Notes for the terameeting of 12 may 2005

 Present: Yannick, Emmanuel, Fred, Henry, Mireille, Anthony, Anis, Chiara

 Hardware, software Fred

 Fred describes various forum software available; the idea is that we will use such a tool for low-traffic mailing lists (such as the SExtractor mailing list); it is not clear if we will use such a system for the teralist.

 by Tuesday evening the MyBB software should be installed; on Friday we will try to make decision whether this tool is suitable.

 New machines: the configuration of the "efigi" machine has been decided

 fcix2 / efigi machine will be ordered now

 There is a possiblity that the EFIGI grant could pay Anthony's salary for six months, but to do that there has to be a transfer of funds between TERAPIX/EFIGI of around 15,000 euro. Action to be done by Yannick, Fred and Emmanuel by next Monday.

 mix8 / mix9 will be ordered once we have verified that this fund transfer between EFIGI and TERAPIX can be carried out.

Hardware status
 mix1 is unstable and the machine should no be used

 mix5 is currently running with four slower processors. AMD has published a press release indicating there might be a problem with the quality control for these faster processors; the problem is that we no longer have these (possibly defective) processors; they have been returned to option+. How to proceed? One suggest was to ask option+ to perform the checks for us. Fred will investigate

 Henry's test, using swarp, he ran on mix5 will be done on mix6 as well to check whether mix6 is stable or not.

 pix7: there is a problem with the filesystem; Fred suggests that the filesystem should be reformatted. Unfortunately there is data on this machine which is currently being used for processing on other nodes

 Fred will investigate the instability of automount. Why automount is down so often? Is it a well-known problem? Can it be replaced by another daemon? Fred will summarize on May 19.

Status SPICA

 There are problems with processing some PI data because the weight-maps have not been correctly generated by quality-fits

 Tully data currently being processed

 Note added: SPICA ran the wrong way on new PI data. All Qfits-in have to be done again for LinWhai, Tully, Robin, L02/L05 data

 Around 30% of the very-wide images have been processed. How can we carry out automatic quality control of all these images? An important point as currently they are all being inspected by eye

 Emmanuel will add 2d matching statistics which might help to find badly-matching images which are identified by Henry's procedure. Emmanuel paln is to compute the first and second moments of the astr-*error2d.* plots in order to provide offset with respect to center and elongation of both internal and external error plots.

WIRCAM

 Chiara explains the current status of the processing and tests carried out with the COSMOS data. Henry and Chiara galaxy counts are not similar. They have to figure out the origin of differences.

 Postcript added after the meeting: Yannick has sent a formal request to Loic Albert to supply us with data for which the sky background has NOT been subtracted. There is a currently a problem with Loic's sky-subtraction method for deep images like COSMOS, and it is likely that we will have this problem for other deep stacks.

 Chiara is still learning Prerepix

Panorapix
 Anis has implemented the shader control of contrast and brightness

 There is some problem with the displaying the image on full floating point; until now, the images have been displayed in full floating point.

 Not sure if this will work actually, there is some problem with the available GL extensions on each card

 Deadline: 27 May images will be correctly displayed in 32 bit.

Scamp

 Emmanuel has done a lot of work on the pattern matching and needs to finish this before continuing with the tasks requested by Yannick.

EFIGI galaxy classification

 Emmanuel and Anthony demonstrates the web interface. A lot of scatter between the different classifiers.