This is an archive of the discontinued LLVM Phabricator instance.

Differential D133219

[mlir][sparse] extend codegen test cases with an additional step after storage expansion
ClosedPublic

Authored by Peiming on Sep 2 2022, 11:07 AM.

Details

Reviewers
aartbik
wrengr
bixia
Commits
rG662ee93cbbaa: [mlir][sparse] extend codegen test cases with an additional step after storage…

Diff Detail

Event Timeline

Peiming created this revision.Sep 2 2022, 11:07 AM
Herald added a reviewer: aartbik. · View Herald TranscriptSep 2 2022, 11:07 AM
Herald added a project: Restricted Project. · View Herald Transcript
Herald added subscribers: anlunx, bzcheeseman, sdasgup3 and 19 others. · View Herald Transcript
Peiming requested review of this revision.Sep 2 2022, 11:07 AM
Herald added a project: Restricted Project. · View Herald TranscriptSep 2 2022, 11:07 AM
Herald added subscribers: stephenneuendorffer, nicolasvasilache. · View Herald Transcript
Peiming added reviewers: wrengr, bixia.Sep 2 2022, 11:07 AM
Peiming added a parent revision: D133175: [mlir][sparse] add conversion rules for storage_get/set/callOp.
aartbik accepted this revision.Sep 2 2022, 11:20 AM

Nice! One nit, but good to go in (after dep revision of course)

mlir/test/Dialect/SparseTensor/codegen.mlir
43

here and below, can you use a // for the empty lines between CODEGEN/STORAGE, so the block as a whole "stays connecedt" with the function

This revision is now accepted and ready to land.Sep 2 2022, 11:20 AM
Peiming updated this revision to Diff 457649.Sep 2 2022, 11:23 AM

Address comments

Peiming marked an inline comment as done.Sep 2 2022, 11:23 AM
This revision was landed with ongoing or failed builds.Sep 2 2022, 11:34 AM
Closed by commit rG662ee93cbbaa: [mlir][sparse] extend codegen test cases with an additional step after storage… (authored by Peiming). · Explain Why
This revision was automatically updated to reflect the committed changes.
Peiming added a commit: rG662ee93cbbaa: [mlir][sparse] extend codegen test cases with an additional step after storage….
Harbormaster completed remote builds in B184860: Diff 457649.Sep 2 2022, 11:44 AM

Revision Contents

PathSize
mlir/
test/
Dialect/
SparseTensor/
202 lines

Diff 457654

mlir/test/Dialect/SparseTensor/codegen.mlir

// RUN: mlir-opt %s --sparse-tensor-codegen --canonicalize --cse | FileCheck %s // RUN: mlir-opt %s --sparse-tensor-codegen --canonicalize --cse | FileCheck %s --check-prefix=CHECK-CODEGEN
// RUN: mlir-opt %s --sparse-tensor-codegen --sparse-tensor-storage-expansion --canonicalize --cse | FileCheck %s --check-prefix=CHECK-STORAGE
#SparseVector = #sparse_tensor.encoding<{ #SparseVector = #sparse_tensor.encoding<{
dimLevelType = [ "compressed" ], dimLevelType = [ "compressed" ],
indexBitWidth = 64, indexBitWidth = 64,
pointerBitWidth = 32 pointerBitWidth = 32
}> }>
#Dense2D = #sparse_tensor.encoding<{ #Dense2D = #sparse_tensor.encoding<{
Show All 20 Lines#DCSR = #sparse_tensor.encoding<{
pointerBitWidth = 32 pointerBitWidth = 32
}> }>
#Dense3D = #sparse_tensor.encoding<{ #Dense3D = #sparse_tensor.encoding<{
dimLevelType = [ "dense", "dense", "dense" ], dimLevelType = [ "dense", "dense", "dense" ],
dimOrdering = affine_map<(i, j, k) -> (k, i, j)> dimOrdering = affine_map<(i, j, k) -> (k, i, j)>
}> }>
// CHECK-LABEL: func @sparse_nop( // CHECK-CODEGEN-LABEL: func @sparse_nop(
// CHECK-SAME: %[[A:.*]]: tuple<memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>)
// CHECK: return %[[A]] : tuple<memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> // CHECK-CODEGEN: return %[[A]] : tuple<memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>
//
aartbikUnsubmitted
Done
ReplyInline Actions

