Unpredicted behiver of code

Hope It cleares my problem and sulution.

No, it doesn't. You're being rather excessively vague.

As a matter of fact, the startup code(s) provided by Keil all initialize the SP to the start of the unused portion of IDATA, as it should be. The stack grows upward from there, until it hits the wrap-around barrier at 0xff. You modify that mechanism at your own peril, especially if you don't know exactly what you're doing.

Hope It cleares my problem and sulution.

No, it doesn't. You're being rather excessively vague.

As a matter of fact, the startup code(s) provided by Keil all initialize the SP to the start of the unused portion of IDATA, as it should be. The stack grows upward from there, until it hits the wrap-around barrier at 0xff. You modify that mechanism at your own peril, especially if you don't know exactly what you're doing.

I did change that file.
The code at first look as a zero mem util.
As I wanted to avoid xdata loop calls and save some time I avoided the A51 with a bypass.

This was a mistake as the SP was not updated.

So, form this post I recommend to carefully remove the XDATA zero calls only if you want to save running time and not cancel it out all together.

Hop now its clear.

The code at first look as a zero mem util.

That's because zeroing static variables is one of several jobs the startup file has.

As I wanted to avoid xdata loop calls and save some time

Doing that will hardly ever save any time. The startup code only loops over XDATA if you use XDATA in your program --- it's clever enough as-is to disable the entire loop if you don't use any XDATA. And unless you're comfortable with your code working with random garbage data, you need to initialize those variables at some point. And whichever way you eventually do that --- it will quite certainly take longer than the way the startup code would have done it.

...as it should be

Very questionable extension to the sentence. There is no rule that says it MUST be put there. It is only a reasonably sensible option that Keil (amongst others) have chosen.

You modify that mechanism at your own peril, especially if you don't know exactly what you're doing.

Absoultely. Someone who knows what they're might have reason to make these changes. It is unlikely that a beginner could conceive a true advantage.

Hi,

The reason to bypass the code was to save time and avoid xdata reset as it was mapped to already initialized HW.
The way to do it is to remark the movx operation and not jump directly to start.

That said,
Thank you all for the replies.

The way to do it is to remark the movx operation and not jump directly to start.

May I remark that I do not know what you mean. comment out??? delete??? something else???

Abuse of old BASIC terminology - REMing lines, as in adding the REM(ark) keyword in front of them to deactivate them.

Since BASIC doesn't have the same huge market share anymore, it's way better to use a more neutral expression like "commenting out" or something similar.

:) Old is relative

Ok then - comment out

Well, old and old. Todays BASIC without line numbers looks very much different from the original BASIC available in home computers with 4kB RAM ;)

But old in the sense that a lot of developers have either forgotten - or never heard about - the use of REM to comment out code lines, because the world have exploded into a huge number of newer programming languages. The old C standard didn't even have a method to comment out a single line - the // one-line comments was a late inherritance from C++.