This is the mail archive of the gcc@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]
Other format: [Raw text]

how difficult to make UNITS_PER_WORD 1


Hello,

I try to make estimation of time and efforts to implement support of
one CPU in gcc 4.x.

The main problem, as can I see, that CPU can only access memory  using
 32bit word as unit.

I try to play with ggx:
http://spindazzle.org/ggx/

and force it to work with memory like above mentioined CPU:

#define UNITS_PER_WORD 1
#define BITS_PER_UNIT 32
#define BITS_PER_WORD 32
#define POINTER_SIZE 32

but this cause a segfault in builtin_define_float_constants.

The old implementation for this CPU based on 3.3.x branch uses trick like this:
#define QImode SImode

but now it does not work.

I define Pmode to QImode but this also not helps, it still catch segfault.

Can anybody with deep knowledge in gcc, describe a little,
the main steps to force compiler generate code for UINTS_PER_WORD==1 target?


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