From: Ralf Jung Date: Thu, 10 Mar 2022 18:18:48 +0000 (-0500) Subject: dont singe those poor integers X-Git-Url: https://git.ralfj.de/web.git/commitdiff_plain/78e44b8e3962fbd2c8e89369b5b73ccd39035d6f?ds=inline dont singe those poor integers thanks Drew! --- diff --git a/personal/_posts/2020-12-14-provenance.md b/personal/_posts/2020-12-14-provenance.md index 4e95665..ee3f5b6 100644 --- a/personal/_posts/2020-12-14-provenance.md +++ b/personal/_posts/2020-12-14-provenance.md @@ -88,7 +88,7 @@ All involved optimizations need to exactly agree on what is and is not UB, to en This is exactly what we also expect from the specification of a programming language such as C, which is why I think we should consider compiler IRs as proper programming languages in their own right, and specify them with the same diligence as we would specify "normal" languages.[^ub-difference] Sure, no human is going to write many programs in LLVM IR, so their syntax barely matters, but clang and rustc produce LLVM IR programs all the time, and as we have seen understanding the exact rules governing the behavior of programs is crucial to ensuring that the optimizations LLVM performs do not change program behavior. -[^cheat]: If now you feel like we somehow cheated, since we can always translate the program from C to LLVM IR, optimize there, and translate back, consider this: translating from LLVM IR to C is really hard! In particular, singed integer addition in LLVM IR can *not* be translated into signed integer addition in C, since the former is well-defined with `poison` result in case of overflow, but the latter says overflow is UB. C has strictly more UB than LLVM IR (for integer arithmetic), which makes translation in one direction easy, while the other direction is hard. +[^cheat]: If now you feel like we somehow cheated, since we can always translate the program from C to LLVM IR, optimize there, and translate back, consider this: translating from LLVM IR to C is really hard! In particular, signed integer addition in LLVM IR can *not* be translated into signed integer addition in C, since the former is well-defined with `poison` result in case of overflow, but the latter says overflow is UB. C has strictly more UB than LLVM IR (for integer arithmetic), which makes translation in one direction easy, while the other direction is hard. [^ub-difference]: In particular, two different variants of the IR with different rules for UB are really *two different programming languages*. A program that is well-defined in one language may have UB in another, so great care needs to be taken when the program is moved from being governed by one set of rules to another.