Message boards : Number crunching : Status and Plans, 20th August, 2013
Message board moderation

To post messages, you must log in.

AuthorMessage
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 857
Credit: 1,619,050
RAC: 0
Message 25709 - Posted: 30 Aug 2013, 12:50:35 UTC

Status and Plans, 20th August, 2013 Sorry for lack of recent news. Been very busy, running fast to stay in the same place. A new version of SixTrack will be introduced in the next few days; just doing some final tests on the BOINC SixTrack test project. This version fixes a nasty bug of mine whch I am hoping will explain a few cases where I have replicated but different results.
We shall see. There is also some new physics which will bring new work. Plans are to implement the return of all results in a zipped tar file instead of just fort.10. This will allow us to split each case into a series of runs of 100,000 turns (say) which will be much better for work sharing and will allow us to extend runs to 10 million turns or more. (I just remind you that the time estimate for each run is an estimate of the maximum time; if all particles are lost, and that can happen very quickly at high amplitudes, then the time taken and the associated credit will be much less.
I believe our system of credits is fair; no credit inflation and based on the work actually done.)

I shall try and get some recent papers with BOINC based results onto the WWW.

I am (finally?) trying to get ready to publish on results reproducibility. I will need to run more tests and I am trying to figure out a "drop box" system so that I can ask you to run these tests on all sorts of hardware/software combinations.
It would not be practical to create BOINC executables for every test or to make new executables to be installed each time.
GPU tests are delayed again but would also be interesting for these reproducibility tests. I am now getting identical results on Windows/Linux/Mac on "any" PC hardware with 5 different Fortran compilers at any level of optimisation (apart from compiler bugs).
Still haven't got a Fortran 2003 compiler (in 2013!) which would greatly simplify matters.

However as usual I always give priority to the service and continued thanks as the system is churning away well
and we are getting lots of results. Eric.

ID: 25709 · Report as offensive     Reply Quote
Dirk Broer

Send message
Joined: 20 Sep 05
Posts: 31
Credit: 1,212,091
RAC: 10
Message 25713 - Posted: 30 Aug 2013, 15:35:34 UTC - in response to Message 25709.  

ID: 25713 · Report as offensive     Reply Quote
Robert H

Send message
Joined: 22 Mar 13
Posts: 5
Credit: 8,896
RAC: 0
Message 25714 - Posted: 30 Aug 2013, 16:26:46 UTC - in response to Message 25713.  
Last modified: 30 Aug 2013, 16:27:17 UTC

Dirk:

Link not working, try again!
ID: 25714 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 798
Credit: 644,770,056
RAC: 231,829
Message 25715 - Posted: 30 Aug 2013, 17:20:59 UTC

ID: 25715 · Report as offensive     Reply Quote
[AF>FAH-Addict.net]toTOW

Send message
Joined: 9 Oct 10
Posts: 77
Credit: 3,671,357
RAC: 0
Message 25718 - Posted: 30 Aug 2013, 18:41:35 UTC

Glad to see that the project is still active that new stuff is coming soon :)
ID: 25718 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 857
Credit: 1,619,050
RAC: 0
Message 25735 - Posted: 2 Sep 2013, 19:00:59 UTC

Thanks for the pointers to Fortran 2003 status.
(I am out of touch.) Probably the most important for me is
the ROUND= specifier. gfortran seems to be progressing well.
I can't switch in a hurry but it would seem ideal to use a free
compiler on our volunteer project especially after all the
effort fighting ifort. :-) I need to step back a bit and come up
with a decent plan. I need to sort out the code generation option (SSE
etc) and just compare performance. Another issue is that gfortran and gcclibs
are outwith my control at CERN. It would be great to have gfortran on Linux,
Windows and MAC I suppose I could install my own
versions though. Food for thought and thanks again. Eric.
ID: 25735 · Report as offensive     Reply Quote
Robert H

Send message
Joined: 22 Mar 13
Posts: 5
Credit: 8,896
RAC: 0
Message 25762 - Posted: 5 Sep 2013, 13:48:24 UTC - in response to Message 25735.  

Thanks for the pointers to Fortran 2003 status.
(I am out of touch.) Probably the most important for me is
the ROUND= specifier. gfortran seems to be progressing well.
I can't switch in a hurry but it would seem ideal to use a free
compiler on our volunteer project especially after all the
effort fighting ifort. :-) I need to step back a bit and come up
with a decent plan. I need to sort out the code generation option (SSE
etc) and just compare performance. Another issue is that gfortran and gcclibs
are outwith my control at CERN. It would be great to have gfortran on Linux,
Windows and MAC I suppose I could install my own
versions though. Food for thought and thanks again. Eric.


Hmm..there is something called GNU compiler and then there is something called OpenMP in Gfortran.

And yes there are free compilers.

But i really understand your question, so i think you should make a phonecall to Canonical or Red Hat.

:)
ID: 25762 · Report as offensive     Reply Quote
Richard Haselgrove

Send message
Joined: 27 Oct 07
Posts: 186
Credit: 3,297,640
RAC: 0
Message 25763 - Posted: 5 Sep 2013, 15:44:14 UTC - in response to Message 25762.  

there is something called OpenMP in Gfortran.

Isn't OpenMP used in multi-processor (multi-threaded) applications?

I see output like this in Milkyway's multithreaded (N-Body) applications:

Using OpenMP 4 max threads on a system with 4 processors
ID: 25763 · Report as offensive     Reply Quote

Message boards : Number crunching : Status and Plans, 20th August, 2013


©2024 CERN