Size changes, round 1

Richard Kenner kenner@vlsi1.ultra.nyu.edu
Mon Feb 21 02:29:00 GMT 2000


    The reason I would prefer that this be dealt with using some kind of
    abstracted data type is that it is extremely likely that in the future
    we will want to, again, change the way we represent offsets in structures.

But I don't think that offsets in structures will end up being done the
same way as sizes.  The reason is that we don't currently support variable
bit position, but we often have offsets that are variable and not a multiple
of a byte.  So my proposal was to have offsets be represented as a
sizetype byte offset *plus* a bit offset, which must currently be constant.

I think that's the only way to avoid 64-bit run-time computations for
records with variable-sized objects.


More information about the Gcc-patches mailing list