xref: /llvm-project/clang/test/CodeGen/AArch64/neon-fp16fml.c (revision 207e5ccceec8d3cc3f32723e78f2a142bc61b07d)

// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
// RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +v8.2a -target-feature +neon -target-feature +fp16fml \
// RUN: -disable-O0-optnone -emit-llvm -o - %s | opt -S -passes=mem2reg | FileCheck %s

// REQUIRES: aarch64-registered-target

// Test AArch64 Armv8.2-A FP16 Fused Multiply-Add Long intrinsics

#include <arm_neon.h>

// Vector form

// CHECK-LABEL: @test_vfmlal_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x half> [[C:%.*]] to <8 x i8>
// CHECK-NEXT:    [[VFMLAL_LOW3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlal.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[C]])
// CHECK-NEXT:    ret <2 x float> [[VFMLAL_LOW3_I]]
//
float32x2_t test_vfmlal_low_f16(float32x2_t a, float16x4_t b, float16x4_t c) {
  return vfmlal_low_f16(a, b, c);
}

// CHECK-LABEL: @test_vfmlsl_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x half> [[C:%.*]] to <8 x i8>
// CHECK-NEXT:    [[VFMLSL_LOW3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlsl.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[C]])
// CHECK-NEXT:    ret <2 x float> [[VFMLSL_LOW3_I]]
//
float32x2_t test_vfmlsl_low_f16(float32x2_t a, float16x4_t b, float16x4_t c) {
  return vfmlsl_low_f16(a, b, c);
}

// CHECK-LABEL: @test_vfmlal_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x half> [[C:%.*]] to <8 x i8>
// CHECK-NEXT:    [[VFMLAL_HIGH3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlal2.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[C]])
// CHECK-NEXT:    ret <2 x float> [[VFMLAL_HIGH3_I]]
//
float32x2_t test_vfmlal_high_f16(float32x2_t a, float16x4_t b, float16x4_t c) {
  return vfmlal_high_f16(a, b, c);
}

// CHECK-LABEL: @test_vfmlsl_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x half> [[C:%.*]] to <8 x i8>
// CHECK-NEXT:    [[VFMLSL_HIGH3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlsl2.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[C]])
// CHECK-NEXT:    ret <2 x float> [[VFMLSL_HIGH3_I]]
//
float32x2_t test_vfmlsl_high_f16(float32x2_t a, float16x4_t b, float16x4_t c) {
  return vfmlsl_high_f16(a, b, c);
}

// CHECK-LABEL: @test_vfmlalq_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x half> [[C:%.*]] to <16 x i8>
// CHECK-NEXT:    [[VFMLAL_LOW3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlal.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[C]])
// CHECK-NEXT:    ret <4 x float> [[VFMLAL_LOW3_I]]
//
float32x4_t test_vfmlalq_low_f16(float32x4_t a, float16x8_t b, float16x8_t c) {
  return vfmlalq_low_f16(a, b, c);
}

// CHECK-LABEL: @test_vfmlslq_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x half> [[C:%.*]] to <16 x i8>
// CHECK-NEXT:    [[VFMLSL_LOW3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlsl.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[C]])
// CHECK-NEXT:    ret <4 x float> [[VFMLSL_LOW3_I]]
//
float32x4_t test_vfmlslq_low_f16(float32x4_t a, float16x8_t b, float16x8_t c) {
  return vfmlslq_low_f16(a, b, c);
}

// CHECK-LABEL: @test_vfmlalq_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x half> [[C:%.*]] to <16 x i8>
// CHECK-NEXT:    [[VFMLAL_HIGH3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlal2.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[C]])
// CHECK-NEXT:    ret <4 x float> [[VFMLAL_HIGH3_I]]
//
float32x4_t test_vfmlalq_high_f16(float32x4_t a, float16x8_t b, float16x8_t c) {
  return vfmlalq_high_f16(a, b, c);
}

// CHECK-LABEL: @test_vfmlslq_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x half> [[C:%.*]] to <16 x i8>
// CHECK-NEXT:    [[VFMLSL_HIGH3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlsl2.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[C]])
// CHECK-NEXT:    ret <4 x float> [[VFMLSL_HIGH3_I]]
//
float32x4_t test_vfmlslq_high_f16(float32x4_t a, float16x8_t b, float16x8_t c) {
  return vfmlslq_high_f16(a, b, c);
}

// Indexed form

// CHECK-LABEL: @test_vfmlal_lane_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_847:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_847:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8474:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_8475:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84714:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84715:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84724:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84725:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <4 x half> [[C:%.*]], ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i16>, ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[TMP0]], i32 0
// CHECK-NEXT:    store i16 [[VGET_LANE]], ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x i16> [[TMP2]], i32 0
// CHECK-NEXT:    store i16 [[VGET_LANE8]], ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x i16> [[TMP4]], i32 0
// CHECK-NEXT:    store i16 [[VGET_LANE18]], ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <4 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x i16> [[TMP6]], i32 0
// CHECK-NEXT:    store i16 [[VGET_LANE28]], ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <4 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[VECINIT31]] to <8 x i8>
// CHECK-NEXT:    [[VFMLAL_LOW3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlal.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[VECINIT31]])
// CHECK-NEXT:    ret <2 x float> [[VFMLAL_LOW3_I]]
//
float32x2_t test_vfmlal_lane_low_f16(float32x2_t a, float16x4_t b, float16x4_t c) {
  return vfmlal_lane_low_f16(a, b, c, 0);
}

