I am no longer able to reproduce this bug. Please close it and I will assume that I was somehow doing something wrong (or the computer simply needed a reboot). If I am able to reproduce it again, I shall contact you.
Sorry about the apparent false alarm. Best, Keith -----Original Message----- >From: Moritz Muehlenhoff <[email protected]> >Sent: Dec 25, 2008 11:01 AM >To: Keith Godfrey <[email protected]> >Cc: [email protected], [email protected], [email protected] >Subject: Re: 'top' doesn't report multi-core CPU usage properly w/ default >kernel > >reassign 509215 linux-2.6 >thanks > >Keith Godfrey wrote: >> >> Hardware: Dell Optiplex >> CPU: Core 2 quad >> OS: Lenny - fresh install ~Dec 15th >> >> Testing configuration and report: >> I have a pthreads based scientific application that uses ~170% and 250% CPU >> usage when running 2 and 3 threads, respectively. This behavior has been >> consistent on several systems, including MacOS and a previous version of >> Lenny. CPU usage was measured using 'top'. >> >> On the freshly installed Lenny, 'top' reported between 70%-105% usage for >> both 2 and 3 threads. The time required for the application to complete was >> comparable to runtimes on MacOS and a previous version of Lenny (all using >> the same or similar CPUs). >> >> I compiled a new kernel (2.6.27.10), basing it the config stored in /boot, >> incorporating minor additional changes. Under the newly compiled kernel, >> 'top' appears to report correct CPU usage. Application runtime remains the >> same. >> >> It apperas that 'top' is reporting incorrect CPU usage on the default >> kernel. It is also possible that the multi-threading doesn't speed anything >> up in this application (very possible) and that the default kernel wasn't >> properly handling the threads, whereas 'top' is OK. > >top parses /proc/stat, so it might either be that the data from it >is inconsistent in the 2.6.26 kernel or that top fails to parse >the output properly. > >Could you compare the content of /proc/stat from both kernels? > >Cheers, > Moritz -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
