Skip to content

[mlir][sparse] Fix memory leaks (part 2) #81979

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Feb 17, 2024
Merged

[mlir][sparse] Fix memory leaks (part 2) #81979

merged 1 commit into from
Feb 17, 2024

Conversation

matthias-springer
Copy link
Member

This commit fixes memory leaks in sparse tensor integration tests by adding bufferization.dealloc_tensor ops.

Note: Buffer deallocation will be automated in the future with the ownership-based buffer deallocation pass, making dealloc_tensor obsolete (only codegen path, not when using the runtime library).

@matthias-springer
[mlir][sparse] Fix memory leaks (part 2)
1902516 
This commit fixes memory leaks in sparse tensor integration tests by adding `bufferization.dealloc_tensor` ops.

Note: Buffer deallocation will be automated in the future with the ownership-based buffer deallocation pass, making `dealloc_tensor` obsolete (only codegen path, not when using the runtime library).
@llvmbot llvmbot added mlir:sparse Sparse compiler in MLIR mlir labels Feb 16, 2024
@llvmbot
Copy link
Member

llvmbot commented Feb 16, 2024 •
edited
Loading

@llvm/pr-subscribers-mlir-sparse

@llvm/pr-subscribers-mlir

Author: Matthias Springer (matthias-springer)

Changes

This commit fixes memory leaks in sparse tensor integration tests by adding bufferization.dealloc_tensor ops.

Note: Buffer deallocation will be automated in the future with the ownership-based buffer deallocation pass, making dealloc_tensor obsolete (only codegen path, not when using the runtime library).

Full diff: https://github.com/llvm/llvm-project/pull/81979.diff

9 Files Affected:

  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_binary.mlir (+5)
  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nchw_fchw.mlir (+1)
  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nhwc_hwcf.mlir (+4)
  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d.mlir (+6)
  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d_ndhwc_dhwcf.mlir (+3)
  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conversion_element.mlir (+3)
  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_dilated_conv_2d_nhwc_hwcf.mlir (+5)
  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_matmul_slice.mlir (+7-1)
  • (modified) mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_strided_conv_2d_nhwc_hwcf.mlir (+5)
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_binary.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_binary.mlir
index da944363cf75c4..59ecbfdef85043 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_binary.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_binary.mlir
@@ -380,6 +380,7 @@ module {
     %dv = sparse_tensor.convert %arg0 : tensor<?xf64, #SparseVector> to tensor<?xf64>
     %3 = vector.transfer_read %dv[%c0], %d0: tensor<?xf64>, vector<32xf64>
     vector.print %3 : vector<32xf64>
+    bufferization.dealloc_tensor %dv : tensor<?xf64>
     return
   }
 
@@ -394,6 +395,7 @@ module {
     %dv = sparse_tensor.convert %arg0 : tensor<?xi32, #SparseVector> to tensor<?xi32>
     %3 = vector.transfer_read %dv[%c0], %d0: tensor<?xi32>, vector<32xi32>
     vector.print %3 : vector<32xi32>
+    bufferization.dealloc_tensor %dv : tensor<?xi32>
     return
   }
 
@@ -403,6 +405,7 @@ module {
     %dm = sparse_tensor.convert %arg0 : tensor<?x?xf64, #DCSR> to tensor<?x?xf64>
     %1 = vector.transfer_read %dm[%c0, %c0], %d0: tensor<?x?xf64>, vector<4x8xf64>
     vector.print %1 : vector<4x8xf64>
+    bufferization.dealloc_tensor %dm : tensor<?x?xf64>
     return
   }
 
@@ -418,6 +421,7 @@ module {
     %2 = vector.transfer_read %1[%c0], %du: memref<?xf64>, vector<16xf64>
     vector.print %2 : vector<16xf64>
 
+    bufferization.dealloc_tensor %c : tensor<4x4xf64>
     return
   }
 
@@ -433,6 +437,7 @@ module {
     %2 = vector.transfer_read %1[%c0], %du: memref<?xi8>, vector<16xi8>
     vector.print %2 : vector<16xi8>
 
