Skip to content

Arm backend: Cast int64 buffers to int32 #18234

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
AdrianLundell merged 2 commits into from
Mar 18, 2026
+19 −3
Merged

Arm backend: Cast int64 buffers to int32 #18234

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
adf1213
Arm backend: Cast int64 buffers to int32
AdrianLundell Mar 17, 2026
d754bcc
Cast int64 buffers to int32 PR fixes
AdrianLundell Mar 17, 2026
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
4 changes: 3 additions & 1 deletion backends/arm/_passes/insert_int32_casts_after_int64_placeholders.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,8 +89,10 @@ def _insert_placeholder_i64_to_i32_casts(self, graph_module: torch.fx.GraphModul
modified = False
graph = graph_module.graph
for node in graph.nodes:
if node.op != "placeholder":
if node.op not in ("placeholder", "get_attr"):
continue
if "val" not in node.meta:
continue # Ignore submodule get_attrs
node_val = node.meta["val"]
if not self._is_tensor_of_dtype(node_val, torch.int64):
continue
Expand Down
18 changes: 16 additions & 2 deletions backends/arm/test/ops/test_clamp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -257,7 +257,9 @@ def test_clamp_vgf_quant(test_data):
pipeline.run()


aten_op_tensor = "torch.ops.aten.clamp.Tensor"
aten_op_tensor = [
"torch.ops.aten.clamp.Tensor",
]
exir_op_tensor = "executorch_exir_dialects_edge__ops_aten_clamp_Tensor"

test_data_suite_tensor_FP = {
Expand Down Expand Up @@ -413,10 +415,22 @@ def test_clamp_tosa_INT_tensor(test_data):
input_tensor, min_val, max_val = test_data()
model = Clamp(min_val, max_val)

# Check that int64 inputs are cast to int32 in the tfa pipeline
if any(
t.dtype == torch.int64
for t in (input_tensor, min_val, max_val)
if isinstance(t, torch.Tensor)
):
aten_op = aten_op_tensor + [
"torch.ops.dim_order_ops._to_dim_order_copy.default"
]
else:
aten_op = aten_op_tensor

pipeline = TosaPipelineINT[input_t](
model,
(input_tensor,),
aten_op_tensor,
aten_op,
exir_op_tensor,
)
pipeline.run()
Expand Down
Loading