here and below, can you use a // for the empty lines between CODEGEN/STORAGE, so the block as a whole "stays connecedt" with the function

aartbik: here and below, can you use a // for the empty lines between CODEGEN/STORAGE, so the block as a…
// CHECK-STORAGE-LABEL: func @sparse_nop(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<1xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xf64>)
// CHECK-STORAGE: return %[[A0]], %[[A1]], %[[A2]], %[[A3]] : memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>
func.func @sparse_nop(%arg0: tensor<?xf64, #SparseVector>) -> tensor<?xf64, #SparseVector> { func.func @sparse_nop(%arg0: tensor<?xf64, #SparseVector>) -> tensor<?xf64, #SparseVector> {
return %arg0 : tensor<?xf64, #SparseVector> return %arg0 : tensor<?xf64, #SparseVector>
} }
// CHECK-LABEL: func @sparse_nop_cast( // CHECK-CODEGEN-LABEL: func @sparse_nop_cast(
// CHECK-SAME: %[[A:.*]]: tuple<memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf32>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf32>>)
// CHECK: return %[[A]] : tuple<memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf32>> // CHECK-CODEGEN: return %[[A]] : tuple<memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf32>>
//
// CHECK-STORAGE-LABEL: func @sparse_nop_cast(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<1xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xf32>)
// CHECK-STORAGE: return %[[A0]], %[[A1]], %[[A2]], %[[A3]] : memref<1xindex>, memref<?xi32>, memref<?xi64>, memref<?xf32>
func.func @sparse_nop_cast(%arg0: tensor<64xf32, #SparseVector>) -> tensor<?xf32, #SparseVector> { func.func @sparse_nop_cast(%arg0: tensor<64xf32, #SparseVector>) -> tensor<?xf32, #SparseVector> {
%0 = tensor.cast %arg0 : tensor<64xf32, #SparseVector> to tensor<?xf32, #SparseVector> %0 = tensor.cast %arg0 : tensor<64xf32, #SparseVector> to tensor<?xf32, #SparseVector>
return %0 : tensor<?xf32, #SparseVector> return %0 : tensor<?xf32, #SparseVector>
} }
// CHECK-LABEL: func @sparse_nop_cast_3d( // CHECK-CODEGEN-LABEL: func @sparse_nop_cast_3d(
// CHECK-SAME: %[[A:.*]]: tuple<memref<3xindex>, memref<?xf32>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<3xindex>, memref<?xf32>>)
// CHECK: return %[[A]] : tuple<memref<3xindex>, memref<?xf32>> // CHECK-CODEGEN: return %[[A]] : tuple<memref<3xindex>, memref<?xf32>>
//
// CHECK-STORAGE-LABEL: func @sparse_nop_cast_3d(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<3xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xf32>)
// CHECK-STORAGE: return %[[A0]], %[[A1]] : memref<3xindex>, memref<?xf32>
func.func @sparse_nop_cast_3d(%arg0: tensor<10x20x30xf32, #Dense3D>) -> tensor<?x?x?xf32, #Dense3D> { func.func @sparse_nop_cast_3d(%arg0: tensor<10x20x30xf32, #Dense3D>) -> tensor<?x?x?xf32, #Dense3D> {
%0 = tensor.cast %arg0 : tensor<10x20x30xf32, #Dense3D> to tensor<?x?x?xf32, #Dense3D> %0 = tensor.cast %arg0 : tensor<10x20x30xf32, #Dense3D> to tensor<?x?x?xf32, #Dense3D>
return %0 : tensor<?x?x?xf32, #Dense3D> return %0 : tensor<?x?x?xf32, #Dense3D>
} }
// CHECK-LABEL: func @sparse_dense_2d( // CHECK-CODEGEN-LABEL: func @sparse_dense_2d(
// CHECK-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xf64>>)
//
// CHECK-STORAGE-LABEL: func @sparse_dense_2d(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<2xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xf64>) {
// CHECK-STORAGE: return
func.func @sparse_dense_2d(%arg0: tensor<?x?xf64, #Dense2D>) { func.func @sparse_dense_2d(%arg0: tensor<?x?xf64, #Dense2D>) {
return return
} }
