[mpich-discuss] Installation problem mpich2-1.0.6/1.0.7 on LINUX MANDRIVA 2008.1

[Broeders, Cornelis]

After a few small modifications in mcnpx2.6.f this code is now running on MANDRIVA 2008.1 with standard gcc ans stable g95. Last solved problem was parameters for g95.
Please contact me if you need more details.
Best regards,
C. Broeders http://www.cornelis-broeders.de


-----Original Message-----
From: owner-mpich-discuss at mcs.anl.gov on behalf of Broeders, Cornelis
Sent: Thu 10/9/2008 10:19 PM
To: mpich-discuss at mcs.anl.gov
Subject: RE: [mpich-discuss] Installation problem mpich2-1.0.6/1.0.7 on LINUX MANDRIVA 2008.1
 
I am continuing without -g option, but it is somewhat surprising that this -g may lead to code-problems. My experience was to be careful with optimizing (-O, -O2 etc). I hope developers understand reasons.
Could somebody give me a comment on my "summary.xml" problem in my previous mail?
Thank you very  much in advance.
C. Broeders


-----Original Message-----
From: owner-mpich-discuss at mcs.anl.gov on behalf of Rajeev Thakur
Sent: Thu 10/9/2008 8:12 PM
To: mpich-discuss at mcs.anl.gov
Subject: RE: [mpich-discuss] Installation problem mpich2-1.0.6/1.0.7 on LINUX MANDRIVA 2008.1
 
If the 6 failing tests are the "i" versions of I/O tests, you can ignore
them for now.

Rajeev  

> -----Original Message-----
> From: owner-mpich-discuss at mcs.anl.gov 
> [mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of 
> Broeders, Cornelis
> Sent: Thursday, October 09, 2008 5:51 AM
> To: mpich-discuss at mcs.anl.gov
> Subject: AW: [mpich-discuss] Installation problem 
> mpich2-1.0.6/1.0.7 on LINUX MANDRIVA 2008.1
> 
> After your reply I did some more detailed investigations:
> - using "./configure --prefix=/home/opt/mpich2-1.0.7 CC=gcc 
> F77=g95 F90=g95", "make testing" ended with "All 497 tests passed!"
> - adding "--enable-g=all --enable-error-messages=all", result 
> was that 6 tests failed. These options were my initial ones 
> of my inquiry.
> - adding only "--enable-g=all", result is "6 tests failed out of 497"
> My next step now will be to further analyze my mcnpx problem. 
> I had introduced the "--enable-g=all" option to obtain 
> traceback information for this problem.
>  
> Another question is related to the the output file 
> "summary.xml" of "make testing". I do not succeeed to produce 
> a good readable representation. I tried to open the file with 
> firefox in a subdirectoty, together with the provided file 
> "TestResults.xsl", but this fails with "XML Parsing Error: no 
> element found". The postion is after last line in file summary.xml.
> Thank you very much for your support,
> C. Broeders  
> 
> -----Ursprüngliche Nachricht-----
> Von: owner-mpich-discuss at mcs.anl.gov 
> [mailto:owner-mpich-discuss at mcs.anl.gov] Im Auftrag von Rajeev Thakur
> Gesendet: Mittwoch, 8. Oktober 2008 23:39
> An: mpich-discuss at mcs.anl.gov
> Betreff: RE: [mpich-discuss] Installation problem 
> mpich2-1.0.6/1.0.7 on LINUX MANDRIVA 2008.1
> 
> 
> What were the exact parameters you passed to configure, and 
> what was the output printed on the screen when you ran "make 
> testing". If the "i" versions of the I/O tests are giving 
> trouble, you can try configuring with --disable-aio.
> 
> Rajeev
> 
> > -----Original Message-----
> > From: owner-mpich-discuss at mcs.anl.gov
> > [mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of 
> > Broeders, Cornelis
> > Sent: Wednesday, October 08, 2008 3:34 PM
> > To: mpich-discuss at mcs.anl.gov
> > Subject: [mpich-discuss] Installation problem 
> > mpich2-1.0.6/1.0.7 on LINUX MANDRIVA 2008.1
> > 
> > Dear all,
> > during implementation of mcnpx2.6.f on LINUX MANDRIVA 2008.1
> > I encountered problems in case of parallel computing with 
> > mpich2-1.0.6/7 when using 3 processes for 2 processors on one 
> > notebook ACER Travelmate 6592G. The same executable runs fine 
> > with 2 processes, using one processor. Similar errors 
> > occurred for different compiler options: icc/ifort intel 
> > 10.1/9.1. Earlier installations with LINUX SuSE9.3 worked 
> > fine for such cases.
> > After some trouble-shooting, I decided to look into detail in 
> > the testing of mpich2. I observed that all test combinations 
> > ended with errors after "make testing". My last detailed 
> > check was performed with gcc4.2/g95 0.91. I applied the "-i4" 
> > option in g95. The "make testing" command produces a huge 
> > "summary.xml" file, mainly containing lines of type "nest 
> > stack exceeded at mpir_request.c:174". Further analysis 
> > showed that these messages are produced in test programs 
> > "iwrite.." in directories "io" for all compilers(f77, f90, cxx).
> > I am not sure about correlation between my mcnpx and 
> > "iwrir.." problem, but I have to go step by step.
> > Could somebody comment and/or give hints for next steps to 
> > get mcnpx running with 2 processors on one system with recent 
> > LINUX version??
> > Thank you very much in advance for any comment.
> > Best regards,
> > C. Broeders     http://www.cornelis-broeders.de     
> > 
> > 
> 
>