On 19/02/18 17:10, Matthias Felleisen wrote: > >> On Feb 19, 2018, at 9:45 AM, Paulo Matos <pmatos@linki.tools> wrote: >> >> Thanks, interesting to see that fully typed programs run faster though. > > When I say “fully typed” you need to understand that some basic Racket > code may never be typed and is included here. I am referring to the code > that is under your control and I should add the adverb “usually” to this > sentence. >
I understand. > >> From your 4th point, one could assume that the more a program is typed, >> the less crossing there will be and therefore the better the performance >> will get? > > > This is unfortunately not correct as stated. It depends on the order in which > you modify the modules. It will be rare that you pick a path along which every > step improves the performance after you picked the worst possible module to > equip with types. Indeed, I conjecture that such paths don’t exist. Asumu, > the original project lead, considered a similar question and I believe it > came to nothing. > > But hopefully you'd get to a point (even if after some performance worsening) where you'd get a benefit on the path to having your app fully typed. Of course, as you said above if you have many crossings into non typed code (libraries) then you might be in trouble. :-/ > >> >>> If you want more specific information, we can post a journal submission >>> that compares several Racket implementations (6.2, 6.3, and 6.4) on a range >>> of programs using functional to ho object-oriented style. >>> >> >> That would certainly be appreciated. > > > > I have temporarily posted the paper at the following URL: > > http://www.ccs.neu.edu/home/matthias/Tmp/draft-performance-of-tr.pdf > Perfect. Got this. > Ben Greenman is the lead author and has taken this project under his wings. > For any details on the measurements, he’s the man to talk to. In a sense, > what we want is a re-evaluation of this benchmark suite for every release of > Racket so that we know whether we are making progress. We just don’t have the > manpower and machine resources to implement this idea. > Thanks. Will contact him. > — Matthias > -- Paulo Matos -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
