Refurbish correlator online measurement #169

Open
urbach opened this Issue Oct 11, 2012 · 11 comments

Comments

Projects
None yet
3 participants
Contributor

urbach commented Oct 11, 2012

we need to refurbish this, because right now it will always use the pure TMWilson operator for the correlator computation. This is not what we need.

Contributor

urbach commented Oct 12, 2012

we could use the operator struct we anyhow use for invert also in hmc_tm for the online measurements. One would therefore need to specify the operators one would like to have inverted and measured. This would then have effect only for the correlator online measurement, of course.

Owner

kostrzewa commented Nov 6, 2012

Maybe we could also make it write into a single file with a header for each write and a double newline between them? I had to modify the code because the lustre file system in Zeuthen couldn't deal with the many files generated during my high statistics run.

Contributor

urbach commented Dec 9, 2012

in principle we could do that, however, thats always a mess if you restart jobs and redo trajectories. Maybe we could implement the choice?

Owner

kostrzewa commented Dec 12, 2012

I agree with you, although in the current situation the default would be to overwrite data with the same trajectory number, no?

Owner

kostrzewa commented Dec 12, 2012

If we're happy with your changes I will close this, I guess we will know tomorrow.

palao commented Dec 13, 2012

Hi,
Thank you for solving this issue.
As we said in the phone conference I can test it in Fermi (online versus
offline correlators).
But it was not clear to me what do I have to do in order to get the code.
My working copy of the code is cloned from Bartek's:
$ git branch

  • interleaving_hs_mergeNDTwistedClover
    master
    $ git log -n 1
    commit fde90345e53366b0b1ee20796c002a4027040908
    Merge: 8b0505e 1896db1
    Author: Bartosz Kostrzewa bartosz.kostrzewa@desy.de
    Date: Wed Dec 5 17:45:55 2012 +0100

    Merge remote-tracking branch 'urbach/NDTwistedClover' into

    interleaving_hs_mergeNDTwistedClover

    Does the following command bring me the update?
    $ git pull origin interleaving_hs_mergeNDTwistedClover

Thanks in advance and best regards,

David

On Wed, Dec 12, 2012 at 4:39 PM, Bartosz Kostrzewa <notifications@github.com

wrote:

If we're happy with your changes I will close this, I guess we will know
tomorrow.


Reply to this email directly or view it on GitHubhttps://github.com/etmc/tmLQCD/issues/169#issuecomment-11293803.

Owner

kostrzewa commented Dec 13, 2012

On 13/12/12 16:39, palao wrote:

Hi,

Hi David,

Does the following command bring me the update?
$ git pull origin interleaving_hs_mergeNDTwistedClover

yes, that should bring you up to date!

All the best,
Bartek

Owner

kostrzewa commented Dec 14, 2012

Oh, please note that the debug levels have been adjusted. At level 1 you now have only basic output such as total iteration count. At level 2 you have timings for many common function calls (derivatives and such) and at level 3 there is output for every CG iteration. Effectively what used to be level 2 has become level 3.

Owner

kostrzewa commented Dec 14, 2012

Also note that on BG/Q using level 2 or 3 seems to make the code quite a bit slower, probably because the master process is already quite busy and having it produce output just causes everyone else to wait excessively.

Contributor

urbach commented Jan 26, 2013

the bug is fixed, however, maybe we should also allow to measure the kaon mass online!?

Owner

kostrzewa commented Jun 4, 2013

the bug is fixed, however, maybe we should also allow to measure the kaon mass online!?

You mean in the unitary case for 2+1(+1) runs?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment