Re:“Pack it in, mathematicians, someone owes LLVM a million ...
by agent327 on Thursday August 19, 2021 @10:50AM
(#28234547)
There is no side effect. The loop only interacts with variable 'x', which is discarded after the function ends. Since it is not observed, there is no side effect.
That leaves only the loop itself, which can run infinitely long (which would be UB, and the compiler is free to assume that won't happen), or it won't, in which case it will always return true. Thus, like them or not, the rules of C++ allow this particular outcome.
I can't say I like this kind of optimisation. Sure, you can construct neat circus tricks with it like this, but other than that it seems pretty pointless for real-world software, and something that could easily turn a simple mistake into a debugging disaster.
Re:“Pack it in, mathematicians, someone owes LLVM a million ...
by bidirectional on Thursday August 19, 2021 @10:24AM
(#28234194)
The only return statement returns a hardcoded true.
Re:“Pack it in, mathematicians, someone owes LLVM a million ...
by Denvercoder9 on Thursday August 19, 2021 @11:01AM
(#28234697)
For the purposes of the C++ standard, looping forever is not a side-effect. Only I/O calls, volatile accesses and atomic/synchronization operations are.
That soundbite also repeats/paraphrases what thread author Joe Groff said "LLVM considers infinite recursion to be UB,"
But... the key is infinite loops without side effects can be optimized away. Endless loops even without a provable termination (The Halting Problem) are not UB.
But in this case, the value 'x' has a side effect of determining what bool of true/false value to return. That while() loop does not look like it's free of side effects.