3dc621124f
**COO**
`lvlTypes = [ "compressed_nu", "singleton" ]` to `map = (d0, d1) -> (d0
: compressed(nonunique), d1 : singleton)`
`lvlTypes = [ "compressed_nu_no", "singleton_no" ]` to `map = (d0, d1)
-> (d0 : compressed(nonunique, nonordered), d1 : singleton(nonordered))`
**SortedCOO**
`lvlTypes = [ "compressed_nu", "singleton" ]` to `map = (d0, d1) -> (d0
: compressed(nonunique), d1 : singleton)`
**BCOO**
`lvlTypes = [ "dense", "compressed_hi_nu", "singleton" ]` to `map = (d0,
d1, d2) -> (d0 : dense, d1 : compressed(nonunique, high), d2 :
singleton)`
**BCSR**
`lvlTypes = [ "compressed", "compressed", "dense", "dense" ], dimToLvl =
affine_map<(d0, d1) -> (d0 floordiv 2, d1 floordiv 3, d0 mod 2, d1 mod
3)>` to
`map = ( i, j ) ->
( i floordiv 2 : compressed,
j floordiv 3 : compressed,
i mod 2 : dense,
j mod 3 : dense
)`
**Tensor and other supported formats(e.g. CCC, CDC, CCCC)**
Currently, ELL and slice are not supported yet in the new syntax and the
CHECK tests will be updated once printing is set to output the new
syntax.
Previous PRs: #66146, #66309, #66443
90 lines
6.7 KiB
MLIR
90 lines
6.7 KiB
MLIR
//
|
|
// TODO: this test case is temporarily disabled as we are improving zero-cost sparse tensor reshaping.
|
|
// XFAIL: *
|
|
//
|
|
// RUN: mlir-opt %s --linalg-generalize-named-ops --sparsification --cse --canonicalize | FileCheck %s
|
|
|
|
#COO_2D = #sparse_tensor.encoding<{ map = (d0, d1) -> (d0 : compressed(nonunique), d1 : singleton), posWidth = 32, crdWidth = 32 }>
|
|
#COO_3D = #sparse_tensor.encoding<{ map = (d0, d1, d2) -> (d0 : compressed(nonunique), d1 : singleton(nonunique), d2 : singleton), posWidth = 32, crdWidth = 32 }>
|
|
|
|
|
|
// CHECK-LABEL: func.func @sparse_reshape_fused(
|
|
// CHECK-SAME: %[[VAL_0:.*]]: tensor<5x6xf32>,
|
|
// CHECK-SAME: %[[VAL_1:.*]]: tensor<6x2x3xf32, #sparse_tensor.encoding<{ lvlTypes = [ "compressed_nu", "singleton_nu", "singleton" ], posWidth = 32, crdWidth = 32 }>>) -> tensor<?x?x?xf32> {
|
|
// CHECK-DAG: %[[VAL_2:.*]] = arith.constant false
|
|
// CHECK-DAG: %[[VAL_3:.*]] = arith.constant 5 : index
|
|
// CHECK-DAG: %[[VAL_4:.*]] = arith.constant 3 : index
|
|
// CHECK-DAG: %[[VAL_5:.*]] = arith.constant 0 : index
|
|
// CHECK-DAG: %[[VAL_6:.*]] = arith.constant 1 : index
|
|
// CHECK-DAG: %[[VAL_7:.*]] = tensor.empty() : tensor<5x6xf32>
|
|
// CHECK-DAG: %[[VAL_8:.*]] = sparse_tensor.positions %[[VAL_1]] {level = 0 : index}
|
|
// CHECK-DAG: %[[VAL_9:.*]] = sparse_tensor.coordinates %[[VAL_1]] {level = 0 : index}
|
|
// CHECK-DAG: %[[VAL_10:.*]] = sparse_tensor.coordinates %[[VAL_1]] {level = 1 : index}
|
|
// CHECK-DAG: %[[VAL_11:.*]] = sparse_tensor.coordinates %[[VAL_1]] {level = 2 : index}
|
|
// CHECK-DAG: %[[VAL_12:.*]] = sparse_tensor.values %[[VAL_1]]
|
|
// CHECK-DAG: %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_7]] : memref<5x6xf32>
|
|
// CHECK: scf.for %[[VAL_14:.*]] = %[[VAL_5]] to %[[VAL_3]] step %[[VAL_6]] {
|
|
// CHECK: %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_5]]] : memref<?xi32>
|
|
// CHECK: %[[VAL_16:.*]] = arith.extui %[[VAL_15]] : i32 to i64
|
|
// CHECK: %[[VAL_17:.*]] = arith.index_cast %[[VAL_16]] : i64 to index
|
|
// CHECK: %[[VAL_18:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_6]]] : memref<?xi32>
|
|
// CHECK: %[[VAL_19:.*]] = arith.extui %[[VAL_18]] : i32 to i64
|
|
// CHECK: %[[VAL_20:.*]] = arith.index_cast %[[VAL_19]] : i64 to index
|
|
// CHECK: %[[VAL_21:.*]] = scf.while (%[[VAL_22:.*]] = %[[VAL_17]]) : (index) -> index {
|
|
// CHECK: %[[VAL_23:.*]] = arith.cmpi ult, %[[VAL_22]], %[[VAL_20]] : index
|
|
// CHECK: scf.condition(%[[VAL_23]]) %[[VAL_22]] : index
|
|
// CHECK: } do {
|
|
// CHECK: ^bb0(%[[VAL_24:.*]]: index):
|
|
