Arnoldi algorithm with PARPACK. More...

Collaboration diagram for Arnoldi algorithm using parpack:

Files
file	arna.f
	Set of subroutines to solve eigenvalue problem with Arnoldi algorithm using PARPACK/ARPACK.

file	arna_io.f
	Set of checkpointing routines for arna module.

Functions
subroutine	stepper_register ()
	Register Arnoldi ARPACK module. More...

subroutine	stepper_init ()
	Initilise Arnoldi ARPACK module. More...

logical function	stepper_is_initialised ()
	Check if module was initialised. More...

subroutine	stepper_vsolve
	Create Krylov space, get Ritz values and restart stepper phase. More...

subroutine	arna_esolve
	ARPACK postprocessing. More...

subroutine	arna_naupd
	Interface to pdnaupd. More...

subroutine	arna_rst_save
	Write restart files. More...

subroutine	arna_rst_read
	Read from checkpoints. More...

subroutine	arna_write_par (prefix)
	Write procesor independent data. More...

subroutine	arna_mfop
	Write procesor independent variables. More...

subroutine	arna_read_par (prefix)
	Read procesor independent data. More...

subroutine	arna_mfip
	Read procesor independent variables. More...

subroutine	arna_mfov (prefix)
	Write procesor dependent data (long vectors) More...

subroutine	arna_mfiv (prefix)
	Read procesor dependent data (long vectors) More...

subroutine	arna_mfosv (ioflds, nout, offs0, stride, strideB, ur1, ur2, ur3, vect)
	Write single Krylov vector to the file. More...

subroutine	arna_mfisv (iofldr, offs0, stride, strideB, ur1, ur2, ur3, vect)
	Read single Krylov vector from the file. More...

Detailed Description

Arnoldi algorithm with PARPACK.

Description of the method can be found in [4] [5]

Two possible modes are supported:

direct: A*x = lambda*x
inverse: A*x = lambda*M*x, M-mass matrix defining inner product

Inverse mode is preferred as it takes into account inner product Simulation with temperature or passive scalars has to be performed in inverse mode due to speciffic inner product.

ARPACK_DIRECT preprocessing switch between direct and inverse mode.

Attention: This submodule has its own checkpointing routines.

Todo:: This submodule contains binary tar file with modified (restart added) version of PARPACK. This binary file in git is not good idea. Repalce with separate repository for Parpack added by sub.module.

Todo:: Finish description. Add reference to ARPACK

Module interface:

Global interface list:

Interface provided to tstpr:

stepper_register; provided for Time stepper routines
stepper_init; provided for Time stepper routines
stepper_vsolve; provided for Time stepper routines

Global interface dependency:

Interface required:

multiple interfaces from Runtime parameters and Monitoring module modules
io_file_freeid, io_mfo_fname, io_mbyte_open and io_mbyte_close from Input/output module module

Module parameters:

Global parameter list:: Parameters provided by arna module (include file - ARNAD):

Varaible	Type	Runtime parameter	Description
arna_nkrl	integer	_arna:nkrl	Krylov space size
arna_negv	integer	_arna:negv	number of eigenvalues

Global parameter dependency:

Parameters required by power iteration module:

Checkpointing routines :
- _chkpt:readchkpt
- _chkpt:chkpFnumber
Nek5000 PARAM array and logical flags:
- IFHEAT, NSTEPS

Module parameter usage:: [_ARNA] # Runtime paramere section for Arnoldi ARPACK module

NKRL = 50 # Krylov space size

NEGV = 10 # Number of eigenvalues

Function Documentation

◆ arna_esolve()

subroutine arna_esolve

ARPACK postprocessing.

Definition at line 424 of file arna.f.

References copy(), io_file_freeid(), mntr_abort(), mntr_check_abort(), mntr_log(), mntr_logi(), and outpost2().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ arna_mfip()

subroutine arna_mfip

Read procesor independent variables.

Definition at line 420 of file arna_io.f.

References blank(), byte_read, byte_reverse, copy(), and mntr_abort().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ arna_mfisv()

subroutine arna_mfisv	(	integer	iofldr,
		integer*8	offs0,
		integer*8	stride,
		integer*8	strideB,
		real, dimension(lx1lx1lx1*lelt)	ur1,
		real, dimension(lx1lx1lx1*lelt)	ur2,
		real, dimension(lx1lx1lx1*lelt)	ur3,
		real, dimension(arna_ls)	vect
	)

Read single Krylov vector from the file.

Parameters

[in,out]	iofldr	Vector counter
[in]	offs0	global file offset (header +...)
[in]	stride	single vector length
[in]	strideB	space saved for processes with lower nid
[in]	ur1,ur2,ur3	input arrays
[out]	vect	Krylov vector

Remarks: This routine uses global scratch space SCRNS

Definition at line 819 of file arna_io.f.

References byte_set_view(), copy(), mfi_gets(), and mfi_getv().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ arna_mfiv()

subroutine arna_mfiv ( character*3 prefix )

Read procesor dependent data (long vectors)

Parameters

[in] prefix prefix

Remarks: This routine uses global scratch space SCRNS, SCRUZ

Definition at line 654 of file arna_io.f.

References arna_mfisv(), io_init(), io_mbyte_close(), io_mfo_fname(), mfi_prepare(), mntr_check_abort(), mntr_log(), mntr_logi(), and mntr_logr().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ arna_mfop()

subroutine arna_mfop

Write procesor independent variables.

Definition at line 131 of file arna_io.f.

References blank(), byte_write, and copy().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ arna_mfosv()

subroutine arna_mfosv	(	integer	ioflds,
		integer	nout,
		integer*8	offs0,
		integer*8	stride,
		integer*8	strideB,
		real, dimension(lxolxolxo*lelt)	ur1,
		real, dimension(lxolxolxo*lelt)	ur2,
		real, dimension(lxolxolxo*lelt)	ur3,
		real, dimension(arna_ls)	vect
	)

Write single Krylov vector to the file.

Parameters

[in,out]	ioflds	Vector counter
[in]	nout	local number of elements to write
[in]	offs0	global file offset (header +...)
[in]	stride	single vector length
[in]	strideB	space saved for processes with lower nid
[in]	ur1,ur2,ur3	output arrays
[in]	vect	Krylov vector