t.p.northover added a comment. With current CodeGen by the time you reach the trap you have no idea what came before so I think you'd still need a separate trap instruction per failure kind. So the Clang and generic LLVM side would be unaffected.
I suppose on X86 it could save a few bytes in .text and so maybe icache (though again we're talking about roughly 50% bloat for enabling UBSAN at all so this is kind of lost in the noise). On the whole I'm not a fan of inventing a new entirely separate scheme for that. CHANGES SINCE LAST ACTION https://reviews.llvm.org/D89959/new/ https://reviews.llvm.org/D89959 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