// CHECK-LABEL: @test_vfmlal_lane_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_847:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_847:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8474:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_8475:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84714:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84715:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84724:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84725:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <4 x half> [[C:%.*]], ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i16>, ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[TMP0]], i32 1
// CHECK-NEXT:    store i16 [[VGET_LANE]], ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x i16> [[TMP2]], i32 1
// CHECK-NEXT:    store i16 [[VGET_LANE8]], ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x i16> [[TMP4]], i32 1
// CHECK-NEXT:    store i16 [[VGET_LANE18]], ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <4 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x i16> [[TMP6]], i32 1
// CHECK-NEXT:    store i16 [[VGET_LANE28]], ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <4 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[VECINIT31]] to <8 x i8>
// CHECK-NEXT:    [[VFMLAL_HIGH3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlal2.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[VECINIT31]])
// CHECK-NEXT:    ret <2 x float> [[VFMLAL_HIGH3_I]]
//
float32x2_t test_vfmlal_lane_high_f16(float32x2_t a, float16x4_t b, float16x4_t c) {
  return vfmlal_lane_high_f16(a, b, c, 1);
}

// CHECK-LABEL: @test_vfmlalq_lane_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_847:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_847:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8474:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_8475:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84714:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84715:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84724:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84725:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84734:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84735:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84744:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84745:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84754:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84755:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84764:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84765:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <4 x half> [[C:%.*]], ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i16>, ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[TMP0]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE]], ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE8]], ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x i16> [[TMP4]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE18]], ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <8 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x i16> [[TMP6]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE28]], ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <8 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84734]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, ptr [[__REINT_84734]], align 8
// CHECK-NEXT:    [[VGET_LANE38:%.*]] = extractelement <4 x i16> [[TMP8]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE38]], ptr [[__REINT1_84735]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = load half, ptr [[__REINT1_84735]], align 2
// CHECK-NEXT:    [[VECINIT41:%.*]] = insertelement <8 x half> [[VECINIT31]], half [[TMP9]], i32 4
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84744]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i16>, ptr [[__REINT_84744]], align 8
// CHECK-NEXT:    [[VGET_LANE48:%.*]] = extractelement <4 x i16> [[TMP10]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE48]], ptr [[__REINT1_84745]], align 2
// CHECK-NEXT:    [[TMP11:%.*]] = load half, ptr [[__REINT1_84745]], align 2
// CHECK-NEXT:    [[VECINIT51:%.*]] = insertelement <8 x half> [[VECINIT41]], half [[TMP11]], i32 5
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84754]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = load <4 x i16>, ptr [[__REINT_84754]], align 8
// CHECK-NEXT:    [[VGET_LANE58:%.*]] = extractelement <4 x i16> [[TMP12]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE58]], ptr [[__REINT1_84755]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = load half, ptr [[__REINT1_84755]], align 2
// CHECK-NEXT:    [[VECINIT61:%.*]] = insertelement <8 x half> [[VECINIT51]], half [[TMP13]], i32 6
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84764]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load <4 x i16>, ptr [[__REINT_84764]], align 8
// CHECK-NEXT:    [[VGET_LANE68:%.*]] = extractelement <4 x i16> [[TMP14]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE68]], ptr [[__REINT1_84765]], align 2
// CHECK-NEXT:    [[TMP15:%.*]] = load half, ptr [[__REINT1_84765]], align 2
// CHECK-NEXT:    [[VECINIT71:%.*]] = insertelement <8 x half> [[VECINIT61]], half [[TMP15]], i32 7
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP18:%.*]] = bitcast <8 x half> [[VECINIT71]] to <16 x i8>
// CHECK-NEXT:    [[VFMLAL_LOW3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlal.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[VECINIT71]])
// CHECK-NEXT:    ret <4 x float> [[VFMLAL_LOW3_I]]
//
float32x4_t test_vfmlalq_lane_low_f16(float32x4_t a, float16x8_t b, float16x4_t c) {
  return vfmlalq_lane_low_f16(a, b, c, 2);
}

