ParaView 3 Telecon 03-17-2011
paraview meeting notes 3-17-2011
3.10 released? Alan
3.10.1 utkarsh 3.10.0.6
3.10.1 immediate bug fix
alan: turns out we have memory leak in paraview which is fatal--bad juju not yet in bugtracker, hopes by end of tonight will have info easiest way to replicate; 300 files, wiposhield, 100 million cells, 256 cores turn off variable, dataset memory footprint doesn't go down what reader? any. data container, template alan planning to go to smaller dataset with smaller # cores
3.10.0 memory footprint twice 3.8.1 cth or exodus
another place seems to leak; play forward, memory leak grows totalview
greg: plot overlain feature:
--checkbox for all variables vs no variables in display tab
--somehow the view properties when you try to set to autoscaling or fixed range, somehow flaky
3.10.2 new features
discussion: prism 3.10.2
scott's wish list prism bug fixes
cth reader density bug
alan all of isolated prism fixes we could in 3.10.2 all of cth amr cleanups that we know of in clip pass variables NOT change cube axes because it sounds major
prism: right on cusp of really becoming of interest, especially to people such as Jason Loki and his group and work he is doing cth Jason Loki and all cth users--
greg context like this: prism is about to roll out, allegro users will like it, also cth users will like it, turns out most allegro users also know how to run cth, lots of cth users
capability might pull them into paraview general functionality with being able to handle amr data sets
also, with cth; taught paraview specific to cth --discouraging how many places it was hackie extract cth parts
extract cth parts will have guts replaced; will remain same
may want 3 clip planes independently controllable
4.0 too busy with other stuff to think 4.0
ldev paper nathan gorden bell submission, university of colorado
gearing up to run 160k cores
doing isosurfacing in situ transferring geometry to a viz cluster live almost animation speed moving entire data to viz cluster
aim to have in situ library, i/o library, w/o user having to do mpi code