Bug 815213

Summary: perf top reporst 9% in acpi_pm_read
Product: [Fedora] Fedora Reporter: Catalin BOIE <fedora-bugzilla>
Component: kernelAssignee: Kernel Maintainer List <kernel-maint>
Status: CLOSED NOTABUG QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: unspecified    
Version: 16CC: gansalmon, itamar, jonathan, kernel-maint, madhu.chinakonda
Target Milestone: ---Keywords: Reopened
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2012-05-15 12:18:51 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
perf trace number 1 none

Description Catalin BOIE 2012-04-23 06:52:41 UTC
Description of problem:
acpi_pm_read seems to eat too much CPU time according with perf top.

Version-Release number of selected component (if applicable):
3.3.2-1.fc16.x86_64

How reproducible:
perf top

Steps to Reproduce:
1.
2.
3.
  
Actual results:
Too much CPU usage by acpi_pm_read.

Expected results:
Much lower usage for acpi_pm_read.

Additional info:
Maybe is normal, but looks too big and the machine seems a bit sluggish. Hard to say when it started.

Thank you.

Comment 1 Catalin BOIE 2012-04-23 06:54:22 UTC
More info:

This is where time is spent:
99.51 :        ffffffff814a9850:       and    $0xffffff,%eax

Hardware:
Intel(R) Core(TM)2 Duo CPU     E8400  @ 3.00GHz

Comment 2 Dave Jones 2012-05-14 16:50:13 UTC
acpi pm timer is slow. there's no bug here.

Comment 3 Catalin BOIE 2012-05-15 06:11:51 UTC
I can understand that is slow, but 9%?!

I switched to tsc and it eats ~5%. Isn't it still too much?!

I run a perf and I generated the attached file. Seems that radeon eats a lot of resources in detecting DVI output. Monitor is connected on DVI.

radeon_dvi_detect appears a lot of time in the call stack.

Probably should be assigned to radeon.

Comment 4 Catalin BOIE 2012-05-15 06:12:54 UTC
Created attachment 584547 [details]
perf trace number 1