speed up Flash execution

[UB39]

Hello!

I ve a very perfomance criticle application. So I have to speed up my code. Performance  measurements show that there s a big difference if I boot from Flash or from RAM.  Unfortunately I can only boot from RAM during debugging. It seems that the difference between  RAm and Flashh execution fluctuates very much. I ve code fragments with nearly no difference and some with a 50 percent add on if I boot from flash.

So I m very interested why there s such a big fluctuation and how I can reduce the add on for Flash execution.

Nearly all of my time critical code is part of a big loop. So maybe it might be a solution to copy the time critical code from Flash to RAM. But how?

I use CW 10.1 and a K60N512 Tower Module

Many thanks

[BenGartner · Accepted solution]

This appnote describes the process for relocating code to RAM using the linker for Coldfire architecture. The theory is the same for Kinetis. Hopefully this will give you a starting point.

http://cache.freescale.com/files/soft_dev_tools/doc/app_note/AN4329.pdf

[UB39]

Thank you very much for the application note. It helps to reduce the overhead for my most time criticle section from 50 percent to around 30 percent. But there s still an overhead between FLASH and RAM booting although I can see in the *.xMAP file that the relocating works fine. 

Do you ve got any ideas for the reason?

[UB39]

Problem solved.

Many thanks