usually, you don't need that precise of a delay: something in the rough neighbor would typically work.

plus, if you have interrupt,  your delay routines will produce different actual delays.

"num1 = Fcy/40322;     "

I would redefine 40322 so that it can be easily changed when you move the routine to a different chip / architecture.

"            while(--num1);     "

this will create a problem when num1 is calculated to be 0 - from very low Fcy for example. I would use "while (num1--) continue;" instead.

I use a similar approach, except that mine was built up, from delay(cycles); to delay_us(us) which calls delay() and then delay_ms(ms) which calls delay_us(), all of which is based on pre-defined F_CPU.

this allows the code to be fairly easily ported.