// CHECK-LABEL: @test_vfmlalq_lane_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_847:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_847:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8474:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_8475:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84714:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84715:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84724:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84725:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84734:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84735:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84744:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84745:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84754:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84755:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84764:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84765:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <4 x half> [[C:%.*]], ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i16>, ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[TMP0]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE]], ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE8]], ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x i16> [[TMP4]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE18]], ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <8 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x i16> [[TMP6]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE28]], ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <8 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84734]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, ptr [[__REINT_84734]], align 8
// CHECK-NEXT:    [[VGET_LANE38:%.*]] = extractelement <4 x i16> [[TMP8]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE38]], ptr [[__REINT1_84735]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = load half, ptr [[__REINT1_84735]], align 2
// CHECK-NEXT:    [[VECINIT41:%.*]] = insertelement <8 x half> [[VECINIT31]], half [[TMP9]], i32 4
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84744]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i16>, ptr [[__REINT_84744]], align 8
// CHECK-NEXT:    [[VGET_LANE48:%.*]] = extractelement <4 x i16> [[TMP10]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE48]], ptr [[__REINT1_84745]], align 2
// CHECK-NEXT:    [[TMP11:%.*]] = load half, ptr [[__REINT1_84745]], align 2
// CHECK-NEXT:    [[VECINIT51:%.*]] = insertelement <8 x half> [[VECINIT41]], half [[TMP11]], i32 5
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84754]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = load <4 x i16>, ptr [[__REINT_84754]], align 8
// CHECK-NEXT:    [[VGET_LANE58:%.*]] = extractelement <4 x i16> [[TMP12]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE58]], ptr [[__REINT1_84755]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = load half, ptr [[__REINT1_84755]], align 2
// CHECK-NEXT:    [[VECINIT61:%.*]] = insertelement <8 x half> [[VECINIT51]], half [[TMP13]], i32 6
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84764]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load <4 x i16>, ptr [[__REINT_84764]], align 8
// CHECK-NEXT:    [[VGET_LANE68:%.*]] = extractelement <4 x i16> [[TMP14]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE68]], ptr [[__REINT1_84765]], align 2
// CHECK-NEXT:    [[TMP15:%.*]] = load half, ptr [[__REINT1_84765]], align 2
// CHECK-NEXT:    [[VECINIT71:%.*]] = insertelement <8 x half> [[VECINIT61]], half [[TMP15]], i32 7
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP18:%.*]] = bitcast <8 x half> [[VECINIT71]] to <16 x i8>
// CHECK-NEXT:    [[VFMLAL_HIGH3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlal2.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[VECINIT71]])
// CHECK-NEXT:    ret <4 x float> [[VFMLAL_HIGH3_I]]
//
float32x4_t test_vfmlalq_lane_high_f16(float32x4_t a, float16x8_t b, float16x4_t c) {
  return vfmlalq_lane_high_f16(a, b, c, 3);
}

// CHECK-LABEL: @test_vfmlal_laneq_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_850:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_850:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8504:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_8505:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85014:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85015:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85024:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85025:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <8 x half> [[C:%.*]], ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP0]], i32 4
// CHECK-NEXT:    store i16 [[VGETQ_LANE]], ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x i16> [[TMP2]], i32 4
// CHECK-NEXT:    store i16 [[VGETQ_LANE8]], ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x i16> [[TMP4]], i32 4
// CHECK-NEXT:    store i16 [[VGETQ_LANE18]], ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <4 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x i16> [[TMP6]], i32 4
// CHECK-NEXT:    store i16 [[VGETQ_LANE28]], ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <4 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[VECINIT31]] to <8 x i8>
// CHECK-NEXT:    [[VFMLAL_LOW3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlal.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[VECINIT31]])
// CHECK-NEXT:    ret <2 x float> [[VFMLAL_LOW3_I]]
//
float32x2_t test_vfmlal_laneq_low_f16(float32x2_t a, float16x4_t b, float16x8_t c) {
  return vfmlal_laneq_low_f16(a, b, c, 4);
}

// CHECK-LABEL: @test_vfmlal_laneq_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_850:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_850:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8504:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_8505:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85014:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85015:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85024:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85025:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <8 x half> [[C:%.*]], ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP0]], i32 5
// CHECK-NEXT:    store i16 [[VGETQ_LANE]], ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x i16> [[TMP2]], i32 5
// CHECK-NEXT:    store i16 [[VGETQ_LANE8]], ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x i16> [[TMP4]], i32 5
// CHECK-NEXT:    store i16 [[VGETQ_LANE18]], ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <4 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x i16> [[TMP6]], i32 5
// CHECK-NEXT:    store i16 [[VGETQ_LANE28]], ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <4 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[VECINIT31]] to <8 x i8>
// CHECK-NEXT:    [[VFMLAL_HIGH3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlal2.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[VECINIT31]])
// CHECK-NEXT:    ret <2 x float> [[VFMLAL_HIGH3_I]]
//
float32x2_t test_vfmlal_laneq_high_f16(float32x2_t a, float16x4_t b, float16x8_t c) {
  return vfmlal_laneq_high_f16(a, b, c, 5);
}

