Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.
Announcements
FPGA community forums and blogs on community.intel.com are migrating to the new Altera Community and are read-only. For urgent support needs during this transition, please visit the FPGA Design Resources page or contact an Altera Authorized Distributor.
29282 Discussions

Non-deterministic results from ARPACK

greg_rowe
Beginner
488 Views
I've compiled ARPACK using visual fortran and I get slightly different values each time it is run. This could be a slight change in magnitude or a change in sign of eigenvectors. Could this be due to not aligning double arrays to 16 byte boundaries (if so how would I do this)? Please help.
0 Kudos
1 Reply
TimP
Honored Contributor III
488 Views
Quoting - greg.rowe
I've compiled ARPACK using visual fortran and I get slightly different values each time it is run. This could be a slight change in magnitude or a change in sign of eigenvectors. Could this be due to not aligning double arrays to 16 byte boundaries (if so how would I do this)? Please help.
If the differences are due to alignments on 32-bit Windows, you may be able to avoid it by setting /fp:source (on a compiler recent enough to support that switch). If your source is entirely Fortran, about the only obvious control you have over alignment of double precision arrays is to make sure they aren't preceded by data typesother than double complex in COMMON blocks or derived types. Since ifort 8.1, local data should be aligned automatically.
0 Kudos
Reply