[Hadstruct] Minutes of today's call...

[David Richards]

Present

AR,BJ,CC,CM,CK,CE,ER,Guest?,JK,RS,RE,SZ,TK,WM,KO. Apologies: JQ

<https://github.com/JeffersonLab/HadStruc/wiki/Minutes-of-13th-May,-2020#phasing>Phasing

CE and DR described progress at understanding the phasing, and in particular what strategy we should adopt for possible source-sink separation, notably that we only compute for values where t_src - t_sink had decent (to be defined) signal in 2pt. RE noted we should look at principle correlators (which depend on only one time slice) rather than effective mass. CE also noted we adopted a fitting criterion based purely on the ground-state principle correlator.

<https://github.com/JeffersonLab/HadStruc/wiki/Minutes-of-13th-May,-2020#gauge-generation>Gauge generation.

JK gave an update on gauge generation, with the aim of getting sufficient statistics over a large range of Ioffe time, using a machine in CH. TK has been using Bridges for the two-point functions, with perambulators computed on GPUs at JLab. There was discussion of the autocorrelation, both in time slices and between successive configurations. So far, we have been computing 2pt (and 3pt) functions for every source time slice, but this may not be the best strategy going forward given the cost of "measurement" vs gauge generation. The former might be the easiest to accomplish. KO asked whether we could use an SVD on the perambulators, given we have all-to-all perambulators. TK noted we were only computing 16 sink values.

<https://github.com/JeffersonLab/HadStruc/wiki/Minutes-of-13th-May,-2020#two-point-functions>Two-point functions

On bridges, one time source for 16 operators/2 rows and 16 time slices takes around 2.25 hours on a single node, per momenta, on bridges which has 28 cores. Total is around 1.5M core-hours per momentum for the 350 configs so far analyses. We need to make sure that we are using the latest update for colorvec, including the saxby (?) constructions.

<https://github.com/JeffersonLab/HadStruc/wiki/Minutes-of-13th-May,-2020#running>Running

GPDs were running on Frontera with 19 (non-zero) momentum transfers. RE noted we had used 6 percent of our allocation that begin 1st April, so we are a little behind. ER was also running on Frontera on 48^3 lattices.

<https://github.com/JeffersonLab/HadStruc/wiki/Minutes-of-13th-May,-2020#incite>INCITE

DR noted that we had been approached by a couple of groups re. INCITE proposal, include SUNY, UKY and a (subset of) BNL.

<https://github.com/JeffersonLab/HadStruc/wiki/Minutes-of-13th-May,-2020#next-meeting>Next Meeting
20th May, 2020

<https://github.com/JeffersonLab/HadStruc/wiki/_new?wiki%5Bname%5D=_Footer>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/hadstruct/attachments/20200513/96afb5a2/attachment.html>