[Mlir-commits] [mlir] 36c0187 - [mlir][sparse] fix asan issue

[Aart Bik]

Author: Aart Bik
Date: 2022-06-16T14:49:02-07:00
New Revision: 36c01876d79dd962d3aa1f5691dbd3faa437b9a8

URL: https://github.com/llvm/llvm-project/commit/36c01876d79dd962d3aa1f5691dbd3faa437b9a8
DIFF: https://github.com/llvm/llvm-project/commit/36c01876d79dd962d3aa1f5691dbd3faa437b9a8.diff

LOG: [mlir][sparse] fix asan issue

The LinalgElementwiseOpFusion pass has become smarter, and converts
the simple conversion linalg operation into a sparse dialect convert
operation. However, since our current bufferization does not take the
new semantics into consideration, we leak memory of the allocation.
For now, this has been fixed by making the operation less trivial.

Reviewed By: bixia

Differential Revision: https://reviews.llvm.org/D128002

Added: 
    

Modified: 
    mlir/test/Integration/Dialect/SparseTensor/CPU/dense_output.mlir

Removed: 
    


################################################################################
diff  --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/dense_output.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/dense_output.mlir
index 0018c07b342c5..ca02b4073aba9 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/dense_output.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/dense_output.mlir
@@ -23,7 +23,7 @@
     affine_map<(i,j) -> (i,j)>  // X (out)
   ],
   iterator_types = ["parallel", "parallel"],
-  doc = "X(i,j) = A(i,j)"
+  doc = "X(i,j) = A(i,j) * 2"
 }
 
 //
@@ -39,11 +39,12 @@
 // library.
 module {
   //
-  // A kernel that assigns elements from A to X.
+  // A kernel that assigns multiplied elements from A to X.
   //
   func.func @dense_output(%arga: tensor<?x?xf64, #SparseMatrix>) -> tensor<?x?xf64, #DenseMatrix> {
     %c0 = arith.constant 0 : index
     %c1 = arith.constant 1 : index
+    %c2 = arith.constant 2.0 : f64
     %d0 = tensor.dim %arga, %c0 : tensor<?x?xf64, #SparseMatrix>
     %d1 = tensor.dim %arga, %c1 : tensor<?x?xf64, #SparseMatrix>
     %init = bufferization.alloc_tensor(%d0, %d1) : tensor<?x?xf64, #DenseMatrix>
@@ -51,7 +52,8 @@ module {
        ins(%arga: tensor<?x?xf64, #SparseMatrix>)
       outs(%init: tensor<?x?xf64, #DenseMatrix>) {
       ^bb(%a: f64, %x: f64):
-        linalg.yield %a : f64
+        %0 = arith.mulf %a, %c2 : f64
+        linalg.yield %0 : f64
     } -> tensor<?x?xf64, #DenseMatrix>
     return %0 : tensor<?x?xf64, #DenseMatrix>
   }
@@ -78,7 +80,7 @@ module {
     //
     // Print the linearized 5x5 result for verification.
     //
-    // CHECK: ( 1, 0, 0, 1.4, 0, 0, 2, 0, 0, 2.5, 0, 0, 3, 0, 0, 4.1, 0, 0, 4, 0, 0, 5.2, 0, 0, 5 )
+    // CHECK: ( 2, 0, 0, 2.8, 0, 0, 4, 0, 0, 5, 0, 0, 6, 0, 0, 8.2, 0, 0, 8, 0, 0, 10.4, 0, 0, 10 )
     //
     %m = sparse_tensor.values %0
       : tensor<?x?xf64, #DenseMatrix> to memref<?xf64>