MLBEDSW-7648: Fix bug with filter padding in conv2d
* Fix bug that caused filter padding to not be added proportionally
compared to the hardware padding added to IFM.
* Update needed_total_padding function that calculates hardware padding
to also account for the cases in which IFM width is not divisible by
the stride width.
* Update supported ops constraint on strides for conv2d to mark ops with
stride width > 3 and IFM width that is not divisible by the
optimization resize factor as not supported.
* Update unit tests that verify correct functionality when checking
whether ops are supported or not.
Change-Id: I62f14cca890b779ca787a9603fa37c873ad522f8
Signed-off-by: Raul Farkas <raul.farkas@arm.com>
diff --git a/ethosu/vela/graph_optimiser_util.py b/ethosu/vela/graph_optimiser_util.py
index da3fe13..220ba1a 100644
--- a/ethosu/vela/graph_optimiser_util.py
+++ b/ethosu/vela/graph_optimiser_util.py
@@ -185,10 +185,11 @@
def needed_total_padding(input_size, stride, filter_size):
- out_size = (input_size + stride - 1) // stride
- needed_input = (out_size - 1) * stride + filter_size
- total_padding = max(0, needed_input - input_size)
- return total_padding
+ """Compute hardware padding."""
+ if input_size % stride == 0:
+ return max(filter_size - stride, 0)
+
+ return max(filter_size - (input_size % stride), 0)
# Set input/output tensor equivalence to the same id for memory operations