For our scientific studies we need to diagonalise very large matrices, however when matrix size reaches 98304x98304 scalapack throws error
- Code: Select all
xxmr2d:out of memory
This is strange since on our machine we have 400Gb RAM which should be enough for such matrix. Tracing memory usage shows that maximum used memory during the run is only 190Gb.
I am using python scalapack wrapper scalapy. It is linked with scalapack2.1.0 (also tried with 2.0.2 version). Under the hood scalapy calls "pdsyevr".
Lukas
