iraf.net

HeyI have noticed something weird on my system:
I have done a lot of photometry with DAOPHOT, also on large images and very important here is the option that stores the image in cache (which is a daophot option).However, on my new system (Mac OS 10.8.4 64bit IRAF 2.16) this does not seem to work. The algorithm always accesses the disk for reads and writes and the photometry is very slow.On my other IRAF 2.16 installation under 64bit Ubuntu 12.04 (VMware Fusion) this works flawlessly and the photometry is MUCH faster.Any idea what is causing this? cheers

The DAOTEST task will run both with and without the cache set and the logfile produced should show there really is a difference, otherwise I'd suspect the parameters. Note that the DIGIPHOT package as well as a specific task like ALLSTAR both have a 'cache' parameter: By default the task will inherit the value from the package param but these can be set individually, and I can't guarantee some script task won't reset the value.To avoid using old parameter settings, do a new MKIRAF in some temp directory like /tmp so you have an empty uparm and run your test again. If that seems to enable to cache then check your parameters for both the DIGIPHOT package and individual tasks. There haven't specifically been changes to this code in years and no reason to believe the 64-bit work matters unless the memory allocation is actually failing for some reason. If you're certain it isn't a parameter problem please post back and I'll do some more detailed checking.

HiThanks for the quick reply. I have done as you said and the daotest procedure works and the output is fine, but since the test image is so small, it does not really make a difference.I then created a new login.cl and also tried to set the cache parameters for the single tasks to yes but still the photometry is super slow, so I am quite sure that this is not a parameter problem.I would like to see if you can reproduce this problem on your machines.cheers

I'm running OSX 10.8.3 but otherwise have the same system you describe but unfortunately I don't see the problem, the 'cache' parameter makes my tests run 2-3 times faster than when not enabling it.You didn't really give any details about what a "large image' is or how long "slow" is, for a sample 4K squared image I can run ALLSTAR in 3-4 sec with the cache enabled and 10-12 sec otherwise. Are your numbers significantly different? How much memory is on the machine and is it possible something else has most of the memory tied up already? Other than the 'cache' param are there any other parameters you've changed?