NEXT:    ret <4 x float> [[TMP6]]
 //
 float32x4_t test_vfmaq_laneq_f32_0(float32x4_t a, float32x4_t b, float32x4_t v) {
   return vfmaq_laneq_f32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfms_lane_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <2 x float> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[FNEG]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[V:%.*]] 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:    [[FMLA:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[FMLA]], <2 x float> [[LANE]], <2 x float> [[FMLA1]])
 // CHECK-NEXT:    ret <2 x float> [[FMLA2]]
 //
 float32x2_t test_vfms_lane_f32_0(float32x2_t a, float32x2_t b, float32x2_t v) {
   return vfms_lane_f32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmsq_lane_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <4 x float> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[FNEG]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[V:%.*]] 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:    [[FMLA:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
 // CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[FMLA2:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[FMLA]], <4 x float> [[LANE]], <4 x float> [[FMLA1]])
 // CHECK-NEXT:    ret <4 x float> [[FMLA2]]
 //
 float32x4_t test_vfmsq_lane_f32_0(float32x4_t a, float32x4_t b, float32x2_t v) {
   return vfmsq_lane_f32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfms_laneq_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <2 x float> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[FNEG]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP5]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP6:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[LANE]], <2 x float> [[TMP4]], <2 x float> [[TMP3]])
 // CHECK-NEXT:    ret <2 x float> [[TMP6]]
 //
 float32x2_t test_vfms_laneq_f32_0(float32x2_t a, float32x2_t b, float32x4_t v) {
   return vfms_laneq_f32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmsq_laneq_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <4 x float> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[FNEG]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP5]], <4 x float> [[TMP5]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP6:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[LANE]], <4 x float> [[TMP4]], <4 x float> [[TMP3]])
 // CHECK-NEXT:    ret <4 x float> [[TMP6]]
 //
 float32x4_t test_vfmsq_laneq_f32_0(float32x4_t a, float32x4_t b, float32x4_t v) {
   return vfmsq_laneq_f32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmaq_laneq_f64_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP5]], <2 x double> [[TMP5]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP6:%.*]] = call <2 x double> @llvm.fma.v2f64(<2 x double> [[LANE]], <2 x double> [[TMP4]], <2 x double> [[TMP3]])
 // CHECK-NEXT:    ret <2 x double> [[TMP6]]
 //
 float64x2_t test_vfmaq_laneq_f64_0(float64x2_t a, float64x2_t b, float64x2_t v) {
   return vfmaq_laneq_f64(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vfmsq_laneq_f64_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG:%.*]] = fneg <2 x double> [[B:%.*]]
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[FNEG]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP5]], <2 x double> [[TMP5]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP6:%.*]] = call <2 x double> @llvm.fma.v2f64(<2 x double> [[LANE]], <2 x double> [[TMP4]], <2 x double> [[TMP3]])
 // CHECK-NEXT:    ret <2 x double> [[TMP6]]
 //
 float64x2_t test_vfmsq_laneq_f64_0(float64x2_t a, float64x2_t b, float64x2_t v) {
   return vfmsq_laneq_f64(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_lane_s16_0(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vmlal_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_lane_s32_0(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vmlal_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_laneq_s16_0(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vmlal_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_laneq_s32_0(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vmlal_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_high_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_high_lane_s16_0(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vmlal_high_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_high_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_high_lane_s32_0(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vmlal_high_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_high_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_high_laneq_s16_0(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vmlal_high_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_high_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_high_laneq_s32_0(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vmlal_high_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_lane_s16_0(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vmlsl_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_lane_s32_0(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vmlsl_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_laneq_s16_0(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vmlsl_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_laneq_s32_0(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vmlsl_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_high_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_high_lane_s16_0(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vmlsl_high_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_high_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_high_lane_s32_0(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vmlsl_high_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_high_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_high_laneq_s16_0(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vmlsl_high_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_high_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_high_laneq_s32_0(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vmlsl_high_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_lane_u16_0(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vmlal_lane_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_lane_u32_0(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vmlal_lane_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_laneq_u16_0(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vmlal_laneq_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_laneq_u32_0(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vmlal_laneq_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_high_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_high_lane_u16_0(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vmlal_high_lane_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_high_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_high_lane_u32_0(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vmlal_high_lane_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_high_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 int32x4_t test_vmlal_high_laneq_u16_0(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vmlal_high_laneq_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlal_high_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD]]
 //
 int64x2_t test_vmlal_high_laneq_u32_0(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vmlal_high_laneq_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_lane_u16_0(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vmlsl_lane_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_lane_u32_0(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vmlsl_lane_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_laneq_u16_0(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vmlsl_laneq_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_laneq_u32_0(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vmlsl_laneq_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_high_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_high_lane_u16_0(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vmlsl_high_lane_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_high_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_high_lane_u32_0(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vmlsl_high_lane_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_high_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 int32x4_t test_vmlsl_high_laneq_u16_0(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vmlsl_high_laneq_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsl_high_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB]]
 //
 int64x2_t test_vmlsl_high_laneq_u32_0(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vmlsl_high_laneq_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmull_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 int32x4_t test_vmull_lane_s16_0(int16x4_t a, int16x4_t v) {
   return vmull_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 int64x2_t test_vmull_lane_s32_0(int32x2_t a, int32x2_t v) {
   return vmull_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 uint32x4_t test_vmull_lane_u16_0(uint16x4_t a, uint16x4_t v) {
   return vmull_lane_u16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 uint64x2_t test_vmull_lane_u32_0(uint32x2_t a, uint32x2_t v) {
   return vmull_lane_u32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 int32x4_t test_vmull_high_lane_s16_0(int16x8_t a, int16x4_t v) {
   return vmull_high_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 int64x2_t test_vmull_high_lane_s32_0(int32x4_t a, int32x2_t v) {
   return vmull_high_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 uint32x4_t test_vmull_high_lane_u16_0(uint16x8_t a, uint16x4_t v) {
   return vmull_high_lane_u16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 uint64x2_t test_vmull_high_lane_u32_0(uint32x4_t a, uint32x2_t v) {
   return vmull_high_lane_u32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 int32x4_t test_vmull_laneq_s16_0(int16x4_t a, int16x8_t v) {
   return vmull_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 int64x2_t test_vmull_laneq_s32_0(int32x2_t a, int32x4_t v) {
   return vmull_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 uint32x4_t test_vmull_laneq_u16_0(uint16x4_t a, uint16x8_t v) {
   return vmull_laneq_u16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 uint64x2_t test_vmull_laneq_u32_0(uint32x2_t a, uint32x4_t v) {
   return vmull_laneq_u32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 int32x4_t test_vmull_high_laneq_s16_0(int16x8_t a, int16x8_t v) {
   return vmull_high_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 int64x2_t test_vmull_high_laneq_s32_0(int32x4_t a, int32x4_t v) {
   return vmull_high_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I]]
 //
 uint32x4_t test_vmull_high_laneq_u16_0(uint16x8_t a, uint16x8_t v) {
   return vmull_high_laneq_u16(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I]]
 //
 uint64x2_t test_vmull_high_laneq_u32_0(uint32x4_t a, uint32x4_t v) {
   return vmull_high_laneq_u32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlal_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I]]
 //
