What is padding and detecting it...

[Facultas]

Right,

Im trying to discover whether the machine my program is running on is currently using padding, and just to make sure im trying to find the right thing am i right in saying that padding is extra bytes added to the top of the stack that the stack can 'grow into'? And can it also be unused bytes added to a variable that has been put on a stack in order to make it conform to say, a 16 byte alignment for example? If this is right, is there any simple way of checking these?

[RedSon]

Right,

Im trying to discover whether the machine my program is running on is currently using padding, and just to make sure im trying to find the right thing am i right in saying that padding is extra bytes added to the top of the stack that the stack can 'grow into'? If this is right, is there any simple way of checking this?

Padding does mean adding extra bytes, but I have never heard it in reference to the stack. The call stack builds dynamically so adding padding to the stack would be redundant and not useful. Usually applications pad their memory requests and maybe the files that they write.

[JosAH]

Padding (either internal or external) is all about alignment of data. Most
processors want their ints on a four byte boundary and their doubles on either
a four byte boundary or even an eight byte address boundary. You can find out
what padding your processor uses by checking this:

struct {
   char dummy;
      // <--- internal padding here
   double d; // or int i, or short s or whatever
} padding_t;
...
printf("alignment of double: %d\n", offsetof(padding_t, d));

kind regards,

Jos

[RedSon]

Padding (either internal or external) is all about alignment of data. Most
processors want their ints on a four byte boundary and their doubles on either
a four byte boundary or even an eight byte address boundary. You can find out
what padding your processor uses by checking this:

struct {
   char dummy;
      // <--- internal padding here
   double d; // or int i, or short s or whatever
} padding_t;
...
printf("alignment of double: %d\n", offsetof(padding_t, d));

kind regards,

Jos

Good point. Some systems require that memory requests be on page boundaries. So when you request memory you have to request at least one page. You will also need padding for this type of memory request.

[Facultas]

Padding (either internal or external) is all about alignment of data. Most
processors want their ints on a four byte boundary and their doubles on either
a four byte boundary or even an eight byte address boundary. You can find out
what padding your processor uses by checking this:

struct {
   char dummy;
      // <--- internal padding here
   double d; // or int i, or short s or whatever
} padding_t;
...
printf("alignment of double: %d\n", offsetof(padding_t, d));

kind regards,

Jos

Ah this is the same code (which is really useful btw :) ). you put into my alignment post. So padding is basically the unused bytes put in to 'pad' out the memory so that the variable being put into memory is aligned on the correct byte?

[Banfa]

So padding is basically the unused bytes put in to 'pad' out the memory so that the variable being put into memory is aligned on the correct byte?

Yes. However in a structure using bit fields it can also be extra bits put into the structure (memory) so that the next field is aligned to the correct memory boundary.

It is possible (especially in embedded programming where ram tends to be more limited) to waste a significant amount of memory but poorly designing your structures and ending up with a lot of unrequired padding.