Skip to content

[mlir][linalg] Fix FoldReshapeWithGenericOpByCollapsing insertion point #133476

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

Merged
IanWood1 merged 1 commit into from
Mar 28, 2025
Merged

[mlir][linalg] Fix FoldReshapeWithGenericOpByCollapsing insertion point #133476

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1897,6 +1897,9 @@ struct FoldReshapeWithGenericOpByCollapsing
"fusion blocked by control function");
}

// Set the insertion point after `producer` because there could be uses
// of `producer` between it and the `tensor.collapse_shape` op.
rewriter.setInsertionPointAfter(producer);
std::optional<CollapseResult> collapseResult =
collapseOpIterationDims(producer, collapsableIterationDims, rewriter);
if (!collapseResult) {
Expand Down
32 changes: 32 additions & 0 deletions mlir/test/Dialect/Linalg/fuse-with-reshape-by-collapsing.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -798,3 +798,35 @@ func.func @fuse_by_collapsing_change_reshape_order_bubblecollapse(%arg0 : tensor
// CONTROL-SAME: ins(%[[ARG0]],
// CONTROL: %[[COLLAPSE:.+]] = tensor.collapse_shape %[[GENERIC]]
// CONTROL: return %[[COLLAPSE]]

// -----

// Check that new ops are inserted at `%0` because `%0` is also used by `tensor.dim`.
#map0 = affine_map<(d0, d1) -> (d0, d1)>
func.func @fuse_by_collapsing_correct_insertion(%arg0 : tensor<?x?xf32>,
%sz0: index, %sz1: index) -> (tensor<?xf32>, index) {
%c0 = arith.constant 0 : index
%init = tensor.empty(%sz1, %sz0) : tensor<?x?xf32>
%0 = linalg.generic {
indexing_maps = [#map0, #map0],
iterator_types = ["parallel", "parallel"]}
ins(%arg0 : tensor<?x?xf32>)
outs(%init : tensor<?x?xf32>) {
^bb0(%b0 : f32, %b1 : f32):
%out = arith.negf %b0 : f32
linalg.yield %out : f32
} -> tensor<?x?xf32>
%dim = tensor.dim %0, %c0 : tensor<?x?xf32>
%1 = tensor.collapse_shape %0 [[0, 1]] : tensor<?x?xf32> into tensor<?xf32>
return %1, %dim : tensor<?xf32>, index
}

// CHECK-LABEL: func @fuse_by_collapsing_correct_insertion
// CHECK-SAME: %[[ARG0:.+]]: tensor<?x?xf32>
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK: %[[COLLAPSE:.+]] = tensor.collapse_shape %[[ARG0]]
// CHECK: %[[OUT:.+]] = linalg.generic
// CHECK-SAME: ins(%[[COLLAPSE]] : tensor<?xf32>)
// CHECK: %[[EXPANDED:.+]] = tensor.expand_shape %[[OUT]]
// CHECK: %[[DIM:.+]] = tensor.dim %[[EXPANDED]], %[[C0]]
// CHECK: return %[[OUT]], %[[DIM]]