This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH]: Allocate bbs/edges using pool



On Thu, 2 Jan 2003, Andreas Jaeger wrote:

>
> Daniel,
>
> did you bootstrap and regtest the patch?
Yes, on i686-pc-linux-gnu and powerpc-apple-darwin6.3

>On which platforms?  Does it
> give any performance improvements?
Yes, it improves performance between 1 and 10%, depending on platform, on
testcases i have with more than 1000 bb's.

The et-forest patch I submitted weeks ago does more, improving performance
10-15% on testcases i have that consist of mainly generated code (IE where
you end up with dominator computation being 10-20 seconds, it's now 10,
rather than 20).

Even if it gave 0% performance improvement (and no performance loss), it
still removes a lot of ad-hoc code in favor of a standardized interface,
and removes usage of the flow_obstack.

 > > Daniel Berlin <dberlin@dberlin.org> writes:
>
> > 2002-12-30  Daniel Berlin  <dberlin@dberlin.org>
> >
> > 	* cfg.c: Include alloc-pool.h
> > 	(edge_pool): New pool.
> > 	(bb_pool): New pool.
> > 	(first_deleted_edge): Remove.
> > 	(first_deleted_block): Remove.
> > 	(init_flow): Alloc/free the pools.
> > 	(free_edge): Use pools.
> > 	(alloc_block): Ditto.
> > 	(expunge_block): Ditto.
> > 	(cached_make_edge): Ditto.
> > [...]
>
> Andreas
> --
>  Andreas Jaeger
>   SuSE Labs aj@suse.de
>    private aj@arthur.inka.de
>     http://www.suse.de/~aj
>


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]