On Aug 24, 8:48 am, Stuart Halloway <stuart.hallo...@gmail.com> wrote: > > On Aug 24, 6:44 am, Stuart Halloway <stuart.hallo...@gmail.com> wrote: > >> Clojure 1.3's performance improvements will significantly impact perf on > >> some of the benchmarks. If you are trying these out, please try them on > >> both 1.2 and 1.3. > > > Has Clojure 1.3 been released? > > No, but since the num/prim/equiv work specifically targets performance, we > want to collect people's experiences comparing 1.2 and 1.3. This is totally > separate from the benchmark submission process and for our own information. > > >> Also: the benchmarks are totally a numbers game: throw idioms and > >> readability out the window. Clojure 1.3 should be able to match Java > >> performance if you basically write Java-in-Clojure. On Clojure 1.2 you > >> will have to do stranger things to get there. > > > If you choose to throw idioms and readability out the window then > > don't be surprised at the comments that will be made about Clojure. > > Let me reduce the stridency of my previous statement: "throwing things out > the window" is too strong. Benchmarks are a numbers contest, not a beauty > contest. But, there's no saying the resulting programs will be ugly (or even > non-idiomatic). Try things. Measure. It is simply the case that some > idiomatic code (i.e. numeric code with no hints in the body) is faster in > master/1.3 than in 1.2, and that some benchmark-useful things (fns > taking/returning primitives) are available only post 1.2.
Well when Clojure 1.3 is released... The phrase "idiomatic code" often seems to be used to mean - code written in a natural way for that language and as if performance doesn't matter - whereas I seem to have the strange notion that both code written as if performance matters and code written as if performance doesn't matter can be "idiomatic code". http://shootout.alioth.debian.org/u32q/program.php?test=spectralnorm&lang=ghc&id=2 http://shootout.alioth.debian.org/u32q/program.php?test=spectralnorm&lang=ghc&id=4 > > If you have to "do stranger things to get there" with Clojure 1.2 then > > doesn't that simply suggest Clojure 1.2 performance doesn't match Java > > performance? > > If I had wanted to suggest that, I would have said "you can't get there". > With Clojure 1.2, you can get fast programs easily, or screaming-fast > programs with effort. Some of the work in Clojure 1.3 reduces that effort. > > Stu -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
