| Author |
Message |
|---|
|
|
|
I have a CPU monitor (menumeter) on my Macs. On my old G5 power Mac, it showed SAH as "nice". As I understand it, nice activity is low priority the will be interrupted when I am using the computer. I just got a new Mac Pro with Intel CPU's and running Leopard. The Menumeter now shows the SAH activity as "user". Does this mean that Seti processing will now compete with what I am using my computer for? Is this an Intel thing? I haven't had this computer long enough to tell if Seti is slowing it down or not.
____________
 |
|
|
|
|
|
This appears to be a Leopard issue; I haven’t seen it on my Tiger/Core2Duo iMacs. There are some similarly titled threads in this forum. Read those and you’ll know as much as I do …
____________
  |
|
|
|
|
This appears to be a Leopard issue; I haven’t seen it on my Tiger/Core2Duo iMacs. There are some similarly titled threads in this forum. Read those and you’ll know as much as I do …
My G5 was running Tiger also. Let me know if you learn any more about this issue.
____________
|
|
|
|
|
|
I Have a macbook running 10.4 (intel) and all boinc projects I run including seti @ Home show up as "nice" |
|
|
|
|
I have a CPU monitor (menumeter) on my Macs.
Type this command in the terminal and hit return:
ps -auxcl nice | grep boinc
The output should be as follows (I've snipped the unrelated columns):
boinc_pr [snip] 19 - setiathome_5.28_powerpc-apple-darwin
boinc_ma [snip] 0 - boinc
As you can see, seti runs at nice 19, and boinc runs at nice 0...
Give it a try... ;)
(Sometimes the output of third party apps doesn't accurately represent what's really going on.)
____________
 It may not be 1984 but George Orwell sure did see the future . . . |
|
|
|
|
I have a CPU monitor (menumeter) on my Macs.
Type this command in the terminal and hit return:
ps -auxcl nice | grep boinc
The output should be as follows (I've snipped the unrelated columns):
boinc_pr [snip] 19 - setiathome_5.28_powerpc-apple-darwin
boinc_ma [snip] 0 - boinc
As you can see, seti runs at nice 19, and boinc runs at nice 0...
Give it a try... ;)
(Sometimes the output of third party apps doesn't accurately represent what's really going on.)
I don't know didily about the terminal. I did what you told me in the first part, but I didn't get anything that looked like what you have for the output. Here is what I got:
Last login: Wed Jan 23 05:17:19 on ttys000
[cblmdm72-241-51-102:~] robertmerrill% ps -auxcl nice | grep boinc
ps: No user named 'xcl'
ps: illegal argument: nice
usage: ps [-AaCcEefhjlMmrSTvwXx] [-O fmt | -o fmt] [-G gid[,gid...]]
[-g grp[,grp...]] [-u [uid,uid...]]
[-p pid[,pid...]] [-t tty[,tty...]] [-U user[,user...]]
ps [-L]
[cblmdm72-241-51-102:~] robertmerrill%
Did I do something wrong?
____________
|
|
|
|
|
Did I do something wrong?
Hang on... Let me check my Leopard machine... Maybe Apple changed something... [rolls eyes]...
[edit]
Yep, Apple changed something... This command is better, and Leopard friendly too:
ps -axco nice,user,command | grep boinc
The output:
0 boinc_ma boinc
19 boinc_pr setiathome_5.28_powerpc-apple-darwin
Sorry 'bout that... ;)
____________
It may not be 1984 but George Orwell sure did see the future . . . |
|
|
|
|
Did I do something wrong?
Hang on... Let me check my Leopard machine... Maybe Apple changed something... [rolls eyes]...
[edit]
Yep, Apple changed something... This command is better, and Leopard friendly too:
ps -axco nice,user,command | grep boinc
The output:
0 boinc_ma boinc
19 boinc_pr setiathome_5.28_powerpc-apple-darwin
Sorry 'bout that... ;)
OK, I got the same numbers this time, except I got 4,19, since I have seti running on 4 of my 8 cores. Does this mean that all 3 of my CPU monitors are wrong?
____________
|
|
|
|
|
Does this mean that all 3 of my CPU monitors are wrong?
If they differ from that of the terminal output, then I would have to say yes...
ESPECIALLY if they are menu items or widgets... ;)
____________
It may not be 1984 but George Orwell sure did see the future . . . |
|
|
|
|
Does this mean that all 3 of my CPU monitors are wrong?
If they differ from that of the terminal output, then I would have to say yes...
