GR updates: (Heather) [slightly updated since meeting] r48 will include Leon's updates
r49 will have new LDF.
Overlay: (Tracy) has written up plans and current status for the Randoms Overlay project. One aspect with possible repercussions is that Cal Clustering will need to be run in multiple cluster finding mode. (Leon) checked and discovered about 100 places in the code where CalCluster is called. Typically the caller just takes the first cluster, not necessarily the right thing to do when there are multiple clusters. (Philippe) would like to be able to run Overlay with pass 6 as it is now (just one Cal cluster). (Tracy) Clustering is done by a Gaudi tool; it is easy to switch back and forth. (Richard) We should estimate # of events needed to recalculate IRFs. (Philippe) Volunteered (or at least didn't object to being volunteered) to create the input files needed for Overlay.
Reprocessing: [Thanks to Tom for this summary]
Reading from afs: (Warren) There was a problem reading ROOT files from afs; turns out it goes away if the files are copied to scratch (an accidental discovery). (Anders) Will need to reprocess 20-30 runs in L1 to correct for afs corruption.
Batch allocations: (Richard) The Linear Collider Detector folks need more priority on the farm and we can probably afford to give it to them. As of last night, there is a 25% decrease in general queue batch allocation. It's not expected to impact us, but if it does we can complain.
Documentation: (Richard for Chuck) Main focus recently is on Science Analysis and iLat.
ScienceTools: Seth gave the report.
skimmer (Heather) David is proceeding with plans to move the skimmer to a local (for him) repository. Heather will be making a new version soon which will properly rebuild truncated indices in some early runs (the original attempt wasn't quite right).
SCons, RM [Navid kindly provided the following:] I've been mostly working on getting SCons RM compiled for rhel5 and Windows. In theory, this should have been 1 day worth of work but because the rhel5 boxes we have are setup really weird Qt would compile fine on them but fail with really strange errors. It took a week of debugging to realize the problem. I now have a clean installation of rhel5 on a virtual machine that I use to compile Qt. I'm using that compile on the SLAC rhel5 boxes without problems now. I'm getting some other bugs fixed right now to have the RM use correct submit arguments for rhel5. The lsf queues for our rhel4 and rhel5 queues are the same. To obtain the correct machine, resource parameters need to be specified. The new SCons RM is not setting these resource requirements correctly yet. The windows side is a similar problem in that I need to compile Qt for windows in static mode. After some initial problems with getting MySQL binaries and Qt to use the same compiler, I think the current Qt compile in progress will result in a working set of static libraries to use with GoGui and the RM.
External libraries (Emmanuel) Detailed status can be found in Confluence.
Yet another candidate library to be included for python users is LAPACK. (Eric W.) LAPACK is not itself written in Python; it's compiled code, an optimized version of functionality which exists elsewhere implemented in Python. When available, NumPy will use LAPACK and that will speed it up noticeably. However, since LAPACK is compiled, it is more difficult to support on multiple platforms. (Heather) suggests we hold off for the time being.
GoGui (Joanne) There have been a couple small improvements to GoGui in the past week. See the GoGui Confluence page for current status.
The current version of the RHEL4 GoGui executable is now available from SLAC Linux and its use is (mostly) documented. The most important features still lacking on the Linux side are ability to tag and use of an override directory. The latter may be a while in coming since it requires some changes in the SCons infrastructure.
For Windows, the next step is to make static Qt libraries and a statically-linked GoGui (as is already the case for the RHEL4 executable) so that Qt libraries do not have to be installed by users. [As of this afternoon, such a thing is built, but it's huge: 77 MBytes. The comparable Linux executable is 14 MBytes]. Next I'll attempt to implement the plan Navid has laid out for building MS project and solution files.
previous | minutes index | next |