// CHECK-LABEL: @test_vfmlalq_laneq_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_850:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_850:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8504:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_8505:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85014:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85015:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85024:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85025:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85034:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85035:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85044:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85045:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85054:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85055:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85064:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85065:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <8 x half> [[C:%.*]], ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP0]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE]], ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x i16> [[TMP2]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE8]], ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x i16> [[TMP4]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE18]], ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <8 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x i16> [[TMP6]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE28]], ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <8 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85034]], align 16
// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, ptr [[__REINT_85034]], align 16
// CHECK-NEXT:    [[VGETQ_LANE38:%.*]] = extractelement <8 x i16> [[TMP8]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE38]], ptr [[__REINT1_85035]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = load half, ptr [[__REINT1_85035]], align 2
// CHECK-NEXT:    [[VECINIT41:%.*]] = insertelement <8 x half> [[VECINIT31]], half [[TMP9]], i32 4
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85044]], align 16
// CHECK-NEXT:    [[TMP10:%.*]] = load <8 x i16>, ptr [[__REINT_85044]], align 16
// CHECK-NEXT:    [[VGETQ_LANE48:%.*]] = extractelement <8 x i16> [[TMP10]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE48]], ptr [[__REINT1_85045]], align 2
// CHECK-NEXT:    [[TMP11:%.*]] = load half, ptr [[__REINT1_85045]], align 2
// CHECK-NEXT:    [[VECINIT51:%.*]] = insertelement <8 x half> [[VECINIT41]], half [[TMP11]], i32 5
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85054]], align 16
// CHECK-NEXT:    [[TMP12:%.*]] = load <8 x i16>, ptr [[__REINT_85054]], align 16
// CHECK-NEXT:    [[VGETQ_LANE58:%.*]] = extractelement <8 x i16> [[TMP12]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE58]], ptr [[__REINT1_85055]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = load half, ptr [[__REINT1_85055]], align 2
// CHECK-NEXT:    [[VECINIT61:%.*]] = insertelement <8 x half> [[VECINIT51]], half [[TMP13]], i32 6
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85064]], align 16
// CHECK-NEXT:    [[TMP14:%.*]] = load <8 x i16>, ptr [[__REINT_85064]], align 16
// CHECK-NEXT:    [[VGETQ_LANE68:%.*]] = extractelement <8 x i16> [[TMP14]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE68]], ptr [[__REINT1_85065]], align 2
// CHECK-NEXT:    [[TMP15:%.*]] = load half, ptr [[__REINT1_85065]], align 2
// CHECK-NEXT:    [[VECINIT71:%.*]] = insertelement <8 x half> [[VECINIT61]], half [[TMP15]], i32 7
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP18:%.*]] = bitcast <8 x half> [[VECINIT71]] to <16 x i8>
// CHECK-NEXT:    [[VFMLAL_LOW3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlal.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[VECINIT71]])
// CHECK-NEXT:    ret <4 x float> [[VFMLAL_LOW3_I]]
//
float32x4_t test_vfmlalq_laneq_low_f16(float32x4_t a, float16x8_t b, float16x8_t c) {
  return vfmlalq_laneq_low_f16(a, b, c, 6);
}

// CHECK-LABEL: @test_vfmlalq_laneq_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_850:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_850:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8504:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_8505:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85014:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85015:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85024:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85025:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85034:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85035:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85044:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85045:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85054:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85055:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85064:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85065:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <8 x half> [[C:%.*]], ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP0]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE]], ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x i16> [[TMP2]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE8]], ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x i16> [[TMP4]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE18]], ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <8 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x i16> [[TMP6]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE28]], ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <8 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85034]], align 16
// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, ptr [[__REINT_85034]], align 16
// CHECK-NEXT:    [[VGETQ_LANE38:%.*]] = extractelement <8 x i16> [[TMP8]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE38]], ptr [[__REINT1_85035]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = load half, ptr [[__REINT1_85035]], align 2
// CHECK-NEXT:    [[VECINIT41:%.*]] = insertelement <8 x half> [[VECINIT31]], half [[TMP9]], i32 4
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85044]], align 16
// CHECK-NEXT:    [[TMP10:%.*]] = load <8 x i16>, ptr [[__REINT_85044]], align 16
// CHECK-NEXT:    [[VGETQ_LANE48:%.*]] = extractelement <8 x i16> [[TMP10]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE48]], ptr [[__REINT1_85045]], align 2
// CHECK-NEXT:    [[TMP11:%.*]] = load half, ptr [[__REINT1_85045]], align 2
// CHECK-NEXT:    [[VECINIT51:%.*]] = insertelement <8 x half> [[VECINIT41]], half [[TMP11]], i32 5
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85054]], align 16
// CHECK-NEXT:    [[TMP12:%.*]] = load <8 x i16>, ptr [[__REINT_85054]], align 16
// CHECK-NEXT:    [[VGETQ_LANE58:%.*]] = extractelement <8 x i16> [[TMP12]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE58]], ptr [[__REINT1_85055]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = load half, ptr [[__REINT1_85055]], align 2
// CHECK-NEXT:    [[VECINIT61:%.*]] = insertelement <8 x half> [[VECINIT51]], half [[TMP13]], i32 6
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85064]], align 16
// CHECK-NEXT:    [[TMP14:%.*]] = load <8 x i16>, ptr [[__REINT_85064]], align 16
// CHECK-NEXT:    [[VGETQ_LANE68:%.*]] = extractelement <8 x i16> [[TMP14]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE68]], ptr [[__REINT1_85065]], align 2
// CHECK-NEXT:    [[TMP15:%.*]] = load half, ptr [[__REINT1_85065]], align 2
// CHECK-NEXT:    [[VECINIT71:%.*]] = insertelement <8 x half> [[VECINIT61]], half [[TMP15]], i32 7
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP18:%.*]] = bitcast <8 x half> [[VECINIT71]] to <16 x i8>
// CHECK-NEXT:    [[VFMLAL_HIGH3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlal2.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[VECINIT71]])
// CHECK-NEXT:    ret <4 x float> [[VFMLAL_HIGH3_I]]
//
float32x4_t test_vfmlalq_laneq_high_f16(float32x4_t a, float16x8_t b, float16x8_t c) {
  return vfmlalq_laneq_high_f16(a, b, c, 7);
}

// CHECK-LABEL: @test_vfmlsl_lane_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_847:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_847:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8474:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_8475:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84714:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84715:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84724:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84725:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <4 x half> [[C:%.*]], ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i16>, ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[TMP0]], i32 0
// CHECK-NEXT:    store i16 [[VGET_LANE]], ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x i16> [[TMP2]], i32 0
// CHECK-NEXT:    store i16 [[VGET_LANE8]], ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x i16> [[TMP4]], i32 0
// CHECK-NEXT:    store i16 [[VGET_LANE18]], ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <4 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x i16> [[TMP6]], i32 0
// CHECK-NEXT:    store i16 [[VGET_LANE28]], ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <4 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[VECINIT31]] to <8 x i8>
// CHECK-NEXT:    [[VFMLSL_LOW3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlsl.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[VECINIT31]])
// CHECK-NEXT:    ret <2 x float> [[VFMLSL_LOW3_I]]
//
float32x2_t test_vfmlsl_lane_low_f16(float32x2_t a, float16x4_t b, float16x4_t c) {
  return vfmlsl_lane_low_f16(a, b, c, 0);
}

// CHECK-LABEL: @test_vfmlsl_lane_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_847:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_847:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8474:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_8475:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84714:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84715:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84724:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84725:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <4 x half> [[C:%.*]], ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i16>, ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[TMP0]], i32 1
// CHECK-NEXT:    store i16 [[VGET_LANE]], ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x i16> [[TMP2]], i32 1
// CHECK-NEXT:    store i16 [[VGET_LANE8]], ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x i16> [[TMP4]], i32 1
// CHECK-NEXT:    store i16 [[VGET_LANE18]], ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <4 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x i16> [[TMP6]], i32 1
// CHECK-NEXT:    store i16 [[VGET_LANE28]], ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <4 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[VECINIT31]] to <8 x i8>
// CHECK-NEXT:    [[VFMLSL_HIGH3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlsl2.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[VECINIT31]])
// CHECK-NEXT:    ret <2 x float> [[VFMLSL_HIGH3_I]]
//
float32x2_t test_vfmlsl_lane_high_f16(float32x2_t a, float16x4_t b, float16x4_t c) {
  return vfmlsl_lane_high_f16(a, b, c, 1);
}

// CHECK-LABEL: @test_vfmlslq_lane_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_847:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_847:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8474:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_8475:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84714:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84715:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84724:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84725:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84734:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84735:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84744:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84745:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84754:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84755:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84764:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84765:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <4 x half> [[C:%.*]], ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i16>, ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[TMP0]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE]], ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE8]], ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x i16> [[TMP4]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE18]], ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <8 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x i16> [[TMP6]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE28]], ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <8 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84734]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, ptr [[__REINT_84734]], align 8
// CHECK-NEXT:    [[VGET_LANE38:%.*]] = extractelement <4 x i16> [[TMP8]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE38]], ptr [[__REINT1_84735]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = load half, ptr [[__REINT1_84735]], align 2
// CHECK-NEXT:    [[VECINIT41:%.*]] = insertelement <8 x half> [[VECINIT31]], half [[TMP9]], i32 4
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84744]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i16>, ptr [[__REINT_84744]], align 8
// CHECK-NEXT:    [[VGET_LANE48:%.*]] = extractelement <4 x i16> [[TMP10]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE48]], ptr [[__REINT1_84745]], align 2
// CHECK-NEXT:    [[TMP11:%.*]] = load half, ptr [[__REINT1_84745]], align 2
// CHECK-NEXT:    [[VECINIT51:%.*]] = insertelement <8 x half> [[VECINIT41]], half [[TMP11]], i32 5
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84754]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = load <4 x i16>, ptr [[__REINT_84754]], align 8
// CHECK-NEXT:    [[VGET_LANE58:%.*]] = extractelement <4 x i16> [[TMP12]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE58]], ptr [[__REINT1_84755]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = load half, ptr [[__REINT1_84755]], align 2
// CHECK-NEXT:    [[VECINIT61:%.*]] = insertelement <8 x half> [[VECINIT51]], half [[TMP13]], i32 6
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84764]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load <4 x i16>, ptr [[__REINT_84764]], align 8
// CHECK-NEXT:    [[VGET_LANE68:%.*]] = extractelement <4 x i16> [[TMP14]], i32 2
// CHECK-NEXT:    store i16 [[VGET_LANE68]], ptr [[__REINT1_84765]], align 2
// CHECK-NEXT:    [[TMP15:%.*]] = load half, ptr [[__REINT1_84765]], align 2
// CHECK-NEXT:    [[VECINIT71:%.*]] = insertelement <8 x half> [[VECINIT61]], half [[TMP15]], i32 7
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP18:%.*]] = bitcast <8 x half> [[VECINIT71]] to <16 x i8>
// CHECK-NEXT:    [[VFMLSL_LOW3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlsl.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[VECINIT71]])
// CHECK-NEXT:    ret <4 x float> [[VFMLSL_LOW3_I]]
//
float32x4_t test_vfmlslq_lane_low_f16(float32x4_t a, float16x8_t b, float16x4_t c) {
  return vfmlslq_lane_low_f16(a, b, c, 2);
}

// CHECK-LABEL: @test_vfmlslq_lane_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_847:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_847:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8474:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_8475:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84714:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84715:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84724:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84725:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84734:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84735:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84744:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84745:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84754:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84755:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_84764:%.*]] = alloca <4 x half>, align 8
// CHECK-NEXT:    [[__REINT1_84765:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <4 x half> [[C:%.*]], ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i16>, ptr [[__REINT_847]], align 8
// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[TMP0]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE]], ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_847]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i16>, ptr [[__REINT_8474]], align 8
// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE8]], ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8475]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, ptr [[__REINT_84714]], align 8
// CHECK-NEXT:    [[VGET_LANE18:%.*]] = extractelement <4 x i16> [[TMP4]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE18]], ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_84715]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <8 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, ptr [[__REINT_84724]], align 8
// CHECK-NEXT:    [[VGET_LANE28:%.*]] = extractelement <4 x i16> [[TMP6]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE28]], ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_84725]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <8 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84734]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, ptr [[__REINT_84734]], align 8
// CHECK-NEXT:    [[VGET_LANE38:%.*]] = extractelement <4 x i16> [[TMP8]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE38]], ptr [[__REINT1_84735]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = load half, ptr [[__REINT1_84735]], align 2
// CHECK-NEXT:    [[VECINIT41:%.*]] = insertelement <8 x half> [[VECINIT31]], half [[TMP9]], i32 4
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84744]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i16>, ptr [[__REINT_84744]], align 8
// CHECK-NEXT:    [[VGET_LANE48:%.*]] = extractelement <4 x i16> [[TMP10]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE48]], ptr [[__REINT1_84745]], align 2
// CHECK-NEXT:    [[TMP11:%.*]] = load half, ptr [[__REINT1_84745]], align 2
// CHECK-NEXT:    [[VECINIT51:%.*]] = insertelement <8 x half> [[VECINIT41]], half [[TMP11]], i32 5
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84754]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = load <4 x i16>, ptr [[__REINT_84754]], align 8
// CHECK-NEXT:    [[VGET_LANE58:%.*]] = extractelement <4 x i16> [[TMP12]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE58]], ptr [[__REINT1_84755]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = load half, ptr [[__REINT1_84755]], align 2
// CHECK-NEXT:    [[VECINIT61:%.*]] = insertelement <8 x half> [[VECINIT51]], half [[TMP13]], i32 6
// CHECK-NEXT:    store <4 x half> [[C]], ptr [[__REINT_84764]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load <4 x i16>, ptr [[__REINT_84764]], align 8
// CHECK-NEXT:    [[VGET_LANE68:%.*]] = extractelement <4 x i16> [[TMP14]], i32 3
// CHECK-NEXT:    store i16 [[VGET_LANE68]], ptr [[__REINT1_84765]], align 2
// CHECK-NEXT:    [[TMP15:%.*]] = load half, ptr [[__REINT1_84765]], align 2
// CHECK-NEXT:    [[VECINIT71:%.*]] = insertelement <8 x half> [[VECINIT61]], half [[TMP15]], i32 7
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP18:%.*]] = bitcast <8 x half> [[VECINIT71]] to <16 x i8>
// CHECK-NEXT:    [[VFMLSL_HIGH3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlsl2.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[VECINIT71]])
// CHECK-NEXT:    ret <4 x float> [[VFMLSL_HIGH3_I]]
//
float32x4_t test_vfmlslq_lane_high_f16(float32x4_t a, float16x8_t b, float16x4_t c) {
  return vfmlslq_lane_high_f16(a, b, c, 3);
}

