// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
// REQUIRES: riscv-registered-target
// RUN: %clang_cc1 -triple riscv64 -target-feature +v \
// RUN:   -target-feature +experimental-zvfbfa -disable-O0-optnone \
// RUN:   -emit-llvm %s -o - | opt -S -passes=mem2reg | \
// RUN:   FileCheck --check-prefix=CHECK-RV64 %s

#include <riscv_vector.h>

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vv_bf16mf4_f32mf2_tu(
// CHECK-RV64-SAME: <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.nxv1f32.nxv1bf16.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], <vscale x 1 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vv_bf16mf4_f32mf2_tu(vfloat32mf2_t vd,
                                               vbfloat16mf4_t vs2,
                                               vbfloat16mf4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vf_bf16mf4_f32mf2_tu(
// CHECK-RV64-SAME: <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.nxv1f32.nxv1bf16.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vf_bf16mf4_f32mf2_tu(vfloat32mf2_t vd,
                                               vbfloat16mf4_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wv_bf16mf4_f32mf2_tu(
// CHECK-RV64-SAME: <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], <vscale x 1 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wv_bf16mf4_f32mf2_tu(vfloat32mf2_t vd,
                                               vfloat32mf2_t vs2,
                                               vbfloat16mf4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wf_bf16_f32mf2_tu(
// CHECK-RV64-SAME: <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.nxv1f32.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wf_bf16_f32mf2_tu(vfloat32mf2_t vd, vfloat32mf2_t vs2,
                                            __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vv_bf16mf2_f32m1_tu(
// CHECK-RV64-SAME: <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.nxv2f32.nxv2bf16.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], <vscale x 2 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vv_bf16mf2_f32m1_tu(vfloat32m1_t vd,
                                             vbfloat16mf2_t vs2,
                                             vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vf_bf16mf2_f32m1_tu(
// CHECK-RV64-SAME: <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.nxv2f32.nxv2bf16.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vf_bf16mf2_f32m1_tu(vfloat32m1_t vd,
                                             vbfloat16mf2_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wv_bf16mf2_f32m1_tu(
// CHECK-RV64-SAME: <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], <vscale x 2 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wv_bf16mf2_f32m1_tu(vfloat32m1_t vd, vfloat32m1_t vs2,
                                             vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wf_bf16_f32m1_tu(
// CHECK-RV64-SAME: <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.nxv2f32.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wf_bf16_f32m1_tu(vfloat32m1_t vd, vfloat32m1_t vs2,
                                          __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vv_bf16m1_f32m2_tu(
// CHECK-RV64-SAME: <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.nxv4f32.nxv4bf16.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], <vscale x 4 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vv_bf16m1_f32m2_tu(vfloat32m2_t vd, vbfloat16m1_t vs2,
                                            vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vf_bf16m1_f32m2_tu(
// CHECK-RV64-SAME: <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.nxv4f32.nxv4bf16.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vf_bf16m1_f32m2_tu(vfloat32m2_t vd, vbfloat16m1_t vs2,
                                            __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wv_bf16m1_f32m2_tu(
// CHECK-RV64-SAME: <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], <vscale x 4 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wv_bf16m1_f32m2_tu(vfloat32m2_t vd, vfloat32m2_t vs2,
                                            vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wf_bf16_f32m2_tu(
// CHECK-RV64-SAME: <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.nxv4f32.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wf_bf16_f32m2_tu(vfloat32m2_t vd, vfloat32m2_t vs2,
                                          __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vv_bf16m2_f32m4_tu(
// CHECK-RV64-SAME: <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.nxv8f32.nxv8bf16.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], <vscale x 8 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vv_bf16m2_f32m4_tu(vfloat32m4_t vd, vbfloat16m2_t vs2,
                                            vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vf_bf16m2_f32m4_tu(
// CHECK-RV64-SAME: <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.nxv8f32.nxv8bf16.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vf_bf16m2_f32m4_tu(vfloat32m4_t vd, vbfloat16m2_t vs2,
                                            __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wv_bf16m2_f32m4_tu(
// CHECK-RV64-SAME: <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], <vscale x 8 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wv_bf16m2_f32m4_tu(vfloat32m4_t vd, vfloat32m4_t vs2,
                                            vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wf_bf16_f32m4_tu(
// CHECK-RV64-SAME: <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.nxv8f32.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wf_bf16_f32m4_tu(vfloat32m4_t vd, vfloat32m4_t vs2,
                                          __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vv_bf16m4_f32m8_tu(
// CHECK-RV64-SAME: <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.nxv16f32.nxv16bf16.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], <vscale x 16 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vv_bf16m4_f32m8_tu(vfloat32m8_t vd, vbfloat16m4_t vs2,
                                            vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vf_bf16m4_f32m8_tu(
// CHECK-RV64-SAME: <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.nxv16f32.nxv16bf16.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vf_bf16m4_f32m8_tu(vfloat32m8_t vd, vbfloat16m4_t vs2,
                                            __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wv_bf16m4_f32m8_tu(
// CHECK-RV64-SAME: <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], <vscale x 16 x bfloat> [[VS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wv_bf16m4_f32m8_tu(vfloat32m8_t vd, vfloat32m8_t vs2,
                                            vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wf_bf16_f32m8_tu(
// CHECK-RV64-SAME: <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.nxv16f32.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], bfloat [[RS1]], i64 7, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wf_bf16_f32m8_tu(vfloat32m8_t vd, vfloat32m8_t vs2,
                                          __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vv_bf16mf4_f32mf2_tum(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vv_bf16mf4_f32mf2_tum(vbool64_t vm, vfloat32mf2_t vd,
                                                vbfloat16mf4_t vs2,
                                                vbfloat16mf4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vf_bf16mf4_f32mf2_tum(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vf_bf16mf4_f32mf2_tum(vbool64_t vm, vfloat32mf2_t vd,
                                                vbfloat16mf4_t vs2, __bf16 rs1,
                                                size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wv_bf16mf4_f32mf2_tum(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wv_bf16mf4_f32mf2_tum(vbool64_t vm, vfloat32mf2_t vd,
                                                vfloat32mf2_t vs2,
                                                vbfloat16mf4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wf_bf16_f32mf2_tum(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wf_bf16_f32mf2_tum(vbool64_t vm, vfloat32mf2_t vd,
                                             vfloat32mf2_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vv_bf16mf2_f32m1_tum(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vv_bf16mf2_f32m1_tum(vbool32_t vm, vfloat32m1_t vd,
                                              vbfloat16mf2_t vs2,
                                              vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vf_bf16mf2_f32m1_tum(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vf_bf16mf2_f32m1_tum(vbool32_t vm, vfloat32m1_t vd,
                                              vbfloat16mf2_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wv_bf16mf2_f32m1_tum(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wv_bf16mf2_f32m1_tum(vbool32_t vm, vfloat32m1_t vd,
                                              vfloat32m1_t vs2,
                                              vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wf_bf16_f32m1_tum(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wf_bf16_f32m1_tum(vbool32_t vm, vfloat32m1_t vd,
                                           vfloat32m1_t vs2, __bf16 rs1,
                                           size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vv_bf16m1_f32m2_tum(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vv_bf16m1_f32m2_tum(vbool16_t vm, vfloat32m2_t vd,
                                             vbfloat16m1_t vs2,
                                             vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vf_bf16m1_f32m2_tum(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vf_bf16m1_f32m2_tum(vbool16_t vm, vfloat32m2_t vd,
                                             vbfloat16m1_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wv_bf16m1_f32m2_tum(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wv_bf16m1_f32m2_tum(vbool16_t vm, vfloat32m2_t vd,
                                             vfloat32m2_t vs2,
                                             vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wf_bf16_f32m2_tum(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wf_bf16_f32m2_tum(vbool16_t vm, vfloat32m2_t vd,
                                           vfloat32m2_t vs2, __bf16 rs1,
                                           size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vv_bf16m2_f32m4_tum(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vv_bf16m2_f32m4_tum(vbool8_t vm, vfloat32m4_t vd,
                                             vbfloat16m2_t vs2,
                                             vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vf_bf16m2_f32m4_tum(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vf_bf16m2_f32m4_tum(vbool8_t vm, vfloat32m4_t vd,
                                             vbfloat16m2_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wv_bf16m2_f32m4_tum(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wv_bf16m2_f32m4_tum(vbool8_t vm, vfloat32m4_t vd,
                                             vfloat32m4_t vs2,
                                             vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wf_bf16_f32m4_tum(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wf_bf16_f32m4_tum(vbool8_t vm, vfloat32m4_t vd,
                                           vfloat32m4_t vs2, __bf16 rs1,
                                           size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vv_bf16m4_f32m8_tum(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vv_bf16m4_f32m8_tum(vbool4_t vm, vfloat32m8_t vd,
                                             vbfloat16m4_t vs2,
                                             vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vf_bf16m4_f32m8_tum(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vf_bf16m4_f32m8_tum(vbool4_t vm, vfloat32m8_t vd,
                                             vbfloat16m4_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wv_bf16m4_f32m8_tum(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wv_bf16m4_f32m8_tum(vbool4_t vm, vfloat32m8_t vd,
                                             vfloat32m8_t vs2,
                                             vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wf_bf16_f32m8_tum(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wf_bf16_f32m8_tum(vbool4_t vm, vfloat32m8_t vd,
                                           vfloat32m8_t vs2, __bf16 rs1,
                                           size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vv_bf16mf4_f32mf2_tumu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vv_bf16mf4_f32mf2_tumu(vbool64_t vm, vfloat32mf2_t vd,
                                                 vbfloat16mf4_t vs2,
                                                 vbfloat16mf4_t vs1,
                                                 size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vf_bf16mf4_f32mf2_tumu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vf_bf16mf4_f32mf2_tumu(vbool64_t vm, vfloat32mf2_t vd,
                                                 vbfloat16mf4_t vs2, __bf16 rs1,
                                                 size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wv_bf16mf4_f32mf2_tumu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wv_bf16mf4_f32mf2_tumu(vbool64_t vm, vfloat32mf2_t vd,
                                                 vfloat32mf2_t vs2,
