* [PATCH] Disable slow duplicate object tracking in ACPI_DEBUG
@ 2005-08-10 11:32 Andi Kleen
0 siblings, 0 replies; 3+ messages in thread
From: Andi Kleen @ 2005-08-10 11:32 UTC (permalink / raw)
To: acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f
The current SUSE 10.0 beta has CONFIG_ACPI_DEBUG enabled. That lead
to severe slowdowns on several machines. One issue was the different
default for ec_burst_mode. Another one was that the object tracking
would always walk the full list of all objects, which lead to kacpid
needing up to 25% of the total CPU time.
This patch disables the slow list walking. It could be probably
replaced with a hash or similar too, but I don't think the duplicate
object tracking is too useful anaywys so I just disabled it.
-Andi
From: ak-l3A5Bk7waGM@public.gmane.org
Subject: Disable linear search in ACPI debugging
Suse-bugzilla: 102565
Looking through all the memory objects all the time is just too slow.
Don't do it.
Signed-off-by: Andi Kleen <ak-l3A5Bk7waGM@public.gmane.org>
Index: linux-2.6.13rc5/drivers/acpi/utilities/utalloc.c
===================================================================
--- linux-2.6.13rc5.orig/drivers/acpi/utilities/utalloc.c
+++ linux-2.6.13rc5/drivers/acpi/utilities/utalloc.c
@@ -674,6 +674,9 @@ acpi_ut_find_allocation (
{
struct acpi_debug_mem_block *element;
+ /* too slow to be usable */
+ return NULL;
+
ACPI_FUNCTION_ENTRY ();
-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
^ permalink raw reply [flat|nested] 3+ messages in thread
* RE: [PATCH] Disable slow duplicate object tracking in ACPI_DEBUG
@ 2005-08-11 2:14 Brown, Len
[not found] ` <F7DC2337C7631D4386A2DF6E8FB22B30044D213E-N2PTB0HCzHKkrb+BlOpmy7fspsVTdybXVpNB7YpNyf8@public.gmane.org>
0 siblings, 1 reply; 3+ messages in thread
From: Brown, Len @ 2005-08-11 2:14 UTC (permalink / raw)
To: Andi Kleen, acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f
Maybe it is time to figure out why SLES has ACPI_DEBUG enabled.
If it is certain warnings and error messages, then perhaps
we should enable those but not all the other misc debug
and paranoia checks that come with ACPI_DEBUG?
thanks,
-Len
ps. ec-burst should be disabled by default in 2.6.12
and 2.6.13-rc6. there is an option to enable it at boot-time
in -mm, and hopefully we'll get it working there (very) soon
so we can experiment again with ec-burst enabled by default.
>-----Original Message-----
>From: acpi-devel-admin-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
>[mailto:acpi-devel-admin-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org] On Behalf Of Andi Kleen
>Sent: Wednesday, August 10, 2005 7:33 AM
>To: acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
>Subject: [ACPI] [PATCH] Disable slow duplicate object tracking
>in ACPI_DEBUG
>
>
>The current SUSE 10.0 beta has CONFIG_ACPI_DEBUG enabled. That lead
>to severe slowdowns on several machines. One issue was the different
>default for ec_burst_mode. Another one was that the object tracking
>would always walk the full list of all objects, which lead to kacpid
>needing up to 25% of the total CPU time.
>
>This patch disables the slow list walking. It could be probably
>replaced with a hash or similar too, but I don't think the duplicate
>object tracking is too useful anaywys so I just disabled it.
>
>-Andi
>
>From: ak-l3A5Bk7waGM@public.gmane.org
>Subject: Disable linear search in ACPI debugging
>Suse-bugzilla: 102565
>
>Looking through all the memory objects all the time is just too slow.
>Don't do it.
>
>
>Signed-off-by: Andi Kleen <ak-l3A5Bk7waGM@public.gmane.org>
>
>Index: linux-2.6.13rc5/drivers/acpi/utilities/utalloc.c
>===================================================================
>--- linux-2.6.13rc5.orig/drivers/acpi/utilities/utalloc.c
>+++ linux-2.6.13rc5/drivers/acpi/utilities/utalloc.c
>@@ -674,6 +674,9 @@ acpi_ut_find_allocation (
> {
> struct acpi_debug_mem_block *element;
>
>+ /* too slow to be usable */
>+ return NULL;
>+
>
> ACPI_FUNCTION_ENTRY ();
>
>
>
>-------------------------------------------------------
>SF.Net email is Sponsored by the Better Software Conference & EXPO
>September 19-22, 2005 * San Francisco, CA * Development
>Lifecycle Practices
>Agile & Plan-Driven Development * Managing Projects & Teams *
>Testing & QA
>Security * Process Improvement & Measurement *
>http://www.sqe.com/bsce5sf
>_______________________________________________
>Acpi-devel mailing list
>Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
>https://lists.sourceforge.net/lists/listinfo/acpi-devel
>
-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Disable slow duplicate object tracking in ACPI_DEBUG
[not found] ` <F7DC2337C7631D4386A2DF6E8FB22B30044D213E-N2PTB0HCzHKkrb+BlOpmy7fspsVTdybXVpNB7YpNyf8@public.gmane.org>
@ 2005-08-11 2:29 ` Andi Kleen
0 siblings, 0 replies; 3+ messages in thread
From: Andi Kleen @ 2005-08-11 2:29 UTC (permalink / raw)
To: Brown, Len; +Cc: Andi Kleen, acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f
On Wed, Aug 10, 2005 at 10:14:43PM -0400, Brown, Len wrote:
> Maybe it is time to figure out why SLES has ACPI_DEBUG enabled.
> If it is certain warnings and error messages, then perhaps
> we should enable those but not all the other misc debug
> and paranoia checks that come with ACPI_DEBUG?
SLES doesn't have ACPI_DEBUG enabled.
ACPI_DEBUG is just sometimes enabled in the betas because it makes
it easier to track down ACPI bugs - mostly because of the
tracing options which can be then used without recompiling the kernel
(which BTW seem to also cause considerable
overhead. One current case is a system where kacpid takes about 1.7%
of CPU and most of that seems to be just tracing functions)
ACPI_DEBUG_LITE was intended to be a subset that has only the
needed stuff enabled, but somehow that didn't take off.
-Andi
-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2005-08-11 2:29 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-08-11 2:14 [PATCH] Disable slow duplicate object tracking in ACPI_DEBUG Brown, Len
[not found] ` <F7DC2337C7631D4386A2DF6E8FB22B30044D213E-N2PTB0HCzHKkrb+BlOpmy7fspsVTdybXVpNB7YpNyf8@public.gmane.org>
2005-08-11 2:29 ` Andi Kleen
-- strict thread matches above, loose matches on Subject: below --
2005-08-10 11:32 Andi Kleen
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox