Re: [libvirt] [PATCH] Add VIR_DIV_UP to divide memory or storage request sizes with round up

On 01/28/2011 02:14 PM, Matthias Bolte wrote: Much nicer than open-coding it everywhere. Wow - are we THAT guilty of copy-and-paste? I hope you automated some of these edits, rather than doing it all by hand (perl, or even sed+xargs, can be pretty powerful). Hmm - now that we are codifying it, can we make it less prone to overflow issues? That is, if value < max, but (value + size - 1) wraps around, this could result in 0 (for unsigned types) or negative (for signed types - in fact, overflow on signed types is technically undefined behavior), where the old (unsafe) division would have resulted in max/size. Unfortunately, I don't see any way to do that without evaluating value multiple times. Any better ideas than: # define VIR_DIV_UP(value, size) \ (((value) + (size) - 1) < (value) \ ? ((value) / (size)) + 1 \ : ((value) + (size) - 1) / (size) Also, is this the sort of thing where we should require that size be a power of 2 and/or a compile-time constant (and thus that we can use << and >> instead of /)? Then again, if size is a compile time constant (which it looks like in all? of your cases), then the compiler should be smart enough to optimize this (if not, you have a sorry compiler). ACK to everything else, but let's make sure we agree on the final version of VIR_DIV_UP being committed. -- Eric Blake eblake(a)redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org