I've recently noticed that there is a large difference between the context switch times on the i.MX53 QSB and our own device based on the i.MX53. Our device is 4-5 times slower as measured with lat_ctx described in this post: http://www.pengutronix.de/development/kernel/arm-benchmarks-20100729_en.html
Because I see better performance on the QSB, I am wondering if I have something misconfigured somewhere on my board but I do not really know where to look. What kernel configuration options could affect context switch performance? Where should I look to track this issue down?