Its "clean" but it won't do what you want, so its useless. How helpful is "clean" ? It is certainly not "right", because, again, it doesn't do what you want. You don't know HOW to write an interrupt version yet you say putting in an interrupt will make it slower !

You are completely wrong. THINK about it. The servicing of the on and off of the port bits takes MICRO seconds, the interval between the servicing is hundreds of microseconds, the interrupt SPEEDS UP the program. Lets modulate the LEDS at 30Hz, 33mSec, 33,000 usec.

Lets have a timer interrupt firing every 1/255 of 33,000, or every 130 usec. With a 24Meg, straight clock processor, that gives you about 260 instructions between servicings.

Keep 8 registers with your PWM levels, preset by your channel setting routine. Every time you service the interrupt, you have a counter variable in idata which you decrement. Decrement the PWM channel. If the PWMregister is 0 you turn off the bit corresponding to that channel. When the counter variable rolls over to 00 you turn all the outputs back on.

I think its about 5 instructions per channel, 48 instructions in the service routine. So you are in your ISR for about 18% of the time, and that's a fixed quantity for all channels,

;Destroys Accumulator
;Demonstration only. Timer setup is your problem, you may want your PWM ports to work differently. 
       XCH A,r0    ;Get r0
       DEC A
       JNZ nxtr1   ;Is it Zero ? Turn off the bit, else next bit
       clr P2.0 
nxtr1: XCH A,r0    ;PUT DECREMENTED register back !
       XCH A,r1    ; next register
       DEC A
       JNZ nxtr2
       clr P2.1
nxtr2: XCH A,r1
       XCH A,r2
       DEC A.
       JNZ nxtr3
       clr P2.2 
nxtr3: ................

       DEC CTR     ; Reload registers with PWM values
       JNZ ENDi
       MOV r0,PWM0 
       MOV r1,PWM1
       .......
       MOV P2,#255 ; turn all bits on again.

ENDi:  RETI



 

As Kai says, if you use a proper modern variant, you'll have a real PWM generator anyway, and you may even have reduced clock cycles too. The NXP ones will run 6 clocks/ instruction, the Dallas ones manage 1/instruction, so an NXP processor in X2 mode will only have 9% interrupt overhead, a Dallas chip less than 1%.

Your original solution uses 100% cpu time whatever the chip can do. So its dumb.