SX/B Compilation difference between SX18 and SX28

Peter Verkaik

Hi,

I noticed a compilation difference between SX18 and SX28 regarding assigned VAR BYTE (1) locations.
In the attached file I have 19 ram bytes that on the SX18 occupy $0C-$1E and on the SX28 $0D-$1F
and 5 array bytes that occupy $F0-$F4 (forced using __RAM($xx)).
If I use VAR BYTE (1) the array bytes are assigned to these locations for a SX18,
but for SX28, the SX/B compiler tries to assign $FD and higher, resulting in an
EXCEEDING RAM error.

Is there a reason why $F0-$F4 can not or should·not be used for array variables on the SX28 ?
The·SX/B compiler used is 1.51.03

You can select SX18 or SX28 at the top of the file. Now that I forced
the array variables to $F0-$F4 the compilation is always succesful.

regards peter

Bean

Peter,
· I'm not sure why the SX28 won't use $F0 to $F4 ? I don't see any problem.
· I'll check to make sure the beta version acts properly.

P.S. I checked the source code. I assume two memory areas ($F0 to $F4) and ($FD to $FF) to the same array element for the SX28. So area ($F0 to $F4) will never be used.

Thanks for catching that Peter. I will make note of it. Be aware that the next version of SX/B puts these area together. And the locations from $F0 to $F7 are free for the SX18, and $F0 to $F6 are free for the SX28.

Bean.


▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Did you know that 111,111,111 multiplied by 111,111,111 equals 12345678987654321 ?

www.iElectronicDesigns.com



Post Edited (Bean (Hitt Consulting)) : 5/6/2008 12:49:50 PM GMT

Peter Verkaik

Bean,
I also can not reason why SX/B 1.51.03 for the SX28·skips $F0-$F4, but it does.
Luckily I can force it to use $F0-$F4.

Regarding the beta SX/B: when I compile the program with the beta,
the generated video is not good, in the sense that it appears as if
the first 3 rows are skipped, then there is a row, followed by another
skipped row, followed by another row.

xxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxx
tttttttttttttttt
xxxxxxxxxxxxxxxx
tttttttttttttttt

where x is simply blank and t is a character.
I have not compared both listings but it looks the beta
generates code that uses more cycles (or less?) but
significantly different so it shows on the video output.
So I keep SX/B 1.51.03 for this program.

regards peter

Rsadeika

Does this apply to the SX48/52 in anyway? The last time I used my SX52 protoboard with a program that tried to use the BANK directive, I kept getting unexpected results, like it did not find the specific array.

Thanks

Bean

No, the SX18, SX20, SX48, and SX52 are not affected. Only the SX28 devices.

Manually changing the BANK is really touchy in SX/B. It only really makes sense if you are using in-line assembly code.

If you post your code and explain where the trouble area is, I'll take a look at it for you.

Bean.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Did you know that 111,111,111 multiplied by 111,111,111 equals 12345678987654321 ?

www.iElectronicDesigns.com

·