ESPECIALLY if they are menu items or widgets... ;)
One is a menu item, one is a widgets, but the other is Apple's own activity monitor. Here is a link to a screen shot of it.
http://www.geocities.com/bobarny/screenshot_143.jpg
See what I mean?
____________
|
|
|
|
|
See what I mean?
That is odd... You may not know this, but GUI applications such as activity monitor extract their data from the same UNIX commands as the one you just ran... I'd just chalk it up as another Leopard bug...
Remember: Programmers make mistakes and applications break, but the command line never lies... ;)
____________
It may not be 1984 but George Orwell sure did see the future . . . |
|
|
|
|
... You may not know this, but GUI applications such as activity monitor extract their data from the same UNIX commands as the one you just ran...
I believe Activity Monitor is based on top.
Is there a qualitative criterion for “niceâ€, in this context, that’s not necessarily determined by the value of the nice parameter? |
|
|
|
|
I believe Activity Monitor is based on top.
I believe activity monitor is based on a few UNIX commands... But mainly top... ;)
____________
It may not be 1984 but George Orwell sure did see the future . . . |
|
|
|
|
I believe Activity Monitor is based on top.
I believe activity monitor is based on a few UNIX commands... But mainly top... ;)
And if you look at "top" in a Leopard terminal, it does not report Nice processs, only user, idle and system. top in FreeBSD reports nice as well, I suspect that's why activity monitor shows none of them as being nice.
____________
|
|
|
|
|
I believe Activity Monitor is based on top.
I believe activity monitor is based on a few UNIX commands... But mainly top... ;)
And if you look at "top" in a Leopard terminal, it does not report Nice processs, only user, idle and system. top in FreeBSD reports nice as well, I suspect that's why activity monitor shows none of them as being nice.
So, what is the consensus, is Seti really running as nice or user, or don't we know?
Bob
____________
|
|
|
|
|
So, what is the consensus, is Seti really running as nice or user, or don't we know?
Bob
Running at nice level 19 (which is about as low priority as you can get). Activity monitor and top are inherently broken as they do not report nice processes, only user, system and idle.
____________
|
|
|
|
|
So, what is the consensus, is Seti really running as nice or user, or don't we know?
Bob
Running at nice level 19 (which is about as low priority as you can get). Activity monitor and top are inherently broken as they do not report nice processes, only user, system and idle.
Got it. Thanks
____________
|
|
|
|
|
So, what is the consensus, is Seti really running as nice or user, or don't we know?
Bob
Running at nice level 19 (which is about as low priority as you can get). Activity monitor and top are inherently broken as they do not report nice processes, only user, system and idle.
Got it. Thanks
I have been in touch with the guy that created MenuMeter. Here is what he has to say aobut this problem.
The diagnosis in the forum is incorrect and built from bad assumptions,
neither Activity Monitor nor MenuMeters extract their information from 'top'
(or any other commandline tool). Both programs extract the values directly
from the OS X kernel.
The real problem is that Leopard introduced a bug in the kernel's aggregate
scheduler statistics. 'nice' CPU time is no longer accounted for properly.
This is an Apple bug and has been reported to them by several folks
(including me).
'ps' and other commandline tools that report per-process nice levels are not
affected by this bug (because they use the kernel per-process statistics,
not the aggregate numbers affected by the bug).
Hopefully it will be fixed in a future 10.5 update, until then MenuMeters
isn't able to track nice CPU time properly.
____________
|
|
|
|
|
The diagnosis in the forum is incorrect and built from bad assumptions
I'd say the diagnosis in this thread is 100% correct... The GUI applications are not working properly... ;)
(But it was nice (no pun intended) of the developer to chime in and provide us with the reason why.)
____________
It may not be 1984 but George Orwell sure did see the future . . . |
|
|
|
|
The diagnosis in the forum is incorrect and built from bad assumptions
I'd say the diagnosis in this thread is 100% correct... The GUI applications are not working properly... ;)
(But it was nice (no pun intended) of the developer to chime in and provide us with the reason why.)
The problem is not only with the GUI tools. Despite seti are reported by ps to have a 19 (lowest) priority, they are competing with prio 0 tasks. When I run compute bounds tasks (they matlab or idl), they are slower by a factor 4 when seti is running (macpro 4 cores, 4 seti clients). 10.5.2 is still wrong. 10.4.11 is ok: the impact of seti on computations is low, say 10%. There is a problem with 10.5 scheduler.
____________
|
|
|
