home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.sun.misc
- Path: sparky!uunet!fmrco!fmrco!msommer
- From: fmrco!msommer@uunet.uu.net (Mark Sommer)
- Subject: System Mode Swamp on Sparc 2
- Message-ID: <FMRCO!MSOMMER.93Jan7093547@ampere.uunet.uu.net>
- Sender: news@fmrco.uucp
- Organization: Fidelity Management & Research Co.
- Distribution: usa
- Date: Thu, 7 Jan 1993 14:35:47 GMT
- Lines: 68
-
- A program of mine, for some unknown reason, spends an inordinate
- amount of time in system mode. Any insights or suggestions would be
- welcome.
-
- My program calls a Fortran function to solve a large quadratic
- programming problem. The program is extremely computationally
- intensive and it runs very large (>10MB), but my machine has 48MB of
- memory and lots of swap space (>80MB), so I would have expected it to
- spend most of its time in user mode. In fact, according to many
- indicators (such as gprof), it spends most of its time in kernel mode.
- I've included information I think might be relevant.
-
- This is what ps says about my program:
-
- ampere% ps -u
- USER PID %CPU %MEM SZ RSS TT STAT START TIME COMMAND
- msommer 22245 83.3 9.510232 4436 p1 R 15:07 27:51 portopt -p /tmp/port-222
-
- I was first alerted to the problem when I "time"d a script (porto)
- that runs my program (portopt). (Presumably, the "time" command
- tabulates for all porto's child processes.) The 164 page faults
- and 0 swaps seem reasonable to me:
-
- ampere% time porto > opt
- 187.240u 3576.250s 1:17:39.44 80.7% 0+2251k 11+47io 164pf+0w
-
- Pstat says I'm using all my physical memory, but have lots of virtual
- memory left:
-
- ampere% /etc/pstat -s
- 20260k allocated + 10172k reserved = 30432k used, 49844k available
-
- According to "trace -p", my program doesn't do any system calls during
- the bulk of its runtime.
-
- Here's what vmstat says as portopt is running. Notice, there doesn't
- seem to be an unusual amount of paging or swapping, yet the cpu's "sy"
- column is pegged:
-
- procs memory page disk faults cpu
- r b w avm fre si so pi po fr de sr s0 d1 d2 d3 in sy cs us sy id
- 1 0 0 0 2272 10 0 0 0 0 0 0 46 0 0 0 214 264 44 6 94 0
- 1 0 0 0 2272 5 0 0 0 0 0 0 3 0 0 0 114 163 34 7 93 0
- 1 0 0 0 2272 5 0 0 0 0 0 0 0 0 0 0 80 190 42 5 95 0
- 1 0 0 0 2272 10 0 0 0 0 0 0 0 0 0 0 60 181 39 5 95 0
- 1 0 0 0 2272 0 0 0 0 0 0 0 0 0 0 0 65 198 43 5 95 0
- 1 0 0 0 2272 0 0 0 0 0 0 0 0 0 0 0 72 260 53 6 94 0
- 1 0 0 0 2272 10 0 0 0 0 0 0 46 0 0 0 203 180 41 7 93 0
- 1 0 0 0 2272 5 0 0 0 0 0 0 13 0 0 0 131 152 36 8 92 0
- 1 0 0 0 2272 0 0 0 0 0 0 0 0 0 0 0 73 137 34 6 94 0
- 1 0 0 0 2272 0 0 0 0 0 0 0 0 0 0 0 56 141 34 6 94 0
- 1 0 0 0 2272 0 0 0 0 0 0 0 0 0 0 0 51 114 31 6 94 0
- 1 0 0 0 2272 0 0 0 0 0 16 0 0 0 0 0 42 92 26 4 96 0
- 2 0 0 0 2272 10 0 0 0 0 0 0 47 0 0 0 182 154 37 5 95 0
- 1 0 0 0 2272 5 0 0 0 0 0 0 2 0 0 0 126 171 39 6 94 0
- 1 0 0 0 2272 5 0 0 0 0 0 0 0 0 0 0 77 178 41 6 94 0
- 1 0 0 0 2272 10 0 0 0 0 0 0 0 0 0 0 64 184 40 3 97 0
- 1 0 0 0 2272 0 0 0 0 0 0 0 0 0 0 0 51 163 37 4 96 0
- 1 0 0 0 2272 0 0 0 0 0 16 0 0 0 0 0 60 193 41 3 97 0
- 1 0 0 0 2272 10 0 0 0 0 0 0 47 0 0 0 164 140 34 7 93 0
-
-
- Can anyone explain what's going on?
- --
- Mark Sommer
- Fidelity Management and Research
-
-
-