Cantaro sim -quabala !
-
... doesn't mean anything, but it made you look... Hello! I'd like a guru to take a look at this: if I use a pragma pack on a STRUCT like this: #pragma pack(1) // aligment of 1 byte(was 4) struct CODE { unsigned char cType; union { DWORD iValue; FARPROC lpAdress; unsigned char * lpString; }; }; #pragma pack(4) // aligment of 4 bytes I will save out the 3 bytes never used with data alignment of 4 bytes, but does this slow down handling the struct? ( or "Where's the catch?" or "Why isn't everything aligned 1 byte?") Thank You ! Georg Haan (.NL)
-
... doesn't mean anything, but it made you look... Hello! I'd like a guru to take a look at this: if I use a pragma pack on a STRUCT like this: #pragma pack(1) // aligment of 1 byte(was 4) struct CODE { unsigned char cType; union { DWORD iValue; FARPROC lpAdress; unsigned char * lpString; }; }; #pragma pack(4) // aligment of 4 bytes I will save out the 3 bytes never used with data alignment of 4 bytes, but does this slow down handling the struct? ( or "Where's the catch?" or "Why isn't everything aligned 1 byte?") Thank You ! Georg Haan (.NL)
Georg Haan wrote: but does this slow down handling the struct? ( or "Where's the catch?" or "Why isn't everything aligned 1 byte?") Yes, in a big way. I believe most current processor instructions expect 4, 8, 16 or 64 byte alignment. In a world where memory is cheap (or at least cheaper than speed), you should use the default alignment the compiler selects for the appropriate build target. /ravi "There is always one more bug..." http://www.ravib.com ravib@ravib.com
-
Georg Haan wrote: but does this slow down handling the struct? ( or "Where's the catch?" or "Why isn't everything aligned 1 byte?") Yes, in a big way. I believe most current processor instructions expect 4, 8, 16 or 64 byte alignment. In a world where memory is cheap (or at least cheaper than speed), you should use the default alignment the compiler selects for the appropriate build target. /ravi "There is always one more bug..." http://www.ravib.com ravib@ravib.com
Don't ask me why, but with my code, I'm somethimes searching for a CODE with ctype = 6 and iValue = 4, with bytealignment = 1 I can do a while((*(short*) &Ops[p02++].cType)!=(short)((0x100*OPSEP)+4)) where Ops is a CODE array, p02 a counter and OPSEP = #define OPSEP 6 Hmz, Maybe if I add another bogus short and char in front of the ctype... Comments? Thanks. Georg Haan (.NL)
-
Don't ask me why, but with my code, I'm somethimes searching for a CODE with ctype = 6 and iValue = 4, with bytealignment = 1 I can do a while((*(short*) &Ops[p02++].cType)!=(short)((0x100*OPSEP)+4)) where Ops is a CODE array, p02 a counter and OPSEP = #define OPSEP 6 Hmz, Maybe if I add another bogus short and char in front of the ctype... Comments? Thanks. Georg Haan (.NL)
Relying on byte alignment is dangerous and makes for hard to maintain code. /ravi "There is always one more bug..." http://www.ravib.com ravib@ravib.com
-
Relying on byte alignment is dangerous and makes for hard to maintain code. /ravi "There is always one more bug..." http://www.ravib.com ravib@ravib.com
&;