                                                 vbfloat16mf4_t vs1,
                                                 size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wf_bf16_f32mf2_tumu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wf_bf16_f32mf2_tumu(vbool64_t vm, vfloat32mf2_t vd,
                                              vfloat32mf2_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vv_bf16mf2_f32m1_tumu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vv_bf16mf2_f32m1_tumu(vbool32_t vm, vfloat32m1_t vd,
                                               vbfloat16mf2_t vs2,
                                               vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vf_bf16mf2_f32m1_tumu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vf_bf16mf2_f32m1_tumu(vbool32_t vm, vfloat32m1_t vd,
                                               vbfloat16mf2_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wv_bf16mf2_f32m1_tumu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wv_bf16mf2_f32m1_tumu(vbool32_t vm, vfloat32m1_t vd,
                                               vfloat32m1_t vs2,
                                               vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wf_bf16_f32m1_tumu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wf_bf16_f32m1_tumu(vbool32_t vm, vfloat32m1_t vd,
                                            vfloat32m1_t vs2, __bf16 rs1,
                                            size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vv_bf16m1_f32m2_tumu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vv_bf16m1_f32m2_tumu(vbool16_t vm, vfloat32m2_t vd,
                                              vbfloat16m1_t vs2,
                                              vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vf_bf16m1_f32m2_tumu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vf_bf16m1_f32m2_tumu(vbool16_t vm, vfloat32m2_t vd,
                                              vbfloat16m1_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wv_bf16m1_f32m2_tumu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wv_bf16m1_f32m2_tumu(vbool16_t vm, vfloat32m2_t vd,
                                              vfloat32m2_t vs2,
                                              vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wf_bf16_f32m2_tumu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wf_bf16_f32m2_tumu(vbool16_t vm, vfloat32m2_t vd,
                                            vfloat32m2_t vs2, __bf16 rs1,
                                            size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vv_bf16m2_f32m4_tumu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vv_bf16m2_f32m4_tumu(vbool8_t vm, vfloat32m4_t vd,
                                              vbfloat16m2_t vs2,
                                              vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vf_bf16m2_f32m4_tumu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vf_bf16m2_f32m4_tumu(vbool8_t vm, vfloat32m4_t vd,
                                              vbfloat16m2_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wv_bf16m2_f32m4_tumu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wv_bf16m2_f32m4_tumu(vbool8_t vm, vfloat32m4_t vd,
                                              vfloat32m4_t vs2,
                                              vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wf_bf16_f32m4_tumu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wf_bf16_f32m4_tumu(vbool8_t vm, vfloat32m4_t vd,
                                            vfloat32m4_t vs2, __bf16 rs1,
                                            size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vv_bf16m4_f32m8_tumu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vv_bf16m4_f32m8_tumu(vbool4_t vm, vfloat32m8_t vd,
                                              vbfloat16m4_t vs2,
                                              vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vf_bf16m4_f32m8_tumu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vf_bf16m4_f32m8_tumu(vbool4_t vm, vfloat32m8_t vd,
                                              vbfloat16m4_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wv_bf16m4_f32m8_tumu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wv_bf16m4_f32m8_tumu(vbool4_t vm, vfloat32m8_t vd,
                                              vfloat32m8_t vs2,
                                              vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wf_bf16_f32m8_tumu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wf_bf16_f32m8_tumu(vbool4_t vm, vfloat32m8_t vd,
                                            vfloat32m8_t vs2, __bf16 rs1,
                                            size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vv_bf16mf4_f32mf2_mu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vv_bf16mf4_f32mf2_mu(vbool64_t vm, vfloat32mf2_t vd,
                                               vbfloat16mf4_t vs2,
                                               vbfloat16mf4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vf_bf16mf4_f32mf2_mu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vf_bf16mf4_f32mf2_mu(vbool64_t vm, vfloat32mf2_t vd,
                                               vbfloat16mf4_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wv_bf16mf4_f32mf2_mu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wv_bf16mf4_f32mf2_mu(vbool64_t vm, vfloat32mf2_t vd,
                                               vfloat32mf2_t vs2,
                                               vbfloat16mf4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wf_bf16_f32mf2_mu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wf_bf16_f32mf2_mu(vbool64_t vm, vfloat32mf2_t vd,
                                            vfloat32mf2_t vs2, __bf16 rs1,
                                            size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vv_bf16mf2_f32m1_mu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vv_bf16mf2_f32m1_mu(vbool32_t vm, vfloat32m1_t vd,
                                             vbfloat16mf2_t vs2,
                                             vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vf_bf16mf2_f32m1_mu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vf_bf16mf2_f32m1_mu(vbool32_t vm, vfloat32m1_t vd,
                                             vbfloat16mf2_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wv_bf16mf2_f32m1_mu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wv_bf16mf2_f32m1_mu(vbool32_t vm, vfloat32m1_t vd,
                                             vfloat32m1_t vs2,
                                             vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wf_bf16_f32m1_mu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wf_bf16_f32m1_mu(vbool32_t vm, vfloat32m1_t vd,
