Merge release (#398)

* Fix EnergyAndForces tests (#277)

* have them work in debug mode too

* Move factor 4pi out og linear solvers (#278)

* Move some code into PoissonSolverFactory (#279)

* Clean up class Potentials (#280)

* Clean up class Ions, add test for it (#281)

* Add test MD_MVP (#290)

* Clean up code related to DM restart data (#292)

* Write dm (#291)

* Update use of DM in restart

* Remove unused function in Control (#294)

* Change symlink to restart in tests (#295)

* enable tests when old link present

* Extract number empty orbitals from restart file (#296)

* Clean up MD_IonicStepper restart data write (#297)

* Clean up code based on compiler warnings (#299)

* Add getForces for evaluateDMandEnergyAndForces (#300)

* Add check for compatibility MVP/Mehrstellen (#301)

* Add support for Br atom (#302)

* Clean up some HDFrestart functions (#303)

* Fix and test restart single hdf5 file (#305)

* Save Hartree potential for write in restart file (#306)

* Speed-up recently added tests (#309)

* simply use a smaller domain/mesh

* Encapsulate some functions in Potentials (#310)

* some functions were called at random places and confusing

* Remove confusing 0 in naming restart files (#308)

* use added integer only in case of fail/retry

* Add functionalities for extra info in restart file (#312)

* MPI abort (#313)

* Fix error code used by MPI_Abort

* Rho and VH restart  (#311)

* enable restart with consistent rho and VHartree

---------

Co-authored-by: Seung Whan Chung <seung.chung@austin.utexas.edu>

* Strenghten testIons (#315)

* More clean up in class Potentials (#316)

* remove some unused functions
* make use of class MGmol_MPI
* change some function names for clarity
* fix some minor bug in iterative indexes

* Misc code fixes (#317)

* exposed by porting on new environment

* Use unique restart filenames in test HDF5single (#318)

* Swap ions in existing test (#319)

* strengthen testing by swapping ions in testRestartEnergyAndForces

* Code clean up (#320)

* use more const
* use initNuc() instead of moveVnuc()
* remove incorrect assert
* use MGmol_MPI in more cases
* setup g_kbpsi_ inside initNuc()
* rename initNuc() into setupPotentials()

* Added functionalities to set local forces (#321)

* Update mixed precision code (#322)

* enable cmake build with mixed precision
* enable use of BLIS
* isolate blas3 loop rewrite in separate files
* fix misc issues with mixed precision code

* Clean up and fixes Ions (#326)

* reset static index counter ion Ion when needed
* simplify some loops
* make Ions copy constructor private

* Fixes for build without HDF5P (#324)

* Fix testRhoVhRestart (#325)

* Fix a few more issues with class Ions (#328)

* Introduce new constructor for class Ions (#330)

* modify various functions to use argument of type Ions instead of MGmol::ions_

* Add cleanup and MGmolInterface (#331)

* Add function to set local forces (#333)

* based on matching coordinates to local ions

* Fix setLocalForces (#335)

* Fix test WFEnergyAndForces (#336)

* writing restart twice in same directory was failing

* Add mixing option in MVP (#337)

* Update MVP with mixing (#339)

* use tolerance to terminate iterations
* tune verbosity

* Update DensityMatrix class (#342)

* add extrapolation function
* remove unused function
* simplify use of iterative index

* Fixing DM tolerance (#341)

* Fix Density Matrix tolerance read

---------

Co-authored-by: Jean-Luc Fattebert <fattebertj@ornl.gov>

* Rm oversubscribe from CMakeLists.txt (#346)

* already set in build script

* Adapt to new environment on condo @ ORNL (#348)

* Remove unused function

* Clean up use of coords file when restarting

* use coordinates file to override atomic coordinates when using
  restart wavefunctions

