rnk added a comment. I think I was confused in that conversation because I wasn't sure where in the pass pipeline this was going to run. I heard various statements like "lost after ISel" and to me, ISel is like the first step of codegen, it comes before register allocation. What you've done makes sense to me.
As you say, you are reconstructing some debug info by iterating over the spills rather than pulling it directly from the previous passes for handling dbg values, but that's probably fine. I could always be wrong, but I suspect that most stack usage is attributable to escaped allocas for C++ temporary aggregates (strings, etc), and we can be pretty confident that that info is correct. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D135488/new/ https://reviews.llvm.org/D135488 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