                                          vfloat32m1_t vs2, __bf16 rs1,
                                          size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vv_bf16m1_f32m2_mu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vv_bf16m1_f32m2_mu(vbool16_t vm, vfloat32m2_t vd,
                                            vbfloat16m1_t vs2,
                                            vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vf_bf16m1_f32m2_mu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vf_bf16m1_f32m2_mu(vbool16_t vm, vfloat32m2_t vd,
                                            vbfloat16m1_t vs2, __bf16 rs1,
                                            size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wv_bf16m1_f32m2_mu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wv_bf16m1_f32m2_mu(vbool16_t vm, vfloat32m2_t vd,
                                            vfloat32m2_t vs2, vbfloat16m1_t vs1,
                                            size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wf_bf16_f32m2_mu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wf_bf16_f32m2_mu(vbool16_t vm, vfloat32m2_t vd,
                                          vfloat32m2_t vs2, __bf16 rs1,
                                          size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vv_bf16m2_f32m4_mu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vv_bf16m2_f32m4_mu(vbool8_t vm, vfloat32m4_t vd,
                                            vbfloat16m2_t vs2,
                                            vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vf_bf16m2_f32m4_mu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vf_bf16m2_f32m4_mu(vbool8_t vm, vfloat32m4_t vd,
                                            vbfloat16m2_t vs2, __bf16 rs1,
                                            size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wv_bf16m2_f32m4_mu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wv_bf16m2_f32m4_mu(vbool8_t vm, vfloat32m4_t vd,
                                            vfloat32m4_t vs2, vbfloat16m2_t vs1,
                                            size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wf_bf16_f32m4_mu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wf_bf16_f32m4_mu(vbool8_t vm, vfloat32m4_t vd,
                                          vfloat32m4_t vs2, __bf16 rs1,
                                          size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vv_bf16m4_f32m8_mu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vv_bf16m4_f32m8_mu(vbool4_t vm, vfloat32m8_t vd,
                                            vbfloat16m4_t vs2,
                                            vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vf_bf16m4_f32m8_mu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vf_bf16m4_f32m8_mu(vbool4_t vm, vfloat32m8_t vd,
                                            vbfloat16m4_t vs2, __bf16 rs1,
                                            size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wv_bf16m4_f32m8_mu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wv_bf16m4_f32m8_mu(vbool4_t vm, vfloat32m8_t vd,
                                            vfloat32m8_t vs2, vbfloat16m4_t vs1,
                                            size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wf_bf16_f32m8_mu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 7, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wf_bf16_f32m8_mu(vbool4_t vm, vfloat32m8_t vd,
                                          vfloat32m8_t vs2, __bf16 rs1,
                                          size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vv_bf16mf4_f32mf2_rm_tu(
// CHECK-RV64-SAME: <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.nxv1f32.nxv1bf16.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], <vscale x 1 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vv_bf16mf4_f32mf2_rm_tu(vfloat32mf2_t vd,
                                                  vbfloat16mf4_t vs2,
                                                  vbfloat16mf4_t vs1,
                                                  size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vf_bf16mf4_f32mf2_rm_tu(
// CHECK-RV64-SAME: <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.nxv1f32.nxv1bf16.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vf_bf16mf4_f32mf2_rm_tu(vfloat32mf2_t vd,
                                                  vbfloat16mf4_t vs2,
                                                  __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wv_bf16mf4_f32mf2_rm_tu(
// CHECK-RV64-SAME: <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], <vscale x 1 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wv_bf16mf4_f32mf2_rm_tu(vfloat32mf2_t vd,
                                                  vfloat32mf2_t vs2,
                                                  vbfloat16mf4_t vs1,
                                                  size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wf_bf16_f32mf2_rm_tu(
// CHECK-RV64-SAME: <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.nxv1f32.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wf_bf16_f32mf2_rm_tu(vfloat32mf2_t vd,
                                               vfloat32mf2_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vv_bf16mf2_f32m1_rm_tu(
// CHECK-RV64-SAME: <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.nxv2f32.nxv2bf16.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], <vscale x 2 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vv_bf16mf2_f32m1_rm_tu(vfloat32m1_t vd,
                                                vbfloat16mf2_t vs2,
                                                vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vf_bf16mf2_f32m1_rm_tu(
// CHECK-RV64-SAME: <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.nxv2f32.nxv2bf16.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vf_bf16mf2_f32m1_rm_tu(vfloat32m1_t vd,
                                                vbfloat16mf2_t vs2, __bf16 rs1,
                                                size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wv_bf16mf2_f32m1_rm_tu(
// CHECK-RV64-SAME: <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], <vscale x 2 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wv_bf16mf2_f32m1_rm_tu(vfloat32m1_t vd,
                                                vfloat32m1_t vs2,
                                                vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wf_bf16_f32m1_rm_tu(
// CHECK-RV64-SAME: <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.nxv2f32.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wf_bf16_f32m1_rm_tu(vfloat32m1_t vd, vfloat32m1_t vs2,
                                             __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vv_bf16m1_f32m2_rm_tu(
// CHECK-RV64-SAME: <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.nxv4f32.nxv4bf16.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], <vscale x 4 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vv_bf16m1_f32m2_rm_tu(vfloat32m2_t vd,
                                               vbfloat16m1_t vs2,
                                               vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vf_bf16m1_f32m2_rm_tu(
// CHECK-RV64-SAME: <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.nxv4f32.nxv4bf16.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vf_bf16m1_f32m2_rm_tu(vfloat32m2_t vd,
                                               vbfloat16m1_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wv_bf16m1_f32m2_rm_tu(
// CHECK-RV64-SAME: <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], <vscale x 4 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wv_bf16m1_f32m2_rm_tu(vfloat32m2_t vd,
                                               vfloat32m2_t vs2,
                                               vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wf_bf16_f32m2_rm_tu(
// CHECK-RV64-SAME: <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.nxv4f32.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wf_bf16_f32m2_rm_tu(vfloat32m2_t vd, vfloat32m2_t vs2,
                                             __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vv_bf16m2_f32m4_rm_tu(
// CHECK-RV64-SAME: <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.nxv8f32.nxv8bf16.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], <vscale x 8 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vv_bf16m2_f32m4_rm_tu(vfloat32m4_t vd,
                                               vbfloat16m2_t vs2,
                                               vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vf_bf16m2_f32m4_rm_tu(
// CHECK-RV64-SAME: <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.nxv8f32.nxv8bf16.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vf_bf16m2_f32m4_rm_tu(vfloat32m4_t vd,
                                               vbfloat16m2_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wv_bf16m2_f32m4_rm_tu(
// CHECK-RV64-SAME: <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], <vscale x 8 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wv_bf16m2_f32m4_rm_tu(vfloat32m4_t vd,
                                               vfloat32m4_t vs2,
                                               vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wf_bf16_f32m4_rm_tu(
// CHECK-RV64-SAME: <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.nxv8f32.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wf_bf16_f32m4_rm_tu(vfloat32m4_t vd, vfloat32m4_t vs2,
                                             __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vv_bf16m4_f32m8_rm_tu(
// CHECK-RV64-SAME: <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.nxv16f32.nxv16bf16.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], <vscale x 16 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vv_bf16m4_f32m8_rm_tu(vfloat32m8_t vd,
                                               vbfloat16m4_t vs2,
                                               vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vf_bf16m4_f32m8_rm_tu(
// CHECK-RV64-SAME: <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.nxv16f32.nxv16bf16.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vf_bf16m4_f32m8_rm_tu(vfloat32m8_t vd,
                                               vbfloat16m4_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wv_bf16m4_f32m8_rm_tu(
// CHECK-RV64-SAME: <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], <vscale x 16 x bfloat> [[VS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wv_bf16m4_f32m8_rm_tu(vfloat32m8_t vd,
                                               vfloat32m8_t vs2,
                                               vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tu(vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wf_bf16_f32m8_rm_tu(
// CHECK-RV64-SAME: <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.nxv16f32.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], bfloat [[RS1]], i64 0, i64 [[VL]])
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wf_bf16_f32m8_rm_tu(vfloat32m8_t vd, vfloat32m8_t vs2,
                                             __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_wf_tu(vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vv_bf16mf4_f32mf2_rm_tum(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vv_bf16mf4_f32mf2_rm_tum(vbool64_t vm,
                                                   vfloat32mf2_t vd,
                                                   vbfloat16mf4_t vs2,
                                                   vbfloat16mf4_t vs1,
                                                   size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vf_bf16mf4_f32mf2_rm_tum(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vf_bf16mf4_f32mf2_rm_tum(vbool64_t vm,
                                                   vfloat32mf2_t vd,
                                                   vbfloat16mf4_t vs2,
                                                   __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wv_bf16mf4_f32mf2_rm_tum(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wv_bf16mf4_f32mf2_rm_tum(vbool64_t vm,
                                                   vfloat32mf2_t vd,
                                                   vfloat32mf2_t vs2,
                                                   vbfloat16mf4_t vs1,
                                                   size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wf_bf16_f32mf2_rm_tum(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wf_bf16_f32mf2_rm_tum(vbool64_t vm, vfloat32mf2_t vd,
                                                vfloat32mf2_t vs2, __bf16 rs1,
                                                size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vv_bf16mf2_f32m1_rm_tum(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vv_bf16mf2_f32m1_rm_tum(vbool32_t vm, vfloat32m1_t vd,
                                                 vbfloat16mf2_t vs2,
                                                 vbfloat16mf2_t vs1,
                                                 size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vf_bf16mf2_f32m1_rm_tum(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vf_bf16mf2_f32m1_rm_tum(vbool32_t vm, vfloat32m1_t vd,
                                                 vbfloat16mf2_t vs2, __bf16 rs1,
