Virtual RAM Help

[neochrome32]

CCS C 4:130

Ok, here is a pretty much out there idea... now please note:
* SSD1963 is 800x480 24bit colour LCD (i've set it to use 16bit colours only).
NOW this leaves about 400 KBytes to play with.

I've mapped the memory by a simple function, basically I'm drawing pixels OFF the screen (can't see it and it works!).
Since it is 16bit I'm losing 8bits per pixel (memory cell) which to be fair no worries for now.

Now what i want to do is figure out a way to make my MCU "POINT" to the "External ram"

Right as you know

[temtronic]

comment: It sounds like you want to use the LCD modules RAM for data storage for the progam. That isn't a big deal.The LCD datasheet should have a 'memory map' or table describing how to access it.Even the common 2x16 LCD units have that! Writing to non-display memory locations saves the data, it just doesn't get displayed and if enabled, you can read back the LCD ram data. I imagine the LCD memory may be setup as 'blocks' or 'pages' so that several 'screens' or data can be quickly accessed and displayed.PC video cards do this making for fast access even when using QBASIC4.5 so it should be doable. The 'trick' is to read the datasheet,figure out the locations of 'displayable RAM' vs 'hidden RAM', code the algorithms to access it,try with KNOWN data,recode as required,retest,etc.Speed might be an issue depending on how you want to use the 'vitrual RAM' within your program and only testing will tell.

hth
jay

[Ttelmah]

What you seem to want is addressmod.

This allows the CPU to access variables stored in external memory.

Problem is that while it was working a few compilers ago, it has been broken for the last few - for few, read 'large number' - compilers.

It is one of those 'brilliant' features, that CCS need to get working....

Best Wishes

[neochrome32]

@Ttelmah : Thanks for the info, i didn't know CCS had an external Variable stuff as that would have made things so very simply. Hope they fix it soon, but is there anyone that may have written some sort of algorithm.

@temtronic : I'm aware of the ram available, what I'm after is code to make USE of the ram in terms of CPU pointers, and variables.

Is the code still available for making use of External memory ? not just using ext_write(address,value); I'm talking about making it work using real declarations.

??

[neochrome32]

i run 4.130, was the addressmod function broken after this? or before? i can only go back to 4.114

ok seems to be ok ;

problem is, the addressmod doesn't work all that great with strcpy :(

[neochrome32]

addressmod its precicly what i needed!!! THANK YOU !!!

only issue i have is how to use malloc with it too? it works but it didn't seem to work beyond the pics physical ram

sooo need a way around this too

[jeremiah]

You really really really should avoid using malloc or any dynamic memory. The PIC really isn't suited for that typically. There is almost never a need to use it.

[neochrome32]

There is about 256Kb available to me (after altering the addressmod thing) - so messing about with static memory is ok, just seems extremely wasteful.

I'm thinking writing my own malloc system, be a bit slower, BUT why not

[jeremiah]

Dynamic memory is going to be both more wasteful (on average) and slower than static. With Dynamic Memory, a memory pool will need to be allocated as well resources to manage it. Both of those waste memory that could be used for other things.

I'll agree it is a fun thing to do if you are making your own memory management module (I use to play around making virtual functions for structures in CCS...virtual table and all, so I understand the fun factor), but it usually isn't a good idea for anything commercial grade.

[neochrome32]

BOOO... oh well

ok I'm still having issues though

for some reason,

vram int16 test;

test=50;

sprintf (txt, "%lu", test);

this doesn't work, it says not a valid format! lol is the addressmod only able to use Signed stuff? as its screwed up alot of my code :(

scratched away the malloc stuff though, no need for it i guess then

[jeremiah]

might be a compiler bug for your version. I just tried in 5.010 and it compiled fine. Try swapping the order VRAM and int16 around in front of test.

EDIT: however, in version 5.010 addressmod itself doesn't work properly. I emailed a bug report to CCS support the other day, but christmas season, so probably won't hear from them for a bit.

[neochrome32]

i dont have 5.010 to be honest, 4.130 is the best one that worked for me, and i DID find that addressmod appears to be really in favour of SIGNED values!

even structs, i found that

typedef struct {
int16 x,y,z;
int8 a,b,c;
} test;

this was all messed up, it didn't work until i made a,b,c all 16bit

looks like the struct is widened up to the largest number?

and pointers are messy but they do work now!

THANK YOU



EDIT:: i cant seem to use strcpy(txt, "HELLO WORLD")

eg,
FakeRAM char txt[32];

strcpy(txt, "hello");

doesnt seem to work either

i have to run a loop to insert them from a text from a real string!

[neochrome32]

Sorry another question, i cant seem to find any infomation about addressmod limitations

i realise that the addressmod it self can only address up to 64kb or something like (seems to be able to for me right now) but i need more memory

can i use multiple addressmods?