Library routine argument locations

[Andy Drummond]

I am using PCH version 3.142 to compile code for a PIC18F252.

I have a bootstrap loader routine which uses the standard I2C_Read() & I2C_Write() functions. I can make all code reside in the bootstrap area, and all variables too, EXCEPT for the argument to I2C_Write() which gets put into an ad-hoc location (like @I2C_WRITE_1_31763_31764.P1) that shifts whenever I change the program. I2C_Read has a similar problem.

This completely fouls up the system as the bootstrap loader and the new program both use I2C_Write() but they try and pass the argument in via different locations. Which is tragic.

Since the compiler assigns the location of these arguments, is there any way to force it to locate them at known addresses to avoid this problem? Or does anyone have any other helpful suggestions?

[Neutone]

You cant use the compiler I2C functions in both the boot loader and the main program area because they can not be located. If you write your own functions and place them within the bootloader area you could call them from both main and the bootloader because the function location would become static.

[Andy Drummond]

Not so. You can force the location of library routines to be in the boot loader area, and I have done that. The trouble is that whether I use the standard routines or write my own, the ARGUMENTS to the function call are located automatically by the compiler in some convenient hole left after the other memory references are located.
If I write a function mywrite(char fred) { ... } then call it as mywrite(7), the compiler puts the 7 (or whatever) into a location of its own choice and then the function accesses it there. Trouble is that that location could be anywhere depending on all the rest of the memory assignments. And it can change between two compilations with a minor main-program change. Then the bootloader function looks in one place while the caller puts its argument someplace different. Whoo! Disaster!
If I could use one set of routines (in particular the i2c routines) in my boot area and one set in the main program then I guess it would work, but that would waste a lot of code space and I don't know how to do that either!

[Neutone]

So all you need is to locate a passed variable? This seems like something CCS should support if they don't already.

[mpfj]

> This seems like something CCS should support if they don't already.

Yes ... you can use the #locate directive to force parameter locations !!!

For example ...

[Andy Drummond]

Thank you, "mpfj" !

You have hit the nail on the head! I use #locate often but it never dawned on me that I could use it on a function argument. I shall do that now.

As for the scratchpad variables, yes, that is another can of worms. Anyone know how they can be #located???

[ttelmah]

[mpfj]

[Andy Drummond]

Hey, that IS useful. I always thought the line: