Simulation is another place where this comes up – you have a heterogenous collection of event objects and you have a loop that goes through them and handles each one, choosing how to handle them by doing dynamic dispatch on each event. The main thing to do for now to optimize that is make sure that you only pay the cost of a single dispatch by making the handling body a single function so that once you've entered that function via a relatively slow dispatch, at least the body is fast and doesn't require any dispatch.
On Fri, Nov 6, 2015 at 12:50 PM, Cristóvão Duarte Sousa <cris...@gmail.com> wrote: > These experiments were also motivated by the problem of "Scene graph > operations and dispatch > <https://en.wikipedia.org/wiki/Scene_graph#Scene_graph_operations_and_dispatch>", > which can be more or less solved in C++ by using some tricks (e.g., the > "visitor pattern"), but which can be so-much-more-elegantly solved by > multiple dispatch in Julia. > > > On Fri, Nov 6, 2015 at 5:43 PM Cristóvão Duarte Sousa <cris...@gmail.com> > wrote: > >> >> Around 0.000400 seconds (but sometimes 0.001000). >> >> >> >> On Fri, Nov 6, 2015 at 5:39 PM Stefan Karpinski <ste...@karpinski.org> >> wrote: >> >>> I'm curious what the third timing number "yet another experiment" was >>> for the Julia code? I would imagine quite fast. >>> >>> >>> On Fri, Nov 6, 2015 at 12:37 PM, Stefan Karpinski <ste...@karpinski.org> >>> wrote: >>> >>>> Looks like a reasonable benchmark of this. Nice work. There's >>>> definitely a lot of room for improvement of dynamic dispatch. We can't use >>>> classic vtables because of the multiple part of Julia's dispatch – where do >>>> you stick the vtable? Most of the time we get performance by avoiding >>>> dispatch entirely, by pushing the dispatch up as far as possible, but there >>>> are cases like what you've create here, where that doesn't work. >>>> >>>> >>>> On Fri, Nov 6, 2015 at 12:27 PM, Cristóvão Duarte Sousa < >>>> cris...@gmail.com> wrote: >>>> >>>>> Hi, >>>>> >>>>> I've been wondering how Julia dispatching system would compare to the >>>>> C++ virtual functions dispatch one. >>>>> Hence, I write a simple test in both Julia and C++ that measures the >>>>> calling of a function over the elements of both an array of concrete >>>>> objects and another of abstract pointers to objects of derived types. >>>>> >>>>> Here is the code https://gist.github.com/cdsousa/f5d669fe3fba7cf848d8 >>>>> . >>>>> >>>>> The usual timings for C++ in my machine, for the concrete and the >>>>> abstract arrays respectively, are around >>>>> 0.000143 seconds >>>>> 0.000725 seconds >>>>> >>>>> For the Julia code the timings have much more variability, but they >>>>> are around >>>>> 0.000133 seconds >>>>> 0.002414 seconds >>>>> >>>>> This shows that Julia (single) dispatch performance is not that bad >>>>> while it has some room to improvement. >>>>> >>>>> If I'm doing something terribly wrong in these tests, please tell me :) >>>>> >>>>> PS: Thank you all, developers of Julia! >>>>> >>>>> >>>>> >>>> >>>