 int32x4_t test_vqdmlal_lane_s16_0(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vqdmlal_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlal_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I]]
 //
 int64x2_t test_vqdmlal_lane_s32_0(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vqdmlal_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlal_high_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I]]
 //
 int32x4_t test_vqdmlal_high_lane_s16_0(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vqdmlal_high_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlal_high_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I]]
 //
 int64x2_t test_vqdmlal_high_lane_s32_0(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vqdmlal_high_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlsl_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I]]
 //
 int32x4_t test_vqdmlsl_lane_s16_0(int32x4_t a, int16x4_t b, int16x4_t v) {
   return vqdmlsl_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlsl_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I]]
 //
 int64x2_t test_vqdmlsl_lane_s32_0(int64x2_t a, int32x2_t b, int32x2_t v) {
   return vqdmlsl_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I]]
 //
 int32x4_t test_vqdmlsl_high_lane_s16_0(int32x4_t a, int16x8_t b, int16x4_t v) {
   return vqdmlsl_high_lane_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I]]
 //
 int64x2_t test_vqdmlsl_high_lane_s32_0(int64x2_t a, int32x4_t b, int32x2_t v) {
   return vqdmlsl_high_lane_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmull_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I]]
 //
 int32x4_t test_vqdmull_lane_s16_0(int16x4_t a, int16x4_t v) {
   return vqdmull_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmull_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I]]
 //
 int64x2_t test_vqdmull_lane_s32_0(int32x2_t a, int32x2_t v) {
   return vqdmull_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmull_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[A]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I]]
 //
 int32x4_t test_vqdmull_laneq_s16_0(int16x4_t a, int16x8_t v) {
   return vqdmull_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmull_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[A]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I]]
 //
 int64x2_t test_vqdmull_laneq_s32_0(int32x2_t a, int32x4_t v) {
   return vqdmull_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmull_high_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I]]
 //
 int32x4_t test_vqdmull_high_lane_s16_0(int16x8_t a, int16x4_t v) {
   return vqdmull_high_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmull_high_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I]]
 //
 int64x2_t test_vqdmull_high_lane_s32_0(int32x4_t a, int32x2_t v) {
   return vqdmull_high_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmull_high_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I]]
 //
 int32x4_t test_vqdmull_high_laneq_s16_0(int16x8_t a, int16x8_t v) {
   return vqdmull_high_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmull_high_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I]]
 //
 int64x2_t test_vqdmull_high_laneq_s32_0(int32x4_t a, int32x4_t v) {
   return vqdmull_high_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmulh_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULH_LANE_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[VQDMULH_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
 // CHECK-NEXT:    [[VQDMULH_LANE_V2:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.lane.v4i16.v4i16(<4 x i16> [[VQDMULH_LANE_V]], <4 x i16> [[VQDMULH_LANE_V1]], i32 0)
 // CHECK-NEXT:    ret <4 x i16> [[VQDMULH_LANE_V2]]
 //
 int16x4_t test_vqdmulh_lane_s16_0(int16x4_t a, int16x4_t v) {
   return vqdmulh_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmulhq_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V2:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqdmulh.lane.v8i16.v4i16(<8 x i16> [[VQDMULHQ_LANE_V]], <4 x i16> [[VQDMULHQ_LANE_V1]], i32 0)
 // CHECK-NEXT:    ret <8 x i16> [[VQDMULHQ_LANE_V2]]
 //
 int16x8_t test_vqdmulhq_lane_s16_0(int16x8_t a, int16x4_t v) {
   return vqdmulhq_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmulh_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULH_LANE_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[VQDMULH_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
 // CHECK-NEXT:    [[VQDMULH_LANE_V2:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqdmulh.lane.v2i32.v2i32(<2 x i32> [[VQDMULH_LANE_V]], <2 x i32> [[VQDMULH_LANE_V1]], i32 0)
 // CHECK-NEXT:    ret <2 x i32> [[VQDMULH_LANE_V2]]
 //
 int32x2_t test_vqdmulh_lane_s32_0(int32x2_t a, int32x2_t v) {
   return vqdmulh_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmulhq_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
 // CHECK-NEXT:    [[VQDMULHQ_LANE_V2:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmulh.lane.v4i32.v2i32(<4 x i32> [[VQDMULHQ_LANE_V]], <2 x i32> [[VQDMULHQ_LANE_V1]], i32 0)
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULHQ_LANE_V2]]
 //
 int32x4_t test_vqdmulhq_lane_s32_0(int32x4_t a, int32x2_t v) {
   return vqdmulhq_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqrdmulh_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V2:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.lane.v4i16.v4i16(<4 x i16> [[VQRDMULH_LANE_V]], <4 x i16> [[VQRDMULH_LANE_V1]], i32 0)
 // CHECK-NEXT:    ret <4 x i16> [[VQRDMULH_LANE_V2]]
 //
 int16x4_t test_vqrdmulh_lane_s16_0(int16x4_t a, int16x4_t v) {
   return vqrdmulh_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqrdmulhq_lane_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V2:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqrdmulh.lane.v8i16.v4i16(<8 x i16> [[VQRDMULHQ_LANE_V]], <4 x i16> [[VQRDMULHQ_LANE_V1]], i32 0)
 // CHECK-NEXT:    ret <8 x i16> [[VQRDMULHQ_LANE_V2]]
 //
 int16x8_t test_vqrdmulhq_lane_s16_0(int16x8_t a, int16x4_t v) {
   return vqrdmulhq_lane_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqrdmulh_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
 // CHECK-NEXT:    [[VQRDMULH_LANE_V2:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqrdmulh.lane.v2i32.v2i32(<2 x i32> [[VQRDMULH_LANE_V]], <2 x i32> [[VQRDMULH_LANE_V1]], i32 0)
 // CHECK-NEXT:    ret <2 x i32> [[VQRDMULH_LANE_V2]]
 //
 int32x2_t test_vqrdmulh_lane_s32_0(int32x2_t a, int32x2_t v) {
   return vqrdmulh_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqrdmulhq_lane_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
 // CHECK-NEXT:    [[VQRDMULHQ_LANE_V2:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqrdmulh.lane.v4i32.v2i32(<4 x i32> [[VQRDMULHQ_LANE_V]], <2 x i32> [[VQRDMULHQ_LANE_V1]], i32 0)
 // CHECK-NEXT:    ret <4 x i32> [[VQRDMULHQ_LANE_V2]]
 //
 int32x4_t test_vqrdmulhq_lane_s32_0(int32x4_t a, int32x2_t v) {
   return vqrdmulhq_lane_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_lane_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <2 x float> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x float> [[MUL]]
 //
 float32x2_t test_vmul_lane_f32_0(float32x2_t a, float32x2_t v) {
   return vmul_lane_f32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_lane_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <4 x float> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x float> [[MUL]]
 //
 float32x4_t test_vmulq_lane_f32_0(float32x4_t a, float32x2_t v) {
   return vmulq_lane_f32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_laneq_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <2 x float> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x float> [[MUL]]
 //
 float32x2_t test_vmul_laneq_f32_0(float32x2_t a, float32x4_t v) {
   return vmul_laneq_f32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmul_laneq_f64_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to double
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
 // CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <2 x double> [[TMP3]], i32 0
 // CHECK-NEXT:    [[TMP4:%.*]] = fmul double [[TMP2]], [[EXTRACT]]
 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast double [[TMP4]] to <1 x double>
 // CHECK-NEXT:    ret <1 x double> [[TMP5]]
 //
 float64x1_t test_vmul_laneq_f64_0(float64x1_t a, float64x2_t v) {
   return vmul_laneq_f64(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_laneq_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <4 x float> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <4 x float> [[MUL]]
 //
 float32x4_t test_vmulq_laneq_f32_0(float32x4_t a, float32x4_t v) {
   return vmulq_laneq_f32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulq_laneq_f64_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = fmul <2 x double> [[A:%.*]], [[LANE]]
 // CHECK-NEXT:    ret <2 x double> [[MUL]]
 //
 float64x2_t test_vmulq_laneq_f64_0(float64x2_t a, float64x2_t v) {
   return vmulq_laneq_f64(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulx_lane_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x float> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmulx.v2f32(<2 x float> [[A]], <2 x float> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x float> [[VMULX2_I]]
 //
 float32x2_t test_vmulx_lane_f32_0(float32x2_t a, float32x2_t v) {
   return vmulx_lane_f32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulxq_lane_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x float> [[LANE]] to <16 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmulx.v4f32(<4 x float> [[A]], <4 x float> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x float> [[VMULX2_I]]
 //
 float32x4_t test_vmulxq_lane_f32_0(float32x4_t a, float32x2_t v) {
   return vmulxq_lane_f32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulxq_lane_f64_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x double> [[TMP1]], <1 x double> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x double> [[LANE]] to <16 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <2 x double> @llvm.aarch64.neon.fmulx.v2f64(<2 x double> [[A]], <2 x double> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x double> [[VMULX2_I]]
 //
 float64x2_t test_vmulxq_lane_f64_0(float64x2_t a, float64x1_t v) {
   return vmulxq_lane_f64(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulx_laneq_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x float> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmulx.v2f32(<2 x float> [[A]], <2 x float> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x float> [[VMULX2_I]]
 //
 float32x2_t test_vmulx_laneq_f32_0(float32x2_t a, float32x4_t v) {
   return vmulx_laneq_f32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulxq_laneq_f32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x float> [[LANE]] to <16 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmulx.v4f32(<4 x float> [[A]], <4 x float> [[LANE]]) #4
 // CHECK-NEXT:    ret <4 x float> [[VMULX2_I]]
 //
 float32x4_t test_vmulxq_laneq_f32_0(float32x4_t a, float32x4_t v) {
   return vmulxq_laneq_f32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmulxq_laneq_f64_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x double> [[LANE]] to <16 x i8>
 // CHECK-NEXT:    [[VMULX2_I:%.*]] = call <2 x double> @llvm.aarch64.neon.fmulx.v2f64(<2 x double> [[A]], <2 x double> [[LANE]]) #4
 // CHECK-NEXT:    ret <2 x double> [[VMULX2_I]]
 //
 float64x2_t test_vmulxq_laneq_f64_0(float64x2_t a, float64x2_t v) {
   return vmulxq_laneq_f64(a, v, 0);
 }

 // CHECK-LABEL: @test_vmull_high_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I_I_I]]
 //
 int32x4_t test_vmull_high_n_s16(int16x8_t a, int16_t b) {
   return vmull_high_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vmull_high_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I_I_I]]
 //
 int64x2_t test_vmull_high_n_s32(int32x4_t a, int32_t b) {
   return vmull_high_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vmull_high_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I_I_I]]
 //
 uint32x4_t test_vmull_high_n_u16(uint16x8_t a, uint16_t b) {
   return vmull_high_n_u16(a, b);
 }

 // CHECK-LABEL: @test_vmull_high_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I_I_I]]
 //
 uint64x2_t test_vmull_high_n_u32(uint32x4_t a, uint32_t b) {
   return vmull_high_n_u32(a, b);
 }

 // CHECK-LABEL: @test_vqdmull_high_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[A:%.*]], <8 x i16> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I_I_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I_I_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I_I_I]]
 //
 int32x4_t test_vqdmull_high_n_s16(int16x8_t a, int16_t b) {
   return vqdmull_high_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vqdmull_high_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[A:%.*]], <4 x i32> [[A]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I_I_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I_I_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I_I_I]]
 //
 int64x2_t test_vqdmull_high_n_s32(int32x4_t a, int32_t b) {
   return vqdmull_high_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vmlal_high_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    [[ADD_I_I:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I_I_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD_I_I]]
 //
 int32x4_t test_vmlal_high_n_s16(int32x4_t a, int16x8_t b, int16_t c) {
   return vmlal_high_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlal_high_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    [[ADD_I_I:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I_I_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD_I_I]]
 //
 int64x2_t test_vmlal_high_n_s32(int64x2_t a, int32x4_t b, int32_t c) {
   return vmlal_high_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlal_high_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    [[ADD_I_I:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I_I_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD_I_I]]
 //
 uint32x4_t test_vmlal_high_n_u16(uint32x4_t a, uint16x8_t b, uint16_t c) {
   return vmlal_high_n_u16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlal_high_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    [[ADD_I_I:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I_I_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD_I_I]]
 //
 uint64x2_t test_vmlal_high_n_u32(uint64x2_t a, uint32x4_t b, uint32_t c) {
   return vmlal_high_n_u32(a, b, c);
 }

 // CHECK-LABEL: @test_vqdmlal_high_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I_I_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I_I_I]]
 //
 int32x4_t test_vqdmlal_high_n_s16(int32x4_t a, int16x8_t b, int16_t c) {
   return vqdmlal_high_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vqdmlal_high_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I_I_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I_I_I]]
 //
 int64x2_t test_vqdmlal_high_n_s32(int64x2_t a, int32x4_t b, int32_t c) {
   return vqdmlal_high_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsl_high_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    [[SUB_I_I:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I_I_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB_I_I]]
 //
 int32x4_t test_vmlsl_high_n_s16(int32x4_t a, int16x8_t b, int16_t c) {
   return vmlsl_high_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsl_high_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    [[SUB_I_I:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I_I_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB_I_I]]
 //
 int64x2_t test_vmlsl_high_n_s32(int64x2_t a, int32x4_t b, int32_t c) {
   return vmlsl_high_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsl_high_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    [[SUB_I_I:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I_I_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB_I_I]]
 //
 uint32x4_t test_vmlsl_high_n_u16(uint32x4_t a, uint16x8_t b, uint16_t c) {
   return vmlsl_high_n_u16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsl_high_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    [[SUB_I_I:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I_I_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB_I_I]]
 //
 uint64x2_t test_vmlsl_high_n_u32(uint64x2_t a, uint32x4_t b, uint32_t c) {
   return vmlsl_high_n_u32(a, b, c);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <4 x i16> [[VECINIT_I_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[VECINIT3_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I_I]], <4 x i16> [[VECINIT3_I_I]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I_I_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I_I_I]]
 //
 int32x4_t test_vqdmlsl_high_n_s16(int32x4_t a, int16x8_t b, int16_t c) {
   return vqdmlsl_high_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[VECINIT_I_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I_I:%.*]] = insertelement <2 x i32> [[VECINIT_I_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[SHUFFLE_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[VECINIT1_I_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I_I]], <2 x i32> [[VECINIT1_I_I]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I_I_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I_I_I]]
 //
 int64x2_t test_vqdmlsl_high_n_s32(int64x2_t a, int32x4_t b, int32_t c) {
   return vqdmlsl_high_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmul_n_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x float> undef, float [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x float> [[VECINIT_I]], float [[B]], i32 1
 // CHECK-NEXT:    [[MUL_I:%.*]] = fmul <2 x float> [[A:%.*]], [[VECINIT1_I]]
 // CHECK-NEXT:    ret <2 x float> [[MUL_I]]
 //
 float32x2_t test_vmul_n_f32(float32x2_t a, float32_t b) {
   return vmul_n_f32(a, b);
 }

 // CHECK-LABEL: @test_vmulq_n_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x float> undef, float [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x float> [[VECINIT_I]], float [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x float> [[VECINIT1_I]], float [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x float> [[VECINIT2_I]], float [[B]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = fmul <4 x float> [[A:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    ret <4 x float> [[MUL_I]]
 //
 float32x4_t test_vmulq_n_f32(float32x4_t a, float32_t b) {
   return vmulq_n_f32(a, b);
 }

 // CHECK-LABEL: @test_vmulq_n_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x double> undef, double [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x double> [[VECINIT_I]], double [[B]], i32 1
 // CHECK-NEXT:    [[MUL_I:%.*]] = fmul <2 x double> [[A:%.*]], [[VECINIT1_I]]
 // CHECK-NEXT:    ret <2 x double> [[MUL_I]]
 //
 float64x2_t test_vmulq_n_f64(float64x2_t a, float64_t b) {
   return vmulq_n_f64(a, b);
 }

 // CHECK-LABEL: @test_vfma_n_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x float> undef, float [[N:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x float> [[VECINIT_I]], float [[N]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[B]], <2 x float> [[VECINIT1_I]], <2 x float> [[A]]) #4
 // CHECK-NEXT:    ret <2 x float> [[TMP3]]
 //
 float32x2_t test_vfma_n_f32(float32x2_t a, float32x2_t b, float32_t n) {
   return vfma_n_f32(a, b, n);
 }

 // CHECK-LABEL: @test_vfma_n_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <1 x double> undef, double [[N:%.*]], i32 0
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <1 x double> [[VECINIT_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = call <1 x double> @llvm.fma.v1f64(<1 x double> [[B]], <1 x double> [[VECINIT_I]], <1 x double> [[A]]) #4
 // CHECK-NEXT:    ret <1 x double> [[TMP3]]
 //
 float64x1_t test_vfma_n_f64(float64x1_t a, float64x1_t b, float64_t n) {
   return vfma_n_f64(a, b, n);
 }

 // CHECK-LABEL: @test_vfmaq_n_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x float> undef, float [[N:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x float> [[VECINIT_I]], float [[N]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x float> [[VECINIT1_I]], float [[N]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x float> [[VECINIT2_I]], float [[N]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[VECINIT3_I]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[B]], <4 x float> [[VECINIT3_I]], <4 x float> [[A]]) #4
 // CHECK-NEXT:    ret <4 x float> [[TMP3]]
 //
 float32x4_t test_vfmaq_n_f32(float32x4_t a, float32x4_t b, float32_t n) {
   return vfmaq_n_f32(a, b, n);
 }

 // CHECK-LABEL: @test_vfms_n_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG_I:%.*]] = fneg <2 x float> [[B:%.*]]
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x float> undef, float [[N:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x float> [[VECINIT_I]], float [[N]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[FNEG_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x float> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[FNEG_I]], <2 x float> [[VECINIT1_I]], <2 x float> [[A]]) #4
 // CHECK-NEXT:    ret <2 x float> [[TMP3]]
 //
 float32x2_t test_vfms_n_f32(float32x2_t a, float32x2_t b, float32_t n) {
   return vfms_n_f32(a, b, n);
 }

 // CHECK-LABEL: @test_vfms_n_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG_I:%.*]] = fneg <1 x double> [[B:%.*]]
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <1 x double> undef, double [[N:%.*]], i32 0
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[FNEG_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <1 x double> [[VECINIT_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = call <1 x double> @llvm.fma.v1f64(<1 x double> [[FNEG_I]], <1 x double> [[VECINIT_I]], <1 x double> [[A]]) #4
 // CHECK-NEXT:    ret <1 x double> [[TMP3]]
 //
 float64x1_t test_vfms_n_f64(float64x1_t a, float64x1_t b, float64_t n) {
   return vfms_n_f64(a, b, n);
 }

 // CHECK-LABEL: @test_vfmsq_n_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FNEG_I:%.*]] = fneg <4 x float> [[B:%.*]]
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x float> undef, float [[N:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x float> [[VECINIT_I]], float [[N]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x float> [[VECINIT1_I]], float [[N]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x float> [[VECINIT2_I]], float [[N]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[FNEG_I]] to <16 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[VECINIT3_I]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[FNEG_I]], <4 x float> [[VECINIT3_I]], <4 x float> [[A]]) #4
 // CHECK-NEXT:    ret <4 x float> [[TMP3]]
 //
 float32x4_t test_vfmsq_n_f32(float32x4_t a, float32x4_t b, float32_t n) {
   return vfmsq_n_f32(a, b, n);
 }

 // CHECK-LABEL: @test_vmul_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i16> [[A:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL_I]]
 //
 int16x4_t test_vmul_n_s16(int16x4_t a, int16_t b) {
   return vmul_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vmulq_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x i16> [[VECINIT3_I]], i16 [[B]], i32 4
 // CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x i16> [[VECINIT4_I]], i16 [[B]], i32 5
 // CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x i16> [[VECINIT5_I]], i16 [[B]], i32 6
 // CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x i16> [[VECINIT6_I]], i16 [[B]], i32 7
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <8 x i16> [[A:%.*]], [[VECINIT7_I]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL_I]]
 //
 int16x8_t test_vmulq_n_s16(int16x8_t a, int16_t b) {
   return vmulq_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vmul_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <2 x i32> [[A:%.*]], [[VECINIT1_I]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL_I]]
 //
 int32x2_t test_vmul_n_s32(int32x2_t a, int32_t b) {
   return vmul_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vmulq_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i32> [[VECINIT1_I]], i32 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i32> [[VECINIT2_I]], i32 [[B]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i32> [[A:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL_I]]
 //
 int32x4_t test_vmulq_n_s32(int32x4_t a, int32_t b) {
   return vmulq_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vmul_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i16> [[A:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    ret <4 x i16> [[MUL_I]]
 //
 uint16x4_t test_vmul_n_u16(uint16x4_t a, uint16_t b) {
   return vmul_n_u16(a, b);
 }

 // CHECK-LABEL: @test_vmulq_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x i16> [[VECINIT3_I]], i16 [[B]], i32 4
 // CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x i16> [[VECINIT4_I]], i16 [[B]], i32 5
 // CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x i16> [[VECINIT5_I]], i16 [[B]], i32 6
 // CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x i16> [[VECINIT6_I]], i16 [[B]], i32 7
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <8 x i16> [[A:%.*]], [[VECINIT7_I]]
 // CHECK-NEXT:    ret <8 x i16> [[MUL_I]]
 //
 uint16x8_t test_vmulq_n_u16(uint16x8_t a, uint16_t b) {
   return vmulq_n_u16(a, b);
 }

 // CHECK-LABEL: @test_vmul_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <2 x i32> [[A:%.*]], [[VECINIT1_I]]
 // CHECK-NEXT:    ret <2 x i32> [[MUL_I]]
 //
 uint32x2_t test_vmul_n_u32(uint32x2_t a, uint32_t b) {
   return vmul_n_u32(a, b);
 }

 // CHECK-LABEL: @test_vmulq_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i32> [[VECINIT1_I]], i32 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i32> [[VECINIT2_I]], i32 [[B]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i32> [[A:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    ret <4 x i32> [[MUL_I]]
 //
 uint32x4_t test_vmulq_n_u32(uint32x4_t a, uint32_t b) {
   return vmulq_n_u32(a, b);
 }

 // CHECK-LABEL: @test_vmull_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[A]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I_I]]
 //
 int32x4_t test_vmull_n_s16(int16x4_t a, int16_t b) {
   return vmull_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vmull_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[A]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I_I]]
 //
 int64x2_t test_vmull_n_s32(int32x2_t a, int32_t b) {
   return vmull_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vmull_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[A]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VMULL2_I_I]]
 //
 uint32x4_t test_vmull_n_u16(uint16x4_t a, uint16_t b) {
   return vmull_n_u16(a, b);
 }

 // CHECK-LABEL: @test_vmull_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[A]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VMULL2_I_I]]
 //
 uint64x2_t test_vmull_n_u32(uint32x2_t a, uint32_t b) {
   return vmull_n_u32(a, b);
 }

