Migration from one chip set to another

[Guest]

I developed a two chip system using my ICD-S40 with the 16F877A and a 16F84A on a breadboard. I use a RS232 driven LCD for a monitor and a 4 pushbutton input. The system works just fine under the breadboard conditions.

I modified the 16F877A software and ported it to a 16F648A. I modified the 16F84A software and ported it to a 12C508A. I made all of the requisite changes to pin assignements and to the fuses.

I build a prototype using the target PICs. All has been wrung out and all solder mishaps have been corrected. Unfortunately, the prototype does not work properly. About 70% of the functionality is OK. I have been driven to distraction to find what is wrong.

I know that the 16F877A has 8K available. The 16F648a has 4K available. All 4K is being used! Yes, that includes the fragments. If I try to change a delay like delay_us(50) to delay_us(100) I get an error message saying I am out of memory. If I change the delay_us(50) to delay_us(40), all is OK. So I know I am really tight.

Does anyone have any ideas I can try out? Is there something in the differences between the architectures of the two devices that are not being addressed by the compiler?

I won't post the code as it is 14 pages of C code for the two chips.

I am using the PCW compiler, version 3.190 and the Mach X programmer with the latest driver.

[PCM programmer]

Here's an old post that I wrote, that explains how to go through the .LST
file and look for inefficient code. Once you notice those areas in the
program, you can re-write the C source code so it uses less ROM.
http://web.archive.org/web/20020228162153/http://www.pic-c.com/forum/old/messages/1777.html


If you're using the #separate directive, read this thread which talks
about the risks, if the compiler uses too many stack levels:
http://www.ccsinfo.com/forum/viewtopic.php?t=27717

[alchazz]

Thanx, I'll look into this.