Download presentation
Presentation is loading. Please wait.
Published byElla Webb Modified over 9 years ago
1
CAM Process, Redesign, and Plans Tom Henderson February 4, 2003
2
Outline Who am I? CAM check-in process New physics-chemistry interfaces Plans
3
Who Am I? I don’t know, I just arrived recently… Formerly with NOAA FSL –Scalable Modeling System (SMS) Directive-based SPMD Explicit FDA, nesting, spectral
4
CAM Check-in Process Goal –Reduce risk for IPCC deadline Schedule –Begin immediately –Relax after IPCC delivery
5
CAM Check-in Process Approach –Serialize check-ins to dev branch CAM CRB approves all changes Eliminates “race to avoid merging” –Standardize test procedures test-model.pl blackforest, chinook –Add anchorage-Lahey next week
6
CAM Check-in Process Two types of changes –“Science” changes Change model results Follow policy at http://www.cgd.ucar.edu/~cam/CRB_Policy.shtml –“Structural” changes Bit-for-bit or roundoff difference Simpler process, especially for bfb
7
CAM Check-in Process Approach –Split development branches Retain “dev” branch Add new “science” branch This will begin when new aerosols are checked in
8
CAM Check-in Process Approach –Display all proposed changes on web page Check-in order, etc. http://www.cgd.ucar.edu/~cam/cam_checkins/ … and now for a brief tour …
9
CAM Check-in Process Improvements –Integrate check-in forms with the cvs tag form Some information is duplicated now –Upgrade to CGI form Currently just a file to edit –Integrate or replace with a real project management tool
10
New Physics-Chemistry Interfaces Designed by Brian Eaton and Byron Boville Goals: –Make it much easier to add a new physics package to CAM –Make it much easier to use other existing physics packages with CAM Follow accepted design standards (CMIWG)
11
New Physics-Chemistry Interfaces Approach –Define responsibilities of a physics package Do not change model state directly Return tendencies instead Conserve certain vertical integrals Provide specified interfaces for –Registration / initialization –Time step advance
12
New Physics-Chemistry Interfaces Approach –Reduce coupling (dependencies) between physics package Encapsulate routines and data structures into modules Limit module interactions to interfaces Use “blackboard” for communication between physics packages and to/from I/O –“physics buffer”
13
New Physics-Chemistry Interfaces Status –Just getting started…
14
Plans Short-term –Public release of CAM 2.0.2 Tedious manual gutting done Resolving testing problems on Linux/PGI Release “soon” (?)
15
Plans Short-term –New physics interfaces Implement physics interfaces for turbulence/vertical diffusion Propose improvements for “blackboard” and phys_grid Extend to other physics packages –Improve performance “low hanging fruit”
16
Plans Long-term –Significant science changes –Dynamics interface re-design –Re-design of system tests –Streamline release process “make release” –Use a real project management suite –…
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.