[Bug target/21316] [3.4 Regression] pointer lookup cache misses more frequent

[kpmassey at gmail dot com]

--- Additional Comments From kpmassey at gmail dot com  2005-05-04 01:20 
---
Subject: Re:  [3.4 Regression] pointer lookup cache misses more frequent

Actually, no.  Gentoo doesn't yet have 4.0.0 in portage, and being a
novice, I am reluctant to build myself.

On 4 May 2005 01:16:33 -, pinskia at gcc dot gnu dot org
<[EMAIL PROTECTED]> wrote:
> 
> --- Additional Comments From pinskia at gcc dot gnu dot org  2005-05-04 
> 01:16 ---
> I could not reproduce this on a pentium4.
> 
> --
> 
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21316
> 
> --- You are receiving this mail because: ---
> You reported the bug, or are watching the reporter.
>


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21316

[Bug c++/21316] pointer lookup cache misses more frequent

[kpmassey at gmail dot com]

--- Additional Comments From kpmassey at gmail dot com  2005-05-01 13:26 
---
Created an attachment (id=8783)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8783&action=view)
.s file from my compilation


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21316

[Bug c++/21316] pointer lookup cache misses more frequent

[kpmassey at gmail dot com]

--- Additional Comments From kpmassey at gmail dot com  2005-05-01 13:25 
---
Created an attachment (id=8782)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8782&action=view)
.ii file from my compilation


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21316

[Bug c++/21316] pointer lookup cache misses more frequent

[kpmassey at gmail dot com]

--- Additional Comments From kpmassey at gmail dot com  2005-05-01 13:23 
---
Created an attachment (id=8781)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8781&action=view)
source file that illustrates problem


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21316

[Bug c++/21316] New: pointer lookup cache misses more frequent

[kpmassey at gmail dot com]

The sample code runs 30-50% slower on my machine when compiled with 3.4.3 as
opposed to 3.3.4.  It creates some objects and then references them via pointer
lookup, so my best guess is that somehow 3.4.3 binaries have many more cache 
misses.

I have noticed the severe regression on AMD Athlon 3200 running linux 2.6, with
compile options -O3 -march=k8.  Other options (e.g. -O2 instead) don't help.

-- 
   Summary: pointer lookup cache misses more frequent
   Product: gcc
   Version: 3.4.3
Status: UNCONFIRMED
  Severity: normal
  Priority: P2
 Component: c++
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: kpmassey at gmail dot com
CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21316