ALIGNMENT REVISITED 2

[puzzlecracker]

2)

32 bit machine
--------------
1st block: i+pad = 4 bytes
2rd block: s1+pad = 4 bytes
3rd block" c+pad = 4 bytes
4th block: d+ pad = 4 bytes
5th block: j+pad = 4 bytes
---------------------------------------
TOTAL: 20 bytes


64 bit Machine
----------------
1st block: i + pad = 8 byte
2nd block: si + pad = 8 byte
3rd block: c + pad = 8 byte
4ht block: d + pad = 8 byte
5ht block: j + pad = 8 byte


---------------------------------
TOTAL: = 40 bytes



Why would it be differnt from the previous example?


would explain how to use sizeof to determine that as well?
what is POD and ALIGNED ADDRESS?




what is POD?
Thanks

 

[Victor Bazarov]

[...]
what is POD?

Try to learn to use groups.google.com to look up answers that
have already been given multiple times.

 

[EventHelix.com]

The following article should help:

http://www.eventhelix.com/RealtimeMantra/ByteAlignmentAndOrdering.htm

Deepa

 

[Gary Labowitz]

The following article should help:

http://www.eventhelix.com/RealtimeMantra/ByteAlignmentAndOrdering.htm

Interesting. A dumb question: is there anything in standards that prohibits
compiler from rearranging the order of variables declared in a struct or
class? I have worked on compilers that rearranged variables to avoid padding
(old S/360). It wouldn't seem prudent, but perhaps there are compiler
switches for this, or not.

 

[Victor Bazarov]

Interesting. A dumb question: is there anything in standards that prohibits
compiler from rearranging the order of variables declared in a struct or
class? I have worked on compilers that rearranged variables to avoid padding
(old S/360). It wouldn't seem prudent, but perhaps there are compiler
switches for this, or not.

Yes, there is. 9.2/12: "Nonstatic data members of a (non-union) class
declared without an intervening access-specifier are allocated so that
later members have higher addresses within a class object."

V

 

[Gary Labowitz]

Yes, there is. 9.2/12: "Nonstatic data members of a (non-union) class
declared without an intervening access-specifier are allocated so that
later members have higher addresses within a class object."

Thank you. Then &b-&a makes sense for b declared after a. And sizeof(a)
would equal (&b-&a) if no padding were added?

 

[Victor Bazarov]

Gary Labowitz wrote:
[..] A dumb question: is there anything in standards that
prohibits

compiler from rearranging the order of variables declared in a struct or
class? I have worked on compilers that rearranged variables to avoid

padding

(old S/360). It wouldn't seem prudent, but perhaps there are compiler
switches for this, or not.

Yes, there is. 9.2/12: "Nonstatic data members of a (non-union) class
declared without an intervening access-specifier are allocated so that
later members have higher addresses within a class object."

Thank you. Then &b-&a makes sense for b declared after a. And sizeof(a)
would equal (&b-&a) if no padding were added?

I don't see why not. Keep in mind that both 'a' and 'b' have to be in the
same access "section". If they are in different sections or if they have
the same access but there is another section between them, the &b-&a rule
doesn't apply any more.

V

 

[EventHelix.com]

I haven't come across any compilers that reorder the member
variables of a structure.

However, most compilers do have options to choose between
efficiency and memory requirement. For example, if accessing
a 32 integer is efficient at 4 byte boundaries, the default
option would be to align it that way. The compiler may let to align
it at (say) a 1 byte boundary.

This is applicable only on platforms that will allow misaligned
transfers.

Deepa