qleenju
commented
2 years ago Regarding the LZA module, I have found that as long as the results do not overflow, the leading zero anticipation of the sum of two inputs is generally correct, with occasional deviations of 1, which can be corrected in the process of normalization.
MaoPann
hello: I have some questions about how LZA module works? As leading zero anticipator, it serves to predict the leading zero counts of result while adding, so that normalization can be directly implemented after addition is completed. However, in the module "fmac.sv", the input of LZA is A_LZA_D and B_LZA_D, instead of Csa_sum_D and Csa_carry_D. What's more, the value of A_LZA_D is always consistent with Sum_pos_D, which is the sum of Csa_sum_D and Csa_carry_D, and the value of B_LZA_D is always 0. In this case, the leading zero anticipation can not play a role at all, even though the out of fmac is correct. I would also like to ask about the exact role of "LZA" module, since I found that it is not always accurate in predicting the leading zero of the sum of two inputs.
Looking forward to your response. Best wishes! Qiong Li, a student from Nanjing University.