> > * Elixir and the BEAM VM are awesome at many things, but I suspect (from > experience not evidence) that the defun version is still faster than the > elixir version.
In Clojure, the defun version is not the default or idiomatic way to write functions. I kind of expected it to be slower. Maybe if the core team felt it was good enough to be the default, they could make internal changes to the core to optimize stuff (pure speculation here). On the other hand, skimming through the elixir documentation gives me the feeling that pattern matched functions is the default way to write. As a first class language feature, I find it hard to believe it is slower (compared to what in Elixir?). Or do you mean the Beam VM in general is slower than the JVM? On Sunday, 6 September 2015 02:04:17 UTC+5:30, Rob Lally wrote: > > Out of interest, I ran the benchmarks as is, and got more or less the same > results - 15x. Then I tried upgrading the defun dependencies - clojure, > core.match and tools.macro - all of which have newer versions, and then > running the benchmarks without leiningen’s jvm-opts and in a trampolined > repl. The results are better (see below). Still not great - but down from > 15x to 10x. > > That said: > > * I’m not sure I’d care: for most applications the overhead of function > dispatch is probably not the bottleneck. > * Elixir and the BEAM VM are awesome at many things, but I suspect (from > experience not evidence) that the defun version is still faster than the > elixir version. > > > Rob > > --- > > user=> (bench (accum-defn 10000)) > WARNING: Final GC required 2.590098761776679 % of runtime > Evaluation count : 429360 in 60 samples of 7156 calls. > Execution time mean : 139.664539 µs > Execution time std-deviation : 4.701755 µs > Execution time lower quantile : 134.451108 µs ( 2.5%) > Execution time upper quantile : 150.214646 µs (97.5%) > Overhead used : 1.565276 ns > > Found 5 outliers in 60 samples (8.3333 %) > low-severe 5 (8.3333 %) > Variance from outliers : 20.5880 % Variance is moderately inflated by > outliers > > user=> (bench (accum-defun 10000)) > Evaluation count : 44940 in 60 samples of 749 calls. > Execution time mean : 1.361631 ms > Execution time std-deviation : 40.489537 µs > Execution time lower quantile : 1.333474 ms ( 2.5%) > Execution time upper quantile : 1.465123 ms (97.5%) > Overhead used : 1.565276 ns > > Found 9 outliers in 60 samples (15.0000 %) > low-severe 1 (1.6667 %) > low-mild 8 (13.3333 %) > Variance from outliers : 17.3434 % Variance is moderately inflated by > outliers > > --- > > > On 5 Sep 2015, at 05:16, Amith George <strid...@gmail.com <javascript:>> > wrote: > > Nice. Hadn't heard of it before. It looks interesting. The criterium > benchmark is kinda disappointing though. The pattern matched function took > nearly 15x the time of the normal function. > > > -- 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 --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
