[Mlir-commits] [mlir] f4e8f44 - [mlir][bufferize] Fix enclosing repetitive region computation

[Matthias Springer]

Author: Matthias Springer
Date: 2022-10-07T10:37:04+09:00
New Revision: f4e8f44811e25fd5547d879aafafd621386eb927

URL: https://github.com/llvm/llvm-project/commit/f4e8f44811e25fd5547d879aafafd621386eb927
DIFF: https://github.com/llvm/llvm-project/commit/f4e8f44811e25fd5547d879aafafd621386eb927.diff

LOG: [mlir][bufferize] Fix enclosing repetitive region computation

The wrong function overload was called.

Differential Revision: https://reviews.llvm.org/D135342

Added: 
    

Modified: 
    mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp
    mlir/test/Dialect/SCF/one-shot-bufferize-analysis.mlir

Removed: 
    


################################################################################
diff  --git a/mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp b/mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp
index 100fc7767994a..e56d986cd4968 100644
--- a/mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp
+++ b/mlir/lib/Dialect/Bufferization/Transforms/OneShotAnalysis.cpp
@@ -386,7 +386,7 @@ getCommonEnclosingRepetitiveRegion(ArrayRef<Value> values,
     return None;
   Region *r = getEnclosingRepetitiveRegion(values.front(), options);
   for (Value value : values.drop_front())
-    if (getEnclosingRepetitiveRegion(value) != r)
+    if (getEnclosingRepetitiveRegion(value, options) != r)
       return None;
   return r;
 }

diff  --git a/mlir/test/Dialect/SCF/one-shot-bufferize-analysis.mlir b/mlir/test/Dialect/SCF/one-shot-bufferize-analysis.mlir
index bde8ac7bb5aa5..783adcee60b1b 100644
--- a/mlir/test/Dialect/SCF/one-shot-bufferize-analysis.mlir
+++ b/mlir/test/Dialect/SCF/one-shot-bufferize-analysis.mlir
@@ -599,3 +599,34 @@ func.func @write_to_same_tensor_in_loop_in_place(
 
   return %r0 : tensor<?xf32>
 }
+
+// -----
+
+// This is a regression test. Everything can bufferize in-place because %7 and
+// %arg1 are in the same repetitive region.
+
+// CHECK-LABEL: func @same_enclosing_repetitive_region
+func.func @same_enclosing_repetitive_region(%2: tensor<320xf32>,
+                                            %3: tensor<320x10240xf32>)
+  -> tensor<320xf32>
+{
+  %c0 = arith.constant 0 : index
+  %cst = arith.constant -0.000000e+00 : f32
+  %c320 = arith.constant 320 : index
+  %4 = scf.foreach_thread (%arg0) in (%c320) shared_outs(%arg1 = %2) -> (tensor<320xf32>) {
+    // CHECK: tensor.extract_slice {{.*}} {__inplace_operands_attr__ = ["true", "none"]}
+    %5 = tensor.extract_slice %3[%arg0, 0] [1, 10240] [1, 1]  : tensor<320x10240xf32> to tensor<1x10240xf32>
+    // CHECK: tensor.extract_slice {{.*}} {__inplace_operands_attr__ = ["true", "none"]}
+    %6 = tensor.extract_slice %arg1[%arg0] [1] [1] : tensor<320xf32> to tensor<1xf32>
+    // CHECK: linalg.fill {__inplace_operands_attr__ = ["none", "true"]}
+    %7 = linalg.fill ins(%cst : f32) outs(%6 : tensor<1xf32>) -> tensor<1xf32>
+    // CHECK: linalg.fill {__inplace_operands_attr__ = ["none", "true"]}
+    %8 = linalg.fill ins(%cst : f32) outs(%7 : tensor<1xf32>) -> tensor<1xf32>
+
+    scf.foreach_thread.perform_concurrently {
+      // CHECK: tensor.parallel_insert_slice {{.*}} {__inplace_operands_attr__ = ["true", "true", "none"]}
+      tensor.parallel_insert_slice %8 into %arg1[%arg0] [1] [1] : tensor<1xf32> into tensor<320xf32>
+    }
+  } {thread_dim_mapping = []}
+  return %4 : tensor<320xf32>
+}