Pointers problem

Arye Lerner

Hello,

I run into a problem using pointers on an array. In my application I try to burn SPI FLASH with program I downloaded into array in SDRAM (this application is strictly for testing). This is a pseudo code of a program:

int FLASH_Buffer[40000];

int *ReadPtr, *WritePtr;

WritePtr = FLASH_Buffer;

for loop {

write to FLASH;

read from FLASH;

validate;

WritePtr +=64;
ReadPtr +=64;
}

The problem is that the pointers advanced in steps of 256 instead of 64, as defined in program. I could change the number from 64 to 16, but I need to understand why does this happens. Both buffer and ponter to budder are integers.

Thank You

over 15 years ago

0 L. Kurylo over 15 years ago

Prodigy 40 points

Hello,

you have pointers to int type which size is 4 bytes.

consider this:

WritePtr = FLASH_Buffer;

WritePtr += 1; //this now points to next int

(WritePtr == &(Flash_Buffer[1])) which is next int -> 4 bytes further in the memory

If you have pointers don't use them as table indexes

this is WRONG:

something = FLASH_Buffer[ReadPtr]; this is WRONG

You should use pointers like this:

something = *ReadPtr;

*WritePtr = something;

or declare indexes to table

int ReadIndex, WriteIndex; //NOT pointers and use them like this

something = FLASH_Buffer[ReadIndex];

FLASH_Buffer[WriteIndex] = something;

ReadIndex and WriteIndex would be incremented as you've expected by one for each int.

Pointers on the other side are addresses and each int takes 4bytes hence inrement by 4.

0 Arye Lerner over 15 years ago in reply to L. Kurylo

Expert 1130 points

Hi,

I am aware of those rules, that is actually what I was counting on.

instead of using buffer with index e.g: buffer[index], i declare a pointer to buffer of same type e.g:

int * pointer;

int buffer[40000];

pointer=buffer;

and use pointer to perform read/write e.g. value = *pointer. That suppose to save execution time for CPU.

The problem occurs when I declare pointer of type integer, that is 4 bytes for this particular format, and trying to advance it by 64 integer values i.e. pointer+=64.

New pointer value suppose to point to the 64-th integer in array since, as You said, pointer is an integer, instead it it points to 256-th integer in array.

I suspect that it has somthing to do with DATA ALIGNMENT or the way data is packed.

I am using a big arrays - 40000 integers, maybe this causing a problem.

Regards

Arye lerner

0 Mariana over 15 years ago in reply to Arye Lerner

TI__Mastermind 24340 points

Hi Arye,

Arye Lerner said:

The problem occurs when I declare pointer of type integer, that is 4 bytes for this particular format, and trying to advance it by 64 integer values i.e. pointer+=64.

New pointervalue suppose to point to the 64-th integer in array since, as You said, pointer is an integer, instead it it points to 256-th integer in array.

I suspect that it has somthing to do with DATA ALIGNMENT or the way data is packed.

Whenever you increment a pointer type "int", it will increment 4 bytes in the memory -sizeof(int). It should not increment 32 bytes like you said - I never saw that happen. If your big array is local, make sure to move it to global and re-test it - you could have a stack problem, or even a broken pointer, etc...

0 Arye Lerner over 15 years ago in reply to Mariana

Expert 1130 points

Thank You, Mariana!

0 Mark Moldavsky over 15 years ago in reply to Arye Lerner

Prodigy 180 points

Arye Lerner said:
That suppose to save execution time for CPU.

If you use optimization option 3 you get the same assambly code exectly

Mark

Processors

Processors forum

Pointers problem