Myths about Integer Overflow in Rust

Re:Myths about Integer Overflow in Rust by Manishearth on Friday April 29, 2016 @02:22PM (#11597675)

> Overflow is defined by the language as an illegal operation

Is it? What do you mean by illegal here? Overflow can't cause UB madness in Rust like it does in C; the compiler does not optimize loops assuming no overflow.

[ Parent | on HN ]

Re:Myths about Integer Overflow in Rust by vardump on Friday April 29, 2016 @05:42PM (#11598960)

UB can be confusing, but please remember it also means a lot of very reasonable optimization opportunities.

Say in data flow analysis, when the compiler can assume "(value + 1) > value" to be always true, a whole expensive branch can be removed.

If the overflow is never going to happen anyways, the savings are considerable (expensive branch eliminated) and the result is always correct.

[ Parent | on HN ]

Re:Myths about Integer Overflow in Rust by Manishearth on Friday April 29, 2016 @06:03PM (#11599062)

I know that UB can be useful for optimizations. Where did I assert otherwise?

I'm saying that overflow is not undefined in Rust (the very first section of the blog post says this). I made this same mistake a few days ago and thought it was undefined (and implementation-defined to panic on debug), but I was wrong. Overflow is well defined in Rust.

[ Parent | on HN ]