This is the mail archive of the
mailing list for the Java project.
Re: [PATCH] unit-at-a-time for java
- From: Jan Hubicka <jh at suse dot cz>
- To: Jeff Sturm <jsturm at one-point dot com>
- Cc: gcc-patches at gcc dot gnu dot org, java-patches at gcc dot gnu dot org, jh at suse dot cz
- Date: Tue, 22 Jul 2003 06:56:53 +0200
- Subject: Re: [PATCH] unit-at-a-time for java
- References: <Pine.LNX.firstname.lastname@example.org>
> I thought I'd give this a try, since it enables effective cross-module
> inlining for gcj. Following is the patch I'm testing. There are two
> parts: a patchlet to cgraphunit.c, and the java frontend changes and
> bugfixes I needed to make -funit-at-a-time work.
> Jan, I'm not sure what the comment regarding current_function_decl means.
> However the java inliner requires this to be set to the outermost calling
> function. I don't know what the impact to C/C++ is. (If that won't work
> I suppose we could introduce a new variable.)
Your change will work. the inliner just avoids inlining of
current_function_decl so when the current_funcntion_decl is set to
something that we want to inline we lose. setting it to the function
itself is OK as we don't do recursive inlining yet.
> Calling walk_tree_with_duplicates turned out to be necessary to avoid
> infinite recursion when gathering costs.
this seems to be fine to me too.
thanks for doing all the work! I planned to do it myself sometime but I
had dificulties to find the time so it is great to see it done!