Bug Description
Hi, when we performing correctness checks on the division ISA using a formal tool, we discovered that in the serdiv module, when encountering the case of -1/1, the answer is incorrectly computed as 0 instead of the expected correct value of -1.
We have also investigated this issue on the original master of CVA6 and found that someone has already proposed a solution and made improvements to address this problem. You can refer to:
https://github.com/openhwgroup/cva6/issues/421
The following is a comparison between the error results obtained from checking the DUV using JasperGold FPV and the golden answer we wrote for SVA.
Is there an existing CVA6 bug for div?
Bug Description Hi, when we performing correctness checks on the division ISA using a formal tool, we discovered that in the serdiv module, when encountering the case of -1/1, the answer is incorrectly computed as 0 instead of the expected correct value of -1.
We have also investigated this issue on the original master of CVA6 and found that someone has already proposed a solution and made improvements to address this problem. You can refer to: https://github.com/openhwgroup/cva6/issues/421
The following is a comparison between the error results obtained from checking the DUV using JasperGold FPV and the golden answer we wrote for SVA.