Motifps and performance
- To: mathgroup at yoda.physics.unc.edu
- Subject: Motifps and performance
- From: David Harrison <harrison%faraday at yoda.physics.unc.edu>
- Date: Tue, 2 Feb 1993 10:18:16 -0500
A couple of weeks ago I mailed to this group a report of serious performance problems on our HP9000/750 that seemed to be related to multiple motifps processes running on our various X-terminals. We believe we have found the cause of the problem. If probably affects all HP's running MMA 2.1 and perhaps all computers running motifps. The problem is that the file /usr/lib/X11/XKeysymDB gets overwritten by the install process of MMA with a 7.5 Megabyte monster. This causes all motifps processes to have at least that size; it also causes all Motif processes to also have that lower bound. Memory goes to zero as motifps processes are spawned. I didn't twig earlier because it doesn't show up as swaps. Instead it shows up as the page outs mirroring the page ins as the system tries to find some memory. Also, we don't run a lot of Motif applications so I was fingering motifps as a probable culprit, when it is a MMA version of a standard support file that was the problem. We have restored the original /usr/lib/X11/XKeysymDB file shipped by HP (size = 2908 bytes), the motifps processes now have a reasonable size, our page outs are down to virtually zero, and the functionality of motifps seems unaffected, at least for our NCD 17c and HP 700/RX X-terminals. -- David Harrison | "Why do three tones make a Dept. of Physics, Univ. of Toronto | triad and not a triangle?" Inet: harrison at faraday.physics.utoronto.ca | -- Mach -- David Harrison. Office: MP124 Phone: x2977