forked from OSchip/llvm-project
125 lines
7.8 KiB
MLIR
125 lines
7.8 KiB
MLIR
// NOTE: Assertions have been autogenerated by utils/generate-test-checks.py
|
|
|
|
// The script is designed to make adding checks to
|
|
// a test case fast, it is *not* designed to be authoritative
|
|
// about what constitutes a good test! The CHECK should be
|
|
// minimized and named to reflect the test intent.
|
|
|
|
// RUN: mlir-opt %s -sparsification="vectorization-strategy=2 vl=8" -canonicalize | \
|
|
// RUN: FileCheck %s
|
|
|
|
#SparseVector = #sparse_tensor.encoding<{
|
|
dimLevelType = ["compressed"]
|
|
}>
|
|
|
|
#trait_1d = {
|
|
indexing_maps = [
|
|
affine_map<(i) -> (i)>, // a
|
|
affine_map<(i) -> (i)> // x (out)
|
|
],
|
|
iterator_types = ["parallel"],
|
|
doc = "X(i) = a(i) op i"
|
|
}
|
|
|
|
// CHECK-LABEL: func @sparse_index_1d_conj(
|
|
// CHECK-SAME: %[[VAL_0:.*]]: tensor<8xi64, #sparse_tensor.encoding<{{{.*}}}>>) -> tensor<8xi64> {
|
|
// CHECK-DAG: %[[VAL_1:.*]] = arith.constant dense<0> : vector<8xi64>
|
|
// CHECK-DAG: %[[VAL_2:.*]] = arith.constant dense<0> : vector<8xindex>
|
|
// CHECK-DAG: %[[VAL_3:.*]] = arith.constant 8 : index
|
|
// CHECK-DAG: %[[VAL_4:.*]] = arith.constant 1 : index
|
|
// CHECK-DAG: %[[VAL_5:.*]] = arith.constant 0 : i64
|
|
// CHECK-DAG: %[[VAL_6:.*]] = arith.constant 0 : index
|
|
// CHECK-DAG: %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_6]] : tensor<8xi64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
|
|
// CHECK-DAG: %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_6]] : tensor<8xi64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
|
|
// CHECK-DAG: %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<8xi64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xi64>
|
|
// CHECK-DAG: %[[VAL_10:.*]] = memref.alloc() : memref<8xi64>
|
|
// CHECK-DAG: linalg.fill ins(%[[VAL_5]] : i64) outs(%[[VAL_10]] : memref<8xi64>)
|
|
// CHECK-DAG: %[[VAL_11:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex>
|
|
// CHECK-DAG: %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_4]]] : memref<?xindex>
|
|
// CHECK: scf.for %[[VAL_13:.*]] = %[[VAL_11]] to %[[VAL_12]] step %[[VAL_3]] {
|
|
// CHECK: %[[VAL_14:.*]] = affine.min #map0(%[[VAL_13]]){{\[}}%[[VAL_12]]]
|
|
// CHECK: %[[VAL_15:.*]] = vector.create_mask %[[VAL_14]] : vector<8xi1>
|
|
// CHECK: %[[VAL_16:.*]] = vector.maskedload %[[VAL_8]]{{\[}}%[[VAL_13]]], %[[VAL_15]], %[[VAL_2]] : memref<?xindex>, vector<8xi1>, vector<8xindex> into vector<8xindex>
|
|
// CHECK: %[[VAL_17:.*]] = vector.maskedload %[[VAL_9]]{{\[}}%[[VAL_13]]], %[[VAL_15]], %[[VAL_1]] : memref<?xi64>, vector<8xi1>, vector<8xi64> into vector<8xi64>
|
|
// CHECK: %[[VAL_18:.*]] = arith.index_cast %[[VAL_16]] : vector<8xindex> to vector<8xi64>
|
|
// CHECK: %[[VAL_19:.*]] = arith.muli %[[VAL_17]], %[[VAL_18]] : vector<8xi64>
|
|
// CHECK: vector.scatter %[[VAL_10]]{{\[}}%[[VAL_6]]] {{\[}}%[[VAL_16]]], %[[VAL_15]], %[[VAL_19]] : memref<8xi64>, vector<8xindex>, vector<8xi1>, vector<8xi64>
|
|
// CHECK: }
|
|
// CHECK: %[[VAL_20:.*]] = bufferization.to_tensor %[[VAL_10]] : memref<8xi64>
|
|
// CHECK: return %[[VAL_20]] : tensor<8xi64>
|
|
// CHECK: }
|
|
func.func @sparse_index_1d_conj(%arga: tensor<8xi64, #SparseVector>) -> tensor<8xi64> {
|
|
%init = linalg.init_tensor [8] : tensor<8xi64>
|
|
%r = linalg.generic #trait_1d
|
|
ins(%arga: tensor<8xi64, #SparseVector>)
|
|
outs(%init: tensor<8xi64>) {
|
|
^bb(%a: i64, %x: i64):
|
|
%i = linalg.index 0 : index
|
|
%ii = arith.index_cast %i : index to i64
|
|
%m1 = arith.muli %a, %ii : i64
|
|
linalg.yield %m1 : i64
|
|
} -> tensor<8xi64>
|
|
return %r : tensor<8xi64>
|
|
}
|
|
|
|
// CHECK-LABEL: func @sparse_index_1d_disj(
|
|
// CHECK-SAME: %[[VAL_0:.*]]: tensor<8xi64, #sparse_tensor.encoding<{{{.*}}}>>) -> tensor<8xi64> {
|
|
