Error 71 ... Out of ROM, A segment or the program is too lar

[guy]

Ok, now it's my turn:
PIC16F887 (8K words program memory)

*** Error 71 "vlv2.c" Line 100(1,2): Out of ROM, A segment or the program is too large ReadTempSensor
Seg 00800-00FFF, 022C left, need 0070D
Seg 01000-017FF, 043F left, need 0070D
Seg 01800-01EFF, 0700 left, need 0070D
Seg 00000-00003, 0000 left, need 0070D Reserved
Seg 00004-00050, 0000 left, need 0070D Reserved
Seg 00051-007FF, 0003 left, need 0070D

I already split many of the functions and added #SEPARATE. It seems I'm almost there (0x70D needed, 0x700 left) but it's elusive - I split a little more and everything changes again...
Since I have plenty of non-consecutive memory left my question is this:
Does the linker do a good work in fitting the code blocks into available memory blocks, or do I need to help it with #ORG ? I can't see the function sizes until it compiles

[PCM programmer]

[Ttelmah]

The compiler does a very good job of fitting stuff in. Remove something so it compiles, then identify what function it is that is this large in one piece.

[guy]

when I started I thought it was easy. It gets worse and worse.
I #separate all of the large functions and even split functions into sub functions.

Also look what I get now (after commenting out a call to a function in order to compile) - it needs TWO different block sizes...? What am I missing?

[guy]

I tried commenting out each major function, write down the ROM usage % in order to know the size of the function (since I can't compile when all functions are used).
It seems that when I 'fix' something with #separate the compiler moves on to the next challenge which could be a bigger block, hence the growing block size.
And at the end of the day when I sum up the percentages of usage, the whole program would take 113% which explains a few things... No use but migrating...
Good night people and thanks for the tips.

[Ttelmah]

The other thing is to look at 'why' the code is so large?. Something silly like using fp maths when it is not needed, can easily use up all the memory space in chips that are perfectly capable of handling the job required....

[guy]

after I finished sulking about writing unnecessarily large code I remove FP conversion from Celcius to F and the code fit. No need for migration!

One big surprise is that a #SEPARATE statement caused the code NOT TO FIT. Removing it enabled the compilation. Could be that a function *doesn't* have to fit inside a page. Back in the days when I wrote in Assembly I could continue a function beyond a page limit (PC doesn't roll over) as long as I didn't GOTO without handling the pages. What do you think??

With #SEPARATE before handleCcommand :

[temtronic]

re: C to F conversion

f=(C*2)-(.1C)+32

say C is 20
20 * 2 = 40 //this is a simple rotate left
40 - 4 is 36 //just take 10s digit
36+32 = 68//simple add

I do this in my head all the time as Canada is 'supposed' to be metric but I'm old and like *F !

Don't know about cold(-) temps.. when it's that cold I stay inside !


You could probably go through most functions and 'tweak' them to gain some more code space,but time IS money...you don't live forever, AND there HAS to be a pin compatible PIC with more in it for just a few pennies more.

Jay

[guy]

I needed 0.1F accuracy so my code was:

[Ttelmah]

%w and %Lw work with signed numbers.

It's worth understanding that the reason to avoid #separate, is that if used for functions that are called only once, it uses extra space. Splitting the code means extra calls and returns, and moving the data to temporary storage for the call. The reason to use it, is to allow the code block to be subdivided.

This is why it can become such a complex balancing act... :(

[guy]

I used printf("%1.1Lw",signed16) but on negative numbers a bug causes trailing zeros - it produces 000000000.1 for -1 (compiler version 5.048).
This could be solved by changing to printf("%2.1Lw",signed16) which correctly outputs -0.1 and works well from -32768 to 32767.
%w rules!! Thanks Ttelmah.

[Ttelmah]

The reason for this, is that you are overflowing the width.

In C, the field width includes the decimal point. %1.1, specifies to print in a field that is not wide enough to take any value....
While the field will automatically expand, if the number is bigger than will fit, it has to be able to get the decimal part in.

[TimothyCarter]

Your best bet is probably just to simply re-arrange the order of the functions (as has already been mentioned). This has worked for me a few times.

Of course, even better - if time permits - is to shrink the functions down as much as possible.
_________________
“Computer science education cannot make anybody an expert programmer any more than studying brushes and pigment can make somebody an expert painter.”
- Eric S. Raymond