Re: [PATCH] Fix VRP handling of __builtin_ffs* (PR tree-optimization/61725)

2014-07-08 Thread Richard Biener
On Tue, 8 Jul 2014, Jakub Jelinek wrote: > Hi! > > When writing this code, I've been assuming __builtin_ffs* argument > is unsigned, which is the case of popcount also handled there, and > many other builtins (clz, ctz, ...), except that clrsb has signed argument. > But in reality for some reason

[PATCH] Fix VRP handling of __builtin_ffs* (PR tree-optimization/61725)

2014-07-08 Thread Jakub Jelinek
Hi! When writing this code, I've been assuming __builtin_ffs* argument is unsigned, which is the case of popcount also handled there, and many other builtins (clz, ctz, ...), except that clrsb has signed argument. But in reality for some reason ffs has signed argument and some time ago Marek has e