Skip to content

Fix HiFi conv2d_nhwc operator for depthwise case (again) #18559

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
meta-codesync merged 1 commit into from
Mar 30, 2026
Merged

Fix HiFi conv2d_nhwc operator for depthwise case (again) #18559

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
22 changes: 21 additions & 1 deletion backends/cadence/hifi/operators/op_quantized_conv2d_nhwc_out.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -300,7 +300,7 @@ void xa_opt_quantized_conv2d_nhwc(
return;
}

if (groups == input_channels) {
if (is_depthwise) {
WORD32 channels_multiplier = out_channels / input_channels;

scratch_size = xa_nn_conv2d_depthwise_getsize(
Expand Down Expand Up @@ -359,6 +359,26 @@ void xa_opt_quantized_conv2d_nhwc(
return;
}
}

// Fallback to generic grouped conv for cases not handled by nnlib
// (e.g. grouped conv with 4D weight where is_depthwise is false)
::impl::generic::native::quantized_conv2d_nhwc_per_tensor_out(
ctx,
input,
weight,
bias,
stride,
padding,
dilation,
groups,
in_zero_point,
weight_zero_point,
bias_scale,
output_scale,
output_zero_point,
0, // out_multiplier (unused)
0, // out_shift (unused)
out);
}

void quantized_conv2d_nhwc(
Expand Down
Loading