mickey@illogica.UUCP (Michael Thompson) (10/21/85)
Has anyone got pathalias to work on machines with 16 bit address space? It seems to work properly with small maps (pathalias -v asia.japan), but with large maps (pathalias -v usa.ca.n) it core dumps. The reason I suspect a memory allocation problem is that I noticed that it does a malloc of a new table while it still has the old table allocated, and this table can get pretty large. thanks - Michael Thompson {ucbvax,decwrl}!dual!vecpyr!altos86!illogica!mickey
honey@down.FUN (Peter Honeyman) (10/24/85)
you bet the hash table can get big -- on down it gets up to 70 kbytes. you're wrong, however, in assuming that the old table is wasted when a new table is allocated -- the old table is re-used for node and link structs, which are the real space hogs. e.g., i have input describing 10,000 hosts and 20,000 links. it takes a whole big pile of space to represent the graph. (to be precise, a node costs 32 bytes and a link costs 14 bytes. call it a cool half megabyte.) i have no urge to hack up a disk-based version of pathalias, and i don't recommend that anyone else try. sorry, 16 bitters, you're out of luck. peter ps: incidentally, i recently cut pathalias' space usage by 20% or so, but i don't plan to post a new version for at least 4 months. at that time, i may even have reasonable support for internet domains. (i love to watch the internet writhe. it makes me hot.)
lat@druil.UUCP (TepperL) (10/24/85)
> i have no urge to hack up a disk-based version of pathalias, and i don't > recommend that anyone else try. sorry, 16 bitters, you're out of luck. > > peter Well, not necessarily out of luck. A 16-bit site need only find a 32-bit neighbor that would be willing to process the maps for them. -- Larry Tepper {ihnp4 | allegra}!drutx!druil!lat +1-303-538-1759 Back to you, Walter.
dave@lsuc.UUCP (David Sherman) (10/27/85)
In article <1631@druil.UUCP> lat@druil.UUCP (TepperL) writes: >> i have no urge to hack up a disk-based version of pathalias, and i don't >> recommend that anyone else try. sorry, 16 bitters, you're out of luck. >> >> peter > >Well, not necessarily out of luck. A 16-bit site need only find a >32-bit neighbor that would be willing to process the maps for them. It's not so simple, and the problem is worse than Peter indicates. lsuc is a 32-bit machine with 1 megabyte of memory, of which UNIX takes 300K. Pathalias will not run on the current "mod.map"s in the remaining ~700K. Dave Sherman The Law Society of Upper Canada Toronto -- { ihnp4!utzoo pesnta utcs hcr decvax!utcsri } !lsuc!dave
honey@down.FUN (Peter Honeyman) (10/27/85)
trying ... i managed recently to cut the memory use by 20-25%, but i'm not done profiling this new version. (i sent it to dave anyway). peter
ahby@meccts.UUCP (Shane P. McCarron) (10/30/85)
In article <610@down.FUN> honey@down.FUN (Peter Honeyman) writes: >i have no urge to hack up a disk-based version of pathalias, and i don't >recommend that anyone else try. sorry, 16 bitters, you're out of luck. > > peter We have a '16-bit' machine (Plexus P35), and pathalias runs just fine! Sure, it requires about 680k of memory, but as long as you have that around, there are no problems. Maybe I have an inadequate understanding of the problem. -- Shane P. McCarron Minnesota Educational Computing Corporation - Technical Services UUCP ihnp4!dicomed!meccts!ahby