// CHECK-DAG: %[[VAL_1:.*]] = arith.constant dense<[0, 1, 2, 3, 4, 5, 6, 7]> : vector<8xindex>
|
|
// CHECK-DAG: %[[VAL_2:.*]] = arith.constant 1 : index
|
|
// CHECK-DAG: %[[VAL_3:.*]] = arith.constant 0 : i64
|
|
// CHECK-DAG: %[[VAL_4:.*]] = arith.constant 8 : index
|
|
// CHECK-DAG: %[[VAL_5:.*]] = arith.constant 0 : index
|
|
// CHECK-DAG: %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_5]] : tensor<8xi64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
|
|
// CHECK-DAG: %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_5]] : tensor<8xi64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xindex>
|
|
// CHECK-DAG: %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<8xi64, #sparse_tensor.encoding<{{{.*}}}>> to memref<?xi64>
|
|
// CHECK-DAG: %[[VAL_9:.*]] = memref.alloc() : memref<8xi64>
|
|
// CHECK-DAG: linalg.fill ins(%[[VAL_3]] : i64) outs(%[[VAL_9]] : memref<8xi64>)
|
|
// CHECK-DAG: %[[VAL_10:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
|
|
// CHECK-DAG: %[[VAL_11:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_2]]] : memref<?xindex>
|
|
// CHECK: %[[VAL_12:.*]]:2 = scf.while (%[[VAL_13:.*]] = %[[VAL_10]], %[[VAL_14:.*]] = %[[VAL_5]]) : (index, index) -> (index, index) {
|
|
// CHECK: %[[VAL_15:.*]] = arith.cmpi ult, %[[VAL_13]], %[[VAL_11]] : index
|
|
// CHECK: scf.condition(%[[VAL_15]]) %[[VAL_13]], %[[VAL_14]] : index, index
|
|
// CHECK: } do {
|
|
// CHECK: ^bb0(%[[VAL_16:.*]]: index, %[[VAL_17:.*]]: index):
|
|
// CHECK: %[[VAL_18:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_16]]] : memref<?xindex>
|
|
// CHECK: %[[VAL_19:.*]] = arith.cmpi eq, %[[VAL_18]], %[[VAL_17]] : index
|
|
// CHECK: scf.if %[[VAL_19]] {
|
|
// CHECK: %[[VAL_20:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_16]]] : memref<?xi64>
|
|
// CHECK: %[[VAL_21:.*]] = arith.index_cast %[[VAL_17]] : index to i64
|
|
// CHECK: %[[VAL_22:.*]] = arith.addi %[[VAL_20]], %[[VAL_21]] : i64
|
|
// CHECK: memref.store %[[VAL_22]], %[[VAL_9]]{{\[}}%[[VAL_17]]] : memref<8xi64>
|
|
// CHECK: } else {
|
|
// CHECK: %[[VAL_23:.*]] = arith.index_cast %[[VAL_17]] : index to i64
|
|
// CHECK: memref.store %[[VAL_23]], %[[VAL_9]]{{\[}}%[[VAL_17]]] : memref<8xi64>
|
|
// CHECK: }
|
|
// CHECK: %[[VAL_24:.*]] = arith.cmpi eq, %[[VAL_18]], %[[VAL_17]] : index
|
|
// CHECK: %[[VAL_25:.*]] = arith.addi %[[VAL_16]], %[[VAL_2]] : index
|
|
// CHECK: %[[VAL_26:.*]] = arith.select %[[VAL_24]], %[[VAL_25]], %[[VAL_16]] : index
|
|
// CHECK: %[[VAL_27:.*]] = arith.addi %[[VAL_17]], %[[VAL_2]] : index
|
|
// CHECK: scf.yield %[[VAL_26]], %[[VAL_27]] : index, index
|
|
// CHECK: }
|
|
// CHECK: scf.for %[[VAL_28:.*]] = %[[VAL_29:.*]]#1 to %[[VAL_4]] step %[[VAL_4]] {
|
|
// CHECK: %[[VAL_30:.*]] = affine.min #map1(%[[VAL_28]])
|
|
// CHECK: %[[VAL_31:.*]] = vector.create_mask %[[VAL_30]] : vector<8xi1>
|
|
// CHECK: %[[VAL_32:.*]] = vector.broadcast %[[VAL_28]] : index to vector<8xindex>
|
|
// CHECK: %[[VAL_33:.*]] = arith.addi %[[VAL_32]], %[[VAL_1]] : vector<8xindex>
|
|
// CHECK: %[[VAL_34:.*]] = arith.index_cast %[[VAL_33]] : vector<8xindex> to vector<8xi64>
|
|
// CHECK: vector.maskedstore %[[VAL_9]]{{\[}}%[[VAL_28]]], %[[VAL_31]], %[[VAL_34]] : memref<8xi64>, vector<8xi1>, vector<8xi64>
|
|
// CHECK: }
|
|
// CHECK: %[[VAL_35:.*]] = bufferization.to_tensor %[[VAL_9]] : memref<8xi64>
|
|
// CHECK: return %[[VAL_35]] : tensor<8xi64>
|
|
// CHECK: }
|
|
func.func @sparse_index_1d_disj(%arga: tensor<8xi64, #SparseVector>) -> tensor<8xi64> {
|
|
%init = linalg.init_tensor [8] : tensor<8xi64>
|
|
%r = linalg.generic #trait_1d
|
|
ins(%arga: tensor<8xi64, #SparseVector>)
|
|
outs(%init: tensor<8xi64>) {
|
|
^bb(%a: i64, %x: i64):
|
|
%i = linalg.index 0 : index
|
|
%ii = arith.index_cast %i : index to i64
|
|
%m1 = arith.addi %a, %ii : i64
|
|
linalg.yield %m1 : i64
|
|
} -> tensor<8xi64>
|
|
return %r : tensor<8xi64>
|
|
}
|