VTune driver causes machine crash

[kkboyanov]

Hello,

Since the installation of VTune on my server (PowerEdge R710 with dualXeon E5540 CPU @ 2.53GHz, SMT on, running Scientific Linux 2.6.18-194.3.1.el5)

I am experiencing system crashes while running sampling in VTune GUI mode with either a lot of events to be observed (>40-50) or when I run an application for an extended period of time (like High energy physics simulations or the SPEC benchmarks).

The onlymeaningfulerror message I have seen in the logs comes from the System Event Log of the service processor:

PCIE Fatal Err: Critical Event sensor, bus fatal error (Bus 0 Device 0 Function 0) was asserted

I can run the examples which come with VTune without a problem in Eclipse GUI, and when running the insmod-vtune script there are no warnings or errors shown. Also, when running smaller programs (smaller runtime) and sampling for a small amount of events I get no crashes.

Can it be that Iaccidentallytry to access PMU features not present on the processor used?

What is youopinionon the source of this error? Any thoughts and recommendations are highly appreciated!

Best regards,

Konstantin Boyanov

[Peter_W_Intel]

Hi,

1. The error message can be found in http://support.dell.com/support/edocs/software/smdrac3/idrac/idrac1.11/en/ug/html/chap13.htm, it indicates "Fatal error is detected on the PCIE bus."

2. If you ran application with unsupported PMU event(s), sampling will report it then stop.

3. Can you run same amount of events with smaller application? Can you run small amount of events with bigger application?

4. 40-50 events are too much for one sampling activity. Sampling will separate these events into several sampling sessions. I don't know if there is the problem - if sampling stops one session, and terminates bigger application...quickly another sampling session invokes bigger application again.

My suggestion is to separateone sampling activity (with 40-50 events) into several sampling activities (with a few events).

Regards, Peter