                                                 size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wv_bf16mf2_f32m1_rm_tum(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wv_bf16mf2_f32m1_rm_tum(vbool32_t vm, vfloat32m1_t vd,
                                                 vfloat32m1_t vs2,
                                                 vbfloat16mf2_t vs1,
                                                 size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wf_bf16_f32m1_rm_tum(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wf_bf16_f32m1_rm_tum(vbool32_t vm, vfloat32m1_t vd,
                                              vfloat32m1_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vv_bf16m1_f32m2_rm_tum(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vv_bf16m1_f32m2_rm_tum(vbool16_t vm, vfloat32m2_t vd,
                                                vbfloat16m1_t vs2,
                                                vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vf_bf16m1_f32m2_rm_tum(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vf_bf16m1_f32m2_rm_tum(vbool16_t vm, vfloat32m2_t vd,
                                                vbfloat16m1_t vs2, __bf16 rs1,
                                                size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wv_bf16m1_f32m2_rm_tum(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wv_bf16m1_f32m2_rm_tum(vbool16_t vm, vfloat32m2_t vd,
                                                vfloat32m2_t vs2,
                                                vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wf_bf16_f32m2_rm_tum(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wf_bf16_f32m2_rm_tum(vbool16_t vm, vfloat32m2_t vd,
                                              vfloat32m2_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vv_bf16m2_f32m4_rm_tum(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vv_bf16m2_f32m4_rm_tum(vbool8_t vm, vfloat32m4_t vd,
                                                vbfloat16m2_t vs2,
                                                vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vf_bf16m2_f32m4_rm_tum(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vf_bf16m2_f32m4_rm_tum(vbool8_t vm, vfloat32m4_t vd,
                                                vbfloat16m2_t vs2, __bf16 rs1,
                                                size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wv_bf16m2_f32m4_rm_tum(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wv_bf16m2_f32m4_rm_tum(vbool8_t vm, vfloat32m4_t vd,
                                                vfloat32m4_t vs2,
                                                vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wf_bf16_f32m4_rm_tum(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wf_bf16_f32m4_rm_tum(vbool8_t vm, vfloat32m4_t vd,
                                              vfloat32m4_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vv_bf16m4_f32m8_rm_tum(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vv_bf16m4_f32m8_rm_tum(vbool4_t vm, vfloat32m8_t vd,
                                                vbfloat16m4_t vs2,
                                                vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vf_bf16m4_f32m8_rm_tum(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vf_bf16m4_f32m8_rm_tum(vbool4_t vm, vfloat32m8_t vd,
                                                vbfloat16m4_t vs2, __bf16 rs1,
                                                size_t vl) {
  return __riscv_vfwsub_vf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wv_bf16m4_f32m8_rm_tum(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wv_bf16m4_f32m8_rm_tum(vbool4_t vm, vfloat32m8_t vd,
                                                vfloat32m8_t vs2,
                                                vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tum(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wf_bf16_f32m8_rm_tum(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 2)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wf_bf16_f32m8_rm_tum(vbool4_t vm, vfloat32m8_t vd,
                                              vfloat32m8_t vs2, __bf16 rs1,
                                              size_t vl) {
  return __riscv_vfwsub_wf_tum(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vv_bf16mf4_f32mf2_rm_tumu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vv_bf16mf4_f32mf2_rm_tumu(vbool64_t vm,
                                                    vfloat32mf2_t vd,
                                                    vbfloat16mf4_t vs2,
                                                    vbfloat16mf4_t vs1,
                                                    size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vf_bf16mf4_f32mf2_rm_tumu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vf_bf16mf4_f32mf2_rm_tumu(vbool64_t vm,
                                                    vfloat32mf2_t vd,
                                                    vbfloat16mf4_t vs2,
                                                    __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wv_bf16mf4_f32mf2_rm_tumu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wv_bf16mf4_f32mf2_rm_tumu(vbool64_t vm,
                                                    vfloat32mf2_t vd,
                                                    vfloat32mf2_t vs2,
                                                    vbfloat16mf4_t vs1,
                                                    size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wf_bf16_f32mf2_rm_tumu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wf_bf16_f32mf2_rm_tumu(vbool64_t vm, vfloat32mf2_t vd,
                                                 vfloat32mf2_t vs2, __bf16 rs1,
                                                 size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vv_bf16mf2_f32m1_rm_tumu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vv_bf16mf2_f32m1_rm_tumu(vbool32_t vm, vfloat32m1_t vd,
                                                  vbfloat16mf2_t vs2,
                                                  vbfloat16mf2_t vs1,
                                                  size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vf_bf16mf2_f32m1_rm_tumu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vf_bf16mf2_f32m1_rm_tumu(vbool32_t vm, vfloat32m1_t vd,
                                                  vbfloat16mf2_t vs2,
                                                  __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wv_bf16mf2_f32m1_rm_tumu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wv_bf16mf2_f32m1_rm_tumu(vbool32_t vm, vfloat32m1_t vd,
                                                  vfloat32m1_t vs2,
                                                  vbfloat16mf2_t vs1,
