michael@stb.uu.net (Michael Gersten) (11/18/89)
Ok, what do the SZ and RSZ fields really mean? The best I can figure out is that SZ is total size of a program in 4K blocks (fits very closely with what 'size file' tells me), and RSZ is an indication of how much is in memory (as opposed to swapped out). The problem? RSZ can be larger than SZ. Programs that are not used for ages are not 100% swapped out. Even when multiple compress/uncompress's are started up (memory hogs). So whats the story? And how does ps report shared mode programs (or does the UNIX-PC support shared text?)? Michael
jcm@mtunb.ATT.COM (was-John McMillan) (11/20/89)
In article <1989Nov18.041254.11009@stb.uu.net> michael@stb.uu.net (Michael Gersten) writes: >Ok, what do the SZ and RSZ fields really mean? > >The best I can figure out is that SZ is total size of a program in 4K >blocks (fits very closely with what 'size file' tells me), and RSZ is >an indication of how much is in memory (as opposed to swapped out). Good guesses. SZ is the sum (p_dsize + p_ssize + p_tsize) [DATA+STACK+TEXT]. - This IGNORES Shared Library [TEXT+DATA] ! RSZ is the resident subset of the larger amount. >The problem? RSZ can be larger than SZ. Programs that are not used for >ages are not 100% swapped out. Even when multiple compress/uncompress's >are started up (memory hogs). No PROBLEM, n'est ce pas? Swapping is declasse. It DOES happen -- you might be able to see it if you try those multiple *compresses with just .5 MB 8) -- but it's pretty rare. >So whats the story? And how does ps report shared mode programs (or >does the UNIX-PC support shared text?)? As noted above -- gee, 'hope I'm correct -- Shared Text is reported for EACH process sharing it: They EACH require it, right? (Yes, Virginia, there IS a Shared Text.) Shared LIBRARY is more than a bit queer: 5 pages DATA 5 pages BSS 27 pages TEXT (Yeah! The /bin/file data is in odd order!) and it's hidden in the RSZ data. (The concept of Shared Library was a marginal hack onto the creaking Berkeley fragments that make up the memory management of the 3B1.) john mcmillan -- att!mtunb!jcm -- speaking for SELF, not THEM