// CHECK-LABEL: @test_vfmlsl_laneq_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_850:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_850:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8504:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_8505:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85014:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85015:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85024:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85025:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <8 x half> [[C:%.*]], ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP0]], i32 4
// CHECK-NEXT:    store i16 [[VGETQ_LANE]], ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x i16> [[TMP2]], i32 4
// CHECK-NEXT:    store i16 [[VGETQ_LANE8]], ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x i16> [[TMP4]], i32 4
// CHECK-NEXT:    store i16 [[VGETQ_LANE18]], ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <4 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x i16> [[TMP6]], i32 4
// CHECK-NEXT:    store i16 [[VGETQ_LANE28]], ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <4 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[VECINIT31]] to <8 x i8>
// CHECK-NEXT:    [[VFMLSL_LOW3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlsl.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[VECINIT31]])
// CHECK-NEXT:    ret <2 x float> [[VFMLSL_LOW3_I]]
//
float32x2_t test_vfmlsl_laneq_low_f16(float32x2_t a, float16x4_t b, float16x8_t c) {
  return vfmlsl_laneq_low_f16(a, b, c, 4);
}

// CHECK-LABEL: @test_vfmlsl_laneq_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_850:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_850:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8504:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_8505:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85014:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85015:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85024:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85025:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <8 x half> [[C:%.*]], ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP0]], i32 5
// CHECK-NEXT:    store i16 [[VGETQ_LANE]], ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x i16> [[TMP2]], i32 5
// CHECK-NEXT:    store i16 [[VGETQ_LANE8]], ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x i16> [[TMP4]], i32 5
// CHECK-NEXT:    store i16 [[VGETQ_LANE18]], ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <4 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x i16> [[TMP6]], i32 5
// CHECK-NEXT:    store i16 [[VGETQ_LANE28]], ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <4 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[A:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[VECINIT31]] to <8 x i8>
// CHECK-NEXT:    [[VFMLSL_HIGH3_I:%.*]] = call <2 x float> @llvm.aarch64.neon.fmlsl2.v2f32.v4f16(<2 x float> [[A]], <4 x half> [[B]], <4 x half> [[VECINIT31]])
// CHECK-NEXT:    ret <2 x float> [[VFMLSL_HIGH3_I]]
//
float32x2_t test_vfmlsl_laneq_high_f16(float32x2_t a, float16x4_t b, float16x8_t c) {
  return vfmlsl_laneq_high_f16(a, b, c, 5);
}

// CHECK-LABEL: @test_vfmlslq_laneq_low_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_850:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_850:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8504:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_8505:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85014:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85015:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85024:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85025:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85034:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85035:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85044:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85045:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85054:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85055:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85064:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85065:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <8 x half> [[C:%.*]], ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP0]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE]], ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x i16> [[TMP2]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE8]], ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x i16> [[TMP4]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE18]], ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <8 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x i16> [[TMP6]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE28]], ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <8 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85034]], align 16
// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, ptr [[__REINT_85034]], align 16
// CHECK-NEXT:    [[VGETQ_LANE38:%.*]] = extractelement <8 x i16> [[TMP8]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE38]], ptr [[__REINT1_85035]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = load half, ptr [[__REINT1_85035]], align 2
// CHECK-NEXT:    [[VECINIT41:%.*]] = insertelement <8 x half> [[VECINIT31]], half [[TMP9]], i32 4
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85044]], align 16
// CHECK-NEXT:    [[TMP10:%.*]] = load <8 x i16>, ptr [[__REINT_85044]], align 16
// CHECK-NEXT:    [[VGETQ_LANE48:%.*]] = extractelement <8 x i16> [[TMP10]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE48]], ptr [[__REINT1_85045]], align 2
// CHECK-NEXT:    [[TMP11:%.*]] = load half, ptr [[__REINT1_85045]], align 2
// CHECK-NEXT:    [[VECINIT51:%.*]] = insertelement <8 x half> [[VECINIT41]], half [[TMP11]], i32 5
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85054]], align 16
// CHECK-NEXT:    [[TMP12:%.*]] = load <8 x i16>, ptr [[__REINT_85054]], align 16
// CHECK-NEXT:    [[VGETQ_LANE58:%.*]] = extractelement <8 x i16> [[TMP12]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE58]], ptr [[__REINT1_85055]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = load half, ptr [[__REINT1_85055]], align 2
// CHECK-NEXT:    [[VECINIT61:%.*]] = insertelement <8 x half> [[VECINIT51]], half [[TMP13]], i32 6
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85064]], align 16
// CHECK-NEXT:    [[TMP14:%.*]] = load <8 x i16>, ptr [[__REINT_85064]], align 16
// CHECK-NEXT:    [[VGETQ_LANE68:%.*]] = extractelement <8 x i16> [[TMP14]], i32 6
// CHECK-NEXT:    store i16 [[VGETQ_LANE68]], ptr [[__REINT1_85065]], align 2
// CHECK-NEXT:    [[TMP15:%.*]] = load half, ptr [[__REINT1_85065]], align 2
// CHECK-NEXT:    [[VECINIT71:%.*]] = insertelement <8 x half> [[VECINIT61]], half [[TMP15]], i32 7
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP18:%.*]] = bitcast <8 x half> [[VECINIT71]] to <16 x i8>
// CHECK-NEXT:    [[VFMLSL_LOW3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlsl.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[VECINIT71]])
// CHECK-NEXT:    ret <4 x float> [[VFMLSL_LOW3_I]]
//
float32x4_t test_vfmlslq_laneq_low_f16(float32x4_t a, float16x8_t b, float16x8_t c) {
  return vfmlslq_laneq_low_f16(a, b, c, 6);
}

