GMP ref count

[Hans Aberg]

I experimented with a quick fix in order to put in all data in the _mp_d
field of __mpz_struct and __mpf_struct.

In file memory.c, I put in

size_t _mp_d_offset     = 4 * sizeof(mp_limb_t);

#define malloc(size)  ((char*)malloc(size + _mp_d_offset) + _mp_d_offset)
#define realloc(oldptr, new_size)  ((char*)realloc((char*)oldptr -
_mp_d_offset, new_size + _mp_d_offset) + _mp_d_offset)
#define free(ptr)  (free((char*)ptr - _mp_d_offset))

in order to reserve memory, and the __mpz_struct and __mpf_struct were
replaced by

typedef struct
{
  mp_limb_t *_mp_d;             /* Pointer to all.  */
} __mpz_struct;

#define _mp_alloc  _mp_d[-3]
#define _mp_count  _mp_d[-2]
#define _mp_size   _mp_d[-1]

typedef struct
{
  mp_limb_t *_mp_d;             /* Pointer to all.  */
} __mpf_struct;

#define _mp_prec  _mp_d[-4]
#define _mp_exp   _mp_d[-3]

However, the code then breaks (as expected) in places such as

mpz_init (mpz_ptr x)
{
  x->_mp_alloc = 1;   /* Reference before allocation */
  x->_mp_d = (mp_ptr) (*_mp_allocate_func) (BYTES_PER_MP_LIMB);
  x->_mp_size = 0;
}

-- As I am not sufficiently familiar with the code, I decided to not now
try to identify the places where __mpz_struct and __mpf_struct fields are
accessed before the _mp_d field has been allocated.

However, if this could be done, one could make an experimental C++ wrap
using a ref count, to see how well it times relative the C version.

  Hans Aberg