[erlang-questions] Erlang vs IO

Imre Palik <>
Wed Sep 3 16:27:23 CEST 2008


It is me messing up something, or the IO in erlang is quite slow?  I run the following program on a largish file:


problem_loop(0, _, Acc) ->
    io:fwrite("~w~n", [Acc]),
problem_loop(I, D, A) ->
    {ok, [V]} = io:fread([], "~d"),
        0 == V rem D ->
            problem_loop(I - 1, D, A + 1);
        true ->
            problem_loop(I - 1, D, A)

run() ->
    {ok, [Iter, Div]} = io:fread([], "~d ~d"),
    problem_loop(Iter, Div, 0).

Then both of my cores started to run at 100%, which I find weird in itself with an IO bound problem.  Also, I compared the runtime with the following program written in scheme

#!/usr/bin/guile -s

(define (intest iter div acc)
  (if (= iter 0)
      (display acc)
      (if (= 0 (remainder (read) div))
          (intest (- iter 1) div (+ acc 1))
          (intest (- iter 1) div acc))))

(intest (read) (read) 0)

The scheme implementation was 10 times as fast as the erlang, and it only used one of the cores.

How can I get a decent IO performance out of erlang?  Am I making some mistakes in my erlang code? Should I upgrade my erlang system? (I am running R12B on windows)



More information about the erlang-questions mailing list