* Add test restart with override atomic coordinates

* Add some missing include files

* required for gcc12

* Print MVP timers

* Fix initialization for Dirichelt BC

* was not correct for other than Gaussian functions initialization

* Modify two utils scripts

* Link tests with Lapack

* Fix mixed-precision to use ps-projectors in double

* Template axpy functions to better use BLAS1

* Replace Li PBE potential with ONCV

* Tune 2 pythons scripts for better looking output

* Store wavefunctions in double in KB objects

* this allows to used optimized dot functions in mixed precision

* Remove two unused functions

* Simplify and remove unused code

* Script to get matrix of wavefunctions from hdf5

* Opimize code a little bit

* Rm unused/duplicated functions in LocGridOrbitals

* Replace some ptr with vector in HDFrestart

* Replicated Projected Matrices on CPU (#367)

* Implement option of using ReplicatedMatrix on CPU

* Add timers for handwritten linear algebra loops

* Skip data reset before assigning arrays w ghosts
* costly, and not necessary since data is assigned right after

* Add timers in BlockVector class

* Optimize preconditioner implementation

* conversion between data type and storage type was taking a substantial
  amount of time

* Differentiate timers in GridFuncVector based on datatype

* Rm redundant function

* Do not use syrk anymore

* make double precision consistent with mixed-precision

* Nosubdivx extended (#376)

* Assume subdivx=1 in ExtendedGridOrbitals

* Fix several compiler warnings (#377)

* Fix several compiler warnings

* Fix some stdout content (#378)

* Fix header of asci files generated by read_hdf5 (#379)

* Tune timers (#380)

* Move Hloc Timer to not include ghosts value filling
* Remove double counted Timer in KBPsiMatrixSparse
* Remove unused code in KBPsiMatrixSparse

* Pass in Hamiltonian object to functions that need it (#381)

* Template preconditioner (#382)

* Use more shared_ptr in OrbitalsPreconditioning

* Template OrbitalsPreconditioning on P datatype

* Use local Hphi in Davidson (#383)

* Residual compute (#384)

* Refactor H application and R computation

* Rm optional args to Hamiltonian::addHlocal2matrix()

* Fix bug in preconditioner setup (#385)

* affecting Davidson only

* Fix potential iterative index (#386)

* was failing in some assert

* Refactor Hartree_CG (#387)

* Template Hartree_CG on preconditioner data type

* Enable runtime precision selection for Hatree_CG preconditioner

* Preconditioning precision (#388)

* Redesign OrbitalsPreconditioning

* Make orbitals preconditioner precision a runtime option

* Template manage_memory on datatype (#389)

* New DotProduct family of classes (#390)

* Template GridOrbitals classes (#391)

* Add test for MVP ReplicatedMatrix (#392)

* Some fixes in Hamiltonian and its usage in MVP (#393)

* Avoid recompute complete H*Phi in MVP and Davidson (#394)

* Use delta potential in Davidson and MVP

* Bug fix in Poisson PCG Mehrstellen r.h.s. (#396)

* Add Convergence tolerance option for Poisson problem and some minor cleanup. (#395)

* Add input option for PCG convergence tolerance.

* Change default options for PCG

* Renamed CG to PCG in input files.

* Updated build script for LC - build.pel

* Resolve errors

* Minor change to scripts

* Minor fix in names

---------

Co-authored-by: Jean-Luc Fattebert <fattebertj@ornl.gov>
Co-authored-by: Seung Whan Chung <seung.chung@austin.utexas.edu>
Co-authored-by: Daniel Osei-Kuffuor <oseikuffuor1@llnl.gov>

Author: 4 people

CMakeLists.txt

@@ -151,7 +151,7 @@ if(USE_LIBROM)
   find_package(libROM REQUIRED)
 
   if(libROM_FOUND)
-    set(MGMOL_HAS_LIBROM 1)
+    add_definitions(-DMGMOL_HAS_LIBROM)
   endif(libROM_FOUND)
 endif(USE_LIBROM)
 
@@ -233,6 +233,7 @@ FortranCInterface_HEADER(
         DGEMM SGEMM DSYMM DSYRK SSYRK DTRMM DTRSM STRSM
         #lapack
         DSYEV DPOTRF DPOTRS DPOTRI DSYGST DTRTRS DPOCON DSYGV DLANGE
+        DGETRF DGETRS DLACPY
 )
 