 // CHECK-LABEL: @test_vqdmull_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[A]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I_I:%.*]] = bitcast <4 x i32> [[VQDMULL_V2_I_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULL_V2_I_I]]
 //
 int32x4_t test_vqdmull_n_s16(int16x4_t a, int16_t b) {
   return vqdmull_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vqdmull_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULL_V2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[A]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[VQDMULL_V3_I_I:%.*]] = bitcast <2 x i64> [[VQDMULL_V2_I_I]] to <16 x i8>
 // CHECK-NEXT:    ret <2 x i64> [[VQDMULL_V2_I_I]]
 //
 int64x2_t test_vqdmull_n_s32(int32x2_t a, int32_t b) {
   return vqdmull_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vqdmulh_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULH_V2_I_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.v4i16(<4 x i16> [[A]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[VQDMULH_V3_I_I:%.*]] = bitcast <4 x i16> [[VQDMULH_V2_I_I]] to <8 x i8>
 // CHECK-NEXT:    ret <4 x i16> [[VQDMULH_V2_I_I]]
 //
 int16x4_t test_vqdmulh_n_s16(int16x4_t a, int16_t b) {
   return vqdmulh_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vqdmulhq_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x i16> [[VECINIT3_I]], i16 [[B]], i32 4
 // CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x i16> [[VECINIT4_I]], i16 [[B]], i32 5
 // CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x i16> [[VECINIT5_I]], i16 [[B]], i32 6
 // CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x i16> [[VECINIT6_I]], i16 [[B]], i32 7
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[VECINIT7_I]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_V2_I_I:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqdmulh.v8i16(<8 x i16> [[A]], <8 x i16> [[VECINIT7_I]]) #4
 // CHECK-NEXT:    [[VQDMULHQ_V3_I_I:%.*]] = bitcast <8 x i16> [[VQDMULHQ_V2_I_I]] to <16 x i8>
 // CHECK-NEXT:    ret <8 x i16> [[VQDMULHQ_V2_I_I]]
 //
 int16x8_t test_vqdmulhq_n_s16(int16x8_t a, int16_t b) {
   return vqdmulhq_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vqdmulh_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMULH_V2_I_I:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqdmulh.v2i32(<2 x i32> [[A]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[VQDMULH_V3_I_I:%.*]] = bitcast <2 x i32> [[VQDMULH_V2_I_I]] to <8 x i8>
 // CHECK-NEXT:    ret <2 x i32> [[VQDMULH_V2_I_I]]
 //
 int32x2_t test_vqdmulh_n_s32(int32x2_t a, int32_t b) {
   return vqdmulh_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vqdmulhq_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i32> [[VECINIT1_I]], i32 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i32> [[VECINIT2_I]], i32 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[VECINIT3_I]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_V2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmulh.v4i32(<4 x i32> [[A]], <4 x i32> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[VQDMULHQ_V3_I_I:%.*]] = bitcast <4 x i32> [[VQDMULHQ_V2_I_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULHQ_V2_I_I]]
 //
 int32x4_t test_vqdmulhq_n_s32(int32x4_t a, int32_t b) {
   return vqdmulhq_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vqrdmulh_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULH_V2_I_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.v4i16(<4 x i16> [[A]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[VQRDMULH_V3_I_I:%.*]] = bitcast <4 x i16> [[VQRDMULH_V2_I_I]] to <8 x i8>
 // CHECK-NEXT:    ret <4 x i16> [[VQRDMULH_V2_I_I]]
 //
 int16x4_t test_vqrdmulh_n_s16(int16x4_t a, int16_t b) {
   return vqrdmulh_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vqrdmulhq_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x i16> undef, i16 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x i16> [[VECINIT_I]], i16 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x i16> [[VECINIT1_I]], i16 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I]], i16 [[B]], i32 3
 // CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x i16> [[VECINIT3_I]], i16 [[B]], i32 4
 // CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x i16> [[VECINIT4_I]], i16 [[B]], i32 5
 // CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x i16> [[VECINIT5_I]], i16 [[B]], i32 6
 // CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x i16> [[VECINIT6_I]], i16 [[B]], i32 7
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[VECINIT7_I]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_V2_I_I:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqrdmulh.v8i16(<8 x i16> [[A]], <8 x i16> [[VECINIT7_I]]) #4
 // CHECK-NEXT:    [[VQRDMULHQ_V3_I_I:%.*]] = bitcast <8 x i16> [[VQRDMULHQ_V2_I_I]] to <16 x i8>
 // CHECK-NEXT:    ret <8 x i16> [[VQRDMULHQ_V2_I_I]]
 //
 int16x8_t test_vqrdmulhq_n_s16(int16x8_t a, int16_t b) {
   return vqrdmulhq_n_s16(a, b);
 }

 // CHECK-LABEL: @test_vqrdmulh_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQRDMULH_V2_I_I:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqrdmulh.v2i32(<2 x i32> [[A]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[VQRDMULH_V3_I_I:%.*]] = bitcast <2 x i32> [[VQRDMULH_V2_I_I]] to <8 x i8>
 // CHECK-NEXT:    ret <2 x i32> [[VQRDMULH_V2_I_I]]
 //
 int32x2_t test_vqrdmulh_n_s32(int32x2_t a, int32_t b) {
   return vqrdmulh_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vqrdmulhq_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i32> undef, i32 [[B:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i32> [[VECINIT_I]], i32 [[B]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i32> [[VECINIT1_I]], i32 [[B]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i32> [[VECINIT2_I]], i32 [[B]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[VECINIT3_I]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_V2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqrdmulh.v4i32(<4 x i32> [[A]], <4 x i32> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[VQRDMULHQ_V3_I_I:%.*]] = bitcast <4 x i32> [[VQRDMULHQ_V2_I_I]] to <16 x i8>
 // CHECK-NEXT:    ret <4 x i32> [[VQRDMULHQ_V2_I_I]]
 //
 int32x4_t test_vqrdmulhq_n_s32(int32x4_t a, int32_t b) {
   return vqrdmulhq_n_s32(a, b);
 }

 // CHECK-LABEL: @test_vmla_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i16> [[B:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <4 x i16> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD_I]]
 //
 int16x4_t test_vmla_n_s16(int16x4_t a, int16x4_t b, int16_t c) {
   return vmla_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlaq_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x i16> [[VECINIT3_I]], i16 [[C]], i32 4
 // CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x i16> [[VECINIT4_I]], i16 [[C]], i32 5
 // CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x i16> [[VECINIT5_I]], i16 [[C]], i32 6
 // CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x i16> [[VECINIT6_I]], i16 [[C]], i32 7
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <8 x i16> [[B:%.*]], [[VECINIT7_I]]
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <8 x i16> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD_I]]
 //
 int16x8_t test_vmlaq_n_s16(int16x8_t a, int16x8_t b, int16_t c) {
   return vmlaq_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vmla_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <2 x i32> [[B:%.*]], [[VECINIT1_I]]
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <2 x i32> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD_I]]
 //
 int32x2_t test_vmla_n_s32(int32x2_t a, int32x2_t b, int32_t c) {
   return vmla_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlaq_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i32> [[VECINIT1_I]], i32 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i32> [[VECINIT2_I]], i32 [[C]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i32> [[B:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <4 x i32> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD_I]]
 //
 int32x4_t test_vmlaq_n_s32(int32x4_t a, int32x4_t b, int32_t c) {
   return vmlaq_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmla_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i16> [[B:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <4 x i16> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD_I]]
 //
 uint16x4_t test_vmla_n_u16(uint16x4_t a, uint16x4_t b, uint16_t c) {
   return vmla_n_u16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlaq_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x i16> [[VECINIT3_I]], i16 [[C]], i32 4
 // CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x i16> [[VECINIT4_I]], i16 [[C]], i32 5
 // CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x i16> [[VECINIT5_I]], i16 [[C]], i32 6
 // CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x i16> [[VECINIT6_I]], i16 [[C]], i32 7
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <8 x i16> [[B:%.*]], [[VECINIT7_I]]
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <8 x i16> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD_I]]
 //
 uint16x8_t test_vmlaq_n_u16(uint16x8_t a, uint16x8_t b, uint16_t c) {
   return vmlaq_n_u16(a, b, c);
 }

 // CHECK-LABEL: @test_vmla_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <2 x i32> [[B:%.*]], [[VECINIT1_I]]
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <2 x i32> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD_I]]
 //
 uint32x2_t test_vmla_n_u32(uint32x2_t a, uint32x2_t b, uint32_t c) {
   return vmla_n_u32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlaq_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i32> [[VECINIT1_I]], i32 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i32> [[VECINIT2_I]], i32 [[C]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i32> [[B:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <4 x i32> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD_I]]
 //
 uint32x4_t test_vmlaq_n_u32(uint32x4_t a, uint32x4_t b, uint32_t c) {
   return vmlaq_n_u32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlal_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD_I]]
 //
 int32x4_t test_vmlal_n_s16(int32x4_t a, int16x4_t b, int16_t c) {
   return vmlal_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlal_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD_I]]
 //
 int64x2_t test_vmlal_n_s32(int64x2_t a, int32x2_t b, int32_t c) {
   return vmlal_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlal_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <4 x i32> [[A:%.*]], [[VMULL2_I_I]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD_I]]
 //
 uint32x4_t test_vmlal_n_u16(uint32x4_t a, uint16x4_t b, uint16_t c) {
   return vmlal_n_u16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlal_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[ADD_I:%.*]] = add <2 x i64> [[A:%.*]], [[VMULL2_I_I]]
 // CHECK-NEXT:    ret <2 x i64> [[ADD_I]]
 //
 uint64x2_t test_vmlal_n_u32(uint64x2_t a, uint32x2_t b, uint32_t c) {
   return vmlal_n_u32(a, b, c);
 }