+    bufferization.dealloc_tensor %c : tensor<4x4xi8>
     return
   }
 
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nchw_fchw.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nchw_fchw.mlir
index c7e4ffab7ab3b1..95ce4f1bf48d5d 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nchw_fchw.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nchw_fchw.mlir
@@ -207,6 +207,7 @@ func.func @entry() {
   bufferization.dealloc_tensor %out2D_nhwc : tensor<?x?x?x?xf32>
   bufferization.dealloc_tensor %out2D_nhwc_CCCD : tensor<?x?x?x?xf32>
   bufferization.dealloc_tensor %out2D_nhwc_CCCC : tensor<?x?x?x?xf32>
+  bufferization.dealloc_tensor %dense_ret :tensor<?x?x?x?xf32>
 
   bufferization.dealloc_tensor %in2D_nhwc_CCCC : tensor<?x?x?x?xf32, #CCCC>
   bufferization.dealloc_tensor %in2D_nhwc_CCCD : tensor<?x?x?x?xf32, #CDCD>
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nhwc_hwcf.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nhwc_hwcf.mlir
index afa6a0c0ebbc7e..d0fbce7146fe57 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nhwc_hwcf.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_2d_nhwc_hwcf.mlir
@@ -227,5 +227,9 @@ func.func @entry() {
   bufferization.dealloc_tensor %CDCD_ret : tensor<?x?x?x?xf32, #CDCD>
   bufferization.dealloc_tensor %DCCD_ret : tensor<?x?x?x?xf32, #DCCD>
 
+  bufferization.dealloc_tensor %1 : tensor<?x?x?x?xf32>
+  bufferization.dealloc_tensor %2 : tensor<?x?x?x?xf32>
+  bufferization.dealloc_tensor %3 : tensor<?x?x?x?xf32>
+
   return
 }
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d.mlir
index 451d2b87694614..f0a26dc46b056e 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d.mlir
@@ -348,5 +348,11 @@ func.func @entry() {
   bufferization.dealloc_tensor %CDC_ret : tensor<?x?x?xf32, #CDC>
   bufferization.dealloc_tensor %DDC_ret : tensor<?x?x?xf32, #DDC>
   bufferization.dealloc_tensor %DCC_ret : tensor<?x?x?xf32, #DCC>
+
+  bufferization.dealloc_tensor %1 : tensor<?x?x?xf32>
+  bufferization.dealloc_tensor %2 : tensor<?x?x?xf32>
+  bufferization.dealloc_tensor %3 : tensor<?x?x?xf32>
+  bufferization.dealloc_tensor %4 : tensor<?x?x?xf32>
+
   return
 }
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d_ndhwc_dhwcf.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d_ndhwc_dhwcf.mlir
index 2c4b96804ccc70..346a1436928971 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d_ndhwc_dhwcf.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conv_3d_ndhwc_dhwcf.mlir
@@ -249,5 +249,8 @@ func.func @entry() {
   bufferization.dealloc_tensor %CCCCC_ret : tensor<?x?x?x?x?xf32, #CCCCC>
   bufferization.dealloc_tensor %CDCDC_ret : tensor<?x?x?x?x?xf32, #CDCDC>
 
+  bufferization.dealloc_tensor %1 : tensor<?x?x?x?x?xf32>
+  bufferization.dealloc_tensor %2 : tensor<?x?x?x?x?xf32>
+
   return
 }
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conversion_element.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conversion_element.mlir
index 65a37bc8e731e5..81f68366be28c6 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conversion_element.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conversion_element.mlir
@@ -106,6 +106,9 @@ module {
     bufferization.dealloc_tensor %s1 : tensor<2x3x4xf64, #Tensor1>
     bufferization.dealloc_tensor %s2 : tensor<2x3x4xf64, #Tensor2>
     bufferization.dealloc_tensor %s3 : tensor<2x3x4xf64, #Tensor3>
+    bufferization.dealloc_tensor %d1 : tensor<2x3x4xf32>
+    bufferization.dealloc_tensor %d2 : tensor<2x3x4xf32>
+    bufferization.dealloc_tensor %d3 : tensor<2x3x4xf32>
 
     return
   }
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_dilated_conv_2d_nhwc_hwcf.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_dilated_conv_2d_nhwc_hwcf.mlir
index af7ee43647581e..b4d40ae084015a 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_dilated_conv_2d_nhwc_hwcf.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_dilated_conv_2d_nhwc_hwcf.mlir
@@ -144,5 +144,10 @@ func.func @entry() {
   bufferization.dealloc_tensor %filter2D_nhwc_CDCC : tensor<?x?x?x?xf32, #CDCC>
   bufferization.dealloc_tensor %in2D_nhwc_CCCC : tensor<?x?x?x?xf32, #CCCC>
   bufferization.dealloc_tensor %in2D_nhwc_CDCC : tensor<?x?x?x?xf32, #CDCC>
+
+  bufferization.dealloc_tensor %dense_ret : tensor<?x?x?x?xf32>
+  bufferization.dealloc_tensor %CCCC_ret : tensor<?x?x?x?xf32>
+  bufferization.dealloc_tensor %CDCC_ret : tensor<?x?x?x?xf32>
+
   return
 }
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_matmul_slice.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_matmul_slice.mlir
index 85d51931db6c21..96c8a30ade8e42 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_matmul_slice.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_matmul_slice.mlir
@@ -264,7 +264,13 @@ module {
     %du = tensor.cast %r : tensor<4x4xf64> to tensor<*xf64>
     call @printMemrefF64(%du) : (tensor<*xf64>) -> ()
 
-    // Releases resources (we do not need to deallocate slices).
+    // Releases resources.
+    bufferization.dealloc_tensor %c2 : tensor<4x4xf64>
+    bufferization.dealloc_tensor %c3 : tensor<4x4xf64>
+    bufferization.dealloc_tensor %c4 : tensor<4x4xf64>
+    bufferization.dealloc_tensor %c4_coo : tensor<4x4xf64>
+    bufferization.dealloc_tensor %c4_dyn : tensor<4x4xf64>
+    bufferization.dealloc_tensor %d : tensor<4x4xf64>
     bufferization.dealloc_tensor %b1 : tensor<8x4xf64, #CSR>
     bufferization.dealloc_tensor %t1 : tensor<8x8xf64, #CSR>
     bufferization.dealloc_tensor %b1_coo : tensor<8x4xf64, #COO>
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_strided_conv_2d_nhwc_hwcf.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_strided_conv_2d_nhwc_hwcf.mlir
index 8ee80045afc760..5184083f665d56 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_strided_conv_2d_nhwc_hwcf.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_strided_conv_2d_nhwc_hwcf.mlir
@@ -143,5 +143,10 @@ func.func @entry() {
   bufferization.dealloc_tensor %filter2D_nhwc_CDCC : tensor<?x?x?x?xf32, #CDCC>
   bufferization.dealloc_tensor %in2D_nhwc_CCCC : tensor<?x?x?x?xf32, #CCCC>
   bufferization.dealloc_tensor %in2D_nhwc_CDCC : tensor<?x?x?x?xf32, #CDCC>
+
+  bufferization.dealloc_tensor %dense_ret : tensor<?x?x?x?xf32>
+  bufferization.dealloc_tensor %CCCC_ret : tensor<?x?x?x?xf32>
+  bufferization.dealloc_tensor %CDCC_ret : tensor<?x?x?x?xf32>
+
   return
 }

@matthias-springer matthias-springer merged commit ccc20b4 into llvm:main Feb 17, 2024
@joker-eph
Copy link
Collaborator

Is the test suite leak-clean by now @matthias-springer ?

@matthias-springer
Copy link
Member Author

There are still a few left. I'm cleaning them up on the side.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@aartbik aartbik aartbik approved these changes

@yinying-lisa-li yinying-lisa-li yinying-lisa-li approved these changes

@PeimingLiu PeimingLiu Awaiting requested review from PeimingLiu PeimingLiu is a code owner

Assignees
No one assigned
Labels
mlir:sparse Sparse compiler in MLIR mlir
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@matthias-springer @llvmbot @joker-eph @aartbik @yinying-lisa-li