Error in multiplexed ip core

Support
1

I wrote a 3*3 convolutional ip core, the ip is correct. I ues it to perform calculations. When the input is 4 channels and the output is 16 channels, only the first channel results are correct, the rest of the 15 channels results are messed up. What is the reason for this?

The code is as follows:
def RunConv3(Conv3, feature_in, weight, bias, feature_out, ch_in, ch_out, size, act):
Conv3.write(0x10, feature_in.physical_address);
Conv3.write(0x18, feature_in.physical_address);
Conv3.write(0x20, feature_in.physical_address);
Conv3.write(0x28, feature_in.physical_address);
Conv3.write(0x30, weight.physical_address);
Conv3.write(0x38, weight.physical_address);
Conv3.write(0x40, weight.physical_address);
Conv3.write(0x48, weight.physical_address);
Conv3.write(0x50, bias.physical_address);
Conv3.write(0x58, feature_out.physical_address);
Conv3.write(0x60, feature_out.physical_address);
Conv3.write(0x68, feature_out.physical_address);
Conv3.write(0x70, feature_out.physical_address);
Conv3.write(0x78, ch_in);
Conv3.write(0x80, ch_out);
Conv3.write(0x88, size);
Conv3.write(0x90, act);
Conv3.write(0, (Conv3.read(0) & 0x80) | 0x01);
tp= Conv3.read(0)
while not ((tp>>1)&0x1):
tp= Conv3.read(0)

Convblock_w_1 = allocate(shape=(16433), cacheable=0, dtype=np.int16)
Convblock_bais_1 = allocate(shape=(16), cacheable=0, dtype=np.int16)
Convblock_out_1 = allocate(shape=(16432*432), cacheable=0, dtype=np.int16)

RunConv3(Conv3, image, Convblock_w_1, Convblock_bais_1, Convblock_out_1, 4, 16, 432, 1)

111
1111741×352 180 KB

2

The ip is correct, as I’ve had no errors in subsequent reuse of it.
I’ve also disabled the cache.

283802ddf09865c00e6d4eecb623ecd
283802ddf09865c00e6d4eecb623ecd823×265 29.2 KB

3

Hi,
For the 4 first, it is using image as feature_in. Is Convblock_w_1 similar than image, of the same type or dimension?

4

Yes, I’ve carefully checked that Convblock_w_1 is no problem.

5

If I understand correctly, you created an IP block in Vivado to do a convolutional operation. Then you run another python code via a terminal.
Why are there semicolons at each Conv3.write line?