Question for DF about the A20 Code

[DynatOS]

I've noticed with your A20 Wait routines you use jz/jnz then jmp... like below...

Code: Select all

a20wait:
.l0:    in      al,0x64
        test    al,2
        jz      .l2
        jmp     .l0
.l2:    ret


a20wait2:
.l0:    in      al,0x64
        test    al,1
        jnz     .l2
        jmp     .l0
.l2:    ret

Is there any particular reason for this? Is the extra jump designed to eat some clock cycles up (like the 65535 time-out loop)?

Just wondering, as then you could redesign them as such...

Code: Select all

a20wait:
        in      al,0x64
        test    al,2
        jnz     a20wait
ret


a20wait2:
        in      al,0x64
        test    al,1
        jz      a20wait2
ret

I wouldn't want to deviate from your code if there is a particular design approach you took. Any advisement would be appreciated.

PS: Thanks for the great routine

[Candy]

You're optimizing a delay loop.

[Kemp]

A sign of insanity as noted when I mentioned something else ;D Though I agree, this one is a fairly obvious optimisation, not even that really as I would have wrote it the suggested way originally. If there was a particular reason I would be quite interested as it would be something I would have missed as well.

[DynatOS]

You're optimizing a delay loop.

It is less a question about optimization and more about code clarity. If there isn't any reason to indirectly jump (and I can't see the near jump offering that much in the way of clock cycles), then there is no reason for me to use it

Just a "bad" force of habit I perform when I go through my code for clean-ups/commenting/refreshing and come across things like this

As I said, this was a question directed towards DF, and he is the only one with the correct answer as he is the only one who *really* knows his own design intentions... obviousness or not. Thanks for attempting to assist anyhow