Skip to content
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

[GPU] Add check of multiple axis broadcasting in is_valid_fusion() when dynamic shape #28252

Merged
merged 2 commits into from
Jan 6, 2025
Merged

[GPU] Add check of multiple axis broadcasting in is_valid_fusion() when dynamic shape #28252

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
67d8cc8
add check of multiple axis broadcasting in is_valid_fusion()
wilson-seok Jan 2, 2025
bbf6c49
update condition and add func test
wilson-seok Jan 3, 2025
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
13 changes: 13 additions & 0 deletions src/plugins/intel_gpu/src/graph/primitive_inst.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2658,6 +2658,19 @@ bool primitive_inst::is_valid_fusion() const {
if (fd.is_type<eltwise>())
can_broadcast = ov::PartialShape::broadcast_merge_into(merged_shape, outer_dep_pshape, fd.typed_desc<eltwise>()->broadcast_spec);

// Check if broadcast happens more than single axis.
// Current FUSED_OP_LOAD macro cannot support broadcast on dynamic dimension.
if (can_broadcast == true && (merged_shape.is_static() && outer_dep_pshape.is_static()) &&
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are these two conditions needed? At runtime, shapes should be static always.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You are right, I removed it. Thanks!

outer_dep.first->_is_dynamic == true && merged_shape.rank().get_length() == outer_dep_pshape.rank().get_length()) {
Copy link
Contributor

@yeonbok yeonbok Jan 2, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

outer_dep.first->_is_dynamic

Do we need this condition? even though it is not dynamic, the fusion is not allowed isn't it? And also this function is not called for static primitive.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You are right. the is_valid_fusion() is called in dynamic shape case. I removed it. Thanks!

uint8_t broadcast_more_than_single_axis = 0;
for (int64_t i = 0; i < merged_shape.rank().get_length(); i++) {
if (merged_shape.get_shape().at(i) != outer_dep_pshape.get_shape().at(i))
broadcast_more_than_single_axis++;
}
if (broadcast_more_than_single_axis > 1)
can_broadcast = false;
}

#ifdef ENABLE_ONEDNN_FOR_GPU
// WA for OneDNN binary add fusions: we need to broadcast batch dimension to avoid situation with
// batch dimension mismatch in OneDNN tensor descriptors as follow:
Expand Down
Loading