Commit 3d880b65 authored by Andreas Marek's avatar Andreas Marek
Browse files

Partial roll-back of ELPA_2015.02.001 release

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
parent d98eb8b5
......@@ -16,7 +16,8 @@ The only major change (which results in point a) is in the ELPA-2
part of the library for real matrices:
the fully blocked QR decomposition has been moved from the development part
to the release!
to the release! However, this is still experimental, since there is still
a bug. For some matrix sizes the results are wrong
It is now possible to use this QR decomposition by either setting an
environment variable "ELPA_QR_DECOMPOSITION" to "yes", or to call the
......@@ -36,6 +37,10 @@ c)
The configure procedure was adapted to be more consistent. No compiler omptimization
flags are set automatically anymore, this is up to the user at build time
The checks for OpenMP in the configure have been improved
Any incompatibles to previous version?
......@@ -646,6 +646,11 @@ function solve_evp_real_2stage(na, nev, a, lda, ev, q, ldq, nblk, &
useQRActual = useQREnvironment
if (useQRActual) then
print *,"Carefull you use the experimental feature QR-decomposition"
print *,"it is possible that the results are wrong"
if (present(THIS_REAL_ELPA_KERNEL_API)) then
! user defined kernel via the optional argument in the API call
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment