So I started a while ago doing some z80 asm, particulary to improve some project of mine

I wanted then to divide HL by C at one point. So I read topic about division, and what I get so far is :

http://pastebin.com/9epPYk4D

But it seems that the value return sometimes is 0 and I can't get why.
Do you have any idea ?

I won't have time to look at your code in depth until later this evening, but have you inspected the 16-bit by 8-bit division routines on the z80 Bits page to see how their method compares to yours? You'll have trouble finding routines as optimized as theirs, so I'd strongly consider using them unless you really want to reinvent them as a learning experience.

Mmmmmhhh didn't see this page. Seems pretty good.
.... It seems by the way that they use the same way for the division than I did :p

Strange that I have a bug then.

EDIT : finally spotted the bug. The first itération is not optimized as it have to be --' add hl,hl can't be removed.

EDIT2 : seems however, if c>160, it always return 0

So back with my question

I need to optimze a particular piece of code, wich is abusing of jp :

Code:


Basically, it do abs(a+e), but even with a overflow, it still get the correct result
(a=-64, e=-100, get a=164).

I need to optimize for speed, since this code is reiterated a lot of time. (and size don't really matter).
Thanks for help :]

It looks like the code needs more than optimization, it needs its logic corrected. It incorrectly negates the result in cases like 100+64, 100+-64, and -64+100. Basically, the current logic negates the result if S|C (sign or carry), but I believe the correct logic is S^V (sign xor overflow). So correct code might look like this, which, in a few minutes of thinking, I can't find any way to further optimize:


Code: