Re: [Sbcl-devel] PPC changes for character_branch

[Christophe R. <cs...@ca...>]

Brian Mastenbrook <bma...@cs...> writes:

> On Sep 3, 2004, at 8:43 PM, Brian Mastenbrook wrote:
>
>> Is this really the right thing to do? I haven't been following the
>> character branch but this looks to me like simple-base-string is now
>> a string of characters, not just base-chars. I am making these
>> changes on the sparc backend right now and this was a bit of a red
>> flag to me.
>
> Oops. "we're never going to want to distinguish the CHARACTERness vs
> BASE-CHARness of characters by their widetags, because we can do it
> based on their CHAR-CODE; thus, calling the primitive type and storage
> classes BASE-CHAR is unneccesarily confusing." I guess this makes
> sense in that context, since base-char and character are the same
> storage class.

I sent Brian mail explaining more-or-less this in a little more
detail: for the interested, here's the text of my reply:

  They're right, I think: that's because the types that the backends
  deal with aren't lisp types (where BASE-CHAR and CHARACTER are
  distinct) but "primitive types", or possibly representational types:
  to the underlying system, there is no difference in representation
  between a base-char and a character -- they share a lowtag/widetag,
  but are different in the immediate data.  So, from the point of view
  of the vm, the primitive type that CHAR on a base-string returns is
  a CHARACTER, and it needs to fit in a CHARACTER-REG: we rely on the
  typechecking in layers above to enforce the constraint that only
  BASE-CHARs fit in a BASE-STRING.

Cheers,

Christophe