29 lines · plain
1// RUN: mlir-opt %s \2// RUN: | mlir-opt -gpu-lower-to-nvvm-pipeline="cubin-format=%gpu_compilation_format allow-pattern-rollback=0" \3// RUN: | mlir-runner \4// RUN: --shared-libs=%mlir_cuda_runtime \5// RUN: --shared-libs=%mlir_runner_utils \6// RUN: --shared-libs=%mlir_c_runner_utils \7// RUN: --entry-point-result=void \8// RUN: | FileCheck %s9 10// CHECK: 200011module attributes {gpu.container_module} {12 func.func @main() {13 %c1 = arith.constant 1 : index14 %c0 = arith.constant 0 : index15 %c1000_i32 = arith.constant 1000 : i3216 %memref = gpu.alloc host_shared () : memref<1xi32>17 memref.store %c1000_i32, %memref[%c1] : memref<1xi32>18 gpu.launch blocks(%arg0, %arg1, %arg2) in (%arg6 = %c1, %arg7 = %c1, %arg8 = %c1) threads(%arg3, %arg4, %arg5) in (%arg9 = %c1, %arg10 = %c1, %arg11 = %c1) {19 %1 = memref.load %memref[%c1] : memref<1xi32>20 %2 = arith.addi %1, %1 : i3221 memref.store %2, %memref[%c1] : memref<1xi32>22 gpu.terminator23 }24 %0 = memref.load %memref[%c1] : memref<1xi32>25 vector.print %0 : i3226 return27 }28}29