This is the mail archive of the
mailing list for the GCC project.
Re: Inlining on trees
- To: Mark Mitchell <mark at codesourcery dot com>, "Jeffrey A. Law" <law at cygnus dot com>, Jim Wilson <wilson at cygnus dot com>
- Subject: Re: Inlining on trees
- From: Richard Henderson <rth at cygnus dot com>
- Date: Wed, 3 Nov 1999 21:42:10 -0800
- Cc: gcc at gcc dot gnu dot org
- References: <19991103152632V.email@example.com>
On Wed, Nov 03, 1999 at 03:26:32PM -0800, Mark Mitchell wrote:
> Our customer has asked us, therefore, do inlining at the tree level in
> C++, and in such a way as to avoid this problem.
I would like to see such tree-based inlining completely replace
the current integrate.c. I would like to not see us hang on to
the old integrator indefinitely, nuking it sooner rather than later.
Suppose, for the sake of argument, that Cygnus over the next few
weeks converts the C front end to do functions-as-trees as well.
Out of the official GCC tree, that would leave Fortran, Java and
Chill without the ability to inline in any form.
I'm not terribly worried about Fortran; I don't think it will be
that difficult to convert once the interface settles down. It
appears to currently work by building up it's own representation
of the function then converting that en masse to trees and thence
to RTL. Java, out of necessity, already does whole translation
units as trees internally, and so should be trivial to update.
Would this be acceptible to folks?