on Sat Sep 17 2011, "Littlefield, Tyler"
On 9/17/2011 1:06 PM, Dmitriy Matison wrote:
Hello,
Would you, please, tell in more detail about the overhead in binary size? What are approximate numbers of it? And how badly the compilation time decreases? I used boost::bind a lot, but haven't noticed that the overhead makes so much difference. Opposite example would be valuable.
I switched to boost::bind because I use an event system, as I pointed out earlier. I liked being able to bind to methods and static members, rather than just global functions with a function pointer. The overhead could very well be with boost::function, since every time I add a typedef I get a lot of overhead there. I can't rip it out and show exactly how much compilation time decreases because it's an integral part of my system now. I don't think the compilation time has went up by a huge factor, but the binary size certainly has. This is both with -O3 and -O0, using g++.
Thanks, Dmitriy.
You can measure how much binary size overhead each bind costs you by replacing it with a hand-rolled function object that does the same thing, and comparing object code size. These are usually fairly easy (if tedious) to write. It's just a struct whose constructor stores everything you're currently passing to the bind that isn't a placeholder, and whose function call operator invokes the underlying object. -- Dave Abrahams BoostPro Computing http://www.boostpro.com