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: share all int csts

From: Nathan Sidwell <nathan at codesourcery dot com>
To: Mark Mitchell <mark at codesourcery dot com>
Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>, Zack Weinberg <zack at codesourcery dot com>
Date: Thu, 28 Oct 2004 09:38:50 +0100
Subject: Re: share all int csts
Organization: Codesourcery LLC
References: <417E534A.6000907@codesourcery.com> <4180305C.7090507@codesourcery.com>

Mark Mitchell wrote:

Nathan Sidwell wrote:
This patch implements a hash table to back the INT_CST vector caches.
When an integer is small, it is still placed in the type's vector cache.
When it is large, it is placed in this hash table.
There was no difference in unchecked bootstrap time on darwin.
It shows a small speedup ~0.2% on qt and similarly a small increase
for geralds PR 8361 testcase.
booted & tested on i686-pc-linux-gnu, ok?
Yes, assuming that constants with TREE_OVERFLOW don't end up in there. (I see the hash function doesn't check for that.) The hard-coded 1024


TREE_OVERFLOW cannot currently be set on a shared (small) int, so we're
ok for big ones too.

for the initial hash table size isn't very attractive, but it doesn't seem very --param-worthy, either.

nathan

--
Nathan Sidwell    ::   http://www.codesourcery.com   ::     CodeSourcery LLC
nathan@codesourcery.com    ::     http://www.planetfall.pwp.blueyonder.co.uk

References:
- share all int csts
  - From: Nathan Sidwell
- Re: share all int csts
  - From: Mark Mitchell

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