                                                  size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wf_bf16_f32m1_rm_tumu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wf_bf16_f32m1_rm_tumu(vbool32_t vm, vfloat32m1_t vd,
                                               vfloat32m1_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vv_bf16m1_f32m2_rm_tumu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vv_bf16m1_f32m2_rm_tumu(vbool16_t vm, vfloat32m2_t vd,
                                                 vbfloat16m1_t vs2,
                                                 vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vf_bf16m1_f32m2_rm_tumu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vf_bf16m1_f32m2_rm_tumu(vbool16_t vm, vfloat32m2_t vd,
                                                 vbfloat16m1_t vs2, __bf16 rs1,
                                                 size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wv_bf16m1_f32m2_rm_tumu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wv_bf16m1_f32m2_rm_tumu(vbool16_t vm, vfloat32m2_t vd,
                                                 vfloat32m2_t vs2,
                                                 vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wf_bf16_f32m2_rm_tumu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wf_bf16_f32m2_rm_tumu(vbool16_t vm, vfloat32m2_t vd,
                                               vfloat32m2_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vv_bf16m2_f32m4_rm_tumu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vv_bf16m2_f32m4_rm_tumu(vbool8_t vm, vfloat32m4_t vd,
                                                 vbfloat16m2_t vs2,
                                                 vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vf_bf16m2_f32m4_rm_tumu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vf_bf16m2_f32m4_rm_tumu(vbool8_t vm, vfloat32m4_t vd,
                                                 vbfloat16m2_t vs2, __bf16 rs1,
                                                 size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wv_bf16m2_f32m4_rm_tumu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wv_bf16m2_f32m4_rm_tumu(vbool8_t vm, vfloat32m4_t vd,
                                                 vfloat32m4_t vs2,
                                                 vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wf_bf16_f32m4_rm_tumu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wf_bf16_f32m4_rm_tumu(vbool8_t vm, vfloat32m4_t vd,
                                               vfloat32m4_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vv_bf16m4_f32m8_rm_tumu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vv_bf16m4_f32m8_rm_tumu(vbool4_t vm, vfloat32m8_t vd,
                                                 vbfloat16m4_t vs2,
                                                 vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vf_bf16m4_f32m8_rm_tumu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vf_bf16m4_f32m8_rm_tumu(vbool4_t vm, vfloat32m8_t vd,
                                                 vbfloat16m4_t vs2, __bf16 rs1,
                                                 size_t vl) {
  return __riscv_vfwsub_vf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wv_bf16m4_f32m8_rm_tumu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wv_bf16m4_f32m8_rm_tumu(vbool4_t vm, vfloat32m8_t vd,
                                                 vfloat32m8_t vs2,
                                                 vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_tumu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wf_bf16_f32m8_rm_tumu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 0)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wf_bf16_f32m8_rm_tumu(vbool4_t vm, vfloat32m8_t vd,
                                               vfloat32m8_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_wf_tumu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vv_bf16mf4_f32mf2_rm_mu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vv_bf16mf4_f32mf2_rm_mu(vbool64_t vm,
                                                  vfloat32mf2_t vd,
                                                  vbfloat16mf4_t vs2,
                                                  vbfloat16mf4_t vs1,
                                                  size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_vf_bf16mf4_f32mf2_rm_mu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.mask.nxv1f32.nxv1bf16.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_vf_bf16mf4_f32mf2_rm_mu(vbool64_t vm,
                                                  vfloat32mf2_t vd,
                                                  vbfloat16mf4_t vs2,
                                                  __bf16 rs1, size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wv_bf16mf4_f32mf2_rm_mu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], <vscale x 1 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.nxv1bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], <vscale x 1 x bfloat> [[VS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wv_bf16mf4_f32mf2_rm_mu(vbool64_t vm,
                                                  vfloat32mf2_t vd,
                                                  vfloat32mf2_t vs2,
                                                  vbfloat16mf4_t vs1,
                                                  size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 1 x float> @test_vfwsub_wf_bf16_f32mf2_rm_mu(
// CHECK-RV64-SAME: <vscale x 1 x i1> [[VM:%.*]], <vscale x 1 x float> [[VD:%.*]], <vscale x 1 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 1 x float> @llvm.riscv.vfwsub.w.mask.nxv1f32.bf16.i64(<vscale x 1 x float> [[VD]], <vscale x 1 x float> [[VS2]], bfloat [[RS1]], <vscale x 1 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 1 x float> [[TMP0]]
//
vfloat32mf2_t test_vfwsub_wf_bf16_f32mf2_rm_mu(vbool64_t vm, vfloat32mf2_t vd,
                                               vfloat32mf2_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vv_bf16mf2_f32m1_rm_mu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vv_bf16mf2_f32m1_rm_mu(vbool32_t vm, vfloat32m1_t vd,
                                                vbfloat16mf2_t vs2,
                                                vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_vf_bf16mf2_f32m1_rm_mu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.mask.nxv2f32.nxv2bf16.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_vf_bf16mf2_f32m1_rm_mu(vbool32_t vm, vfloat32m1_t vd,
                                                vbfloat16mf2_t vs2, __bf16 rs1,