 FortranCInterface_HEADER(

drivers/check_input.cc

@@ -70,11 +70,13 @@ int main(int argc, char** argv)
     {
         MGmolInterface* mgmol;
         if (ct.isLocMode())
-            mgmol = new MGmol<LocGridOrbitals>(global_comm, *MPIdata::sout,
-                input_filename, lrs_filename, constraints_filename);
+            mgmol = new MGmol<LocGridOrbitals<ORBDTYPE>>(global_comm,
+                *MPIdata::sout, input_filename, lrs_filename,
+                constraints_filename);
         else
-            mgmol = new MGmol<ExtendedGridOrbitals>(global_comm, *MPIdata::sout,
-                input_filename, lrs_filename, constraints_filename);
+            mgmol = new MGmol<ExtendedGridOrbitals<ORBDTYPE>>(global_comm,
+                *MPIdata::sout, input_filename, lrs_filename,
+                constraints_filename);
 
         *MPIdata::sout << " Input parameters OK\n";
 

drivers/example1.cc

@@ -87,11 +87,13 @@ int main(int argc, char** argv)
 
         MGmolInterface* mgmol;
         if (ct.isLocMode())
-            mgmol = new MGmol<LocGridOrbitals>(global_comm, *MPIdata::sout,
-                input_filename, lrs_filename, constraints_filename);
+            mgmol = new MGmol<LocGridOrbitals<ORBDTYPE>>(global_comm,
+                *MPIdata::sout, input_filename, lrs_filename,
+                constraints_filename);
         else
-            mgmol = new MGmol<ExtendedGridOrbitals>(global_comm, *MPIdata::sout,
-                input_filename, lrs_filename, constraints_filename);
+            mgmol = new MGmol<ExtendedGridOrbitals<ORBDTYPE>>(global_comm,
+                *MPIdata::sout, input_filename, lrs_filename,
+                constraints_filename);
 
         if (MPIdata::onpe0)
         {

examples/AOMM/Si216.cfg

@@ -15,7 +15,7 @@ lz=30.78
 [Potentials]
 pseudopotential=pseudo.Si
 [Poisson]
-solver=CG
+solver=PCG
 [Run]
 type=QUENCH
 [Quench]

examples/C200H272_1.05gcc/mgmol_md.cfg

@@ -16,7 +16,7 @@ lz=30.568
 pseudopotential=pseudo.C_pbe
 pseudopotential=pseudo.H_pbe
 [Poisson]
-solver=CG
+solver=PCG
 [Run]
 type=MD
 [MD]

examples/C200H272_1.05gcc/mgmol_opt.cfg

@@ -16,7 +16,7 @@ lz=30.568
 pseudopotential=pseudo.C_pbe
 pseudopotential=pseudo.H_pbe
 [Poisson]
-solver=CG
+solver=PCG
 [Run]
 type=MD
 [MD]

examples/C200H272_1.05gcc/mgmol_quench.cfg

@@ -16,7 +16,7 @@ lz=30.568
 pseudopotential=pseudo.C_pbe
 pseudopotential=pseudo.H_pbe
 [Poisson]
-solver=CG
+solver=PCG
 [Run]
 type=QUENCH
 [Quench]

examples/Cu32/mvp.cfg

@@ -15,7 +15,7 @@ lz=13.6
 [Potentials]
 pseudopotential=pseudo.Cu_ONCVPSP_LDA
 [Poisson]
-solver=CG
+solver=PCG
 [Run]
 type=QUENCH
 [Quench]

examples/D144localPotentials/mgmol_md.cfg

@@ -15,7 +15,7 @@ lz=21.2406
 [Potentials]
 pseudopotential=pseudo.H
 [Poisson]
-solver=CG
+solver=PCG
 [Run]
 type=MD
 [MD]

examples/D144localPotentials/mgmol_quench.cfg

@@ -15,7 +15,7 @@ lz=21.2406
 [Potentials]
 pseudopotential=pseudo.H
 [Poisson]
-solver=CG
+solver=PCG
 [Run]
 type=QUENCH
 [Quench]