Leo --
[[ Caveat Reador: Extremely dynamic stuff is a pet issue of mine. Keep
your favorite halide handy. ]]
> > You need to account for the possibility that the number of ops in an
oplib
> > could change over its versions,
>
> This does invalidate the PBC, as it's currently done via fingerprinting.
The per-op approach makes fingerprinting obsolete, which is another
reason I'm for it.
> > I think this needs to be done at the op level, not at the oplib level
(as
> > I've
> > detailed before). I believe op{info,func} lookup by name is fast
enough
> > that this can be done as a preamble without too much trouble.
>
> Not really necessary and too expensive IMHO. The language/ops will
> stabilize. Having major changes in opsfiles will invalidate PBCs, as
> e.g. a change from gcc 2.x to 3.x invalidates C++ object files.
I disagree that it is too expensive, but I expect it will require
hard data to settle the matter. Since this is my pet issue, I
expect you won't be surprised when I say invalidating PBC files
isn't necessary, and therefore we shouldn't feel obligated to
follow past practice in that regard.
[...]
Regards,
-- Gregor
