622765f976
- Seems based on whether NVPTX backend is enabled or not, this call can have the range() attribute or not. So make this check more permissive.
77 lines
3.3 KiB
C
77 lines
3.3 KiB
C
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
|
|
// RUN: %clang_cc1 -internal-isystem %S/Inputs/include \
|
|
// RUN: -internal-isystem %S/../../lib/Headers/cuda_wrappers \
|
|
// RUN: -internal-isystem %S/../../lib/Headers/ \
|
|
// RUN: -fcuda-is-device -triple nvptx64 -emit-llvm %s -o - \
|
|
// RUN: | FileCheck %s --check-prefix=CUDA
|
|
//
|
|
// RUN: %clang_cc1 -internal-isystem %S/Inputs/include \
|
|
// RUN: -internal-isystem %S/../../lib/Headers/cuda_wrappers \
|
|
// RUN: -internal-isystem %S/../../lib/Headers/ \
|
|
// RUN: -fcuda-is-device -triple amdgcn -emit-llvm %s -o - \
|
|
// RUN: | FileCheck %s --check-prefix=HIP
|
|
//
|
|
// RUN: %clang_cc1 -internal-isystem %S/Inputs/include \
|
|
// RUN: -internal-isystem %S/../../lib/Headers/ \
|
|
// RUN: -cl-std=CL3.0 -triple amdgcn -emit-llvm %s -o - \
|
|
// RUN: | FileCheck %s --check-prefix=OPENCL
|
|
//
|
|
// RUN: %clang_cc1 -internal-isystem %S/Inputs/include \
|
|
// RUN: -internal-isystem %S/../../lib/Headers/ -cl-std=CL3.0 \
|
|
// RUN: -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa \
|
|
// RUN: -fopenmp-is-target-device -triple amdgcn -emit-llvm %s -o - \
|
|
// RUN: | FileCheck %s --check-prefix=OPENMP
|
|
//
|
|
// RUN: %clang_cc1 -internal-isystem %S/Inputs/include \
|
|
// RUN: -std=c89 -internal-isystem %S/../../lib/Headers/ \
|
|
// RUN: -triple amdgcn-amd-amdhsa -emit-llvm %s -o - \
|
|
// RUN: | FileCheck %s --check-prefix=C89
|
|
|
|
#define _DEFAULT_FN_ATTRS __attribute__((always_inline))
|
|
#include <gpuintrin.h>
|
|
|
|
#ifdef __device__
|
|
__device__ int foo() { return __gpu_thread_id_x(); }
|
|
#else
|
|
// CUDA-LABEL: define dso_local i32 @foo(
|
|
// CUDA-SAME: ) #[[ATTR0:[0-9]+]] {
|
|
// CUDA-NEXT: [[ENTRY:.*:]]
|
|
// CUDA-NEXT: [[TMP0:%.*]] = call {{.*}}i32 @llvm.nvvm.read.ptx.sreg.tid.x()
|
|
// CUDA-NEXT: ret i32 [[TMP0]]
|
|
//
|
|
// HIP-LABEL: define dso_local i32 @foo(
|
|
// HIP-SAME: ) #[[ATTR0:[0-9]+]] {
|
|
// HIP-NEXT: [[ENTRY:.*:]]
|
|
// HIP-NEXT: [[RETVAL_I:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// HIP-NEXT: [[RETVAL:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// HIP-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr
|
|
// HIP-NEXT: [[RETVAL_ASCAST_I:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL_I]] to ptr
|
|
// HIP-NEXT: [[TMP0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
|
|
// HIP-NEXT: ret i32 [[TMP0]]
|
|
//
|
|
// OPENCL-LABEL: define dso_local i32 @foo(
|
|
// OPENCL-SAME: ) #[[ATTR0:[0-9]+]] {
|
|
// OPENCL-NEXT: [[ENTRY:.*:]]
|
|
// OPENCL-NEXT: [[TMP0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
|
|
// OPENCL-NEXT: ret i32 [[TMP0]]
|
|
//
|
|
// OPENMP-LABEL: define hidden i32 @foo(
|
|
// OPENMP-SAME: ) #[[ATTR0:[0-9]+]] {
|
|
// OPENMP-NEXT: [[ENTRY:.*:]]
|
|
// OPENMP-NEXT: [[TMP0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
|
|
// OPENMP-NEXT: ret i32 [[TMP0]]
|
|
//
|
|
// C89-LABEL: define dso_local i32 @foo(
|
|
// C89-SAME: ) #[[ATTR0:[0-9]+]] {
|
|
// C89-NEXT: [[ENTRY:.*:]]
|
|
// C89-NEXT: [[RETVAL_I:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// C89-NEXT: [[RETVAL:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// C89-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr
|
|
// C89-NEXT: [[RETVAL_ASCAST_I:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL_I]] to ptr
|
|
// C89-NEXT: [[TMP0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
|
|
// C89-NEXT: ret i32 [[TMP0]]
|
|
//
|
|
int foo() { return __gpu_thread_id_x(); }
|
|
#pragma omp declare target to(foo)
|
|
#endif
|