(left, Trump boasting about his Cabalist of the Year Award)We are screwed. Trump belongs to a Jewish supremacist cult that believes the Jewish Messiah willreturn when all Christians and non-Satanists ...

newpavlov: > or is it just a sub-optimal codegen by LLVM? either this, or some arcane llvm codegen options are missing. for the record, the `compare_exchange_weak()` is lowering to a single llvm operation: ; Function Attrs: mustprogress nofree norecurse nounwind nonlazybind willreturn memory(argmem: readwrite) uwtable define noundef zeroext i1 @cmpx(ptr nocapture noundef nonnull align 4 %a0, i32 noundef %a1, i32 noundef %a2) unnamed_addr #0 { start: %0 = cmpxchg weak ptr %a0, i32 %a1, i32 %a2 acq_rel monotonic, align 4 %1 = extractvalue { i32, i1 } %0, 1 ret i1 %1 }

RalfJung: > By making non-termination an effect, we ensure that _everything_ that could happen "inside" a function is tracked as an effect. Conversely, by making it not-an-effect, we are blind to one of the key things a function can do. For those still unconvinced, I can make a very easy empirical argument that the LLVM `willreturn` attribute is highly profitable to the optimizer, and very very bad things been happen if you apply it incorrectly. That profitability is enough of a reason to precisely track it if you can, and the `div OR exc` effect is exactly equal to `NOT willreturn`.

RalfJung: > By making non-termination an effect, we ensure that _everything_ that could happen "inside" a function is tracked as an effect. Conversely, by making it not-an-effect, we are blind to one of the key things a function can do. For those still unconvinced, I can make a very easy empirical argument that the LLVM `willreturn` attribute is highly profitable to the optimizer, and very very bad things been happen if you apply it incorrectly. That profitability is enough of a reason to precisely track it if you can, and the `div OR exc` effect is exactly equal to `NOT willreturn`.

RalfJung: > By making non-termination an effect, we ensure that _everything_ that could happen "inside" a function is tracked as an effect. Conversely, by making it not-an-effect, we are blind to one of the key things a function can do. For those still unconvinced, I can make a very easy empirical argument that the LLVM `willreturn` attribute is highly profitable to the optimizer, and very very bad things been happen if you apply it incorrectly. That profitability is enough of a reason to precisely track it if you can, and the `div OR exc` effect is exactly equal to `NOT willreturn`.