This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH to enable inlining-on-trees
- To: Mark Mitchell <mark at codesourcery dot com>
- Subject: Re: PATCH to enable inlining-on-trees
- From: Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>
- Date: Tue, 7 Dec 1999 23:37:23 +0100 (MET)
- cc: gcc-patches at gcc dot gnu dot org
On Sun, 5 Dec 1999, Mark Mitchell wrote:
> This patch turns on the tree-based inliner in C++ patch. I wrote a
> blurb (checked in as wwwdocs/htdocs/news/inlining.html); I'll leave it
> to Gerald to link to that, if he decides that's appropriate.
I definitely think it is and will add that to our main page shortly. :-)
How about the following?
CodeSourcery, LLC has contributed a new inliner for C++.
As a result, memory consumption and compile-time may be dramatically
reduced for programs that make heavy use of templates.
Now to the bad news, however, namely the reason why I substituted
"may" instead of "will" in the paragraph above. :-(
Some of our code, for which we abandoned using gcc snapshots a few
months ago because these became essentially unusable, still is causing
quite some troubles:
GCC 2.95.1 ~51MB 0:50
GCC 2.96 19991122 ~120MB 1:15 Ouch, that hurts.
GCC 2.96 19991207 ~120MB 1:15 No change, unfortunately.
Measured on a 360MHz Ultra sparc-sun-solaris2.6 with -O3, similiar
behavior on i386-unknown-freebsdelf3.3.
Would you like to see (preprocessed) source?
Gerald
--
Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/