MLBEDSW-4839: Fix issues with Elementwise IFM/OFM overlap
- Fixed typo with not using ifm.mem_type
- Fixed bug with using ifm1 properties when only ifm2 is a potential match
- Removed restriction on not considering SHL and SHR for overlap
- Removed some dead reshape code
Signed-off-by: Tim Hall <tim.hall@arm.com>
Change-Id: Id9bcc3c2b3ee9ac7b6276187d3e2f513b4acd4b5
diff --git a/ethosu/vela/pass_packing.py b/ethosu/vela/pass_packing.py
index 518b243..b28f4eb 100644
--- a/ethosu/vela/pass_packing.py
+++ b/ethosu/vela/pass_packing.py
@@ -348,8 +348,7 @@
ps.ifm_shapes.append(op.ifm_shapes[0])
elif ps.ifm_tensor == op.ifm2:
ps.ifm_shapes.append(op.ifm_shapes[1])
- for op in input_ops_list + [primary_op]:
- if op.run_on_npu:
+
if ps.ifm2_tensor == op.ifm:
ps.ifm_shapes.append(op.ifm_shapes[0])
elif ps.ifm2_tensor == op.ifm2: