Various Software Topics
Murrough Landon - 28 November 2001
http://www.hep.ph.qmul.ac.uk/~landon/talks
Overview
0pt
0pt
0pt
- Recent work on run control
- Run types and run parameters
- Test vector organisation
- Software packages and CMT
- Schedule...
Run Control (1)
Recent Non Work
0pt
0pt
0pt
- Not much since the joint meeting (where I reported
adding use of the parts manager to create hardware
modules)
- The above changes are still not in CVS (things are
still in flux a bit and I want to tag the first try).
- Use of "#include hdmc/Dir/file.h" requires extra
link in HDMC directory tree.
Run Control (2)
Calibration Sequencer
0pt
0pt
0pt
- But I have successfully tested a ``calibration/test
sequencer''.
- This is adapted from an Online software
example (part of their internal test suite).
- Sends a sequence of pause/resume commands to the root
run controller, ending with a stop command.
- The idea is that the other controllers can take the
relevant actions for the next step in the calibration
or test run.
- So far only a proof of principle: it just has hard
coded waits for each step. What do we actually want?
Eg software L1A commands to TTCvi? Control of other
external logic?
- So far run separately by hand. Should be started
automatically for specific run types by a suitable
run controller.
- Should also be controllable via run parameters
(stored in IS, set via IGUI).
Run Types and Run Parameters
Run Types
0pt
0pt
0pt
- Physics (and maybe video RAM)
- Calibration and test sequences
- Should various tests, eg loading playback in different
places (PPM, CPM, etc) be separate run types?
- Or a single playback run type with sophisticated test
vector choice as a run parameter?
Run Parameters
0pt
0pt
0pt
- Not too complicated (GUI code is tedious!)
- Enable/disable modules described in full database?
- Some global modes for modules (eg BCmux)?
- Calibration/test sequence control (n.steps, trigger
method, time or n.triggers per step, etc)
Software Packages and CMT
How to organise our packages
0pt
0pt
0pt
- ATLAS has finally decided to adopt CMT. Online group
is now trying it seriously.
- I intend to try it seriously (but have not done so yet).
- We do need some conventions for interworking between
different packages. CMT (and ATLAS) impose some choices.
- Eg "#include package/header.h" style. This implies
suitable directory structure (not yet done for my packages).
- CMT also expects consist use of CVS tags. ATLAS suggest
pkg-vv-rr-pp, which vv the major version,
rr the release, pp a patch.
- Online is developing a working model - but dont yet have
guidelines for the librarian or how to make a binary release:
http://akazarov.home.cern.ch/akazarov/cmt
- CMT home page (manual and tutorials) are at:
http://www.lal.in2p3.fr/SI/CMT/CMT.htm
Schedule, Workplan, Timescales
Gantt chart
0pt
0pt
0pt
- After many repeated suggestions, I have now tried to enter
information about our tasklist and timescale into a tool
which produces Gantt charts.
(Mac shareware to start with which was quick and easy
to get hold of, but isnt very sophisticated).
- Should probably use eg MS Project (does that ensure workers
dont work more than 100% of their available time?)
- Preliminary chart is doubtless very unrealistic.
It also has many overlaps and gaps, etc.
Murrough Landon (m.p.j.landon@qmul.ac.uk)