R-1.0.0 and Alpha-linux (RH 6.1) - R-help

hansel@hansel.moorhead.msus.edu

Wed, Mar 1, 2000 5:47 PM #

The release compiles and installs cleanly, but aborts with the message,
Floating point exception.  Before I start messing with the debugger, has
anyone gotten R to run on an Alpha-linux, Red Hat 6.1?

Thank you.

Mark Hansel
PO Box 41
Moorhead State University
Moorhead, MN 56563
ph: 218-236-2039 fax: 218-236-2593
email: hansel at mnstate.edu
web: http://wwwcj.moorhead.msus.edu


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

hansel@hansel.moorhead.msus.edu

Sun, Mar 5, 2000 4:45 PM #

R-1.0.0 still will not run on Alpha-Linux (RH 6.1).

I have two LX164 machines. One runs RH 6.0, the other RH 6.1. Only
the RH 6.1 gives problems, at least as far as I have tested.

First, R-1.0.0 compiles, installs and runs on an LX164 RH 6.0. All the
checks ran without complaint.

On the LX164 RH 6.1, it compiles and installs. Running it from the command
line or running the checks returns a floating point exception. The check
(make check) output filetests stops with only the floating point
exception.
([ctest-Ex.Rout] reports Error 136) and then
Error 2 as make backs up to the top level.

The error occurs after initialization output and somewhere before a prompt
appears.

On the 6.1 machine, I tried compiling optim.c by hand and relinking (with
make) as suggested, as suggested by one developer. Same crash, same place.

FWIW: This same pattern occurs with the netscape compaq released for alpha
Linux. It runs on RH 6.0 and not on RH 6.1 and crashes with the floating
point exception.

There are no compiler errors for R on either platform.

Compiler is egcs 2.91.66 (from gcc -v) (This is the egcs-1.1.2 release). 

The RH install is pure vanilla at this point. --No upgrades and no
libraries installed that are relevant.

(I have private mail to the effect that one other alpha machines (DS10)
work fine with Rh 6.1.)

Mark Hansel
PO Box 41
Moorhead State University
Moorhead, MN 56563
ph: 218-236-2039 fax: 218-236-2593
email: hansel at mnstate.edu
web: http://wwwcj.moorhead.msus.edu


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Brad McNeney

Sun, Mar 5, 2000 11:50 PM #

I just resubscribed to r-help after being unsubscribed for a while so
apologies in advance if I missed part of this. 

I was having some trouble compiling R using gcc on alpha osf5 (and am
still having some trouble with the native C compiler), but found a
suggestion from Greg Snow to try compiling with the -mieee-with-inexact
flag instead of the default -ieee_with_inexact. I get a similar result to
yours (floating point exception just before the prompt) when I compile
with the default, but it seems to work fine with -mieee-with-inexact
(passes all tests). Don't know if this is relevant to Alpha Linux, but
might be worth a try if you haven't already.

On Sun, 5 Mar 2000 hansel at hansel.moorhead.msus.edu wrote:

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Peter Dalgaard

Mon, Mar 6, 2000 1:37 AM #

hansel at hansel.moorhead.msus.edu writes:

This is pretty darn odd... The quick-exit syndrome with an FP error
can only occur from misconfiguration of the IEEE options I think, but
why would that be different between 6.0 and 6.1?

2 ideas:

1. what do you get from running tools/config.guess? If it doesn't
   match alpha*linux in configure.in, you'd certainly get in trouble.

2. are we using the proper form of the ieee option? configure.in sets
   -mieee (do you see that in compiles too?), but it may be that it
   wants -mieep-fp, -mieee-with-inexact or whatnot... "gcc --help -v
   2>&1 | grep ieee" works with gcc 2.95, but not with egcs 1.1.2 on
   intel, but you could try.

Well, three ideas: does the 6.0 binary run on the 6.1 machine?

O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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

Mon, Mar 6, 2000 5:21 AM #

Peter Dalgaard BSA <p.dalgaard at biostat.ku.dk> writes:

According to the documentation, if we would use -ieee_with_no_inexact
for the DEC cc compiler then we should use -mieee with gcc.

`-mieee'
     The Alpha architecture implements floating-point hardware
     optimized for maximum performance.  It is mostly compliant with
     the IEEE floating point standard.  However, for full compliance,
     software assistance is required.  This option generates code fully
     IEEE compliant code *except* that the INEXACT FLAG is not
     maintained (see below).  If this option is turned on, the CPP
     macro `_IEEE_FP' is defined during compilation.  The option is a
     shorthand for: `-D_IEEE_FP -mfp-trap-mode=su -mtrap-precision=i
     -mieee-conformant'.  The resulting code is less efficient but is
     able to correctly support denormalized numbers and exceptional
     IEEE values such as not-a-number and plus/minus infinity.  Other
     Alpha compilers call this option `-ieee_with_no_inexact'.

Using -mieee-with-inexact is slightly stronger.  According to the
documentation it should not result in changes that affect that
quick-exit.

I did successfully compile and test R-1.0.0 on a DEC Alpha running
Digital Unix 4.0d with gcc 2.95.2 and the -mieee flag.
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._