On Sunday, December 04, 2011 23:41:08 dsimcha wrote: > Thanks for the benchmark. I ended up deciding to just create a second > function, rt_finalize_gc, that gets rid of a whole bunch of cruft that > isn't necessary in the GC case. I think it's worth the small amount of > code duplication it creates. Here are the results of my efforts so far: > https://github.com/dsimcha/druntime/wiki/GC-Optimizations-Round-2 . > I've got one other good idea that I think will shave a few seconds off > the Tree1 benchmark if I don't run into any unforeseen obstacles in > implementing it.
It's best to avoid code duplication in general, but I'd argue that if only a small amount of code duplication is required to get a big gain from the GC, it's well worth it. In general, anything we can do to improve the GC is worthwhile, since it's _the_ place where D risks being inefficient in comparison to languages such as C++. - Jonathan M Davis
