In that case, is it worth it to enable USE=valgrind globally? vin., 26 mai 2023, 10:44 Sam James <s...@gentoo.org> a scris:
> > Sam James <s...@gentoo.org> writes: > > > [[PGP Signed Part:Undecided]] > > > > Sam James <s...@gentoo.org> writes: > > > >> Alexe Stefan <stefanalex...@gmail.com> writes: > >> > >>> Does enabling USE=valgrind impact runtime performance in any way? > >> > >> A very small amount because it adds a check at runtime for whether > >> the application is running under Valgrind. The compiler may be able > >> to optimise this a bit if it can determine it's unlikely (if the > >> Valgrind headers don't already do this, they should probably mark > >> them as cold functions). > > > > Sorry, I should say, they're hot functions but the if()s for them > > should be unlikely()'d. > > Alexander Monakov reached out and pointed out this isn't right - they're > actually branchless and just encoded in nops, so there's really > essentially no runtime cost. > > >
