Skip to contentSkip to conversation

EXT2 vs EXT3 System calls via oprofile using contest 0.34

BENCHMARK Shawn Starr 3 participants 4 messages Sep 19, 2002 – Sep 24, 2002

lore
Share & tools
lore
#0 lore raw

Each test ran for about 15 seconds:

EXT2 kernel calls
==========

c022e080 6533     2.96616     __make_request         /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c013df50 6592     2.99295     __block_prepare_write   /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c012e880 7765     3.52552     generic_file_write_nolock /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c0175020 8588     3.89919     ext2_get_branch         /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

Top 3:

c013cf70 8678     3.94005     get_hash_table          /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c01752e0 10185    4.62427     ext2_get_block          /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c022f8c0 14293    6.48941     elevator_linus_merge    /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

EXT3 kernel calls
==========

c015f1c0 10928    3.00962     ext3_new_block          /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c016bb40 11863    3.26713     journal_dirty_metadata  /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c016fac0 12292    3.38527     journal_cancel_revoke   /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c022f8c0 15519    4.27401     elevator_linus_merge    /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

Top 3:

c013cf70 16380    4.51113     get_hash_table          /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c016b090 22883    6.30209     do_get_write_access     /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux
c0164910 26375    7.2638      ext3_do_update_inode    /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

Test preformed
=========

part of the contest package:

2) launch oprofile (if you use gui oprof_start &)

1) prepare commands
export tmpfile=foobar.log

before youn run io_halfmem click on "start profiler" run io_halfmem

wait 15 seconds

op_time -l which will dump the calls and time.

#1 lore raw

Hi,

Quoted 5 lines · Shawn Starr

EXT3 kernel calls
==========

Top 3:
c013cf70 16380 4.51113 get_hash_table /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

Same as ext2...

Quoted 1 line

c016b090 22883 6.30209 do_get_write_access /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

That's an inevitable penalty from the way ext3 does journaling --- you get two copies of data if a filesystem operation updates a block that is still being journaled (because we need to snapshot the old copy to write to the journal). That's done when ext3 notifies an intent to modify the old block, so all those copies show up in do_get_write_access.

Quoted 1 line

c0164910 26375 7.2638 ext3_do_update_inode /lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

I've got a fix for excessive CPU time spent here.

--Stephen

#2 lore raw
Quoted 5 lines
Quoted 3 lines

c0164910 26375 7.2638 ext3_do_update_inode
/lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

I've got a fix for excessive CPU time spent here.

Could you pass that around? or is it not ready for general consumption... ? Thanks.

------
Michael Cohen

Signature · 1 line
Anyone who quotes me in their sig is an idiot. -- Rusty Russell.
#3 lore raw

If the patch is 'stable' I'd like to bench/test it.

Quoted 11 lines · Michael Cohen
Quoted 4 lines
Quoted 2 lines

c0164910 26375 7.2638 ext3_do_update_inode
/lib/modules/2.4.20-pre7-rmap14a-xfs-uml-shawn12d/build/vmlinux

I've got a fix for excessive CPU time spent here.

Could you pass that around? or is it not ready for general consumption... ? Thanks.

------
Michael Cohen