🚀 go-pugleaf

In article <1993Jan23.042626.991@verdix.com> scotty@verdix.com (Scott R. Chilcote) writes:
>
>Is it legal to address this memory in C as follows?
>
>#define ON_CHIP 0x809800

>   float *oc_ptr = (float *) ON_CHIP;

>Our device-specific ANSI C compiler lets us do this in order to address the
>on-chip memory, without a warning.  The emulator shows that it works.  I've
>looked in several textbooks, however, and I've never seen mention of setting
>a pointer equal to a constant in order to define a specific address.  Even the
>new K & R book doesn't mention if this is legal.

It is certainly "legal" in the sense that it is syntactically correct,
but the results of doing this are implementation-defined.  (Your
compiler manual should somewhere have a paragraph explaining what
precisely it defines this expression to mean.)  The ANSI C standard
*does* guarantee (I believe, not having a copy handy) that there is
some flavor of integer that can be cast back and forth to a pointer
type.  (Different kinds of pointers may require different kinds of
integers.)  I do not believe that you are guaranteed to get the same
integer back; i.e., (unsigned long)(void *)0xfe100004 is not
guaranteed to be == to 4262461444UL.

Some people prefer to use a union for type punning; i.e.:

union Pointer {
  void *vp;
  unsigned long ul;
  void (*vfp)();
};

However, I don't believe that this really buys you anything, although
it may eliminate warning messages on some compilers.  (As someone said
in this newsgroup, "A compiler is free to issue a diagnostic if it's
after five on Friday, or it's raining outside.")  The experts will
probably weigh in soon enough.

-GAWollman

-- 
Garrett A. Wollman   | Shashish is simple, it's discreet, it's brief. ... 
wollman@emba.uvm.edu | Shashish is the bonding of hearts in spite of distance.
uvm-gen!wollman      | It is a bond more powerful than absence.  We like people
UVM disagrees.       | who like Shashish.  - Claude McKenzie + Florent Vollant