// CHECK-LABEL: @test_vfmlslq_laneq_high_f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[__REINT_850:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_850:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_8504:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_8505:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85014:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85015:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85024:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85025:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85034:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85035:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85044:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85045:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85054:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85055:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[__REINT_85064:%.*]] = alloca <8 x half>, align 16
// CHECK-NEXT:    [[__REINT1_85065:%.*]] = alloca i16, align 2
// CHECK-NEXT:    store <8 x half> [[C:%.*]], ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, ptr [[__REINT_850]], align 16
// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP0]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE]], ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[TMP1:%.*]] = load half, ptr [[__REINT1_850]], align 2
// CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <8 x half> poison, half [[TMP1]], i32 0
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, ptr [[__REINT_8504]], align 16
// CHECK-NEXT:    [[VGETQ_LANE8:%.*]] = extractelement <8 x i16> [[TMP2]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE8]], ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load half, ptr [[__REINT1_8505]], align 2
// CHECK-NEXT:    [[VECINIT11:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP3]], i32 1
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, ptr [[__REINT_85014]], align 16
// CHECK-NEXT:    [[VGETQ_LANE18:%.*]] = extractelement <8 x i16> [[TMP4]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE18]], ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[TMP5:%.*]] = load half, ptr [[__REINT1_85015]], align 2
// CHECK-NEXT:    [[VECINIT21:%.*]] = insertelement <8 x half> [[VECINIT11]], half [[TMP5]], i32 2
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, ptr [[__REINT_85024]], align 16
// CHECK-NEXT:    [[VGETQ_LANE28:%.*]] = extractelement <8 x i16> [[TMP6]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE28]], ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load half, ptr [[__REINT1_85025]], align 2
// CHECK-NEXT:    [[VECINIT31:%.*]] = insertelement <8 x half> [[VECINIT21]], half [[TMP7]], i32 3
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85034]], align 16
// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, ptr [[__REINT_85034]], align 16
// CHECK-NEXT:    [[VGETQ_LANE38:%.*]] = extractelement <8 x i16> [[TMP8]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE38]], ptr [[__REINT1_85035]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = load half, ptr [[__REINT1_85035]], align 2
// CHECK-NEXT:    [[VECINIT41:%.*]] = insertelement <8 x half> [[VECINIT31]], half [[TMP9]], i32 4
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85044]], align 16
// CHECK-NEXT:    [[TMP10:%.*]] = load <8 x i16>, ptr [[__REINT_85044]], align 16
// CHECK-NEXT:    [[VGETQ_LANE48:%.*]] = extractelement <8 x i16> [[TMP10]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE48]], ptr [[__REINT1_85045]], align 2
// CHECK-NEXT:    [[TMP11:%.*]] = load half, ptr [[__REINT1_85045]], align 2
// CHECK-NEXT:    [[VECINIT51:%.*]] = insertelement <8 x half> [[VECINIT41]], half [[TMP11]], i32 5
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85054]], align 16
// CHECK-NEXT:    [[TMP12:%.*]] = load <8 x i16>, ptr [[__REINT_85054]], align 16
// CHECK-NEXT:    [[VGETQ_LANE58:%.*]] = extractelement <8 x i16> [[TMP12]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE58]], ptr [[__REINT1_85055]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = load half, ptr [[__REINT1_85055]], align 2
// CHECK-NEXT:    [[VECINIT61:%.*]] = insertelement <8 x half> [[VECINIT51]], half [[TMP13]], i32 6
// CHECK-NEXT:    store <8 x half> [[C]], ptr [[__REINT_85064]], align 16
// CHECK-NEXT:    [[TMP14:%.*]] = load <8 x i16>, ptr [[__REINT_85064]], align 16
// CHECK-NEXT:    [[VGETQ_LANE68:%.*]] = extractelement <8 x i16> [[TMP14]], i32 7
// CHECK-NEXT:    store i16 [[VGETQ_LANE68]], ptr [[__REINT1_85065]], align 2
// CHECK-NEXT:    [[TMP15:%.*]] = load half, ptr [[__REINT1_85065]], align 2
// CHECK-NEXT:    [[VECINIT71:%.*]] = insertelement <8 x half> [[VECINIT61]], half [[TMP15]], i32 7
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
// CHECK-NEXT:    [[TMP18:%.*]] = bitcast <8 x half> [[VECINIT71]] to <16 x i8>
// CHECK-NEXT:    [[VFMLSL_HIGH3_I:%.*]] = call <4 x float> @llvm.aarch64.neon.fmlsl2.v4f32.v8f16(<4 x float> [[A]], <8 x half> [[B]], <8 x half> [[VECINIT71]])
// CHECK-NEXT:    ret <4 x float> [[VFMLSL_HIGH3_I]]
//
float32x4_t test_vfmlslq_laneq_high_f16(float32x4_t a, float16x8_t b, float16x8_t c) {
  return vfmlslq_laneq_high_f16(a, b, c, 7);
}