Hi all,

                The problems we experienced during ADF running under linux were due to
the wrong version of the g77/gcc couple. The latest version of g77/gcc
doesn't work, you have to move to an older one, g77-0.5.19/gcc-2.7.2.1
preferably.

                It is interesting to see that other linux compilers are more efficient
than the GNU compilers, as Jochen stated.

                Regards to all,

                                        Reinaldo

>Dear ADF-Linux users,
>
>during the last weeks several Linux users posted to this list
>and reported compilation and/or run time error messages with ADF.
>Typical run-time errors occured during the CREATE run:
>
><Aug05-98> <12:40:30> ADF 2.2.1 RunTime: Aug05-98 12:40:30
> <Aug05-98> <12:40:30> Molybdenum (IV, 3d frozen)
> <Aug05-98> <12:40:30> RunType : CREATE
> <Aug05-98> <12:40:30> Net Charge: 0 (Nuclei minus Electrons)
> <Aug05-98> <12:40:30> Symmetry : ATOM
> <Aug05-98> <12:40:30> WARNING: dSTEBZ (1) inaccurate
> <Aug05-98> <12:40:30> WARNING: dSTEBZ (1) inaccurate
> <Aug05-98> <12:40:30> WARNING: dSTEBZ (1) inaccurate
> <Aug05-98> <12:40:30> WARNING: dSTEBZ (1) inaccurate
> <Aug05-98> <12:40:31> >>>> CORORT
> <Aug05-98> <12:40:31> SINGULAR MATRIX, NO INVERSE
> <Aug05-98> <12:40:31> WARNING: not all scratch files were closed
> <Aug05-98> <12:40:31> archec: can't find IB2TIDS
> <Aug05-98> <12:40:31> END
>
>If you also get such errors with ADF+Linux they are possibly due to
>g77-optimization problems.
>
>Reducing the optimization flags in the $ADFHOME/Makeflags file
>to "-O2" should be sufficient in most cases to obtain a correctly
>running version of ADF.
>
>If this does not help you might have a buggy version of "g77".
>
>I use the g77/gcc from the "Pentium Compiler Group" without
>problems. This compiler can do additional pentium-specific
>optimizations and leads to approx. 10% shorter cpu-times of
>ADF jobs on my Pentium 166 compared with a "normal"-g77 compiled
>binary (even though I could compile with "-O3" with the original g77
>contained in my Linux distribution).
>
>You can download binaries of this compiler
>from
>
>http://www.iti.cs.tu-bs.de/soft/www.goof.com/pcg/binaries-linux.html
>or
>http://www.gcc.ml.org/
>
>You need to install the packages
>pgcc-1.0.3.linux86.tar.gz
>pg77-1.0.3.linux86.tar.gz
>
>My Makeflags file looks like the following:
>
>FCPARS=g77
>FFPARS=-O -w -fno-second-underscore
>LFPARS=
>
>CC=gcc
>CFLAGS=-Dpentium_linux -fno-second-underscore -mpentium -O6
>CDBFLAGS=-g -Dpentium_linux
>
>FC=g77
>O0V0= -w -fno-second-underscore
>O1V0=-O2 -w -mpentium -fno-second-underscore
>O2V0=-O2 -w -mpentium -fno-second-underscore -funroll-loops -ffast-math
>O0V1= -w -fno-second-underscore
>O1V1=-O2 -w -mpentium -fno-second-underscore
>O2V1=-O2 -w -mpentium -fno-second-underscore -funroll-loops -ffast-math
>LFLAGS=
>
>etc ...
>
>On the web-page
>
>http://ichsg06.unifr.ch/~rbruyn/Miscellaneous/bench_adf.html
>
>Raf Bruyndonckx presents some benchmarks concerning ADF and Linux
>(Thanks to Elke Hunstock for this information).
>It is a nice demonstration of the good price/performance-ratio of
>Pentium PC's.
>
>Best regards,
>Jochen
>--
>
>--------------------------------------------------
> Jochen Autschbach
> Priv.: Office:
> Sassenweg 24 Universitaet-GH-Siegen
> 57074 Siegen Theoretische Chemie
> Germany Postfach 101240
> D-57068 Siegen-Weidenau
> Germany
> e-Mail: jochen_at_aristoteles.chemie.uni-siegen.de
> Phone : +049-(271)-740-3373 (off. )
> +049-(271)-682070 (priv.)
> Fax : +049-(271)-7402330
>--------------------------------------------------
>
Received on 1998-08-19 21:54:11