| 1 |
// RUN: fir-opt --split-input-file --external-name-interop %s | FileCheck %s |
| 2 |
|
| 3 |
module @mod attributes {gpu.container_module} { |
| 4 |
|
| 5 |
gpu.module @cuda_device_mod { |
| 6 |
gpu.func @_QPfoo() kernel { |
| 7 |
fir.call @_QPthreadfence() fastmath<contract> : () -> () |
| 8 |
gpu.return |
| 9 |
} |
| 10 |
func.func private @_QPthreadfence() attributes {cuf.proc_attr = #cuf.cuda_proc<device>} |
| 11 |
} |
| 12 |
|
| 13 |
func.func @test() -> () { |
| 14 |
%0 = llvm.mlir.constant(0 : i64) : i64 |
| 15 |
%1 = llvm.mlir.constant(0 : i32) : i32 |
| 16 |
gpu.launch_func @cuda_device_mod::@_QPfoo blocks in (%0, %0, %0) threads in (%0, %0, %0) : i64 dynamic_shared_memory_size %1 |
| 17 |
return |
| 18 |
} |
| 19 |
|
| 20 |
// CHECK-LABEL: gpu.func @foo_() |
| 21 |
// CHECK: fir.call @threadfence_() |
| 22 |
// CHECK: func.func private @threadfence_() |
| 23 |
// CHECK: gpu.launch_func @cuda_device_mod::@foo_ |
| 24 |
|
| 25 |
} |