treewide: Enable CLIC virtualization

[ezelioli]

No description provided.

[bluewww]

Do you remember why you had to introduce addr_tmp? Sim bug?

[ezelioli]

Not sure anymore unfortunately. Looking at the code it could be also simply for readability, but I am open to comments on that.

[ezelioli]

@bluewww @alex96295 I found a new problem when synthesizing this version of the CLIC with a different synthesis compiler. It seems that the problem is related to the parametric defines, which are not seen as constants in the evaluation of the if else within the for loop in line 669. The synthesis tools requires the LHS of the comparison within the if to be a constant, breaking the elaboration.

I pushed a fix which is not really elegant but seems to work for all synthesis tools tried and does not require unrolling the entire for loop, which would be way worse. Ideally I think we could move some of the localparams in clic.sv to clic_pkg.sv, however due to the fact that most of them rely on parameters passed to the clic top module this is not possible at the moment. Suggestions on style changes are welcome :)

[alex96295]

ideally better to add this to common cells if it is of general use, can be done later

[ezelioli]

will do

[alex96295]

LGTM

[alex96295]

@bluewww @alex96295 I found a new problem when synthesizing this version of the CLIC with a different synthesis compiler. It seems that the problem is related to the parametric defines, which are not seen as constants in the evaluation of the if else within the for loop in line 669. The synthesis tools requires the LHS of the comparison within the if to be a constant, breaking the elaboration.

I pushed a fix which is not really elegant but seems to work for all synthesis tools tried and does not require unrolling the entire for loop, which would be way worse. Ideally I think we could move some of the localparams in clic.sv to clic_pkg.sv, however due to the fact that most of them rely on parameters passed to the clic top module this is not possible at the moment. Suggestions on style changes are welcome :)

I think for the time being we can leave it like this. I remember that Synopsys DC was able to digest the version you just modified (what was not able to digest was the original version with a case switch). For the sake of the projects depending on this IPs, I propose we merge this and we open an issue right now as a reminder to improve this logic. It can also be that meanwhile the tools that do not support this (Yosys?) will be improved as well

[ezelioli]

Fixed history. Ready to merge.