Thread: Structure padding

Am I right to assume that compilers will pad out to a multiple of 4 on x86 platforms?

Is there a way, such a macro, to determine the size?

This is compiler dependent, but usually on x86, a char will be 1-byte aligned, shorts 2-byte aligned, ints/floats 4-byte aligned.

I assume most compilers do. But I am pretty sure you can cook up a way to investigate this.

First idea that comes to my mind, is to try creating a struct which obviously would be padded if padding is done to 4 byte boundary. For example:

typedef struct SPadTest
{
char byte1;
char byte2;
unsigned long long int eightbytesPropably;
char yetAnotherByte;
}SPadTest;

Now sizeof(SPadTest) should give you an idea.

and

Code:

unsigned int i;
SPadTest test;
memset(&test,0,sizeof(SPadTest));
test.byte1=0xFF;
test.byte2=0xFF;
test.eightbytesPropably=0xFFFFFFFFFFFFFFFF;
test.yetAnotherByte=0xFF;

for(i=0;i<sizeof(SPadTest);i++)
{
    printf("0x%x",*(((char *)&test)+i));
}

should give you a strong hint.

There's also #pragma(s) to set desired padding, like #pragma pack() - but that probably is compiler dependant too.

So if you need to ensure padding with compilers you do not own, then I guess it is best to try inventing reliable way to detect padding using sizeof() and some test structs. (Or then someone will tell you the macro/exact answer - although I have not heard of such macro - especially any such macro that would reliably work on all compilers )