On Aug 22, 2013, at 19:43 , Steve Mills <smi...@makemusic.com> wrote:
> On Aug 22, 2013, at 12:31:55, Thomas Wetmore <t...@verizon.net> > wrote: > >> Pre-allocation doesn't really matter as long as the re-allocations, whenever >> they occur, respect the capacity argument. > > Sure they do. If you don't preallocate, but instead keep appending, and the > pointer needs to grow with every append, it *could* reallocate with every > append. But that’s not how it’s typically implemented, you usually get some sort of exponential growth ( 2x or optimally something around 1.5x ), so amortized re-allocation cost on growth tends to be dominated by other factors. Using the NSData variant of the algorithm I presented earlier and averaging over 4 runs each, I get a 2.5% difference. Considering the factor 10-50 or more from the other optimizations, that’s pretty negligible, and was also significantly less than the variance between runs (so we would need a whole lot more runs to see whether the difference is actually real, chances are it isn’t). Despite the fact that I am somewhat performance-obsessed, I tend to not pre-allocate. Cheers, Marcel _______________________________________________ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
