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: Speedup count_or_remove_death_notes


In message <Pine.LNX.4.58.0311251123000.8408@D209.suse.de>, Gerald Pfeifer writ
es:
 >On Mon, 24 Nov 2003 law@redhat.com wrote:
 >> This is not strictly a bugfix, so I have not checked it into the mainline
 >> sources.  It has been checked into the tree-ssa branch.
 >
 >We also have bad performance regressions on mainline -- in fact, PR8361
 >originally was created to track these...
 >
 >> Mark -- if you want me to check in this change, let me know.
 >
 >...so I do recommend this patch for mainline as well.  It addresses a
 >high priority PR and regression compared to previous releases.
Be aware that most (all?) of the benefit will be seen when ENABLE_CHECKING
is turned on as the problematic calls occur only when we've got checking
enabled.

So I wouldn't expect significant improvements for release code since we
turn off checking when we make releases.

jeff


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