[Apbs-users] Re: version 0.4.0 installation: two issues and a question...

[Nathan B. <ba...@bi...>]

Hi Chris --

Thanks for your compile comments; Todd and I will take a look at this
and try to streamline the process a bit.

I'm curious about the "exponential dependence" on SDENS -- can you
post the results?  If there is some well-defined convergence that's
robust from system to system, one could imagine coming up with a
Richardson-like extrapolation scheme to do exactly what you propose...
 This shouldn't be too hard for us to implement and could potentially
be a very nice addition to APBS!

Thanks!

-- Nathan

On 12/16/05, Chang, Christopher <Chr...@nr...> wrote:
>
>
> Dr. Baker,
>
>    I downloaded and installed the latest source version of APBS yesterday=
 on
> our local cluster, based on AMD Opteron processors. I used the PGI
> compilers, and I seem to remember an indication that feedback from users =
of
> these was welcome. The two minor compilation issues I ran into should be
> compiler-independent, though.
>
> (1) After configure/make/make install of MALOC with MPI enabled and BLAS
> disabled (I linked against the ATLAS libraries), the APBS configure with
> command
>
> ---------------
> ./configure --prefix=3D${FETK_PREFIX} --with-blas=3D"-L${BLAS_DIR} -lblas=
" >&
> configure1.log &
> ---------------
>
> failed with message
>
> ---------------
> checking for /copt/mpich.pgi/1.2.6-fPIC/lib/libmpich.a...
> yes
> checking for /copt/mpich.pgi/1.2.6-fPIC/lib/libmpich.so...
> no
> checking for main in -lmpich... no
> checking for main in -lmpi... no
> MALOC was compiled with MPI enabled
> MALOC needs to link against the MPI library
> Try defining FETK_MPI_INCLUDE and FETK_MPI_LIBRARY
> configure: error: MPI library or header is missing!!
> ----------------
>
> which was associated with lines in config.log:
>
> ----------------
> configure:10372: checking for
> /copt/mpich.pgi/1.2.6-fPIC/lib/libmpich.a
> configure:10387: result: yes
> configure:10394: checking for
> /copt/mpich.pgi/1.2.6-fPIC/lib/libmpich.so
> configure:10409: result: no
> configure:10417: checking for main in -lmpich
> configure:10443: pgcc -o conftest -O2 -fPIC -tp amd64 -fastsse
> -I/copt/mpich.pgi/1.2.6-fPIC/include
> -L/copt/mpich.pgi/1.2.6-fPIC/lib
> -L/copt/mpich.pgi/1.2.6-fPIC/lib conftest.c -lmpich
> /copt/mpich.pgi/1.2.6-fPIC/lib  >&5
>
> conftest.c:
> File with unknown suffix passed to linker: /copt/mpich.pgi/1.2.6-fPIC/lib
> /copt/mpich.pgi/1.2.6-fPIC/lib: file not recognized: Is a directory
> -----------------
>
> which I traced back to line 10426 (I believe) in file configure:
>
> -----------------
> LIBS=3D"-lmpich ${mpi_lib_path} $LIBS"
> -----------------
>
> Commenting out and changing to
>
> -----------------
> # LIBS=3D"-lmpich $LIBS"
> -----------------
>
> seemed to clear this up.
>
> (2) After APBS configuration, make crashes with an error about an unexpec=
ted
> token ")". This seems to arise from someone's rather daring use of "@)]" =
as
> a variable name occuring in the Makefile. A quick global replacement in
> Makefile to "${@)]}" led to completion of make and make install.
>
> (3) Now the question: there is a new parameter "SDENS" in v. 0.4.0, for
> which a value of 10.0 is suggested. I tried runs with 10, 20, 30 and 40 o=
n
> my system, and solvation energy appears to follow a clean exponential
> dependence vs. SDENS. Is it sensible to use this functional dependence to
> extrapolate to infinite SDENS?
>
>  I'm subscribed to apbs-users now, so if the question isn't too naive, fe=
el
> free to reply via that forum. :-) Hope the build observations are useful.
>
> Thanks,
>
> Chris
>
> Christopher H. Chang, Ph.D.
> Research Associate
> National Renewable Energy Laboratory
> 1617 Cole Blvd., Mail Stop 1608
> Golden, CO  80401
> Phone (303) 275-3751
> Fax (303) 275-4007


--
Assistant Professor, Dept. of Biochemistry and Molecular Biophysics
Center for Computational Biology, Washington University in St. Louis
Web: http://cholla.wustl.edu/