 // CHECK-LABEL: @test_vqdmlal_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I_I]]
 //
 int32x4_t test_vqdmlal_n_s16(int32x4_t a, int16x4_t b, int16_t c) {
   return vqdmlal_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vqdmlal_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I_I]]
 //
 int64x2_t test_vqdmlal_n_s32(int64x2_t a, int32x2_t b, int32_t c) {
   return vqdmlal_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmls_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i16> [[B:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB_I]]
 //
 int16x4_t test_vmls_n_s16(int16x4_t a, int16x4_t b, int16_t c) {
   return vmls_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsq_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x i16> [[VECINIT3_I]], i16 [[C]], i32 4
 // CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x i16> [[VECINIT4_I]], i16 [[C]], i32 5
 // CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x i16> [[VECINIT5_I]], i16 [[C]], i32 6
 // CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x i16> [[VECINIT6_I]], i16 [[C]], i32 7
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <8 x i16> [[B:%.*]], [[VECINIT7_I]]
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB_I]]
 //
 int16x8_t test_vmlsq_n_s16(int16x8_t a, int16x8_t b, int16_t c) {
   return vmlsq_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vmls_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <2 x i32> [[B:%.*]], [[VECINIT1_I]]
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB_I]]
 //
 int32x2_t test_vmls_n_s32(int32x2_t a, int32x2_t b, int32_t c) {
   return vmls_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsq_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i32> [[VECINIT1_I]], i32 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i32> [[VECINIT2_I]], i32 [[C]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i32> [[B:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB_I]]
 //
 int32x4_t test_vmlsq_n_s32(int32x4_t a, int32x4_t b, int32_t c) {
   return vmlsq_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmls_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i16> [[B:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB_I]]
 //
 uint16x4_t test_vmls_n_u16(uint16x4_t a, uint16x4_t b, uint16_t c) {
   return vmls_n_u16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsq_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <8 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <8 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <8 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <8 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[VECINIT4_I:%.*]] = insertelement <8 x i16> [[VECINIT3_I]], i16 [[C]], i32 4
 // CHECK-NEXT:    [[VECINIT5_I:%.*]] = insertelement <8 x i16> [[VECINIT4_I]], i16 [[C]], i32 5
 // CHECK-NEXT:    [[VECINIT6_I:%.*]] = insertelement <8 x i16> [[VECINIT5_I]], i16 [[C]], i32 6
 // CHECK-NEXT:    [[VECINIT7_I:%.*]] = insertelement <8 x i16> [[VECINIT6_I]], i16 [[C]], i32 7
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <8 x i16> [[B:%.*]], [[VECINIT7_I]]
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB_I]]
 //
 uint16x8_t test_vmlsq_n_u16(uint16x8_t a, uint16x8_t b, uint16_t c) {
   return vmlsq_n_u16(a, b, c);
 }

 // CHECK-LABEL: @test_vmls_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <2 x i32> [[B:%.*]], [[VECINIT1_I]]
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB_I]]
 //
 uint32x2_t test_vmls_n_u32(uint32x2_t a, uint32x2_t b, uint32_t c) {
   return vmls_n_u32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsq_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i32> [[VECINIT1_I]], i32 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i32> [[VECINIT2_I]], i32 [[C]], i32 3
 // CHECK-NEXT:    [[MUL_I:%.*]] = mul <4 x i32> [[B:%.*]], [[VECINIT3_I]]
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB_I]]
 //
 uint32x4_t test_vmlsq_n_u32(uint32x4_t a, uint32x4_t b, uint32_t c) {
   return vmlsq_n_u32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsl_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> [[B]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB_I]]
 //
 int32x4_t test_vmlsl_n_s16(int32x4_t a, int16x4_t b, int16_t c) {
   return vmlsl_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsl_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> [[B]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB_I]]
 //
 int64x2_t test_vmlsl_n_s32(int64x2_t a, int32x2_t b, int32_t c) {
   return vmlsl_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsl_n_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> [[B]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <4 x i32> [[A:%.*]], [[VMULL2_I_I]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB_I]]
 //
 uint32x4_t test_vmlsl_n_u16(uint32x4_t a, uint16x4_t b, uint16_t c) {
   return vmlsl_n_u16(a, b, c);
 }

 // CHECK-LABEL: @test_vmlsl_n_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VMULL2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> [[B]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <2 x i64> [[A:%.*]], [[VMULL2_I_I]]
 // CHECK-NEXT:    ret <2 x i64> [[SUB_I]]
 //
 uint64x2_t test_vmlsl_n_u32(uint64x2_t a, uint32x2_t b, uint32_t c) {
   return vmlsl_n_u32(a, b, c);
 }

 // CHECK-LABEL: @test_vqdmlsl_n_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <4 x i16> undef, i16 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <4 x i16> [[VECINIT_I]], i16 [[C]], i32 1
 // CHECK-NEXT:    [[VECINIT2_I:%.*]] = insertelement <4 x i16> [[VECINIT1_I]], i16 [[C]], i32 2
 // CHECK-NEXT:    [[VECINIT3_I:%.*]] = insertelement <4 x i16> [[VECINIT2_I]], i16 [[C]], i32 3
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i16> [[VECINIT3_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[VECINIT3_I]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I_I]]
 //
 int32x4_t test_vqdmlsl_n_s16(int32x4_t a, int16x4_t b, int16_t c) {
   return vqdmlsl_n_s16(a, b, c);
 }

 // CHECK-LABEL: @test_vqdmlsl_n_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[VECINIT_I:%.*]] = insertelement <2 x i32> undef, i32 [[C:%.*]], i32 0
 // CHECK-NEXT:    [[VECINIT1_I:%.*]] = insertelement <2 x i32> [[VECINIT_I]], i32 [[C]], i32 1
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i32> [[VECINIT1_I]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[VECINIT1_I]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I_I]]
 //
 int64x2_t test_vqdmlsl_n_s32(int64x2_t a, int32x2_t b, int32_t c) {
   return vqdmlsl_n_s32(a, b, c);
 }