                                                size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wv_bf16mf2_f32m1_rm_mu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], <vscale x 2 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.nxv2bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], <vscale x 2 x bfloat> [[VS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wv_bf16mf2_f32m1_rm_mu(vbool32_t vm, vfloat32m1_t vd,
                                                vfloat32m1_t vs2,
                                                vbfloat16mf2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 2 x float> @test_vfwsub_wf_bf16_f32m1_rm_mu(
// CHECK-RV64-SAME: <vscale x 2 x i1> [[VM:%.*]], <vscale x 2 x float> [[VD:%.*]], <vscale x 2 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 2 x float> @llvm.riscv.vfwsub.w.mask.nxv2f32.bf16.i64(<vscale x 2 x float> [[VD]], <vscale x 2 x float> [[VS2]], bfloat [[RS1]], <vscale x 2 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 2 x float> [[TMP0]]
//
vfloat32m1_t test_vfwsub_wf_bf16_f32m1_rm_mu(vbool32_t vm, vfloat32m1_t vd,
                                             vfloat32m1_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vv_bf16m1_f32m2_rm_mu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vv_bf16m1_f32m2_rm_mu(vbool16_t vm, vfloat32m2_t vd,
                                               vbfloat16m1_t vs2,
                                               vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_vf_bf16m1_f32m2_rm_mu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.mask.nxv4f32.nxv4bf16.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_vf_bf16m1_f32m2_rm_mu(vbool16_t vm, vfloat32m2_t vd,
                                               vbfloat16m1_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wv_bf16m1_f32m2_rm_mu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], <vscale x 4 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.nxv4bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], <vscale x 4 x bfloat> [[VS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wv_bf16m1_f32m2_rm_mu(vbool16_t vm, vfloat32m2_t vd,
                                               vfloat32m2_t vs2,
                                               vbfloat16m1_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 4 x float> @test_vfwsub_wf_bf16_f32m2_rm_mu(
// CHECK-RV64-SAME: <vscale x 4 x i1> [[VM:%.*]], <vscale x 4 x float> [[VD:%.*]], <vscale x 4 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x float> @llvm.riscv.vfwsub.w.mask.nxv4f32.bf16.i64(<vscale x 4 x float> [[VD]], <vscale x 4 x float> [[VS2]], bfloat [[RS1]], <vscale x 4 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 4 x float> [[TMP0]]
//
vfloat32m2_t test_vfwsub_wf_bf16_f32m2_rm_mu(vbool16_t vm, vfloat32m2_t vd,
                                             vfloat32m2_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vv_bf16m2_f32m4_rm_mu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vv_bf16m2_f32m4_rm_mu(vbool8_t vm, vfloat32m4_t vd,
                                               vbfloat16m2_t vs2,
                                               vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_vf_bf16m2_f32m4_rm_mu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.mask.nxv8f32.nxv8bf16.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_vf_bf16m2_f32m4_rm_mu(vbool8_t vm, vfloat32m4_t vd,
                                               vbfloat16m2_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wv_bf16m2_f32m4_rm_mu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], <vscale x 8 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.nxv8bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], <vscale x 8 x bfloat> [[VS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wv_bf16m2_f32m4_rm_mu(vbool8_t vm, vfloat32m4_t vd,
                                               vfloat32m4_t vs2,
                                               vbfloat16m2_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 8 x float> @test_vfwsub_wf_bf16_f32m4_rm_mu(
// CHECK-RV64-SAME: <vscale x 8 x i1> [[VM:%.*]], <vscale x 8 x float> [[VD:%.*]], <vscale x 8 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 8 x float> @llvm.riscv.vfwsub.w.mask.nxv8f32.bf16.i64(<vscale x 8 x float> [[VD]], <vscale x 8 x float> [[VS2]], bfloat [[RS1]], <vscale x 8 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 8 x float> [[TMP0]]
//
vfloat32m4_t test_vfwsub_wf_bf16_f32m4_rm_mu(vbool8_t vm, vfloat32m4_t vd,
                                             vfloat32m4_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vv_bf16m4_f32m8_rm_mu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vv_bf16m4_f32m8_rm_mu(vbool4_t vm, vfloat32m8_t vd,
                                               vbfloat16m4_t vs2,
                                               vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_vv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_vf_bf16m4_f32m8_rm_mu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x bfloat> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.mask.nxv16f32.nxv16bf16.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x bfloat> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_vf_bf16m4_f32m8_rm_mu(vbool4_t vm, vfloat32m8_t vd,
                                               vbfloat16m4_t vs2, __bf16 rs1,
                                               size_t vl) {
  return __riscv_vfwsub_vf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wv_bf16m4_f32m8_rm_mu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], <vscale x 16 x bfloat> [[VS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.nxv16bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], <vscale x 16 x bfloat> [[VS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wv_bf16m4_f32m8_rm_mu(vbool4_t vm, vfloat32m8_t vd,
                                               vfloat32m8_t vs2,
                                               vbfloat16m4_t vs1, size_t vl) {
  return __riscv_vfwsub_wv_mu(vm, vd, vs2, vs1, __RISCV_FRM_RNE, vl);
}

// CHECK-RV64-LABEL: define dso_local <vscale x 16 x float> @test_vfwsub_wf_bf16_f32m8_rm_mu(
// CHECK-RV64-SAME: <vscale x 16 x i1> [[VM:%.*]], <vscale x 16 x float> [[VD:%.*]], <vscale x 16 x float> [[VS2:%.*]], bfloat noundef [[RS1:%.*]], i64 noundef [[VL:%.*]]) #[[ATTR0]] {
// CHECK-RV64-NEXT:  [[ENTRY:.*:]]
// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 16 x float> @llvm.riscv.vfwsub.w.mask.nxv16f32.bf16.i64(<vscale x 16 x float> [[VD]], <vscale x 16 x float> [[VS2]], bfloat [[RS1]], <vscale x 16 x i1> [[VM]], i64 0, i64 [[VL]], i64 1)
// CHECK-RV64-NEXT:    ret <vscale x 16 x float> [[TMP0]]
//
vfloat32m8_t test_vfwsub_wf_bf16_f32m8_rm_mu(vbool4_t vm, vfloat32m8_t vd,
                                             vfloat32m8_t vs2, __bf16 rs1,
                                             size_t vl) {
  return __riscv_vfwsub_wf_mu(vm, vd, vs2, rs1, __RISCV_FRM_RNE, vl);
}