About the calculation in "tx/rx_checksum.v"

[linjw16]

Hello !
I have read the module tx_checksum & rx_checksum and checked the simulated wave. I am wondering what checksum algorithm the author is using?

I presume it is adding 4B (32bits) data up to 2B (first 17 bits with overflow and then add up to 16 bits). And do this iteratively on the whole frame, i.e. accumulating all this 2B together. In that way, a register should be used to store the accumulated sum, it might be "sum_acc_reg" in line 395 in "tx_checksum.v". But the assignment is not the way I thought, is there some misunderstandings from me?

line 395

else begin
            // sum_acc_reg <= sum_acc_reg+sum_acc_temp;     // linjw: might be wrong? 
            sum_acc_reg <= sum_acc_temp;
        end

[alexforencich]

See line 378

[linjw16]

Got it !
sum_acc_temp is used to record the accumulated sum, while sum_acc_reg is used to record the last sum_acc_temp.
I had been confused between the two for a while. Thanks.

[alexforencich]

Not exactly. sum_acc_temp is not a register, it is a temporary variable. So the sum is being accumulated in sum_acc_reg, but the previous value of sum_acc_reg is represented in sum_acc_temp a few lines before sum_acc_reg is updated.