IA64 bootstrap problem for Ada

[Zack Weinberg]

kenner@vlsi1.ultra.nyu.edu (Richard Kenner) writes:

>     This is a somewhat new thing.  It isn't inconsistent.  It means the
>     mode has 96 bits of value but takes up 16*BITS_PER_UNIT (128) bits of
>     storage.  Think TYPE_SIZE versus TYPE_PRECISION.  BImode has been like
>     this since its invention.  XFmode being like this is new.  Other modes
>     are likely to take advantage of this in the future.
>
> But you didn't answer my second (and more important) question: what
> should TYPE_SIZE, TYPE_SIZE_UNIT, and TYPE_ALIGN be for such a thing?

Sorry, I missed the second question.

For most purposes these modes need to be regarded as having their full
storage size.  Thus, TYPE_SIZE, TYPE_SIZE_UNIT, and TYPE_ALIGN should
be consistent with GET_MODE_SIZE and GET_MODE_ALIGNMENT, not with
GET_MODE_BITSIZE.

It is especially important that GET_MODE_ALIGNMENT be honored, because
the IA64 ABIs require us to override the normal computation of
alignment.

zw