1// RUN: mlir-spirv-cpu-runner %s -e main --entry-point-result=void --shared-libs=%mlir_runner_utils_dir/libmlir_runner_utils%shlibext,%spirv_wrapper_library_dir/libmlir_test_spirv_cpu_runner_c_wrappers%shlibext 2 3// CHECK: [8, 8, 8, 8, 8, 8] 4module attributes { 5 gpu.container_module, 6 spv.target_env = #spv.target_env< 7 #spv.vce<v1.0, [Shader], [SPV_KHR_variable_pointers]>, 8 #spv.resource_limits< 9 max_compute_workgroup_invocations = 128, 10 max_compute_workgroup_size = [128, 128, 64]>> 11} { 12 gpu.module @kernels { 13 gpu.func @double(%arg0 : memref<6xi32>, %arg1 : memref<6xi32>) 14 kernel attributes { spv.entry_point_abi = #spv.entry_point_abi<local_size = dense<[1, 1, 1]>: vector<3xi32>>} { 15 %factor = arith.constant 2 : i32 16 17 %i0 = arith.constant 0 : index 18 %i1 = arith.constant 1 : index 19 %i2 = arith.constant 2 : index 20 %i3 = arith.constant 3 : index 21 %i4 = arith.constant 4 : index 22 %i5 = arith.constant 5 : index 23 24 %x0 = memref.load %arg0[%i0] : memref<6xi32> 25 %x1 = memref.load %arg0[%i1] : memref<6xi32> 26 %x2 = memref.load %arg0[%i2] : memref<6xi32> 27 %x3 = memref.load %arg0[%i3] : memref<6xi32> 28 %x4 = memref.load %arg0[%i4] : memref<6xi32> 29 %x5 = memref.load %arg0[%i5] : memref<6xi32> 30 31 %y0 = arith.muli %x0, %factor : i32 32 %y1 = arith.muli %x1, %factor : i32 33 %y2 = arith.muli %x2, %factor : i32 34 %y3 = arith.muli %x3, %factor : i32 35 %y4 = arith.muli %x4, %factor : i32 36 %y5 = arith.muli %x5, %factor : i32 37 38 memref.store %y0, %arg1[%i0] : memref<6xi32> 39 memref.store %y1, %arg1[%i1] : memref<6xi32> 40 memref.store %y2, %arg1[%i2] : memref<6xi32> 41 memref.store %y3, %arg1[%i3] : memref<6xi32> 42 memref.store %y4, %arg1[%i4] : memref<6xi32> 43 memref.store %y5, %arg1[%i5] : memref<6xi32> 44 gpu.return 45 } 46 } 47 func.func @main() { 48 %input = memref.alloc() : memref<6xi32> 49 %output = memref.alloc() : memref<6xi32> 50 %four = arith.constant 4 : i32 51 %zero = arith.constant 0 : i32 52 %input_casted = memref.cast %input : memref<6xi32> to memref<?xi32> 53 %output_casted = memref.cast %output : memref<6xi32> to memref<?xi32> 54 call @fillI32Buffer(%input_casted, %four) : (memref<?xi32>, i32) -> () 55 call @fillI32Buffer(%output_casted, %zero) : (memref<?xi32>, i32) -> () 56 57 %one = arith.constant 1 : index 58 gpu.launch_func @kernels::@double 59 blocks in (%one, %one, %one) threads in (%one, %one, %one) 60 args(%input : memref<6xi32>, %output : memref<6xi32>) 61 %result = memref.cast %output : memref<6xi32> to memref<*xi32> 62 call @printMemrefI32(%result) : (memref<*xi32>) -> () 63 return 64 } 65 66 func.func private @fillI32Buffer(%arg0 : memref<?xi32>, %arg1 : i32) 67 func.func private @printMemrefI32(%ptr : memref<*xi32>) 68} 69