1. My aim is to be able to check for thread stalling in my multi-threaded application. I debugged my way to a potential thread stall case (which took over a month). My expectation from vtune is to be able to find same problem in short period. Thus equipping us with other potential such problems.
2. My application is heavy; has to process packets real time; there are 10 threads running (4 processor CPU), 2 of these threads perform device read and device write respectively. I have kept the input packet rate to minimum, packet content processing is negligible.
3. When I use attach to process method of running vtune analyzer, my application gets killed with setI2C error in driver. The same application executable on running independently runs properly. Introduction of vtune analyzer kills it.
4. I was wondering if it is possible to analyse selected modules alone so as to reduce overhead. But as above link states it is not possible. I have tried unselecting collect spin and signal options, it does not work.
5. Is there any other way in which I will be able to run my application and keep vtune attached to it for sometime atleast?. Wondering if we can use cpu_affinity or some other settings which help to do this.