NagyDonat wrote: > An interesting thing could be to track in the analyzer the number of > suppressed underflows (i.e.: the number of times we resorted to hacking with > this solution). That way, we could have more information about how far we > deviate from the precise modelling. This could provide a useful metric in the > future and may not belong strictly to this patch.
Hm, that could be a good idea, but I'm lazy and I feel that when I need to measure the effects of this suppression, I can just temporarily disable it on a private measurement branch. (This is not a generic measurement that needs to be monitored constantly, it will become relevant just once or twice when we try to make the modelling more precise.) https://github.com/llvm/llvm-project/pull/127117 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
