about gcc compiler option

Nicholas Leippe nick at byu.edu
Mon Nov 28 08:53:06 MST 2005


On Monday 28 November 2005 07:55 am, Dave Smith wrote:
> Indeed, even on modern x86 systems gcc may pad the 1-byte "bool" with 3
> unused bytes to word-align it, especially if it is the last member of a
> struct. Not sure what this would do on a 64-bit system...
>
> For example, unless the -fpack-struct is specified, this structure will
> occupy 8 bytes (as reported by sizeof):
>
> struct {
>     int i;
>     char c;
> } foo;
>

It gets even better.  Since the char* type was historically used to specify a 
pointer to void for memory allocation, any malloc() must return a pointer to 
a memory region that is properly aligned to store any possible structure.  
Eg, on alpha that requires double-word alignment, two consecutive calls to 
malloc(1) will return two pointers offset by at least 8 bytes (the 7 in 
between are basically wasted).

Last I checked this was the case for both malloc() in c and new char[] in c++.

This allows you to do the following in C++:

char* s = new char[sizeof(MyStruct)];
MyStruct* m = static_cast<MyStruct*>(s);
new (m) MyStruct(<args>); // calls MyStruct's ctor with this == m



-- 
Respectfully,

Nicholas Leippe
Sales Team Automation, LLC
1335 West 1650 North, Suite C
Springville, UT  84663 +1 801.853.4090
http://www.salesteamautomation.com



More information about the PLUG mailing list