 // CHECK-LABEL: @test_vmla_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD]]
 //
 uint16x4_t test_vmla_lane_u16_0(uint16x4_t a, uint16x4_t b, uint16x4_t v) {
   return vmla_lane_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlaq_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD]]
 //
 uint16x8_t test_vmlaq_lane_u16_0(uint16x8_t a, uint16x8_t b, uint16x4_t v) {
   return vmlaq_lane_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmla_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD]]
 //
 uint32x2_t test_vmla_lane_u32_0(uint32x2_t a, uint32x2_t b, uint32x2_t v) {
   return vmla_lane_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlaq_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 uint32x4_t test_vmlaq_lane_u32_0(uint32x4_t a, uint32x4_t b, uint32x2_t v) {
   return vmlaq_lane_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmla_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD]]
 //
 uint16x4_t test_vmla_laneq_u16_0(uint16x4_t a, uint16x4_t b, uint16x8_t v) {
   return vmla_laneq_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlaq_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD]]
 //
 uint16x8_t test_vmlaq_laneq_u16_0(uint16x8_t a, uint16x8_t b, uint16x8_t v) {
   return vmlaq_laneq_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmla_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD]]
 //
 uint32x2_t test_vmla_laneq_u32_0(uint32x2_t a, uint32x2_t b, uint32x4_t v) {
   return vmla_laneq_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlaq_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 uint32x4_t test_vmlaq_laneq_u32_0(uint32x4_t a, uint32x4_t b, uint32x4_t v) {
   return vmlaq_laneq_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlal_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I]]
 //
 int32x4_t test_vqdmlal_laneq_s16_0(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vqdmlal_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlal_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I]]
 //
 int64x2_t test_vqdmlal_laneq_s32_0(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vqdmlal_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlal_high_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I]]
 //
 int32x4_t test_vqdmlal_high_laneq_s16_0(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vqdmlal_high_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlal_high_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I]]
 //
 int64x2_t test_vqdmlal_high_laneq_s32_0(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vqdmlal_high_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmls_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB]]
 //
 uint16x4_t test_vmls_lane_u16_0(uint16x4_t a, uint16x4_t b, uint16x4_t v) {
   return vmls_lane_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsq_lane_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB]]
 //
 uint16x8_t test_vmlsq_lane_u16_0(uint16x8_t a, uint16x8_t b, uint16x4_t v) {
   return vmlsq_lane_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmls_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB]]
 //
 uint32x2_t test_vmls_lane_u32_0(uint32x2_t a, uint32x2_t b, uint32x2_t v) {
   return vmls_lane_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsq_lane_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 uint32x4_t test_vmlsq_lane_u32_0(uint32x4_t a, uint32x4_t b, uint32x2_t v) {
   return vmlsq_lane_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmls_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB]]
 //
 uint16x4_t test_vmls_laneq_u16_0(uint16x4_t a, uint16x4_t b, uint16x8_t v) {
   return vmls_laneq_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsq_laneq_u16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB]]
 //
 uint16x8_t test_vmlsq_laneq_u16_0(uint16x8_t a, uint16x8_t b, uint16x8_t v) {
   return vmlsq_laneq_u16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmls_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB]]
 //
 uint32x2_t test_vmls_laneq_u32_0(uint32x2_t a, uint32x2_t b, uint32x4_t v) {
   return vmls_laneq_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vmlsq_laneq_u32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 uint32x4_t test_vmlsq_laneq_u32_0(uint32x4_t a, uint32x4_t b, uint32x4_t v) {
   return vmlsq_laneq_u32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlsl_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I]]
 //
 int32x4_t test_vqdmlsl_laneq_s16_0(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vqdmlsl_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlsl_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I]]
 //
 int64x2_t test_vqdmlsl_laneq_s32_0(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vqdmlsl_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I]]
 //
 int32x4_t test_vqdmlsl_high_laneq_s16_0(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vqdmlsl_high_laneq_s16(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> zeroinitializer
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I]]
 //
 int64x2_t test_vqdmlsl_high_laneq_s32_0(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vqdmlsl_high_laneq_s32(a, b, v, 0);
 }

 // CHECK-LABEL: @test_vqdmulh_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V2:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.laneq.v4i16.v8i16(<4 x i16> [[VQDMULH_LANEQ_V]], <8 x i16> [[VQDMULH_LANEQ_V1]], i32 0)
 // CHECK-NEXT:    ret <4 x i16> [[VQDMULH_LANEQ_V2]]
 //
 int16x4_t test_vqdmulh_laneq_s16_0(int16x4_t a, int16x8_t v) {
   return vqdmulh_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmulhq_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V2:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqdmulh.laneq.v8i16.v8i16(<8 x i16> [[VQDMULHQ_LANEQ_V]], <8 x i16> [[VQDMULHQ_LANEQ_V1]], i32 0)
 // CHECK-NEXT:    ret <8 x i16> [[VQDMULHQ_LANEQ_V2]]
 //
 int16x8_t test_vqdmulhq_laneq_s16_0(int16x8_t a, int16x8_t v) {
   return vqdmulhq_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmulh_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V2:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqdmulh.laneq.v2i32.v4i32(<2 x i32> [[VQDMULH_LANEQ_V]], <4 x i32> [[VQDMULH_LANEQ_V1]], i32 0)
 // CHECK-NEXT:    ret <2 x i32> [[VQDMULH_LANEQ_V2]]
 //
 int32x2_t test_vqdmulh_laneq_s32_0(int32x2_t a, int32x4_t v) {
   return vqdmulh_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqdmulhq_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V2:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmulh.laneq.v4i32.v4i32(<4 x i32> [[VQDMULHQ_LANEQ_V]], <4 x i32> [[VQDMULHQ_LANEQ_V1]], i32 0)
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULHQ_LANEQ_V2]]
 //
 int32x4_t test_vqdmulhq_laneq_s32_0(int32x4_t a, int32x4_t v) {
   return vqdmulhq_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqrdmulh_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V2:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.laneq.v4i16.v8i16(<4 x i16> [[VQRDMULH_LANEQ_V]], <8 x i16> [[VQRDMULH_LANEQ_V1]], i32 0)
 // CHECK-NEXT:    ret <4 x i16> [[VQRDMULH_LANEQ_V2]]
 //
 int16x4_t test_vqrdmulh_laneq_s16_0(int16x4_t a, int16x8_t v) {
   return vqrdmulh_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqrdmulhq_laneq_s16_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V2:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqrdmulh.laneq.v8i16.v8i16(<8 x i16> [[VQRDMULHQ_LANEQ_V]], <8 x i16> [[VQRDMULHQ_LANEQ_V1]], i32 0)
 // CHECK-NEXT:    ret <8 x i16> [[VQRDMULHQ_LANEQ_V2]]
 //
 int16x8_t test_vqrdmulhq_laneq_s16_0(int16x8_t a, int16x8_t v) {
   return vqrdmulhq_laneq_s16(a, v, 0);
 }

 // CHECK-LABEL: @test_vqrdmulh_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V2:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqrdmulh.laneq.v2i32.v4i32(<2 x i32> [[VQRDMULH_LANEQ_V]], <4 x i32> [[VQRDMULH_LANEQ_V1]], i32 0)
 // CHECK-NEXT:    ret <2 x i32> [[VQRDMULH_LANEQ_V2]]
 //
 int32x2_t test_vqrdmulh_laneq_s32_0(int32x2_t a, int32x4_t v) {
   return vqrdmulh_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vqrdmulhq_laneq_s32_0(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V2:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqrdmulh.laneq.v4i32.v4i32(<4 x i32> [[VQRDMULHQ_LANEQ_V]], <4 x i32> [[VQRDMULHQ_LANEQ_V1]], i32 0)
 // CHECK-NEXT:    ret <4 x i32> [[VQRDMULHQ_LANEQ_V2]]
 //
 int32x4_t test_vqrdmulhq_laneq_s32_0(int32x4_t a, int32x4_t v) {
   return vqrdmulhq_laneq_s32(a, v, 0);
 }

 // CHECK-LABEL: @test_vmla_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD]]
 //
 uint16x4_t test_vmla_lane_u16(uint16x4_t a, uint16x4_t b, uint16x4_t v) {
   return vmla_lane_u16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlaq_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD]]
 //
 uint16x8_t test_vmlaq_lane_u16(uint16x8_t a, uint16x8_t b, uint16x4_t v) {
   return vmlaq_lane_u16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmla_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD]]
 //
 uint32x2_t test_vmla_lane_u32(uint32x2_t a, uint32x2_t b, uint32x2_t v) {
   return vmla_lane_u32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlaq_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 uint32x4_t test_vmlaq_lane_u32(uint32x4_t a, uint32x4_t b, uint32x2_t v) {
   return vmlaq_lane_u32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmla_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[ADD]]
 //
 uint16x4_t test_vmla_laneq_u16(uint16x4_t a, uint16x4_t b, uint16x8_t v) {
   return vmla_laneq_u16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlaq_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[ADD]]
 //
 uint16x8_t test_vmlaq_laneq_u16(uint16x8_t a, uint16x8_t b, uint16x8_t v) {
   return vmlaq_laneq_u16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmla_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[ADD]]
 //
 uint32x2_t test_vmla_laneq_u32(uint32x2_t a, uint32x2_t b, uint32x4_t v) {
   return vmla_laneq_u32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlaq_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[ADD]]
 //
 uint32x4_t test_vmlaq_laneq_u32(uint32x4_t a, uint32x4_t b, uint32x4_t v) {
   return vmlaq_laneq_u32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlal_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I]]
 //
 int32x4_t test_vqdmlal_laneq_s16(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vqdmlal_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vqdmlal_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I]]
 //
 int64x2_t test_vqdmlal_laneq_s32(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vqdmlal_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlal_high_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLAL_V3_I]]
 //
 int32x4_t test_vqdmlal_high_laneq_s16(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vqdmlal_high_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vqdmlal_high_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLAL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLAL_V3_I]]
 //
 int64x2_t test_vqdmlal_high_laneq_s32(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vqdmlal_high_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmls_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB]]
 //
 uint16x4_t test_vmls_lane_u16(uint16x4_t a, uint16x4_t b, uint16x4_t v) {
   return vmls_lane_u16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsq_lane_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP1]], <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB]]
 //
 uint16x8_t test_vmlsq_lane_u16(uint16x8_t a, uint16x8_t b, uint16x4_t v) {
   return vmlsq_lane_u16(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmls_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB]]
 //
 uint32x2_t test_vmls_lane_u32(uint32x2_t a, uint32x2_t b, uint32x2_t v) {
   return vmls_lane_u32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmlsq_lane_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 uint32x4_t test_vmlsq_lane_u32(uint32x4_t a, uint32x4_t b, uint32x2_t v) {
   return vmlsq_lane_u32(a, b, v, 1);
 }

