Table ideas (related to zoom meeting 07/17)

callat-qcd / lattedb

Lattice QCD database interface using EspressoDB as the content manager.

https://ithems.lbl.gov/lattedb

BSD 3-Clause "New" or "Revised" License

1 stars 0 forks source link

Table ideas (related to zoom meeting 07/17) #16

Closed ckoerber closed 5 years ago

ckoerber commented 5 years ago

Table by point of the meeting

tables.pdf

Table related points

We want basic building blocks

Factorize solver and actions
ensemble instead of gaugeconfig
valence action and seaquark action (pulled out from propagator)
Where to put run / algorithm parameters
Split physics from computation and join them in a status like table

Wanted Help for computational tables

Useful to implement for production

Data summary
Job generation
"give me all the propagators for light quark masses"

erinaldi commented 5 years ago

I might have miss it but I didn't see a HISQ correlator unless it is included in the meson2pt but then a mixed meson would also be nice. I see a DWFTuning but I don't understand the motivation

cchang5 commented 5 years ago

The correlator tables have a key pointing to a propagator table. The propagator table aggregates all types of propagators, so a meson2pt can store any variant of a two point correlator with 2 quark lines. DWTuning is basically a meson2pt also, but with a column that specifics if the sink is on the domain wall, or in the center of L5. Technically you can add this column to meson2pt, and then specify that every correlator is calculated on the domain-wall, except for like... very few. While that is probably more coherent? that sounds like a pain in practice.

erinaldi commented 5 years ago

I think a better name for DWTuning would be DWF5D

cchang5 commented 5 years ago

One idea is to make a meson2pt5d table that inherits from meson2pt actually, but with an extra l5 column. That might be better than making a whole new table.

Let me be more specific. This will make a new table with an extra l5 column but at the class definition it will be a child.

ckoerber commented 5 years ago

After a brief discussion with @arjunqcd and @ebatz, here are a few random remarks:

Propagator remarks

What is correlator momentum? Is at source or sink? Is it the momentum transfer?
Should mval be in action? Change it to flavor?
Is the smearing part of the propagator or the source/sink? If code wise, the (source) smearing information is encoded in the propagator, it might be more useful to store it here as in the hadron linker

Action remarks

Quark masses in action? a_fm is strongly correlated with beta.

ckoerber commented 5 years ago

For further discussion please continue in issue #20.