Known_Issues

This page contains lists all of the issues that we know about in NiCE and are currently investigating.

These issues are not necessarily bugs. They may represent missing capabilities altogether. Normally, though, they are bugs that we know about and are actively trying to fix. If you see an issue on this page that is affecting you, please do not file a bug report. Instead, send an email to billingsjj <at> ornl <dot> gov and let us know that the issue matters to you.

• Persistence
• Job Launchers
• MultiLauncher
• Geometry Editor
• Reactor Analyzer

Persistence

NiCE does not currently persist the items that are created between runs, but it does have an embedded database ready and waiting. We do not want to deploy this capability until we complete further testing because we want to minimize the chance that the database is corrupted by NiCE. Hopefully we will have this available very soon.

Job Launchers

None of our JobLaunchers allow users to tailor the number of MPI processes and OpenMP threads, even though jobs created with the editor allow default values for these to be specified. At the moment, this is by choice. We are doing some background work and testing to make sure that we implement this properly and we are working with some of our partners to make sure that their codes can handle random settings for parallel launchers.

MultiLauncher

Most of the issues in our MultiLauncher come from the fact that it is heavily threaded. We are currently working the following issues:

• The multilauncher sometimes fails on jobs where multiple remote jobs are executed.

Geometry Editor

Our geometry editor is a can of worms. Eclipse and JME3 don't play so well together, but we have been very successful so far. We are currently aware of the following issues for this piece of code:

• The geometry editor doesn't always load on Windows and Mac.
• The geometry editor throws an exception when re-opening a previously created Geometry with the ItemViewer.

Reactor Analyzer

We are aware of the fact that the ReactorAnalyzer behaves funny if there are no .silo files in the project directory. We are working on a smart way of handling this situation for all Items, not just the ReactorAnalyzer.

The ReactorAnalyzer may take some time to load large NiCE-IO-friendly .h5 files, particularly when there are multiple assemblies. On this desktop (Intel Xeon 3 GHz, 6 GB RAM), it takes approximately 12 seconds to read in each assembly for CASL problem 5 (quarter-core symmetry with 56 assemblies, each 17x17, with 6 data features, 1 axial level). We are looking into optimizing our file structure and access methods to reduce the loading time for such files.