out of rom

[Ringo42]

Using a pic16f877a with PCm 3.225

I can compile my code and the compiler says I'm using 54% ROM. Adding a line of code gives me an
"OUT of ROM, a segment or the program is too large"
error.
I'm assuming it is the segment it is talking about. How do I know how big a "segment" is and how would I fix the error?
the .lst file says largest free fragment = 2048
Is it because a function is too long?
Any thoughts?
Thanks
Ringo
_________________
Ringo Davis

[PCM programmer]

[Ringo42]

Ok, that seems to work. How can I tell how big a function is though? Is there an easy way?
Thanks
Ringo
_________________
Ringo Davis

[PCM programmer]

Supposedly the command line compilers will generate a Statistics file
(.STA) if you specify the +A option. I've never been able to make this
work. I think it's really a PCW or PCWH option only.

You can study the .TRE file, but I've never liked that file. It just seems
like a mess.


If anybody has been able to make a .STA file with the command line
compilers, then please post how to do it.

[treitmey]

This is what I use on pch 3.239 [[ But i have pcwh(the whole shebang)]]
ccsc +FH +T +A +DF +P4 "%f"
%f is repaced with the c file name test.c
top part of *.sta file

[Ringo42]

[teletype-guy]

Hi Ringo:

When you uncomment lines, you are also adding the library function code (eg: strcmp) unless you have called it elsewhere. So you are adding much more code than just the call-and-pass-params stuff.

You are doing some rather large if comparisons. I parse serial strings a different way: the main loop constantly calls a parse function, which checks for a serial char (if no char avail it just exits, and the main loop continues). If a char is available, I drop into a switch statement based on a variable "rx_state" -- rx_state starts out as "IDLE," and in this switch case I check for the first char of the desired string. If I find the desired char, I update the rx_state (eg: GOT_1ST_CHAR) and exit. The next time through, if another char is available, it drops to the next state case to check for the second desired char. Yada, yada, so on and so forth.

In my code, I'm looking for a specific header string, and then grabbing comma-delimited ascii-encoded-decimal numbers, until I hit a CR or LF to terminate the command -- at this point the rx_state goes back to the IDLE state, waiting for the next command. If any error occurs during command parsing (bogus chars, etc), you can flag the error and return to IDLE state.

It's all pretty fast and compact, and can be extended to multiple strings. Just a thought.

gil

[PCM programmer]

You probably could fix the problem by putting #separate on all the
routines that are called by Process_Commands(), but I dislike doing that.
That's because the compiler will no longer automatically limit the stack
usage to the hardware limit of the PIC. For 16F PICs, this is only 8.
You have to check the .LST file, each time after you compile.

Here's another option:
Break up your Process_Commands() function into two pieces.
The first one could be called "Parse_Commands()". It would look
at the incoming command string and decide which one it is, and return
a number indicating which command was received. A 2nd function,
"Do_Command()" would perform the command. Example:


cmd_code = Parse_Commands();

Do_Command(cmd_code);


void Do_Command(int8 cmd)
{
if(cmd == 0) // Null command ?
return;

if(cmd > MAX_CMD_CODE)
return;

switch(cmd)
{
case CMD_RD_SRF:
read_srf08();
break;

case CMD_RD_CMPS03:
read_cmps03();
break;

default:
}

}

I don't guarantee this will fix the problem, but it might work.
Also it's possible that using "if-else if" might work better than
switch-case.

[Ringo42]

In my #int_rda function I grab all the bytes until I get a CR, then I call the function I listed before. I was trying to get the strcmp in there to get it working so I could replace all those letter by letter comparisons with 1 strcmp and hope it is smaller that way. I still have several commands I would to add. I guess I could just check the first letter then call another function (based on that letter) to check the rest.

But I guess the question still remains, is there a way to tell how big a function is without adding stuff to it until it will not compile?

Thanks
RIngo
_________________
Ringo Davis

[PCM programmer]

I just emailed CCS and asked them to enable the Statistics file option
for the command line compilers. If they reply, I'll post the answer to
this thread.

[Ringo42]

[PCM programmer]

[Ringo42]

Thank you very much for the help. I'll definitely put the comment you suggested in the code.
Thanks again
Ringo
_________________
Ringo Davis

[PCM programmer]