[log in to unmask] wrote:
>Since Van is raising a performance issue, I'm wondering if he is referring
>to array arguments. In some of these situations with scalar arguments,
>in-lining ought to deal with the performance issue, and (ideally) shouldn't
>that do away with the deep vs shallow issue?
>
>Tim
If the implementation does inlining, the means of parameter passing
for external calls becomes a moot question. In that case the data
flow analysis (done after the inlining) would eliminate any truly
redundant copying and only create temporary space when its's
really needed to avoid clobering still "live" data. I was only
answering the question I thought he was asking (parameter
passing to independently compiled externals).
--
J. Giles
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|