xazax.hun added a comment. > Make CTU a two phase analysis
At this point maybe it will be a three phase, as we might dump the ASTs/create index in phase 1 :) > During this phase, if we find a foreign function (that could be inlined from > another TU) then we don’t inline that immediately, we rather mark that to be > analysed later. I wonder whether this is the right approach. Not inlining a function can also introduce false positives. While I do see why we do not want to inline ALL functions upfront, I wonder if we actually want to inline small functions. The CSA already has heuristics like this, to always inline really small functions (defined by the number of nodes in the Cfg) despite other cut heuristics and it was found to improve the quality of the results overall. I think doing something similar for CTU might ultimately improve the precision of the analysis. What do you think? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D123773/new/ https://reviews.llvm.org/D123773 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
