This is what I came up with for a 16 bit pulse on RA0. It has 1 execution
cycle resolution and simulates OK, I haven't calculated the overhead but
it seems to be the same regardless of what number is fed to it:

        bsf     PORTA,0 ;set pin high
 do_high:               ;high byte sets the number of 256 cycle delays
        movf    countHI,f
        btfsc   3,2     ;test for zero
        goto    do_low
        movlw   0x53    ;this constant makes it 256 cycle delay
        movwf   temp
        decfsz  temp
        goto    $-1
        decf    countHI
        goto    do_high
 do_low:                ;finish up with the low byte
        bcf     3,0     ;clear carry
        rrf     countLO ;rotate right
        btfss   3,0     ;test LSB, take long way if set
        goto    $+2
        nop
        bcf     3,0     ;same thing again
        rrf     countLO
        btfss   3,0     ;except long way is even longer
        goto    $+4
        nop
        nop
        nop
        decfsz  countLO ;4-cycle loop the rest
        goto    $-2
        bcf     PORTA,0 ;clear the pin
Cheerful regards,

Bob

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.