I was using the uVision debugger with a RV compiled code. And the startup module was wrong with the PLL. So it was thinking the system clock was running at 98MHz instead of 58MHz. So this makes my C code run at 67us for the test.

The Keil C compiler takes 108.7us at 58MHz clock.

I am sure that a bit of hand tweaking could get an improvement. But then I am faily sure that an 8051 single-clocker requires hand tweaking to run at this speed. i.e. no 8051 C compiler would achieve it by itself.

As I said earlier, I have little experience with ARM chips. I am relying on the uVision simulation being correct. (It times my 1ms delay as 1ms.)

All this is pretty irrelevant if the application is not suitable for an ARM.
But it does show that the cheap low-end ARM chips are competitive.

David.


Per,
Have I got the execution times completely wrong? Your post implied that you would have to loop-unroll the whole code. I do not see this.