On 16 June 2014 09:25, Brian Rogoff via Digitalmars-d <[email protected]> wrote: > On Sunday, 15 June 2014 at 20:53:58 UTC, w0rp wrote: >> >> I'm going to try my hand at making a game with 2.066, because I believe >> @nogc is a final piece in a puzzle of making doing that easy. Much like >> writing bare metal D code without the runtime, I'm going to try my hand at >> writing D code with the main function marked as @nogc, because I reckon it's >> going to leave me with a saner set of syntax and semantics than either C or >> C++ in the end, with none of the drawbracks from the stop the world effect >> in a game loop. > > > Good luck, I'm sure a lot of people are interested. > > >> Having said that, I do think there's some kind of brain malfunction on the >> part of games programmers that makes them think "is slow and can't escape >> from" when they hear "garbage collector" > > > There's a similar brain malfunction on the part of GC advocates that makes > them think "GC programs are just as performant as non GC'ed programs at no > cost". (**) > > On the Lang.Next panel, Andrei said something like > > "For the same payload the garbage collected program uses 3 times as much > memory" > > and researcher Emery Berger writes that > > http://people.cs.umass.edu/~emery/plasma/emery/memory-management-studies.html > > "... a good GC can match the performance of a good allocator, but it takes > 5X more space. If physical memory is tight, however, conventional garbage > collectors suffer an order-of-magnitude performance penalty." > > That's not even taking into account the non-determinism of tracing GC or the > issues of finalizers vs destructors or ... > > Being able to turn off GC, and having the libraries not hit the GC, are all > important.
Thank you. I was very worried I was about to start another holy war responding to that post ;) >> and "makes things more complicated and slower" when they hear "template." >> Neither of these things are true. > > > D metaprogramming is a winning feature. But GC has a number of costs as well > as a number of benefits. > > (**) I much prefer working in a GC'ed language with high level features that > require a GC (full closures!) but if resources are tight I think you need to > give up some pleasant features. Closures are so few and infrequent, they could trivially be ARC allocated, and the language dependence on GC would be meaningfully reduced.
