On Thu, Dec 6, 2012 at 2:36 PM, Jay McCarthy <jay.mccar...@gmail.com> wrote: > On Thu, Dec 6, 2012 at 3:32 PM, Neil Toronto <neil.toro...@gmail.com> wrote: >> >> On 12/06/2012 02:08 PM, Sam Tobin-Hochstadt wrote: >>> >>> On Thu, Dec 6, 2012 at 12:44 PM, <ntoro...@racket-lang.org> wrote: >>>> >>>> >>>> | Reimplemented really simple FFI functions (e.g. mpfr-prec, mpfr-exp) >>>> to >>>> | avoid calling overhead >>> >>> >>> If you have meaningful benchmarks where this makes a difference, that >>> may be useful to Matthew, since he recently was working on improving >>> the FFI's code generation. >> >> >> I've got some benchmarks showing via timing loops that pulling a _long >> directly out of an _mpfr takes just over half the time it takes to call >> libmpfr to do it. It's enough to make me want to rewrite simple things like >> `bfnegative?' in Racket. >> >> I won't, though, if Matthew has near-future Big Plans. Or Medium Plans. >> >> I see we have "tests/racket/benchmarks". Should I just drop the code >> there, or is there some kind of procedure or protocol I should follow? > > > Generally that directory is for benchmark Racket against other languages, so > the program has to be cross-compatible. I think you are talking about a > stress test, which can be used to monitor Racket's performance against > itself in the future to catch regressions. That's in tests/racket/stress.
I don't think that's right about the `benchmarks` directory -- the `shootout` directory is not cross-compatible, and we have Typed Racket benchmarks that are mostly used for benchmarking TR against plain R. Sam _________________________ Racket Developers list: http://lists.racket-lang.org/dev
