GitLab

For some matrix/block size combinations the real case of ELPA2
crashes, e.g:

mpiexec -n 1 ./elpa2_test_real 50 50 32

leads to an error message
 ** On entry to DGEMM parameter number  3 had an illegal value
 and a crash.

This only seems to happen with matrix size smaller than 64*64.
he code path responsible for this has been identified, but the problem
tself is not yet solved!

The part of the code, which causes these crashes, has been switched on
as default by Intel in commit fe63372d. The rest of the commit fe63372d
seems to be fine, and is performance critical.

As an intermediate step, the responsible code path is switched off again
as default, this will be changed again once the underlying root cause
has been solved.

The examples, how to invoke ELPA from a c program have been updated.
There are now examples for ELPA1 and ELPA2 both real and complex case.
The test cases are still with less functionality than their Fortran
counter parts, they are just ment as a "proof-of-concept".

- enabling fusing iterations of stage 5 in ELPA2 for every configuration
- Changed reduction bandwidth in ELPA2 to be at least 64
- partial OpenMP parallelization of the QR factorization in bandred_real
- OpenMP parallelization of SYMM
- OpenMP parallelization of SYR2K in bandred_real
- OpenMP parallelization for elpa1_reduce_add_vectors and elpa1_transpose_vectors
- AVX2 support in backtransformation elpa2_kernels (FMA3 instructions introduced with Haswell microarchitecture)

complex cases

Create automatically two independent routines for real and complex
valued matrices

This commit is not ABI compatible

This commit is not ABI compatible, since it changes the interfaces
of some routines

Also, introduce type checking for transpose and reduce_add routines

Andreas Gloess informed us about a memory leak in ELPA, which was
introduced in version 2013.11.008.

This memory leak is removed now again.

Note, that older versions of ELPA will not be fixed right now.

In the test programms "MPI_ABORT" has been used incorrectly.

The blocksize should be intent(in) and not intent(inout), since
it should not be changed by solve_evlp_real_2stage

If the user chooeses parameters for the QR-decomposition which are
not allowed an error has been produced. This error is caught now,
and the library aborts with a message

It is now possible to switch on more debug messages via
the environment variable "ELPA_DEBUG_MESSAGES=yes"

Error in configure test program fixed

If the QR-decomposition is used wrongly (matrix size is not a
multiple of block size) the the execution will abort, in
order to prevent the wrong results, discussed in a previous commit

Debug messages are now available by setting the environment variable
"ELPA_DEBUG_MESSAGES" to "yes".

We found a bug in the QR-decomposition, which appears for
some matrix sizes and produces wrong results!

If the QR decomposition is switched on, an appropiate warning
is shown.

This bug is still under investigation

- cleanup of the file
- add more  (optional) timing information

The qr decomposition is now available as a runtime choice.
Some testing has still to be done

At build time it can be specified that the ELPA test
programs give more detailed timing information, which
is usefull for performace measurements

Build servers recommend against calling "abort" functions in a
library. Thus, from now on, ELPA  returns a success flag, which
is a change of ABI!

In order to force the user to honour this, the ELPA drivers
are converted to functions, which implies an adaptation and
rebuild of the user code. The driver functions return a
logical "success", which, if it is false, indicates an
error within the library.

It is advised to check catch and check this
return value in the user code, which calls ELPA.

- the library version number is changed due to change in ABI
- the INSTALL ascii file is updated
- a RELEASE_NOTES file is introduced

This includes fixes necessary to please gfortran as well as a slight
change in the way the various libraries are tested in configure.ac

Now it is possible
- to choose the kernel (real and complex independently) at run-time
  via environment variables, or
- to specify the kernel (real and complex independently) at runtime
  via specifing the kernel in the call to ELPA

This has a few implications
1) The ELPA 2014.06 release has a change in the API and is thus not
   binary compatible with previous versions
2) if no kernels are specified, a default kernel is choosen
3) if a wrong kernel is specified, a default kernel is choosen

For sake of simplicity it is still possible to build ELPA with
support for only one kernel, as in previous versions. However, it is
still not binary compatible to previous versions

Configure checks whether the Fortran environment module
is available. If yes, the library is build such, that all
messages (from within the library) are printed at the correct
stderr unit. If not, then the stderr unit is set to unit=6

The configure.ac of ELPA_2013.11 is cleaned up a bit
Also, ELPA_2013.11 is copied to ELPA_2014.06 in order to have
the base for the next ELPA release

By mistake the complex kernel with blocking 2 was not called
in parallel if OpenMP was used.

Due to an error in a preprocessor statement, the results for
real matrices were wrong if the kernels "avx-real-block6" or
"avx-real-block4" were chosen. No other kernels are affected.

The test programms always correctly stated that the results for
these kernels are wrong.