Hex Decompile - TABLAT Register Interrupt Push/Pop question

[PointOnePA]

I'm trying to recreate the build environment for an OLD application using the CCS PCH V4.112 compiler on an PIC18F2420 target . I'm doing this to document that I have the source code to match an old .hex file, as a starting point for my code changes. My problem is that the .hex file from MY BUILD contains the following EXTRA PUSH/POP registers in the interrupt.

[Ttelmah]

None of these instructions are pushes or pops....
They are simply saving/restoring the table pointer registers to scratch.

Honestly, if you need to match the old code, the only way to do it, would be to use #INT_GLOBAL, and manually save/restore the registers to match the old code (and poll the interrupt involved). However I would have expected the table pointers to have been saved, otherwise there will be problems as these are used anywhere.

Optimisation has no effect on the registers being saved. All temporary registers are always saved unless you use INT_GLOBAL and do it yourself.

Are you sure the hex was built for the same chip, and with this compiler?. A quick check shows that a basic interrupt handler built on 4.114, does exactly the same save/restore of these three registers.

[PointOnePA]

Ttelmah, thanks for the information and correction on push/pop. I was hoping that there might be a compiler directive to disable this. The documentation that I have and some build files from the same era, including a list file, indicate the compiler was:

CCS PCH C Compiler, Version 4.112, 12345678 15-Oct-10 14:12

The second to last lines of the .HEX file I'm trying to match states:
";PIC18F2420", and the last line indicates CREATED="18-Dec-09".

so I'm sure that I have the right target, but the compiler could be earlier than 4.112. You mentioned "a quick check shows that ....on 4.114". Where did you find that information, or is that from software you have installed? I'm just wonder if I can look at compilers prior to 4.112 to see if they worked the same way.

If I have to, I can replace the existing interrupt dispatcher with custom code to avoid having these table points written to scratch, but if I can do it with the existing V4.112 compiler directives, that would be preferred.

[Ttelmah]

4.112, also saves these registers.

It has to. Otherwise anything using table addressing, will result in register corruption.

[PointOnePA]

Ttelmah, thanks again for the suggestions.

I did some research using archive.org, following suggestions from forum posts, and found that version 4.112 was not used to create this .hex file because it was not available until Sep 19, 2010, so could not have been used to create this 18-Dec-09 .hex file. So I'm going to look at an earlier compiler to see if it resolves the problem. I agree with you that the table registers need to be saved to avoid corruption, but I can see that they are not being saved in this 2009 .hex file that I'm attempt to recreate. The only reason I want to recreate the file is to validate the C source as a known starting place. Once I know that source, I can update the compiler to a more reasonable version.

[Ttelmah]

4.099, produces a smaller interrupt handler, without those specific saves. I'd guess the code itself does not use those registers in the same way.

If you want to build the code, contact CCS. They have an option in their download page for 'old compiler versions'. They should be possibly able to do 4.099 for you.

[PointOnePA]

Ttelmah, thanks again for the comments.

CCS Tech support has been great. They set me up with V4.099 and I'm checking it out now. I'll follow up with a summary once I've completed this process.

This Forum ROCKS!