[Rd] Improvement of SignRank functions

[Ivo Ugrina]

Martin Maechler wrote:
> do you have evidence for your belief?
> i.e. a set of  system.time(.) calls where you see the
> difference?


system.time(dsignrank(17511, 400))
    user  system elapsed
   1.010   0.120   1.145
system.time(dsignrank((0:17511), 400))
    user  system elapsed
    1.25    0.13    1.40
system.time(dsignrank((0:17511), 500))
    user  system elapsed
   2.040   0.220   2.296
system.time(psignrank((0:17511), 600))
    user  system elapsed
  20.670   0.580  21.403
system.time(qsignrank(0.56, 300))
    user  system elapsed
   0.700   0.050   0.753
======================================
system.time(dsignrank(17511, 400))
    user  system elapsed
   0.070   0.000   0.078
system.time(dsignrank((0:17511), 400))
    user  system elapsed
   0.100   0.000   0.104
system.time(dsignrank((0:17511), 500))
    user  system elapsed
   0.160   0.000   0.164
system.time(psignrank((0:17511), 600))
    user  system elapsed
  16.330   0.370  16.729
system.time(qsignrank(0.56, 300))
    user  system elapsed
   0.020   0.010   0.029



system.time(dsignrank((0:20000), 600))
    user  system elapsed
   3.470   0.280   3.745
RAM: ~130MB
======================================
system.time(dsignrank((0:20000), 600))
    user  system elapsed
   0.250   0.010   0.26
RAM: ~1MB



> BTW: If you had a smart idea to *not* use a static 'w' and still
>      be memory efficient,
>      that could lead to make that code "thread-safe", but I am
>      not at all sure this is possible without using
>      "thread-library C code".

I'll look into it.


With respect,
-- 
Ivo Ugrina
ICQ: 47508335 | www.iugrina.com
-------------------------------
baza matematickih pojmova
http://baza.iugrina.com
---------------------------
anime, manga, Japan fanzin
http://yoshi.iugrina.com