eo@ansa.co.uk (Ed Oskiewicz) (07/25/90)
Hello all, I have noticed something similar recently that might be cache thrashing in ST/1.5.10. I converted mg1b, a public domain gnu-emacs look-alike which is my minix editor quite some time ago and it runs chmem'd to 100K it has given me no problems. Recently when trying to write out a 43K file it took a very long time as there was only disc activity about every 30 seconds (like when update calls sync). Normally it would take a couple of seconds - this problem does not occur if the emacs is compiled with the recently posted c68 (using ack libc)! I haven't looked into this in detail yet but will try chmem, my buffer cache is about 120 blocks. I have had little feedback to my shutdown(1m) program so will repost soon incorporating Earl Chew's improvements. If the referees exist could one of you indicate what one has to do to get an acknowledgement that you have received (and are looking at) a submission. Ed (eo@ansa.co.uk)
paula@bcsaic.UUCP (Paul Allen) (07/27/90)
In article <25595@nigel.udel.EDU> eo@ansa.co.uk (Ed Oskiewicz) writes: > >I have had little feedback to my shutdown(1m) program so will repost >soon incorporating Earl Chew's improvements. If the referees exist >could one of you indicate what one has to do to get an >acknowledgement that you have received (and are looking at) a >submission. The referees list hasn't quite worked out the way it was anticipated. It seemed like such a good idea back when there was a new version of 1.5 every time you turned around and we were all madly trying to keep up. Sadly, most of the referees appear to have jobs, families, or other interests that conflict with their basic desire to be guinea pigs for new software. :-) I still think that beta-testing is a Good Thing. Perhaps a better approach would be for the authors of new software to post a description of their program and ask for beta testers. Speaking just for myself, I'm much more likely to make time to test something that I've volunteered for than I am something that just comes over the transom. Comments, anyone? Ed, I've received your shutdown program. It looks interesting and is on my list of things to look at. I probably won't get to it in a timely- enough manner to be of much help as a beta-tester, though. Sorry. Paul Allen -- ------------------------------------------------------------------------ Paul L. Allen | pallen@atc.boeing.com Boeing Advanced Technology Center | ...!uw-beaver!bcsaic!pallen
brucee@runxtsa.runx.oz.au (Bruce Evans) (07/28/90)
In article <25595@nigel.udel.EDU> eo@ansa.co.uk (Ed Oskiewicz) writes: >thrashing in ST/1.5.10. I converted mg1b, a public domain gnu-emacs >... >chmem'd to 100K it has given me no problems. > >Recently when trying to write out a 43K file it took a very long >time as there was only disc activity about every 30 seconds (like Buffer thrashing would probably only occur while reading. Perhaps the write is going through stdio unbuffered, due to malloc failing. This can happen even with lots of memory if the application mallocs everything before stdio has a chance. Success with another compiler and stdio may be mainly luck. The buffer thrashing in 1.5.2 was only alive for a couple of days. -- Bruce Evans Internet: brucee@runxtsa.runx.oz.au UUCP: uunet!runxtsa.runx.oz.au!brucee