[Rd] make check failed for R-1.9.0alpha (2004-03-16) when link against Goto's BLAS

[Prof Brian Ripley]

I suspect you are not going wrong, but rather that the Opteron Goto BLAS 
is not accurate enough on your particular system.  Unfortunately these 
LAPACK routines can be sensitive to the buildup of rounding errors, and 
that is affected by the BLAS in use.

On Tue, 16 Mar 2004, Liaw, Andy wrote:

> Dear all,
> 
> Has anyone seen the same problem?  I tried compiling R-1.9.0 alpha
> (2004-03-16) on our Opteron box running SUSE Linux ES8.  I ran:
> 
> ./configure --enable-R-shlib --with-blas=goto
> 
> and got:
> 
>   Source directory:          .
>   Installation directory:    /usr/local
> 
>   C compiler:                gcc  -m64 -O2 -g -msse2 -march=k8 -Wall
> --pedantic
>   C++ compiler:              g++  -m64 -O2 -g -msse2 -march=k8 -Wall
> --pedantic
>   Fortran compiler:          g77  -m64 -O2 -g -msse2 -march=k8
> 
>   Interfaces supported:      X11, tcltk
>   External libraries:        readline, BLAS(generic)
>   Additional capabilities:   PNG, JPEG
>   Options enabled:           shared library, R profiling
> 
>   Recommended packages:      yes
> 
> (The Goto BLAS is libgoto_opt64-r0.93.so.)
> 
> "make" ran fine, but "make check" failed with tail
> tests/Examples/stats-Ex.Rout.fail giving:
> 
> > ## now see the effect of the explanatory variables
> > data(Seatbelts)
> > X <- Seatbelts[, c("kms", "PetrolPrice", "law")]
> > X[, 1] <- log10(X[, 1]) - 4
> > arima(log10(Seatbelts[, "drivers"]), c(1,0,0),
> +       seasonal = list(order= c(1,0,0)), xreg = X)
> Error in La.svd(x, nu, nv, method) : error code 3 from Lapack routine dgesdd
> Execution halted
> 
> If I do "make clean" and re-configure without the Goto BLAS, then it passes
> make check.  Does anyone know what I'm doing wrong?  Any help much
> appreciated!
> 
> Best,
> Andy
> 
> 
> Andy Liaw, PhD
> Biometrics Research      PO Box 2000, RY33-300     
> Merck Research Labs           Rahway, NJ 07065
> mailto:andy_liaw at merck.com        732-594-0820
> 
> 
> 
> 
> ------------------------------------------------------------------------------
> Notice:  This e-mail message, together with any attachments,...{{dropped}}
> 
> ______________________________________________
> R-devel at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-devel
> 
> 

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595