FreeOrbital SPOSet #4084

Paul-St-Young · 2022-06-28T23:43:06Z

Proposed changes

This PR consolidates the ElectronGas folder into a single SPOSet object "FreeParticle".
This merges real and complex code paths into one file.

While mostly a refactor, this PR has the side effect of enabling simulation of the 2D electron gas, because it reuses KContainer, where ndim was recently introduced #3852.

I updated all the unit tests. One test needed to be dropped: "det_heg.rpa.rs5.xml" in the real build.
The dropped test uses twisted boundary condition, which the real code was never meant to support.

What type(s) of changes does this code introduce?

Refactoring (no functional changes, no api changes)

Does this introduce a breaking change?

Yes

Electron gas with twisted boundary condition can no longer be executed by the real code.
One must compile the complex code to use a twist.

What systems has this change been tested on?

workstation

Checklist

Yes. This PR is up to date with current the current state of 'develop'
No. Code added or changed in the PR has been clang-formatted
Yes. This PR adds tests to cover any new code, or to catch a bug that is being fixed
Yes. Documentation has been added (if appropriate)

also K -> kvecs

ye-luo · 2022-06-28T23:49:30Z

Test this please

ye-luo · 2022-06-28T23:50:48Z

Replace ElectronGas with something without ifdef is long wanted.

prckent · 2022-06-29T00:16:09Z

Thanks.

An easy to address point:

Why not retain the naming of ElectronGas or similar? Do you have a more general non-electron gas usage despite all the electron gas plumbing and formulae?

Repurposing "Particle" here is not ideal because we already have things like ParticleSets which refer to very different things. We are trying to remove these sources of confusion from the codebase.

prckent · 2022-06-29T00:19:07Z

I would like to see deterministic tests added for this code, with very good coverage obtained. This should be easy to achieve given the existing stochastic tests. (On my phone, so apologies if I missed them)

Paul-St-Young · 2022-06-29T00:38:57Z

Why not retain the naming of ElectronGas or similar?

Because the single-particle orbitals for the electron gas can be free, localized (Wigner crystal), or something else.
Do you prefer something like ElectronGasFree, ElectronGasWigner etc?

I would like to see deterministic tests added for this code

Can you give more detail on the kind of tests you are looking for?
Do you mean different particle numbers, different cell shapes, or something else?

Paul-St-Young · 2022-06-29T00:49:00Z

Now I remember: the free orbitals can be used in a simulation with boson or boltzmann statistics.
That's why I did not want to add "ElectronGas" in front of "Free".

prckent · 2022-06-29T15:37:35Z

Paul: the deterministic tests I was referring to use a fixed random seed and a very short run to enable fully deterministic and reproducibile results. Typically they are just a few moves, just enough to hit different code paths. These tests don't demonstrate correctness but do demonstrate consistency. We run them in the CI and recommend users run all the deterministic tests when installing qmcpack (ctest -L deterministic). They are the only way we have found to reliably keep features alive, e.g. ensuring that someone doesn't accidentally break this code . If we have a deterministic test then it will fail in CI and the attempt to break this code in future will fail. We also run sanitizers with the deterministic tests to catch memory leaks etc.

The standard way to make these is to adopt a short test. See e.g. https://github.com/QMCPACK/qmcpack/blob/develop/tests/solids/diamondC_1x1x1_pp/det_qmc_short.in.xml and https://github.com/QMCPACK/qmcpack/blob/develop/tests/solids/diamondC_1x1x1_pp/CMakeLists.txt .The former has

<?xml version="1.0"?>
<simulation>
   <project id="det_qmc_short" series="0">
      <application name="qmcapp" role="molecu" class="serial" version="1.0"/>
   </project>
   <random seed="71"/>

   <qmc method="vmc" move="pbyp">
      <estimator name="LocalEnergy" hdf5="no"/>
      <parameter name="walkers"             >    1               </parameter>
      <parameter name="blocks"              >    3               </parameter>
      <parameter name="steps"               >    3               </parameter>
      <parameter name="subSteps"            >    2               </parameter>
      <parameter name="timestep"            >    0.3             </parameter>
      <parameter name="warmupSteps"         >    3               </parameter>
   </qmc>

prckent · 2022-06-29T15:42:54Z

Some good news is that I was getting ahead of the actual situation - we do have some deterministic tests for the heg, the question is whether all the code paths are covered or not with these changes.

ye-luo · 2022-06-29T15:48:35Z

How about FreeParticleOrbitals?

