Attending: ?JameySharp, ?JoshTriplett, ?TimWelch, IanOsgood
There is a lot of work ahead of us. Tonight we didn't have access to the PPC flight computer, so we discussed plans for the future instead. We will meet again next Tuesday, hopefully with a PPC FlightComputer in front of us. (Please fill in details.)
Uncertainty Team
Work on completing the GAINS prototype. It is currently in Perl, for its matrix math operations, but it doesn't deal with rotation at the moment. See the Uncertainty for a more complete TODO list.
Add more sensors, such as Differential GPS and a magnetometer. Maybe adjust gain/bias for environmental change (temperature).
Architecture
Pain:
- C language limitations, such as hacked-together coroutines
- no microthreads or coroutines
- error handling (such as very slow to determine a disk full error)
- simulator is in C++
- it is hard to unit test
Gain:
- many things are in independent chunks. We should do more of this.
- continue using a state machine
- data flow paradigm
- (reactive programming?)
Programming Language
General concensus was to use a higher level language than C for a lot of the higher level code. But which one?
- Multi-language issues: how to educate new users if the language is obscure? how to interface to/from the other C modules and libraries?
- Compiled Haskell: Jamey has done some prototyping in this language already
- Python: GStreamer, Axon, Pyrex
- ocaml: multi-paradigm but still compiles to fast code
- J is good for matrix stuff (but is it free or easy to read?)
- Esterel can compile single source directly onto a heterogenous distributed system like ours (but is not free)
Methodology
Test First programming would be desireable, as would unit testing in general. This requires us to build more modular code.
Tools
- gcc 4.0
- valgrind
- We have already moved the repository to ?SubVersion (svn)
- cross-compilation for PowerPC architecture
Wiki
With the destruction of LV2, much of the wiki content has become obsolete. We should clean up and update a lot of these pages, such as SoftwareToDo etc. That and root out Twiki content obsoleted by our move to ?MoinMoin.
Simulator
- Separate the graphics from the main simulator
- Add USGS topo maps as ground texture (starting with the Brothers launch site)
- Simulate errors: we need error models for Uncertainty team work.
Action Items
- Ian has a car, so he'll pick up the PPC FlightComputer from Andrew for next week's meeting. (Update: little did Ian know that his car battery had died. By the time he got it fixed, Andrew was in Wisconsin. Bummer.)
- Figure out and document installing and using a PPC cross-compiler toolchain. (Update: Josh has figured a lot of this out and posted directions linked from the Software.)