• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

rc5 running 50% per cpu in linux

I

imhotepmp

Golden Member
First off Im running rh6 and i installed the latest dnet client. Well using the 'top' command it appears that the client is only using 50% of each cpu(I knew something was wrong when the bench only gave 1M k/sec for dual 366's)

How do I get it to use 100% of both?


imhotep(MP)

 
Er... Sounds to me like it's just using one processor... You sure you have an SMP kernel running?
 
There should be something else showing up in your runnig processes that is taking the other half of you CPU time.
 
2 crunchers have been started in the client. I know that the SMP kernel was installed, and looking in sys info it shows both cpus(0 and 1).

Also I should say that I didnt install the second cpu until after i installed linux. Could this be making a differnce?

One last thing, when you run the'top' command does this give you info on one proc or both? Im thinking that it shows info on one and maybe its load balancing on each cpu therefore both clients would use 50%? Possible?

Kilowatt, the only other thing that is taking up any significant spu time is X windows. No other processes are running.


imhotep(MP)

btw how do you determine if youre using an smp kernel in a console?

 
imhotepmp - check your /var/log/dmesg file to see if both processors are being recognized at boot...
 
No, in top it should show both clients at around 97%, at least on my dual 233 it did.

Installing the second CPU after installing Linux wouldn't matter, as long as you're certain you're running an SMP kernel. I'd killall dnetc, wait a few seconds to make sure they've shut down, and try starting dnetc again (you're just starting it once, right?)

If that doesn't help, reboot and try again. If that doesn't do it, you may have to recompile your kernel.

-edit- Oooh... I just made Senior Member 🙂
 
Rendus - that might be unless top is doing like what windud does and only reports "total" CPU usage (ie., total of both processors). That would mean 100% of one processor would equal 50% "total" reported for a dual setup.

[EDIT: Oh and congrats on your seniority!!!! 🙂]
 
checked the log and it shows that two crunchers(a and b) were started

could it be an issue with the client, its beta isnt it? Or maybe its top?



imhotep(MP)

edit: I noticed the keyrate varies from 1-1.99M keys/sec. If its doing 1.99m thats means it is using both. Right?
thanks for the help! The DC forums is definitely the best, IMHO 🙂


 
Open a term box, and type ps aux, that will show you all the processes running for every user, including root.

You should see two instances of dnetc running, if you are for sure running a SMP kernel.
 
You can tell dnetc to start as many threads as you want, and it will regardless of if you have the processors for them or not. If not, they'll all just run on the same processor.

I guess depending on the version of top, it could be telling you the load based on both processors. The problem is, for every process or thread you start for dnetc, it increases your load average by 1.00, so uptime-style load averages won't help here (I've intentionally pushed my load average to over 1024 this way. It rolls over at 1024 🙂)

I don't really have much experience in this area, never had a problem like this myself. Still sounds like a kernel issue to me, but beyond that I'm out of suggestions :/



<< You should see two instances of dnetc running, if you are for sure running a SMP kernel. >>



You can get this effect on a non-SMP kernel too. It's not going to be a sure-fire indicator.

Are you letting dnetc autodetect SMP or are you manually specifying the number of threads to run?
 


<< edit: I noticed the keyrate varies from 1-1.99M keys/sec. If its doing 1.99m thats means it is using both. Right? >>



Heh, sounds that way. 1mkey/sec per processor is about right for a Cel366 from my understanding.

Ack. Just remembered, I think the benchmark only starts one thread, so it'd only test one processor.
 
Make sure your running the latest version of top. I have seen a couple default versions, that when dealing with dual CPU's, they mark the total load as 100%, (not 200%). So when you check your load with top, it will only show 100%. Other commands to check would be uptime, and to check your keyrate.

As for your keyrate flux. Don't know, all depends on what it is doing when you only get 1 MK/sec.

LD
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top