// CHECK-LABEL: func @sparse_row( // CHECK-CODEGEN-LABEL: func @sparse_row(
// CHECK-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>)
//
// CHECK-STORAGE-LABEL: func @sparse_row(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<2xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xf64>) {
// CHECK-STORAGE: return
func.func @sparse_row(%arg0: tensor<?x?xf64, #Row>) { func.func @sparse_row(%arg0: tensor<?x?xf64, #Row>) {
return return
} }
// CHECK-LABEL: func @sparse_csr( // CHECK-CODEGEN-LABEL: func @sparse_csr(
// CHECK-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>)
//
// CHECK-STORAGE-LABEL: func @sparse_csr(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<2xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xf64>) {
// CHECK-STORAGE: return
func.func @sparse_csr(%arg0: tensor<?x?xf64, #CSR>) { func.func @sparse_csr(%arg0: tensor<?x?xf64, #CSR>) {
return return
} }
// CHECK-LABEL: func @sparse_dcsr( // CHECK-CODEGEN-LABEL: func @sparse_dcsr(
// CHECK-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>>)
//
// CHECK-STORAGE-LABEL: func @sparse_dcsr(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<2xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A4:.*4]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A5:.*5]]: memref<?xf64>) {
// CHECK-STORAGE: return
func.func @sparse_dcsr(%arg0: tensor<?x?xf64, #DCSR>) { func.func @sparse_dcsr(%arg0: tensor<?x?xf64, #DCSR>) {
return return
} }
// //
// Querying for dimension 1 in the tensor type can immediately // Querying for dimension 1 in the tensor type can immediately
// fold using the original static dimension sizes. // fold using the original static dimension sizes.
// //
// CHECK-LABEL: func @sparse_dense_3d( // CHECK-CODEGEN-LABEL: func @sparse_dense_3d(
// CHECK-SAME: %[[A:.*]]: tuple<memref<3xindex>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<3xindex>, memref<?xf64>>)
// CHECK: %[[C:.*]] = arith.constant 20 : index // CHECK-CODEGEN: %[[C:.*]] = arith.constant 20 : index
// CHECK: return %[[C]] : index // CHECK-CODEGEN: return %[[C]] : index
//
// CHECK-STORAGE-LABEL: func @sparse_dense_3d(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<3xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xf64>)
// CHECK-STORAGE: %[[C:.*]] = arith.constant 20 : index
// CHECK-STORAGE: return %[[C]] : index
func.func @sparse_dense_3d(%arg0: tensor<10x20x30xf64, #Dense3D>) -> index { func.func @sparse_dense_3d(%arg0: tensor<10x20x30xf64, #Dense3D>) -> index {
%c = arith.constant 1 : index %c = arith.constant 1 : index
%0 = tensor.dim %arg0, %c : tensor<10x20x30xf64, #Dense3D> %0 = tensor.dim %arg0, %c : tensor<10x20x30xf64, #Dense3D>
return %0 : index return %0 : index
} }
// //
// Querying for dimension 1 in the tensor type needs to be permuted // Querying for dimension 1 in the tensor type needs to be permuted
// into querying for dimension 2 in the stored sparse tensor scheme, // into querying for dimension 2 in the stored sparse tensor scheme,
// since the latter honors the dimOrdering. // since the latter honors the dimOrdering.
// //
// CHECK-LABEL: func @sparse_dense_3d_dyn( // CHECK-CODEGEN-LABEL: func @sparse_dense_3d_dyn(
// CHECK-SAME: %[[A:.*]]: tuple<memref<3xindex>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<3xindex>, memref<?xf64>>)
// CHECK: %[[C:.*]] = arith.constant 2 : index // CHECK-CODEGEN: %[[C:.*]] = arith.constant 2 : index
// CHECK: %[[F:.*]] = sparse_tensor.storage_get %[[A]][0] : tuple<memref<3xindex>, memref<?xf64>> to memref<3xindex> // CHECK-CODEGEN: %[[F:.*]] = sparse_tensor.storage_get %[[A]][0] : tuple<memref<3xindex>, memref<?xf64>> to memref<3xindex>