prckent · 2022-06-29T15:55:10Z

Good news, coverage is actually still very decent :-) https://app.codecov.io/gh/QMCPACK/qmcpack/compare/4084/tree/src/QMCWaveFunctions/ElectronGas (You might need to authenticate with your GitHub ID). For simplicity I suggest we put this item on the list of things to verify after this PR is in. The goal would be to craft an extra test or remove some unused legacy code to get the coverage >>80%. If there are vestigial functions from some old usage these should definitely go.

ye-luo

Seems only a few requested change got missed. Very close to merge.

ye-luo · 2022-07-05T03:16:11Z

src/Particle/LongRange/KContainer.h

+  void updateKLists(const ParticleLayout& lattice,
+                    RealType kc,
+                    unsigned ndim,
+                    const PosType& twist = PosType(),


Still missing documentation for twist. Say something about how twist affects KLists.

ye-luo · 2022-07-05T03:16:31Z

src/Particle/LongRange/KContainer.h

@@ -81,7 +81,7 @@ class KContainer : public QMCTraits
   */
  void findApproxMMax(const ParticleLayout& lattice, unsigned ndim);
  /** construct the container for k-vectors */
-  void BuildKLists(const ParticleLayout& lattice, bool useSphere);
+  void BuildKLists(const ParticleLayout& lattice, PosType twist, bool useSphere);


@Paul-St-Young did you miss this one?

ye-luo · 2022-07-05T03:17:07Z

src/Particle/LongRange/StructFact.h

@@ -71,6 +70,9 @@ class StructFact : public QMCTraits
  /// accessor of StorePerParticle
  bool isStorePerParticle() const { return StorePerParticle; }

+  /// accessor of k_lists_
+  const KContainer getKLists() const { return k_lists_; }


Still incorrect. "const KContainer&"

ye-luo · 2022-07-05T03:20:43Z

Test this please

ye-luo · 2022-07-05T03:33:09Z

src/Particle/LongRange/tests/test_kcontainer.cpp

+    for (int ldim=0;ldim<ndim;ldim++)
+    {
+      CHECK(klists.kpts[ik][ldim] == gvecs[ik][ldim]);
+      CHECK(klists.kpts[ik][ldim]*blat == klists.kpts_cart[ik][ldim]);


CI failure are from missing Approx. Please double check all the CHECK. Also this file seems needing clang-format.

ye-luo · 2022-07-05T03:34:00Z

src/QMCWaveFunctions/ElectronGas/FreeOrbital.cpp

+          ((ggg[j2])[la])(lb, la) = -coskr * (kvecs[ik])[la] * (kvecs[ik])[lb] * (kvecs[ik])[la];
+          ((ggg[j1])[la])(la, lb) = ((ggg[j1])[la])(lb, la);
+          ((ggg[j2])[la])(la, lb) = ((ggg[j2])[la])(lb, la);
+          ((ggg[j1])[lb])(la, la) = ((ggg[j1])[la])(lb, la);


My eyes hurt with excessive () around hess and ggg

ye-luo · 2022-07-05T13:52:20Z

Test this please

Paul-St-Young · 2022-07-05T13:52:29Z

@ye-luo sorry I missed some of your requests. They should all be addressed now.

Paul-St-Young · 2022-07-05T13:53:50Z

@PDoakORNL unit tests for the KContainer have been added.
Please see if this is satisfactory now.

ye-luo · 2022-07-05T14:18:46Z

Test this please

prckent

(deleted comments)

Wrong PR

prckent

This now LGTM assuming that all the tests pass. @ye-luo should still sign off. Thanks for all the updates.

