Skip to content

atlas for R1.1.1 and R1.2 different?

7 messages · Marcus Eger, Douglas Bates, Peter Dalgaard +2 more

#
Dear R'ers,

when I e.g. install the integrate2.2-2 package under R1.1.1 (debian2.2)
atlas is linked into the library:


Installing package `integrate' ...
 libs
gcc -shared -lc -shared  -o
/usr/lib/R/library/integrate/libs/integrate.so adapt.o adapt_callback.o
bsrl.o flfm.o nxprt.o symrl.o wht.o -L/usr/local/lib  -lf77blas -lcblas
-latlas  -lg2c -lm -L/usr/lib/gcc-lib/i386-linux/2.95.2 -lm
 R


...however, for R1.2.0 (snapshot ~3 days ago) it looks different:
Installing package `integrate' ...
 libs
gcc -shared  -o integrate.so adapt.o adapt_callback.o bsrl.o flfm.o
nxprt.o symrl.o wht.o -L/usr/local/lib  -lg2c -lm
-L/usr/lib/gcc-lib/i386-linux/2.95.2 -lm
 R

I ask because integrate behaves differently for both R-versions.
  Marcus

--
+-------------------------------------------------------
| Marcus Eger
+--------------------------------------------------------
| E-Mail: eger.m at gmx.de (NEW)
|         marcus.eger at physik.uni-marburg.de (OLD)
| WWW:    http://neuro.physik.uni-marburg.de/~eger (NEW)
+--------------------------------------------------------


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
#
Marcus Eger <marcus.eger at physik.uni-marburg.de> writes:
I don't think the presence or absence of atlas should affect the
results from integrate.  The atlas library is an enhanced, faster
version of the basic linear algebra subroutines (blas) and a few of
the lapack subroutines.  Unless the C code in integrate is calling
these subroutines explicitly, atlas should not affect the results.

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
#
Douglas Bates wrote:

            
An example for the different behaviour is the following:

"prds" <-  function (s,r,sn=1)
1/(pi*sqrt(1-s^2))/(sqrt(2*pi)*sn)*exp(-(r-s)^2/(2*sn^2))


UNDER R1.2:
===========
value        relerr        minpts        lenwrk         ifail
2.735080e-269   7.38275e-08           165            73             0
UNDER R.1.1:
============
value      relerr      minpts      lenwrk       ifail
  0.3146147 0.009182386        1023         283           0

.... which makes more sense.
I'm wondering why the same integrate package under R1.1.1 uses atlas,
but under R1.2 it doesn't.

   Marcus


--
+-------------------------------------------------------
| Marcus Eger
+--------------------------------------------------------
| E-Mail: eger.m at gmx.de (NEW)
|         marcus.eger at physik.uni-marburg.de (OLD)
| WWW:    http://neuro.physik.uni-marburg.de/~eger (NEW)
+--------------------------------------------------------


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
#
Marcus Eger <marcus.eger at physik.uni-marburg.de> writes:
On a Debian 2.2 system running yesterday's snapshot of R-1.2.0 I get
+ 1/(pi*sqrt(1-s^2))/(sqrt(2*pi)*sn)*exp(-(r-s)^2/(2*sn^2))
value      relerr      minpts      lenwrk       ifail 
  0.3146147 0.009182386        1023         283           0 
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
#
Douglas Bates <bates at stat.wisc.edu> writes:
Ditto with RH6.2 

linked with 

gcc -shared  -o integrate.so adapt.o adapt_callback.o bsrl.o flfm.o nxprt.o symrl.o wht.o -L/usr/local/lib  -lg2c -lm -L/usr/lib/gcc-lib/i386-redhat-linux/egcs-2.91.66 -L/usr/i386-redhat-linux/lib -lm

[but I wonder how -L/usr/local/lib got into my LIBPATHS...]
#
The configure detection of fast BLAS libraries is different between
these versions, but I don't think that this is the reason.

I suspect that your R 1.1.1 is from a .deb file which requires package
atlas2, and the R 1.2.0 is compiled from source but cannot find atlas
(look at config.log) because package atlas2-dev is not installed.

Bingo?

-k

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
#
On Tue, 14 Nov 2000, Marcus Eger wrote:
The integrate package never uses atlas. The linker uses atlas under 1.1.1
because R does. This would allow integrate to use atlas if it wanted
to. It doesn't.

	-thomas

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._