SDK • Re: memset isn't faster than loop?

SUB is 1 clock. BCS is taken, 2 clocks. 5+1+2 == 8... sounds about right?

Yep, closer to reality. On RP2350 is worse ... 9 clocks.

You do have a 50 percent overhead too. Increase the number of registers used from 4 to 8 (66 percent)

I think you can use max 7 in this case (one being used as destination address and written back incremented accordingly).

Restrictions In these instructions: reglist and Rn are limited to R0-R7.

Statistics: Posted by gmx — Sun Feb 23, 2025 5:38 am

---