prckent · 2022-07-05T18:12:03Z

Test this please

Paul-St-Young added 9 commits June 28, 2022 14:40

add free particle orbitals

7335258

update heg tests

e58fb64

transfer old header

9e4a4d7

also K -> kvecs

remove ElectronGasOrbital and its Complex

28bb034

occupy half sphere of kvecs in real code

b137c6b

add twist to KContainer

9af84e4

use twist in free-particle orbitals

406b5f1

cut down HEGGrid

c0eb5ce

cut HEG some more

a6c02e0

Paul-St-Young added the refactor label Jun 28, 2022

Paul-St-Young requested review from ye-luo and jtkrogel June 28, 2022 23:43

Paul-St-Young self-assigned this Jun 28, 2022

Paul-St-Young mentioned this pull request Jun 28, 2022

How to resurrect true 2D capability? #3868

Open

5 tasks

Paul-St-Young added 4 commits June 28, 2022 19:58

improve backflow tests

c6d95e6

mark makeClone as override

fa6aacd

document free SPOSet

d83e24e

override createSPOSetFromXML

a7f2c18

Paul-St-Young added 2 commits June 28, 2022 20:31

fix cmake for heg test

b1ec05e

remove twist test in real code

53318ec

Paul-St-Young added 2 commits July 3, 2022 15:46

run KContainer tests

60df854

clang format

199bb85

ye-luo reviewed Jul 5, 2022

View reviewed changes

Paul-St-Young added 6 commits July 5, 2022 08:37

getter return const reference

bbde8ff

fix missing Approx

98ad010

fewer parenthesis

16ebc8b

add license header

ad592eb

Ewald2D license headers

bf73840

clang format

fc38389

ye-luo previously approved these changes Jul 5, 2022

View reviewed changes

one more missed Approx

f4ed10d

Paul-St-Young dismissed ye-luo’s stale review via f4ed10d July 5, 2022 14:11

prckent previously approved these changes Jul 5, 2022

View reviewed changes

prckent self-requested a review July 5, 2022 14:41

prckent approved these changes Jul 5, 2022

View reviewed changes

ye-luo approved these changes Jul 5, 2022

View reviewed changes

PDoakORNL approved these changes Jul 5, 2022

View reviewed changes

Merge branch 'develop' into fporb

2ee528c

prckent enabled auto-merge July 5, 2022 18:12

prckent merged commit 59021cf into QMCPACK:develop Jul 5, 2022

Paul-St-Young deleted the fporb branch July 5, 2022 20:32

Paul-St-Young mentioned this pull request Oct 21, 2022

make two-body Jastrow work in 2D #4289

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FreeOrbital SPOSet #4084

FreeOrbital SPOSet #4084

Paul-St-Young commented Jun 28, 2022 •

edited

Loading

ye-luo commented Jun 28, 2022

ye-luo commented Jun 28, 2022

prckent commented Jun 29, 2022

prckent commented Jun 29, 2022

Paul-St-Young commented Jun 29, 2022 •

edited

Loading

Paul-St-Young commented Jun 29, 2022

prckent commented Jun 29, 2022

prckent commented Jun 29, 2022

ye-luo commented Jun 29, 2022

prckent commented Jun 29, 2022

ye-luo left a comment

ye-luo Jul 5, 2022

ye-luo Jul 5, 2022

ye-luo Jul 5, 2022

ye-luo commented Jul 5, 2022

ye-luo Jul 5, 2022

ye-luo Jul 5, 2022

ye-luo commented Jul 5, 2022

Paul-St-Young commented Jul 5, 2022

Paul-St-Young commented Jul 5, 2022

ye-luo commented Jul 5, 2022

prckent left a comment •

edited

Loading

prckent left a comment

prckent commented Jul 5, 2022

FreeOrbital SPOSet #4084

FreeOrbital SPOSet #4084

Conversation

Paul-St-Young commented Jun 28, 2022 • edited Loading

Proposed changes

What type(s) of changes does this code introduce?

Does this introduce a breaking change?

What systems has this change been tested on?

Checklist

ye-luo commented Jun 28, 2022

ye-luo commented Jun 28, 2022

prckent commented Jun 29, 2022

prckent commented Jun 29, 2022

Paul-St-Young commented Jun 29, 2022 • edited Loading

Paul-St-Young commented Jun 29, 2022

prckent commented Jun 29, 2022

prckent commented Jun 29, 2022

ye-luo commented Jun 29, 2022

prckent commented Jun 29, 2022

ye-luo left a comment

Choose a reason for hiding this comment

ye-luo Jul 5, 2022

Choose a reason for hiding this comment

ye-luo Jul 5, 2022

Choose a reason for hiding this comment

ye-luo Jul 5, 2022

Choose a reason for hiding this comment

ye-luo commented Jul 5, 2022

ye-luo Jul 5, 2022

Choose a reason for hiding this comment

ye-luo Jul 5, 2022

Choose a reason for hiding this comment

ye-luo commented Jul 5, 2022

Paul-St-Young commented Jul 5, 2022

Paul-St-Young commented Jul 5, 2022

ye-luo commented Jul 5, 2022

prckent left a comment • edited Loading

Choose a reason for hiding this comment

prckent left a comment

Choose a reason for hiding this comment

prckent commented Jul 5, 2022

Paul-St-Young commented Jun 28, 2022 •

edited

Loading

Paul-St-Young commented Jun 29, 2022 •

edited

Loading

prckent left a comment •

edited

Loading