llvm/clang/test/Headers/__clang_hip_cmath.hip

// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
// REQUIRES: amdgpu-registered-target

// RUN: %clang_cc1 -include __clang_hip_runtime_wrapper.h      \
// RUN:   -internal-isystem %S/../../lib/Headers/cuda_wrappers \
// RUN:   -internal-isystem %S/Inputs/include \
// RUN:   -triple amdgcn-amd-amdhsa -aux-triple x86_64-unknown-unknown \
// RUN:   -target-cpu gfx906 -emit-llvm %s -fcuda-is-device -O1 -o - \
// RUN:   -D__HIPCC_RTC__ | FileCheck -check-prefix=DEFAULT %s

// Check that we end up with fast math flags set on intrinsic calls
// RUN: %clang_cc1 -include __clang_hip_runtime_wrapper.h \
// RUN:   -internal-isystem %S/../../lib/Headers/cuda_wrappers \
// RUN:   -internal-isystem %S/Inputs/include \
// RUN:   -triple amdgcn-amd-amdhsa -aux-triple x86_64-unknown-unknown \
// RUN:   -target-cpu gfx906 -emit-llvm %s -fcuda-is-device -O1 -menable-no-infs \
// RUN:   -menable-no-nans -o - \
// RUN:   -D__HIPCC_RTC__ | FileCheck -check-prefix=FINITEONLY %s

// DEFAULT-LABEL: @test_fma_f16(
// DEFAULT-NEXT:  entry:
// DEFAULT-NEXT:    [[TMP0:%.*]] = tail call contract noundef half @llvm.fma.f16(half [[X:%.*]], half [[Y:%.*]], half [[Z:%.*]])
// DEFAULT-NEXT:    ret half [[TMP0]]
//
// FINITEONLY-LABEL: @test_fma_f16(
// FINITEONLY-NEXT:  entry:
// FINITEONLY-NEXT:    [[TMP0:%.*]] = tail call nnan ninf contract noundef half @llvm.fma.f16(half [[X:%.*]], half [[Y:%.*]], half [[Z:%.*]])
// FINITEONLY-NEXT:    ret half [[TMP0]]
//
extern "C" __device__ _Float16 test_fma_f16(_Float16 x, _Float16 y,
                                            _Float16 z) {
  return fma(x, y, z);
}

// DEFAULT-LABEL: @test_pow_f16(
// DEFAULT-NEXT:  entry:
// DEFAULT-NEXT:    [[CALL_I:%.*]] = tail call contract noundef half @__ocml_pown_f16(half noundef [[X:%.*]], i32 noundef [[Y:%.*]]) #[[ATTR7:[0-9]+]]
// DEFAULT-NEXT:    ret half [[CALL_I]]
//
// FINITEONLY-LABEL: @test_pow_f16(
// FINITEONLY-NEXT:  entry:
// FINITEONLY-NEXT:    [[CALL_I:%.*]] = tail call nnan ninf contract noundef nofpclass(nan inf) half @__ocml_pown_f16(half noundef nofpclass(nan inf) [[X:%.*]], i32 noundef [[Y:%.*]]) #[[ATTR7:[0-9]+]]
// FINITEONLY-NEXT:    ret half [[CALL_I]]
//
extern "C" __device__ _Float16 test_pow_f16(_Float16 x, int y) {
  return pow(x, y);
}

// DEFAULT-LABEL: @test_fabs_f32(
// DEFAULT-NEXT:  entry:
// DEFAULT-NEXT:    [[TMP0:%.*]] = tail call contract noundef float @llvm.fabs.f32(float [[X:%.*]])
// DEFAULT-NEXT:    ret float [[TMP0]]
//
// FINITEONLY-LABEL: @test_fabs_f32(
// FINITEONLY-NEXT:  entry:
// FINITEONLY-NEXT:    [[TMP0:%.*]] = tail call nnan ninf contract noundef float @llvm.fabs.f32(float [[X:%.*]])
// FINITEONLY-NEXT:    ret float [[TMP0]]
//
extern "C" __device__ float test_fabs_f32(float x) {
  return fabs(x);
}

// DEFAULT-LABEL: @test_sin_f32(
// DEFAULT-NEXT:  entry:
// DEFAULT-NEXT:    [[CALL_I1:%.*]] = tail call contract noundef float @__ocml_sin_f32(float noundef [[X:%.*]]) #[[ATTR8:[0-9]+]]
// DEFAULT-NEXT:    ret float [[CALL_I1]]
//
// FINITEONLY-LABEL: @test_sin_f32(
// FINITEONLY-NEXT:  entry:
// FINITEONLY-NEXT:    [[CALL_I1:%.*]] = tail call nnan ninf contract noundef nofpclass(nan inf) float @__ocml_sin_f32(float noundef nofpclass(nan inf) [[X:%.*]]) #[[ATTR8:[0-9]+]]
// FINITEONLY-NEXT:    ret float [[CALL_I1]]
//
extern "C" __device__ float test_sin_f32(float x) {
  return sin(x);
}

// DEFAULT-LABEL: @test_cos_f32(
// DEFAULT-NEXT:  entry:
// DEFAULT-NEXT:    [[CALL_I1:%.*]] = tail call contract noundef float @__ocml_cos_f32(float noundef [[X:%.*]]) #[[ATTR8]]
// DEFAULT-NEXT:    ret float [[CALL_I1]]
//
// FINITEONLY-LABEL: @test_cos_f32(
// FINITEONLY-NEXT:  entry:
// FINITEONLY-NEXT:    [[CALL_I1:%.*]] = tail call nnan ninf contract noundef nofpclass(nan inf) float @__ocml_cos_f32(float noundef nofpclass(nan inf) [[X:%.*]]) #[[ATTR8]]
// FINITEONLY-NEXT:    ret float [[CALL_I1]]
//
extern "C" __device__ float test_cos_f32(float x) {
  return cos(x);
}

// Check user defined type which can be converted to float and double but not
// specializes __numeric_type will not cause ambiguity diagnostics.
struct user_bfloat16 {
  __host__ __device__ user_bfloat16(float);
  operator float();
  operator double();
};

namespace user_namespace {
  __device__ user_bfloat16 fma(const user_bfloat16 a, const user_bfloat16 b, const user_bfloat16 c) {
    return a;
  }

  __global__ void test_fma() {
    user_bfloat16 a = 1.0f, b = 2.0f;
    fma(a, b, b);
  }
}