AW: [MPICH] Clog->Slog2 - can't use single events?

[Kilian Timmler]

Hi,

> -----Ursprüngliche Nachricht-----
> Von: owner-mpich-discuss at mcs.anl.gov 
> [mailto:owner-mpich-discuss at mcs.anl.gov] Im Auftrag von Anthony Chan
> Gesendet: Donnerstag, 18. Mai 2006 19:06
> An: Kilian Timmler
> Cc: mpich-discuss at mcs.anl.gov
> Betreff: Re: [MPICH] Clog->Slog2 - can't use single events? 
> 
> 
> On Thu, 18 May 2006, Kilian Timmler  wrote:
> 
> > I would like to know whether it is at all possible to 
> convert single CLOG
> > events so that they are shown in jumpshot-4. If it is, I 
> would like to get
> > some idea of how I have to set up the CLOG event so that it 
> is correctly
> > converted into SLOG2 format. Unfortunately I couldn't find 
> any example of
> > this sort of event being used, so I was left with guessing.
> 
> I'm guessing you are using MPE from MPICH1 to create your 
> clog file. 

I'm not using any code from MPE for logging since I just wanted to convert
my own file
to CLOG. I actually tried using what is in mpe-1.9.1/src since it would have
been easy to 
change, but I had problems with it, so I wrote a simple converter from the
information I 
could get about the file format from slog2sdk and mpe-1.9.1 and fpilog.clog.
That worked 
out well enough. 

> The
> old MPE logging in MPICH1 has not been updated for few years. All our
> effort is currently in MPE2 which is shipped with MPICH2.  MPE2 does
> provide user-defined logging API to define/log "single 
> timestamp" event.
> MPE2 logging generates CLOG2 format which is more efficient than CLOG.

That sounds reassuring - I'll switch.

I have seen that one can switch MPI off in MPE, does that allow me to
generate 
trace information for just one node, and do I have to deal with merging
logdata at all?

If I just want to go ahead and improve my simple converter, I have to adjust
it for 
CLOG2. From what I have seen the CLOG2 format requires a peamble of length
1024 bytes, 
is it always 1024 bytes long? 

Currently (with CLOG) I'm just writing down headers and records and make
sure that the 
file is 65536 bytes long. I have to define some states and must place an end
of 
block/file header at the end of my log data the rest is zeros.

I don't really understand how to use blocks especially if the file size is
larger, is this of any 
importance or can I just place an arbitrary number of header/record pairs
into the file 
now (with CLOG2) and finish it of with an end of file header?

> You can download the latest MPE2 from http://www.mcs.anl.gov/perfvis.
> For sample programs to define/log single timestamp event, e.g. see
> mpe2/src/wrappers/test/cpilog.c.  You may also want to check 
> the manpages
> for the newer MPE logging functions.
> 
> A.Chan

I'm tempted to just use MPE on the DSP but I fear that it has some
considerable overhead 
(memory wise too). Also we will probably never have MPI on the DSPs/FPGAs
and the merging of log data will 
have to happen somehow outside the cluster. For that reason I'm mainly
interested in using jumpshot, anything else I may have to do myself. 


Thank you for your help,

Kilian