Source code is from test_conformance/test_tensor_nn.c line:614.
sum = ownApplyWrapRoundingToAccum(fmt, i_val * w_val, wrap, to_ne) + sum;
Looks like ownApplyWrapRoundingToAccum should not be used here. For w_val and i_val of each pixel, you should multiply two numbers together, not apply Wrap.
Source code is from test_conformance/test_tensor_nn.c line:614.
sum = ownApplyWrapRoundingToAccum(fmt, i_val * w_val, wrap, to_ne) + sum;
Looks like ownApplyWrapRoundingToAccum should not be used here. For w_val and i_val of each pixel, you should multiply two numbers together, not apply Wrap.
So we think above code should be changed to
sum = i_val * w_val + sum;
Please take a look and fix it.