// CHECK: %[[L:.*]] = memref.load %[[F]][%[[C]]] : memref<3xindex> // CHECK-CODEGEN: %[[L:.*]] = memref.load %[[F]][%[[C]]] : memref<3xindex>
// CHECK: return %[[L]] : index // CHECK-CODEGEN: return %[[L]] : index
//
// CHECK-STORAGE-LABEL: func @sparse_dense_3d_dyn(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<3xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xf64>)
// CHECK-STORAGE: %[[C:.*]] = arith.constant 2 : index
// CHECK-STORAGE: %[[L:.*]] = memref.load %[[A0]][%[[C]]] : memref<3xindex>
// CHECK-STORAGE: return %[[L]] : index
func.func @sparse_dense_3d_dyn(%arg0: tensor<?x?x?xf64, #Dense3D>) -> index { func.func @sparse_dense_3d_dyn(%arg0: tensor<?x?x?xf64, #Dense3D>) -> index {
%c = arith.constant 1 : index %c = arith.constant 1 : index
%0 = tensor.dim %arg0, %c : tensor<?x?x?xf64, #Dense3D> %0 = tensor.dim %arg0, %c : tensor<?x?x?xf64, #Dense3D>
return %0 : index return %0 : index
} }
// CHECK-LABEL: func @sparse_pointers_dcsr( // CHECK-CODEGEN-LABEL: func @sparse_pointers_dcsr(
// CHECK-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>>)
// CHECK: %[[F:.*]] = sparse_tensor.storage_get %[[A]][3] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xi32> // CHECK-CODEGEN: %[[F:.*]] = sparse_tensor.storage_get %[[A]][3] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xi32>
// CHECK: return %[[F]] : memref<?xi32> // CHECK-CODEGEN: return %[[F]] : memref<?xi32>
//
// CHECK-STORAGE-LABEL: func @sparse_pointers_dcsr(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<2xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A4:.*4]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A5:.*5]]: memref<?xf64>)
// CHECK-STORAGE: return %[[A3]] : memref<?xi32>
func.func @sparse_pointers_dcsr(%arg0: tensor<?x?xf64, #DCSR>) -> memref<?xi32> { func.func @sparse_pointers_dcsr(%arg0: tensor<?x?xf64, #DCSR>) -> memref<?xi32> {
%c = arith.constant 1 : index %c = arith.constant 1 : index
%0 = sparse_tensor.pointers %arg0, %c : tensor<?x?xf64, #DCSR> to memref<?xi32> %0 = sparse_tensor.pointers %arg0, %c : tensor<?x?xf64, #DCSR> to memref<?xi32>
return %0 : memref<?xi32> return %0 : memref<?xi32>
} }
// CHECK-LABEL: func @sparse_indices_dcsr( // CHECK-CODEGEN-LABEL: func @sparse_indices_dcsr(
// CHECK-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>>)
// CHECK: %[[F:.*]] = sparse_tensor.storage_get %[[A]][4] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xi64> // CHECK-CODEGEN: %[[F:.*]] = sparse_tensor.storage_get %[[A]][4] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xi64>
// CHECK: return %[[F]] : memref<?xi64> // CHECK-CODEGEN: return %[[F]] : memref<?xi64>
//
// CHECK-STORAGE-LABEL: func @sparse_indices_dcsr(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<2xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A4:.*4]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A5:.*5]]: memref<?xf64>)
// CHECK-STORAGE: return %[[A4]] : memref<?xi64>
func.func @sparse_indices_dcsr(%arg0: tensor<?x?xf64, #DCSR>) -> memref<?xi64> { func.func @sparse_indices_dcsr(%arg0: tensor<?x?xf64, #DCSR>) -> memref<?xi64> {
%c = arith.constant 1 : index %c = arith.constant 1 : index
%0 = sparse_tensor.indices %arg0, %c : tensor<?x?xf64, #DCSR> to memref<?xi64> %0 = sparse_tensor.indices %arg0, %c : tensor<?x?xf64, #DCSR> to memref<?xi64>
return %0 : memref<?xi64> return %0 : memref<?xi64>
} }
// CHECK-LABEL: func @sparse_values_dcsr( // CHECK-CODEGEN-LABEL: func @sparse_values_dcsr(
