f03826f896
Lower !gpu.async.tokens returned from async.execute regions to events instead of streams. Make !gpu.async.token returned from !async.execute single-use. This allows creating one event per use and destroying them without leaking or ref-counting. Technically we only need this for stream/event-based lowering. I kept the code separate from the rest of the gpu-async-region pass so that we can make this optional or move to a separate pass as needed. Reviewed By: herhut Differential Revision: https://reviews.llvm.org/D96965
22 lines
760 B
MLIR
22 lines
760 B
MLIR
// RUN: mlir-opt %s --gpu-to-llvm | FileCheck %s
|
|
|
|
module attributes {gpu.container_module} {
|
|
|
|
func @foo() {
|
|
// CHECK: %[[t0:.*]] = llvm.call @mgpuStreamCreate
|
|
// CHECK: %[[e0:.*]] = llvm.call @mgpuEventCreate
|
|
// CHECK: llvm.call @mgpuEventRecord(%[[e0]], %[[t0]])
|
|
%t0 = gpu.wait async
|
|
// CHECK: %[[t1:.*]] = llvm.call @mgpuStreamCreate
|
|
// CHECK: llvm.call @mgpuStreamWaitEvent(%[[t1]], %[[e0]])
|
|
// CHECK: llvm.call @mgpuEventDestroy(%[[e0]])
|
|
%t1 = gpu.wait async [%t0]
|
|
// CHECK: llvm.call @mgpuStreamSynchronize(%[[t0]])
|
|
// CHECK: llvm.call @mgpuStreamDestroy(%[[t0]])
|
|
// CHECK: llvm.call @mgpuStreamSynchronize(%[[t1]])
|
|
// CHECK: llvm.call @mgpuStreamDestroy(%[[t1]])
|
|
gpu.wait [%t0, %t1]
|
|
return
|
|
}
|
|
}
|