This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [patches] Re: x86-64 merger part 2 - type sizes


> On Thu, Mar 08, 2001 at 08:40:54PM +0100, Jan Hubicka wrote:
> > + #define BOOL_TYPE_SIZE 8
> > + #define SHORT_TYPE_SIZE 16
> > + #define INT_TYPE_SIZE 32
> > + #define WCHAR_TYPE_SIZE 32
> > + #define MAX_WCHAR_TYPE_SIZE 32
> > + #define FLOAT_TYPE_SIZE 32
> > + #define LONG_TYPE_SIZE (TARGET_64BIT ? 64 : 32)
> > + #define MAX_LONG_TYPE_SIZE 64
> > + #define DOUBLE_TYPE_SIZE 64
> > + #define LONG_LONG_TYPE_SIZE 64
> 
> Why are you defining any but LONG_TYPE_SIZE?
I remember prety clearly, that I run into stubble problems
with definitions of various types according to machine word - for instance
defaults.h says:
#define DOUBLE_TYPE_SIZE (BITS_PER_WORD * 2)
and makes double TFmode.
I've looked at other backends (I think sparc) and seen that it defines all types
so I came to conclusion that it is probably cleanest.

Checking more detaily I need to define DOUBLE/FLOAT/INT/SHORT, since they are
BOTS_PER_WORD based.

I can omit WCHAR and BOOL.  Would you prefer to do that?

Honza

> 
> 
> r~


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]