// CHECK-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>>)
// CHECK: %[[F:.*]] = sparse_tensor.storage_get %[[A]][5] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xf64> // CHECK-CODEGEN: %[[F:.*]] = sparse_tensor.storage_get %[[A]][5] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xf64>
// CHECK: return %[[F]] : memref<?xf64> // CHECK-CODEGEN: return %[[F]] : memref<?xf64>
//
// CHECK-STORAGE-LABEL: func @sparse_values_dcsr(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<2xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A4:.*4]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A5:.*5]]: memref<?xf64>)
// CHECK-STORAGE: return %[[A5]] : memref<?xf64>
func.func @sparse_values_dcsr(%arg0: tensor<?x?xf64, #DCSR>) -> memref<?xf64> { func.func @sparse_values_dcsr(%arg0: tensor<?x?xf64, #DCSR>) -> memref<?xf64> {
%0 = sparse_tensor.values %arg0 : tensor<?x?xf64, #DCSR> to memref<?xf64> %0 = sparse_tensor.values %arg0 : tensor<?x?xf64, #DCSR> to memref<?xf64>
return %0 : memref<?xf64> return %0 : memref<?xf64>
} }
// CHECK-LABEL: func @sparse_dealloc_csr( // CHECK-CODEGEN-LABEL: func @sparse_dealloc_csr(
// CHECK-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>) // CHECK-CODEGEN-SAME: %[[A:.*]]: tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>>)
// CHECK: %[[F0:.*]] = sparse_tensor.storage_get %[[A]][0] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<2xindex> // CHECK-CODEGEN: %[[F0:.*]] = sparse_tensor.storage_get %[[A]][0] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<2xindex>
// CHECK: memref.dealloc %[[F0]] : memref<2xindex> // CHECK-CODEGEN: memref.dealloc %[[F0]] : memref<2xindex>
// CHECK: %[[F1:.*]] = sparse_tensor.storage_get %[[A]][1] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xi32> // CHECK-CODEGEN: %[[F1:.*]] = sparse_tensor.storage_get %[[A]][1] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xi32>
// CHECK: memref.dealloc %[[F1]] : memref<?xi32> // CHECK-CODEGEN: memref.dealloc %[[F1]] : memref<?xi32>
// CHECK: %[[F2:.*]] = sparse_tensor.storage_get %[[A]][2] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xi64> // CHECK-CODEGEN: %[[F2:.*]] = sparse_tensor.storage_get %[[A]][2] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xi64>
// CHECK: memref.dealloc %[[F2]] : memref<?xi64> // CHECK-CODEGEN: memref.dealloc %[[F2]] : memref<?xi64>
// CHECK: %[[F3:.*]] = sparse_tensor.storage_get %[[A]][3] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xf64> // CHECK-CODEGEN: %[[F3:.*]] = sparse_tensor.storage_get %[[A]][3] : tuple<memref<2xindex>, memref<?xi32>, memref<?xi64>, memref<?xf64>> to memref<?xf64>
// CHECK: memref.dealloc %[[F3]] : memref<?xf64> // CHECK-CODEGEN: memref.dealloc %[[F3]] : memref<?xf64>
// CHECK: return // CHECK-CODEGEN: return
//
// CHECK-STORAGE-LABEL: func @sparse_dealloc_csr(
// CHECK-STORAGE-SAME: %[[A0:.*0]]: memref<2xindex>,
// CHECK-STORAGE-SAME: %[[A1:.*1]]: memref<?xi32>,
// CHECK-STORAGE-SAME: %[[A2:.*2]]: memref<?xi64>,
// CHECK-STORAGE-SAME: %[[A3:.*3]]: memref<?xf64>) {
// CHECK-STORAGE: memref.dealloc %[[A0]] : memref<2xindex>
// CHECK-STORAGE: memref.dealloc %[[A1]] : memref<?xi32>
// CHECK-STORAGE: memref.dealloc %[[A2]] : memref<?xi64>
// CHECK-STORAGE: memref.dealloc %[[A3]] : memref<?xf64>
// CHECK-STORAGE: return
func.func @sparse_dealloc_csr(%arg0: tensor<?x?xf64, #CSR>) { func.func @sparse_dealloc_csr(%arg0: tensor<?x?xf64, #CSR>) {
bufferization.dealloc_tensor %arg0 : tensor<?x?xf64, #CSR> bufferization.dealloc_tensor %arg0 : tensor<?x?xf64, #CSR>
return return
} }