 // CHECK-LABEL: @test_vmls_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i16> [[SUB]]
 //
 uint16x4_t test_vmls_laneq_u16(uint16x4_t a, uint16x4_t b, uint16x8_t v) {
   return vmls_laneq_u16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmlsq_laneq_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <8 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <8 x i16> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <8 x i16> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <8 x i16> [[SUB]]
 //
 uint16x8_t test_vmlsq_laneq_u16(uint16x8_t a, uint16x8_t b, uint16x8_t v) {
   return vmlsq_laneq_u16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vmls_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <2 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <2 x i32> [[SUB]]
 //
 uint32x2_t test_vmls_laneq_u32(uint32x2_t a, uint32x2_t b, uint32x4_t v) {
   return vmls_laneq_u32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vmlsq_laneq_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 // CHECK-NEXT:    [[MUL:%.*]] = mul <4 x i32> [[B:%.*]], [[LANE]]
 // CHECK-NEXT:    [[SUB:%.*]] = sub <4 x i32> [[A:%.*]], [[MUL]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB]]
 //
 uint32x4_t test_vmlsq_laneq_u32(uint32x4_t a, uint32x4_t b, uint32x4_t v) {
   return vmlsq_laneq_u32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlsl_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[B]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I]]
 //
 int32x4_t test_vqdmlsl_laneq_s16(int32x4_t a, int16x4_t b, int16x8_t v) {
   return vqdmlsl_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vqdmlsl_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[B]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I]]
 //
 int64x2_t test_vqdmlsl_laneq_s32(int64x2_t a, int32x2_t b, int32x4_t v) {
   return vqdmlsl_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <8 x i16> [[B:%.*]], <8 x i16> [[B]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i16> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[SHUFFLE_I]], <4 x i16> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> [[A]], <4 x i32> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <4 x i32> [[VQDMLSL_V3_I]]
 //
 int32x4_t test_vqdmlsl_high_laneq_s16(int32x4_t a, int16x8_t b, int16x8_t v) {
   return vqdmlsl_high_laneq_s16(a, b, v, 7);
 }

 // CHECK-LABEL: @test_vqdmlsl_high_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> [[B]], <2 x i32> <i32 2, i32 3>
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP1]], <2 x i32> <i32 3, i32 3>
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[SHUFFLE_I]] to <8 x i8>
 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[LANE]] to <8 x i8>
 // CHECK-NEXT:    [[VQDMLAL2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> [[SHUFFLE_I]], <2 x i32> [[LANE]]) #4
 // CHECK-NEXT:    [[VQDMLSL_V3_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> [[A]], <2 x i64> [[VQDMLAL2_I]]) #4
 // CHECK-NEXT:    ret <2 x i64> [[VQDMLSL_V3_I]]
 //
 int64x2_t test_vqdmlsl_high_laneq_s32(int64x2_t a, int32x4_t b, int32x4_t v) {
   return vqdmlsl_high_laneq_s32(a, b, v, 3);
 }

 // CHECK-LABEL: @test_vqdmulh_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V2:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.laneq.v4i16.v8i16(<4 x i16> [[VQDMULH_LANEQ_V]], <8 x i16> [[VQDMULH_LANEQ_V1]], i32 7)
 // CHECK-NEXT:    ret <4 x i16> [[VQDMULH_LANEQ_V2]]
 //
 int16x4_t test_vqdmulh_laneq_s16(int16x4_t a, int16x8_t v) {
   return vqdmulh_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vqdmulhq_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V2:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqdmulh.laneq.v8i16.v8i16(<8 x i16> [[VQDMULHQ_LANEQ_V]], <8 x i16> [[VQDMULHQ_LANEQ_V1]], i32 7)
 // CHECK-NEXT:    ret <8 x i16> [[VQDMULHQ_LANEQ_V2]]
 //
 int16x8_t test_vqdmulhq_laneq_s16(int16x8_t a, int16x8_t v) {
   return vqdmulhq_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vqdmulh_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    [[VQDMULH_LANEQ_V2:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqdmulh.laneq.v2i32.v4i32(<2 x i32> [[VQDMULH_LANEQ_V]], <4 x i32> [[VQDMULH_LANEQ_V1]], i32 3)
 // CHECK-NEXT:    ret <2 x i32> [[VQDMULH_LANEQ_V2]]
 //
 int32x2_t test_vqdmulh_laneq_s32(int32x2_t a, int32x4_t v) {
   return vqdmulh_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vqdmulhq_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    [[VQDMULHQ_LANEQ_V2:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmulh.laneq.v4i32.v4i32(<4 x i32> [[VQDMULHQ_LANEQ_V]], <4 x i32> [[VQDMULHQ_LANEQ_V1]], i32 3)
 // CHECK-NEXT:    ret <4 x i32> [[VQDMULHQ_LANEQ_V2]]
 //
 int32x4_t test_vqdmulhq_laneq_s32(int32x4_t a, int32x4_t v) {
   return vqdmulhq_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vqrdmulh_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V2:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.laneq.v4i16.v8i16(<4 x i16> [[VQRDMULH_LANEQ_V]], <8 x i16> [[VQRDMULH_LANEQ_V1]], i32 7)
 // CHECK-NEXT:    ret <4 x i16> [[VQRDMULH_LANEQ_V2]]
 //
 int16x4_t test_vqrdmulh_laneq_s16(int16x4_t a, int16x8_t v) {
   return vqrdmulh_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vqrdmulhq_laneq_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V2:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqrdmulh.laneq.v8i16.v8i16(<8 x i16> [[VQRDMULHQ_LANEQ_V]], <8 x i16> [[VQRDMULHQ_LANEQ_V1]], i32 7)
 // CHECK-NEXT:    ret <8 x i16> [[VQRDMULHQ_LANEQ_V2]]
 //
 int16x8_t test_vqrdmulhq_laneq_s16(int16x8_t a, int16x8_t v) {
   return vqrdmulhq_laneq_s16(a, v, 7);
 }

 // CHECK-LABEL: @test_vqrdmulh_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A:%.*]] to <8 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    [[VQRDMULH_LANEQ_V2:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqrdmulh.laneq.v2i32.v4i32(<2 x i32> [[VQRDMULH_LANEQ_V]], <4 x i32> [[VQRDMULH_LANEQ_V1]], i32 3)
 // CHECK-NEXT:    ret <2 x i32> [[VQRDMULH_LANEQ_V2]]
 //
 int32x2_t test_vqrdmulh_laneq_s32(int32x2_t a, int32x4_t v) {
   return vqrdmulh_laneq_s32(a, v, 3);
 }

 // CHECK-LABEL: @test_vqrdmulhq_laneq_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    [[VQRDMULHQ_LANEQ_V2:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqrdmulh.laneq.v4i32.v4i32(<4 x i32> [[VQRDMULHQ_LANEQ_V]], <4 x i32> [[VQRDMULHQ_LANEQ_V1]], i32 3)
 // CHECK-NEXT:    ret <4 x i32> [[VQRDMULHQ_LANEQ_V2]]
 //
 int32x4_t test_vqrdmulhq_laneq_s32(int32x4_t a, int32x4_t v) {
   return vqrdmulhq_laneq_s32(a, v, 3);
 }