GitLab

and my_prow (col) values

whether cannon algorithm can be used. If not, switch back to the
original implementation

and passing NULL from the api

on selected mpi ranks using suppress_warnings option

tile size is used in elpa1_tridiag and elpa2_bandred.
Iits value is computed based on block size and least common multiple of
number of processes rows and colums, but then can be enlarged to some
multiple of this value. How much it should be enlarged can now be
influenced by min_tile_size parameter.

instead, blacs context has to be set beforehand

For these options there has still work to be done
before they can be tuned

To be used like this

   class(elpa_t), pointer      :: e
   class(elpa_autotune_t), pointer :: tune_state

   e => elpa_allocate()
   call e%set(...)
   [...]
   assert_elpa_ok(e%setup())

   tune_state => e%autotune_setup(ELPA_AUTOTUNE_FAST, ELPA_AUTOTUNE_DOMAIN_REAL)

   ! Autotuning loop, continues until all combinations have been tried
   do while (e%autotune_step(tune_state))
     ! Do the steps that are representative of your calculation
     call e%eigenvectors(a, ev, z, error)
   end do

   ! Fix best parameters, and de-allocate the autotune object
   call e%autotune_set_best(tune_state)
   call elpa_autotune_deallocate(tune_state)

The blocking factor can be set now with the new interface
set method. The default is 3

Autotuning still has to be implemented.

This closses issue #44

The blocking factor can be set now with the new interface
set method. The default is 3

Autotuning still has to be implemented.

This closses issue #44

Previously you could set up ELPA with nev > na and force a crash.

These can be used internally by

  call self%set_private(name, value)

   or

  call self%get_private(name, value)

but are not valid for the self%set()/get() routines, i.e. are not
visible to users of ELPA.

Since there are at the moment no double-valued entries, it is better to
abort early than to try to walk the zero-length array. It might even
work, but better safe than sorry.

This is a bit cumbersome, as it involves yet another abstract
intermediate type, but necessary due to language limitations.

Setting a bandwdith of the input matrix via

elpa%set("bandwidth",bw)

tells ELPA 2stage, that matrix is already a banded matrix
with bandwidth bw. Then some compute steps can be skipped