Well yeah, I'm trying to make an strstr function (searches for a string within a string then returns the first offset position e.g. searching for "l" in "hello" would return 3). I can do it using all cmp, lea etc, but I wanna do it using the rep mnemonic.

So I was thinking:

- Load the strings with lea or w/e
- Use one of the repeated ones comparing the two strings until they are either equal or one of them equals to zero. Not sure if there is one for doing this :S
- Compare the two strings until they don't match, then compare the needle string byte to 0 (to see if the whole string was matched). If so then return a number, if not then do it again or something.

Is this the right way of doing this? Obviously I don't want a full code sample as it would be pointless, but any pointers? Sorry for all the questions by the way (just it takes me ages to code in assembly so I will waste soo much time doing this).

EDIT by DOS386 2009-05-30 : enhanced subject

It would be really slow to compare the two strings sequentially without aborting immediately when there's no match, and I don't think it's possible with rep to "break" out of it, is there?

Don't you want to use the special string instructions (stos) and registers (edi,esi) for those tasks?

Hmm, I'm thinking that to get the offset of a possible occurance I am going to have to manually do it as cmpsb will increment both values, which I do not want obviously. From there it should be simple.

Push the placement, do a repz cmpsb (the only problem is if both of the strings are the same and have null chars after then then it will carry on incrementing for ages), once it breaks compare the value in the needle string to 0, if it is then it is the occurance, if not then go back to finding an occurance. Or something like that. Thats what I think it should be anyway.

More help for people with i7: http://www.strchr.com/strcmp_and_strlen_using_sse_4.2

Check out the BlackBook chapter 9 for some nice tricks.

nice, revolution, I wasn't aware of that thanks.