3c707310a3
Summary: Some recent support made usage of `__nvvm_reflect` more consistent. We should expose it as a builtin rather than forcing users to externally define the function.
29 lines
1.4 KiB
Common Lisp
29 lines
1.4 KiB
Common Lisp
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 4
|
|
// RUN: %clang_cc1 %s -triple nvptx-unknown-unknown -emit-llvm -O0 -o - | FileCheck %s
|
|
|
|
// CHECK-LABEL: define dso_local zeroext i1 @device_function(
|
|
// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.nvvm.reflect(ptr addrspacecast (ptr addrspace(4) @.str to ptr))
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp uge i32 [[TMP0]], 700
|
|
// CHECK-NEXT: ret i1 [[CMP]]
|
|
//
|
|
bool device_function() {
|
|
return __nvvm_reflect("__CUDA_ARCH") >= 700;
|
|
}
|
|
|
|
// CHECK-LABEL: define dso_local spir_kernel void @kernel_function(
|
|
// CHECK-SAME: ptr addrspace(1) noundef align 4 [[I:%.*]]) #[[ATTR2:[0-9]+]] !kernel_arg_addr_space !4 !kernel_arg_access_qual !5 !kernel_arg_type !6 !kernel_arg_base_type !6 !kernel_arg_type_qual !7 {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[I_ADDR:%.*]] = alloca ptr addrspace(1), align 4
|
|
// CHECK-NEXT: store ptr addrspace(1) [[I]], ptr [[I_ADDR]], align 4
|
|
// CHECK-NEXT: [[CALL:%.*]] = call zeroext i1 @device_function() #[[ATTR3:[0-9]+]]
|
|
// CHECK-NEXT: [[CONV:%.*]] = zext i1 [[CALL]] to i32
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr addrspace(1), ptr [[I_ADDR]], align 4
|
|
// CHECK-NEXT: store i32 [[CONV]], ptr addrspace(1) [[TMP0]], align 4
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
__kernel void kernel_function(__global int *i) {
|
|
*i = device_function();
|
|
}
|