Double checking here only -->
after you upgraded your kernel, you wouldhave *had*to rebuild your sampling drivers pointing to the new kernel source. Either that, or, remove and reinstall the vtserver package.
You did do one of those, yes? I'm guessing you probably removed and reinstalled, but please confirm here.
Message Edited by jdgallag on 10-20-2005 04:02 PM
You're into new territory here. Engineering says please open a premier case on this issue at your first opportunity. They suggest:
1) please run the diagnostic script (default location /opt/intel/vtune/bin/vtqfagent) and be sure to attach the log file to your case
2) Since you're on a custom kernel, also attache the "*.config" file that was used to configure the kernel
3) Also, attach the .tb5 files where results looked OK, and where the results did NOT look OK
Message Edited by jdgallag on 10-27-2005 12:44 PM
I did not find the diag script vtqfagent under /opt/intel/vtune/bin.
I upgraded my RDC to the latest 4.0 and still got the same problem. Two functions used by VTune (prepare_to_wait() and finish_wait()) still used 25% clock ticks.
I heard SEP is a better tool and does not cause so much overhead, is this true?
Message Edited by firstname.lastname@example.org on 12-06-2005 03:39 PM
AH yes, makes sense, you did in fact write you were using the VTune Windows GUI, and I missed that. You'd have to install the full VTune 8.0 Linux product, GUI and all, to get the /opt/intel/vtune/bin/vtqfagent application.
This forum is surely not support: have you opened your premier case yet?
Message Edited by DaveA on 12-09-2005 04:26 PM
If you don't already have an account at Intel Premier Support, please register for an evaluation copy of the VTune analyzer at http://www.intel.com/software/products/global/eval.htm. During your registration, make sure you opt-in to "technical support," this will create an account for you at Premier.
After logging into Premier, click on the "Submit Issue" link and follow the instructions to open a support issue.