Compiler won't allow array[10][10]

[elizabeth ann]

i have a simple question.....


how can i create a 10x10 byte matrix? my compiler does not allow me to create an array with that size and instead goes into error that says:

NOT ENOUGH RAM AVAILABLE or DATA ITEM TOO BIG

my compiler version is CCS C 4.057....and my PIC is 16F877A...

thanks!

[Ttelmah]

On the 16 chips, arrays _must_ fit into a single bank. On the 16F877, there are no 'full size' banks. Instead, it has 96 bytes in the first bank, 80 bytes in the second, and 96 bytes each, in the third and fourth banks. As such, even an array of 100 integers, won't ever fit on this chip.
You need to use two smaller arrays (2*50 byte arrays will potentially fit), or upgrade to a larger chip. If the array contains anything 'larger' than int8 values, the position gets even worse.

Best Wishes

[elizabeth ann]

[Ttelmah]

Have you got #device *=16 selected?. If not, the compiler can _only_ access the 96 bytes in the first bank, and then memory will run out very quickly...
If these are arrays of 8bit integers, and you haven't got much else in use, it should work, with two 50 element arrays.

Best Wishes

[elizabeth ann]

tried inserting that line:

#device *=16

but still can't create a 2x50-sized array...

how about migrating the code to some PIC18 chip? would that be too troublesome or not at all.............?

thanks!

[Ttelmah]

Obviously the ability to create 2* 50 byte arrays, will still depend totally on how much RAM is left. As for migration, it depends completely on how your core is written. If you only use standard CCS functions, then with the header redefined to suit the larger chip, and a few 'extras' (obviously for example, you need to do extra 'setup' commands for any extra peripherals present), translation can be a matter of only a few minutes work. However if you perform your own direct I/O to byte/bit values, or have assembler sections, then these take longer. The 18F4520, is normally a good upgrade route from the 16F877, and also brings the advantage that if you are running off a crystal below 10MHz, you can get an immediate *4 speed gain, by using the H4 fuse as we4ll, to give internal 4* clock multiplication.

Best Wishes

[crystal_lattice]

Hi i've successfully implemented at least five 10x10 int8 arrays on a 16f877, as constant. Not sure what you wanna do but defining it as constant will write it in internal program flash and be read-only, i used it as an look up table.

Regards

PS. if memory serves me correct i saw an old post regarding the combination of smaller arrays to form one large array, not sure what it was used for but recall a discussion on such a topic.

[elizabeth ann]

really? how come i can't compile a simple code as that?
it always halts with the same error..

by read-only you do mean that everything is lost on a power failure, am i right? but i need something that is permanent...stays even after power loss, that is.

nevertheless, would you be kind enough to show an example to start with for a newbie like me ? thanks!

[Matro]

"Read-only" means that the array is ... readable only. You can't write (chage) any value.
You need something non-volatile so you have to use either program memory or EEPROM memory.
The best in your case is probably to use the following method :

[Ttelmah]

Attached below, is a demo program, to show how to access two arrays as if they were one, on this chip. This compiles, works, and only uses just over 30% RAM, on your compiler version. It is setup to use a 20MHz crystal, and output on the serial.

[crystal_lattice]

Matro has done the honors of supplying the code, this array is stored as hardcoded stuff, always same after power up/ not possible to change during normal program execution. unless you use routine to alter program memory...

Regards