1// RUN: mlir-opt %s --sparse-reinterpret-map -sparsification -cse | FileCheck %s 2 3#Dense = #sparse_tensor.encoding<{ 4 map = (d0, d1) -> (d0 : dense, d1 : dense) 5}> 6 7#trait_scale = { 8 indexing_maps = [ 9 affine_map<(i,j) -> (i,j)> // X (out) 10 ], 11 iterator_types = ["parallel", "parallel"], 12 doc = "X(i,j) = X(i,j) * 2.0" 13} 14 15// CHECK-LABEL: func.func @sparse_scale( 16// CHECK-SAME: %[[VAL_0:.*]]: tensor<1x1xf32, #sparse{{[0-9]*}}>) 17// CHECK-DAG: %[[VAL_1:.*]] = arith.constant 0 : index 18// CHECK-DAG: %[[VAL_2:.*]] = arith.constant 2.000000e+00 : f32 19// CHECK: %[[VAL_3:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<1x1xf32, #sparse{{[0-9]*}}> to memref<?xf32> 20// CHECK: %[[VAL_4:.*]] = memref.load %[[VAL_3]]{{\[}}%[[VAL_1]]] : memref<?xf32> 21// CHECK: %[[VAL_5:.*]] = arith.mulf %[[VAL_4]], %[[VAL_2]] : f32 22// CHECK: memref.store %[[VAL_5]], %[[VAL_3]]{{\[}}%[[VAL_1]]] : memref<?xf32> 23// CHECK: %[[VAL_6:.*]] = sparse_tensor.load %[[VAL_0]] : tensor<1x1xf32, #sparse{{[0-9]*}}> 24// CHECK: return %[[VAL_6]] : tensor<1x1xf32, #sparse{{[0-9]*}}> 25func.func @sparse_scale(%argx: tensor<1x1xf32, #Dense>) -> tensor<1x1xf32, #Dense> { 26 %c = arith.constant 2.0 : f32 27 %0 = linalg.generic #trait_scale 28 outs(%argx: tensor<1x1xf32, #Dense>) { 29 ^bb(%x: f32): 30 %1 = arith.mulf %x, %c : f32 31 linalg.yield %1 : f32 32 } -> tensor<1x1xf32, #Dense> 33 return %0 : tensor<1x1xf32, #Dense> 34} 35