Possible to use partial 16it Ints in a struct ?

James B · Guest

I have a message coming from I2C from another PIC that has 8 bytes in it, split into 4 16bit signals, the first 5 bits are unused, the next is the sign bit, then 10 bits for the actual signal.

The binary looks like this:
00000 1 0000111100 == +60

I want to use structs to easily use these.

struct MySignal {

int16 signal:10;
int1 signbit:1;
int void_rest:5;
};

I get the compiler error "Number of bits out of range".

If this doesn't work, is it possible for me to split up an int16 ???

treitmey · Joined: 23 Jan 2004 Posts: 1094 Location: Appleton,WI USA

The error is because you can only use the " : " to break down a int8 value.

Soooo.. If I were you I would use shifts and masks to make the value.

James B · Guest

Well, that solves that.

I was hoping to avoid int32 and shifts, it seemed a struct would faster. Esp. since this is going to be a very repetitive move. And my micro is running pretty slow to save power.

Too bad the : doesn't work for int16.

James B · Guest

Ok, I settled on the fact I'de have to shift the data around, but now I'm having a similar problem with my pointer / struct. This is a little of topic with my last post but.....

The one I posted before just had 2 bytes in it, but my real one setup has two different stucts of 8.

My idea was to use the values returned by my getd(int *data, &type) function and apply whatever struct was appropriate. For example

This are my two structs :

Guest · Posted: Mon Mar 06, 2006 4:24 pm

It is worth perhaps looking at some other ways of doing this. For instance:

Ttelmah · Guest

Beware, that array operations (and pointers), take a relatively large number of operations...
Your declaration syntax should be OK, except the declaration you have posted, has the structure name misdeclared. To access an element in a pointer to a structure, use the '->' declaration, instead of the '.' declaration. You also need to actually declare the structure somewhere, and set the pointer to address it. So:

James B · Guest

I see now, except for one thing. I'm doing this so that I can always use my getd(*data) which returns my *data with 8 new elements and also link structs to it so that I can access the vars by name.

I kind of want to link multiple structs to the same changing data. If that makes any sense.

For example:

getd(data,type);

if (type ==2)
second_struct.data1 = 250;
else
first_struct.data1 = 65000;

I'm not sure but I don't 'think' this would be very memory or cpu intensive.

James B · Guest

I guess so imagine I have two structs and 1 source of data coming in. I want to link both so that I can do this:

struct animals {
int pig;
int cow;
int horse;
int duck;
}

struct equipment {
int tractor;
int truck;
int16 pipes;
}

int *data[8];
...
getd(data,type);
...
if (type == animals)
d = animals.duck;
else
p = equipment.pipes
...

So basically my data is always stored in the data[] but I can symbolicly access it by its type.

I dumbed down the example (obviously) to explain myself better. I'de rather not change the way I collect data in the getd().

This seems like a job for unions but I'm not sure how to aligh to an array multible times.

Thank You,

Ttelmah · Guest

Linking different names, is definately a name for unions.
Basically, if you declare two structure 'names':

James B · Guest

Thank You Ttelmah

I get it, but it seems I am making the program more complex then it needs to be. I was just looking for an easy way to call data by symbolic names but for as much as I'm doing that I may as well just hardcode it in (x = data[5])

It won't be as easy to read and I'm sure to forget thing here or there but it does keep it simple.

Thanks Again!

treitmey · Joined: 23 Jan 2004 Posts: 1094 Location: Appleton,WI USA

Ttelmah · Guest

For some types of operation, the union/structure approach, may be the 'best', but for many (most?) likely embedded applications, careful use of your own direct accesses, and good comments will be better/simpler.
One thing it might be worth considering, is use of the #define.
So:

Ttelmah · Guest

It would help if I had an int8, and int16 declarations in the union. Brain elsewhere...

Best Wishes

James B · Guest

I thought about define and didn't really cling to that since I'de be calling from an array, and like you said that adds a few instructions. I was looking at your post and was like "Huh.... apparently you don't need to give it a data type there", but then came to my senses with your last post Embarassed

I did have an idea that isn't exactly what I wanted to do but.. How about declaring a *data[8], then assigning certain bytes inside it, based on thier memory

Sort of like the #byte command, but is it then required for me to locate *data?

ie:

int *data[8]; //lets say it starts at 0x100;

int &pig = 0x103;

int16 pipes = 0x101;

I figure when I update data it will only be meaningfull depending on the address/name I use, which is kind of what I wanted in the first place, but not really, I'm dumb so I'm comprimising.

Can I do that ? Also, can it be done without having ME locate *data? Or is this a waste of resources ?

Ttelmah · Guest

If you access a _constant_ element in an array, the compiler sorts it all out at compile time, and directly accesses the element (only a couple of instructions). If you use a pointer, or a variable, the maths has to occur at runtime. Hence the define gives quick accesses, but using a pointer doesn't.
Sorry...

Best Wishes