On the Z80, the counter register B was only 8 bits, you could loop 1 to 256 times with DJNZ.
The outer loop was often unrolled, that is the inner loop was repeated, say, 3 times.. Say you wanted to loop 700 times = 256 + 256 +188; load the B register with 188 and jump to the first inner loop, at the end of the first inner loop the B reg is zero, the second inner loop is performed 256 times, and finally the third inner loop is performed 256 times. The same unrolled outer loops can be reused to loop just 200 times, load the B reg with 200 and jump straight to the third inner loop. There was a trade off between RAM data/stack space and ROM code space and execution time. If the inner loop was long, then make it into a subroutine and call it 3 times in a row. It is not clear from the PRU instruction set, but it does not look like you can have a LOOP within a LOOP as the inner LOOP instruction would just overwrite the end address and loop counter of the outer LOOP. You could use the high 16 bits of the register, as an outer loop counter (I've not done PRU assembly before, so could be mistakes) ````asm mov r2.w2, 3+1 mov r2..w0, 65535 OuterLoop: loop EndLoop, r2 // inner loop 65535 times ... EndLoop: sub r2.w2, 1 qbne OuterLoop, r2.w2, 0 ```` On Monday, 8 February 2021 at 19:30:31 UTC Dennis Bieber wrote: > On Sun, 7 Feb 2021 16:29:22 -0800 (PST), in > gmane.comp.hardware.beagleboard.user Tom Stepleton > <[email protected]> wrote: > > > > > >you have a maximum of 65535 repetitions. Very good. What happens though > if > >you want three times that? Can you do > > > > LOOP Exit, 3 > > LDI r10.w0, 0xffff > > LOOP Exit, r10.w0 > > > Off hand, I'd suspect one can NOT nest LOOP instructions. However, a > literal reading of the "operation" section could imply that each LOOP > instruction expands into two data items (counter and top) which may be > allocated/inlined in the code. > > Seems like it would be faster to just write a test program nesting two > LOOP instructions, the outer with a small immediate count, and the inner > reading from a register preset with a similar small count. Maybe have the > inner loop simply increment some counter, and check if the counter has the > result of multiple loops, or just one (and which one, inner or outer). > > > >Bonus question: the manual tells us that "The loop is > >non-interruptible(LOOP)." What does non-interruptible mean---do interrupt > >bits in r31 never get set inside a loop? > > Since the PRU doesn't have asynchronous interrupts by default I'm not > sure what "interruptible" would mean... However, SPRUHF8A indicates that > there is an "interruptible loop" variant. > > -=-=-=- > 5.3.4.3.18 Hardware Loop Assist (LOOP, ILOOP) > > Defines a hardware-assisted loop operation. The loop can be > non-interruptible (LOOP), or can be interruptible based on an external > break signal (ILOOP). The loop operation works by detecting when the > instruction pointer would normal hit the instruction at the designated > target label, and instead decrementing a loop counter and jumping back to > the instruction immediately following the loop instruction. > > Definition: > LOOP LABEL, OP(256) > ILOOP LABEL, OP(256) > -=-=-=- > > Possibly, "interruptible" in this case means via JTAG type debugger. > "LOOP" runs to completion before a debugger can get control, while "ILOOP" > allows debugger to take control within the loop. > > > -- > Dennis L Bieber > > -- For more options, visit http://beagleboard.org/discuss --- You received this message because you are subscribed to the Google Groups "BeagleBoard" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/beagleboard/4dd60612-342c-4779-9cd9-30a62bfe5dd4n%40googlegroups.com.