// CHECK: %[[VAL_25:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_24]]] : memref<?xi32, strided<[?], offset: ?>>
|
|
// CHECK: %[[VAL_26:.*]] = arith.extui %[[VAL_25]] : i32 to i64
|
|
// CHECK: %[[VAL_27:.*]] = arith.index_cast %[[VAL_26]] : i64 to index
|
|
// CHECK: %[[VAL_28:.*]] = scf.while (%[[VAL_29:.*]] = %[[VAL_24]]) : (index) -> index {
|
|
// CHECK: %[[VAL_30:.*]] = arith.cmpi ult, %[[VAL_29]], %[[VAL_20]] : index
|
|
// CHECK: %[[VAL_31:.*]] = scf.if %[[VAL_30]] -> (i1) {
|
|
// CHECK: %[[VAL_32:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_29]]] : memref<?xi32, strided<[?], offset: ?>>
|
|
// CHECK: %[[VAL_33:.*]] = arith.extui %[[VAL_32]] : i32 to i64
|
|
// CHECK: %[[VAL_34:.*]] = arith.index_cast %[[VAL_33]] : i64 to index
|
|
// CHECK: %[[VAL_35:.*]] = arith.cmpi eq, %[[VAL_34]], %[[VAL_27]] : index
|
|
// CHECK: scf.yield %[[VAL_35]] : i1
|
|
// CHECK: } else {
|
|
// CHECK: scf.yield %[[VAL_2]] : i1
|
|
// CHECK: }
|
|
// CHECK: scf.condition(%[[VAL_36:.*]]) %[[VAL_29]] : index
|
|
// CHECK: } do {
|
|
// CHECK: ^bb0(%[[VAL_37:.*]]: index):
|
|
// CHECK: %[[VAL_38:.*]] = arith.addi %[[VAL_37]], %[[VAL_6]] : index
|
|
// CHECK: scf.yield %[[VAL_38]] : index
|
|
// CHECK: }
|
|
// CHECK: %[[VAL_39:.*]] = tensor.extract %[[VAL_0]]{{\[}}%[[VAL_14]], %[[VAL_27]]] : tensor<5x6xf32>
|
|
// CHECK: scf.for %[[VAL_40:.*]] = %[[VAL_24]] to %[[VAL_41:.*]] step %[[VAL_6]] {
|
|
// CHECK: %[[VAL_42:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_40]]] : memref<?xi32, strided<[?], offset: ?>>
|
|
// CHECK: %[[VAL_43:.*]] = arith.extui %[[VAL_42]] : i32 to i64
|
|
// CHECK: %[[VAL_44:.*]] = arith.index_cast %[[VAL_43]] : i64 to index
|
|
// CHECK: %[[VAL_45:.*]] = arith.muli %[[VAL_44]], %[[VAL_4]] : index
|
|
// CHECK: %[[VAL_46:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_40]]] : memref<?xi32, strided<[?], offset: ?>>
|
|
// CHECK: %[[VAL_47:.*]] = arith.extui %[[VAL_46]] : i32 to i64
|
|
// CHECK: %[[VAL_48:.*]] = arith.index_cast %[[VAL_47]] : i64 to index
|
|
// CHECK: %[[VAL_49:.*]] = arith.addi %[[VAL_45]], %[[VAL_48]] : index
|
|
// CHECK: %[[VAL_50:.*]] = tensor.extract %[[VAL_7]]{{\[}}%[[VAL_14]], %[[VAL_49]]] : tensor<5x6xf32>
|
|
// CHECK: %[[VAL_51:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_40]]] : memref<?xf32>
|
|
// CHECK: %[[VAL_52:.*]] = arith.mulf %[[VAL_39]], %[[VAL_51]] : f32
|
|
// CHECK: %[[VAL_53:.*]] = arith.addf %[[VAL_50]], %[[VAL_52]] : f32
|
|
// CHECK: memref.store %[[VAL_53]], %[[VAL_13]]{{\[}}%[[VAL_14]], %[[VAL_49]]] : memref<5x6xf32>
|
|
// CHECK: } {"Emitted from" = "linalg.generic"}
|
|
// CHECK: scf.yield %[[VAL_54:.*]] : index
|
|
// CHECK: } attributes {"Emitted from" = "linalg.generic"}
|
|
// CHECK: } {"Emitted from" = "linalg.generic"}
|
|
// CHECK: %[[VAL_55:.*]] = bufferization.to_tensor %[[VAL_13]] : memref<5x6xf32>
|
|
// CHECK: %[[VAL_56:.*]] = tensor.expand_shape %[[VAL_55]] {{\[\[}}0], [1, 2]] : tensor<5x6xf32> into tensor<5x2x3xf32>
|
|
// CHECK: %[[VAL_57:.*]] = tensor.cast %[[VAL_56]] : tensor<5x2x3xf32> to tensor<?x?x?xf32>
|
|
// CHECK: return %[[VAL_57]] : tensor<?x?x?xf32>
|
|
// CHECK: }
|
|
func.func @sparse_reshape_fused(%arg0: tensor<5x6xf32>, %arg1: tensor<6x2x3xf32, #COO_3D>) -> tensor<?x?x?xf32> {
|
|
%collapsed = tensor.collapse_shape %arg1 [[0], [1, 2]] : tensor<6x2x3xf32, #COO_3D> into tensor<6x6xf32, #COO_2D>
|
|
%0 = tensor.empty() : tensor<5x6xf32>
|
|
%2 = linalg.matmul ins(%arg0, %collapsed : tensor<5x6xf32>, tensor<6x6xf32, #COO_2D>) outs(%0 : tensor<5x6xf32>) -> tensor<5x6xf32>
|
|
%expanded = tensor.expand_shape %2 [[0], [1, 2]] : tensor<5x6xf32> into tensor<5x2x3xf32>
|
|
%ret1 = tensor.cast %expanded : tensor<5x2x3xf32> to tensor<?x?x?xf32>
|
|
return %ret1 : tensor<?x?x?xf32>
|
|
}
|