Message-ID: <x2smog2kzo.fsf@biostat.ku.dk>
Date: 2003-08-05T17:10:02Z
From: Peter Dalgaard
Subject: ^ operation much slower in R 1.7.1 than in R 1.7.0 ???
In-Reply-To: <Pine.A41.4.44.0308050718200.240674-100000@homer01.u.washington.edu>
Thomas Lumley <tlumley@u.washington.edu> writes:
> On Tue, 5 Aug 2003, Philippe Grosjean wrote:
> >
> > OK. But I was wondering if people at R-core team, especially those who
> > worked on Windows specific aspects, would have in mind some changes they did
> > between 1.7.0 and 1.7.1 than can cause this. Since these changes are mainly
> > corrections of bugs, the list is hopefully not so long... and it could help
> > a lot to have these hints. Thanks very much.
>
> I checked that the slowdown really was happening in "^", which is
> Primitive and so fairly direct. I couldn't find anything remotely
> plausible as a cause, though.
>
> My first guess was something involving the methods package and my second
> was a compiler change, so I'm 0-2 so far.
I think we can be fairly sure that the reason is that a new pow()
function was injected in the mingw runtime. You can find the actual
code in
http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/mingw/runtime/mingwex/math/
*why* this code was inserted is anyones guess. Googling threw up some
hint that the MS-supplied pow() was upsetting strtod and causing a
regression test failure with the Gimp. The code looks considerably
different from the kind of stuff used on Linux though, e.g. it
appears not to use FPU transcendentals. I suppose one could ask the
author...
Also, R is generally compiled with -D NO_MATH_INLINES because of some
ancient messup with glibc. I'm not sure whether that is actually still
necessary, but it is forcing the use of the non-inlined pow() function.
--
O__ ---- Peter Dalgaard Blegdamsvej 3
c/ /'_ --- Dept. of Biostatistics 2200 Cph. N
(*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk) FAX: (+45) 35327907