Re: HOST_WIDE_INT

[qinfeng dot zhang at philips dot com]

Although the target machine is 64-bit, but you need not define all data type as 64-bit.



                                                                                                                                                                 
                                                                                                                                                                 
                                                   To:   gcc@gcc.gnu.org                                                                                         
                                                   cc:   (bcc: Qinfeng Zhang/SHA/RESEARCH/PHILIPS)                                                               
                                                   Subject:    HOST_WIDE_INT                                                                                     
                                                                                                                                                                 
               Roberto Belloni                     Classification:                                                                                               
               <rbelloni@yahoo.com>                                                                                                                              
                                                                                                                                                                 
               Sent by:                                                                                                                                          
               gcc-owner@gcc.gnu.org                                                                                                                             
                                                                                                                                                                 
               2003-06-13 18:08                                                                                                                                  
                                                                                                                                                                 
                                                                                                                                                                 




Hello

I'm hoping that someone can help me to better
understand my problem with the macro HOST_WIDE_INT.

I'm working on gcc 3.2.2, I have a 64 bit target
machine and my host
machine is a pentium.
So these are my target macros in local.h:

#define BITS_BIG_ENDIAN 0
#define BYTES_BIG_ENDIAN 0
#define WORDS_BIG_ENDIAN 0

#define BITS_PER_UNIT        64 // <-- NOTE
#define BITS_PER_WORD        64
#define UNITS_PER_WORD       1
#define POINTER_SIZE         64

#define PTRDIFF_TYPE        "int"
#define SIZE_TYPE           "int"
#define POINTER_BOUNDARY     64
#define PARM_BOUNDARY        64
#define STACK_BOUNDARY       64
#define FUNCTION_BOUNDARY    64
#define EMPTY_FIELD_BOUNDARY 64
#define BIGGEST_ALIGNMENT    128
#define STRICT_ALIGNMENT 1

#define INT_TYPE_SIZE 64
#define SHORT_TYPE_SIZE 64
#define LONG_TYPE_SIZE  64
#define LONG_LONG_TYPE_SIZE  64
#define CHAR_TYPE_SIZE  64
#define FLOAT_TYPE_SIZE        64
#define DOUBLE_TYPE_SIZE       64
#define LONG_DOUBLE_TYPE_SIZE  64

First, I have a problem with the function
"assemble_real" in varasm.c because it works only
if BITS_PER_UNIT is 8, 16, or 32 bit and not 64.
So I'm writing the 64 bit case.

void assemble_real (d, mode, align)
     REAL_VALUE_TYPE d;
     enum machine_mode mode;
     unsigned int align;
{
  long data[4];
  long l;
  unsigned int nalign = min_align (align, 32);
  switch (BITS_PER_UNIT)
    {
    case 8:
       //...
     break;

    case 16:
      //...
      break;

    case 32:
      //...
      break;

    case 64:
    // NEW CASE ...!
    break;

    default:
      abort ();
    }
}


Second, the macro REAL_VALUE_TO_TARGET_SINGLE(d, data)
doesn't work because its code is (about)

union {
    REAL_VALUE_TYPE f;
    HOST_WIDE_INT l[2];
} u;

if (sizeof(long) * 2 < sizeof(REAL_VALUE_TYPE))

    abort ();

u.l[0] = u.l[1] = 0;
u.f = d;

if (HOST_FLOAT_WORDS_BIG_ENDIAN ==
FLOAT_WORDS_BIG_ENDIAN)
{
    data[0] = u.l[0];
    data[1] = u.l[1];
}
else
{
     data[1] = u.l[0];
     data[0] = u.l[1];
}


Where data is a long vector and a long is a 32 bit
type.
REAL_VALUE_TYPE and HOST_WIDE_INT are two 64 bit
types.

Q: Why is HOST_WIDE_INT a 64 bit type?
Q: Did I make any errors?

Of course the macro works if I use long instead of
HOST_WIDE_INT.

Thanks for any clarifications.

Bye,

Roberto Belloni


______________________________________________________________________
Yahoo! Mail: 6MB di spazio gratuito, 30MB per i tuoi allegati, l'antivirus, il filtro Anti-spam
http://it.yahoo.com/mail_it/foot/?http://it.mail.yahoo.com/