[PVFS-users] pvfs and btio

[Rajeev Thakur]

Ah, I now remember. You need to apply two patches to MPICH, which are
available from 
  http://www-unix.mcs.anl.gov/mpi/mpich/buglist-tbl.html

You need to apply patch number "rpvfs" and number 4997.
(Note that the cumulative patch file on that page does not include the
patch rpvfs.)

Let me know if it still doesn't work.

Rajeev

P.S.: You can compile for multiple file systems if you want by
specifying them as --with-romio="-file_system=ufs+nfs+pvfs"


>  From: Kevin Makles <KEMakles@lbl.gov>
>  To: Rajeev Thakur <thakur@mcs.anl.gov>, KEMakles@lbl.gov
>  Subject: Re: [PVFS-users] pvfs and btio
>  Date: Wed, 12 Jul 2000 15:16:19 -0700
>  Cc: pvfs-users@beowulf-underground.org
>  
>  On Wed, 12 Jul 2000, Rajeev Thakur wrote:
>  > >  Has anyone been able to get pvfs and romio to work with the btio
>  > > benchmark? If you have I would like to know what steps were taken.
>  >
>  > Kevin,
>  >       Yes, we have run BTIO on ROMIO and PVFS. But I don't recall
>  > having to take any special steps. What problems do you see?
>  >
>  > Rajeev
>  
>  I have been having difficulty in compiling mpich-1.2.0 with romio.  When 
>  romio compiles, nfs + ufs are the only filesystems created.  I assume these 
>  must be the defaults.   These are the options I sent to the configure script.
>  ./configure -c++=pgCC -cc=pgcc -fc=pgf77
>  -rsh=rsh --with-romio="-file_system=pvfs 
>  -cflags=-I/home/makleske/pvfs/include" -lib="-L/home/makleske -lminipvfs" 
>  
>  I even tried single quotes. I also tried compiling mpich with no romio 
>  support then I compiled romio seperately.  Everything compiled nicely but 
>  when I executed btio-full I received. 
>  
>   NAS Parallel I/O Benchmarks -- BTIO Benchmark
>   No input file inputbt.data. Using compiled defaults
>   Output file name:                                /pvfs/btio.full.out
>   BT Benchmark version 2.0
>   Full MPI-IO Implementation, write interval:  5
>   Collective Buffering is Off
>   Size:  12x 12x 12
>   Iterations:  60    dt:   0.010000
>   Number of active processes:     1
>   
>  0 - MPI_TYPE_CREATE_SUBARRAY :
>  [0]  Aborting program !
>  [0] Aborting program!
>  p0_14052:  p4_error: : 134962204      
>  
>  Finally I tried LAM.  It was able to compile romio correctly but I received a 
>  runtime error when running btio-full.  I don't have the error saved but it 
>  had to do with invalid datatypes in ADIOI_Flatten. 
>  -- 
>  Kevin Makles
>  NERSC Future Technologies Group
>  Lawerence Berkeley Laboratory
>