Posted on May 17, 2011 at 09:58I'm not sure if this is 'normal' or not, but it's about on par with what we're seeing. I have measured interrupt latency on an IRQ line (not FIQ) and get about 6 us at a core of 96 MHz. The latency depends on your tool chain and whether you are using the STR9x library. (You may be using up-front code from your compiler vendor, OS vendor, or ST that sets things up nicely so your ISR needs to do very little, but the cost is overhead).
Make sure the timer peripheral you are using is running with a clock divider that suits your application.
Also, keep in mind that if you are measuring latency by toggling a GPIO pin in your ISR there is considerable latency introduced by the GPIO peripheral itself on ARM. I don't have hard numbers to back this up, but I have read that the GPIO pins don't respond as quickly as one might expect. Just an idea.
Rich
