[Users] Problems running BBH example

[Roland Haas]

Hello Krishna, all,

Correction. This is happening on my brand new workstation with 

12th Gen Intel(R) Core(TM) i7-12700

https://en.wikipedia.org/wiki/List_of_Intel_Core_i7_processors#Golden_Cove_+_Gracemont_microarchitecture_(12th_generation)

with performance (16) and efficiency (4) logical cpus.

With this hwloc reports:

core_depth 5
num_cores 12
pu_depth 6
num_pus 20

and the assert num_pus % num_cores fails (b/c only the 8 performance
cores have hyperthreads).

So an newly developed issue in systemtopology's assumptions due to
changing hardware, really.

I have created a ticket for this
(https://bitbucket.org/einsteintoolkit/tickets/issues/2756)
since this needs to be fixed since one would expect the new Intel CPUs
with mixed cores to become more common at least on laptops and
workstations.

Yours,
Roland

> Hello Krishna,
> 
> Turns out I am getting a similar issue myself now when running on my
> (old, well tested) laptop:
> 
> cactus_sim: configs/sim/build/SystemTopology/system_topology.cc:471:
> void {anonymous}::set_bindings(hwloc_topology_t, const
> mpi_host_mapping_t&): Assertion `num_pus % num_cores == 0' failed.
> 
> so something may have changed in hwloc even for standard machines. I
> will see if this is something that can be fixed in code.
> 
> Yours,
> Roland
> 
> > Hello Krishna,
> > 
> > The abort is triggered by the SystemTopology / HWLOC thorns having
> > encountered a processor / core / PU combination that they did not
> > expect. Is there anything in the stdout file?
> > 
> > I am not quite familiar enough with hwloc in particular to really
> > make a good suggestion.
> > 
> > You can certainly try and remove 
> > 
> > SystemTopology
> > 
> > and 
> > 
> > hwloc 
> > 
> > from your parameter file's ActiveThorn line(s). This will avoid the
> > issue but may, depending on your cluster (mis-)configuration, yield
> > non-ideal speeds (basically you must ensure manually that there is
> > at most one thread per CPU core). Carpet has an option
> > 
> > Carpet::set_cpu_affinity = "yes"
> > 
> > that probably gives you almost all you would get from
> > SystemTopology so you can try that one instead.
> > 
> > Can you provide details on the cluster / CPU of the cluster?
> > 
> > Yours,
> > Roland
> >   
> > > Hello,
> > > 
> > > I've downloaded and configured the Einstein Toolkit on my desktop,
> > > but I'm having an issue when I try to run the BBH example provided
> > > here
> > > <https://urldefense.com/v3/__https://einsteintoolkit.org/gallery/bbh/index.html__;!!DZ3fjg!-XFQ4fosXbEkKSRrQrEdwsGzgJCJQRkES951alRCNP9y0etbO9hlBmHws6LOl8wf57_50ErWvtxsYc05tAcu$
> > >    
> > > > and the BNS simulation from here
> > > > <https://urldefense.com/v3/__https://einsteintoolkit.org/gallery/bns/index.html__;!!DZ3fjg!-XFQ4fosXbEkKSRrQrEdwsGzgJCJQRkES951alRCNP9y0etbO9hlBmHws6LOl8wf57_50ErWvtxsYWDkk_Pq$
> > > >    
> > > > >. The job finishes in an instant, producing the attached error
> > > > >file. (I'm using the    
> > > command "simfactory/bin/sim create-submit GW150914_28 --define N
> > > 28 --parfile par/GW150914/GW150914.rpar --procs 24 --walltime
> > > 24:00:00")
> > > 
> > > There seems to be an issue with the 'num_pus' to 'num_cores'
> > > ratio in the "system_topology.cc" file, but the error oddly
> > > persists even when I manually equate them, or cut out line 471
> > > altogether. This error does not appear when I run the tov_ET
> > > example or the HelloWorld one. I'm also able to run this mini BNS
> > > tutorial
> > > <https://urldefense.com/v3/__https://www.youtube.com/watch?v=3euS5yMBFmI__;!!DZ3fjg!-XFQ4fosXbEkKSRrQrEdwsGzgJCJQRkES951alRCNP9y0etbO9hlBmHws6LOl8wf57_50ErWvtxsYZtrb0UG$
> > >    
> > > > without any issues, so I'm a bit baffled. I've attached my .ini
> > > > file and the .cc file along with the error file in case it
> > > > helps.    
> > > Please take a look and let me know if I'm missing something
> > > obvious.
> > > 
> > > PS: My PC has 128GB of memory, a 13900K(24C/32T) processor and
> > > 1TB of storage.
> > > 
> > > Regards,
> > > Praveer    
> > 
> > 
> >   
> 
> 
> 



-- 
My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from http://keys.gnupg.net.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://lists.einsteintoolkit.org/pipermail/users/attachments/20230903/c57a8a90/attachment.sig>