Arc Forumnew | comments | leaders | submitlogin
First half-assed wart performance measurement: it's 1000x slower than arc
3 points by akkartik 5157 days ago | discuss
Here's the test case I ran:

  def fact(n)
    if (iso n 0)
      1
      * n (fact (- n 1))

  def foo(m)
    when (> m 0)
      fact 20
      foo (- m 1)
(wart doesn't have for/repeat loops yet)

arc/racket computes fact.20 in about half a millisecond, and wart currently[1] takes about half a second.

I'm not yet at the point where I'm thinking about performance. The goal is first to build the language to the point where I can build a performance model in space and time for wart programs. Then I'll be able to write deterministic performance tests for wart programs.

[1] http://github.com/akkartik/wart/tree/cb1392875d