Re: [Ext2-devel] Re: [RFC][PATCH 0/2]Extend ext3 filesystem limitfrom 8TB to 16TB

From: Laurent Vivier
Date: Fri Apr 21 2006 - 07:17:37 EST


Le jeu 20/04/2006 Ã 16:39, Laurent Vivier a Ãcrit :
> The functions added by my patch are following and as they are atomic
> (one machine instruction) they are not measurable and don't appears in
> oprofile.
>
> atomic_long_add
> atomic_long_read
> atomic_long_set
> atomic_long_inc

I think, as these commands are atomic/inlined we should measure the time
of the functions modified by the patches.

The functions modified by the patch are:

ext3_free_blocks_sb
ext3_has_free_blocks
ext3_new_block
ext3_put_super
ext3_fill_super
ext3_fill_super
ext3_free_inode
find_group_dir
find_group_orlov
ext3_new_inode
ext3_group_add

If we make a "grep" on tiobench oprofile.out, we have:

atomic_t:

26919 0.0119 vmlinux vmlinux ext3_new_block
2195 9.7e-04 vmlinux vmlinux ext3_free_blocks_sb
1192 5.2e-04 vmlinux vmlinux ext3_has_free_blocks
189 8.3e-05 vmlinux vmlinux ext3_new_inode
40 1.8e-05 vmlinux vmlinux ext3_free_inode
2 8.8e-07 vmlinux vmlinux find_group_orlov

percpu_counter:

16290 0.0067 vmlinux vmlinux ext3_new_block
2075 8.5e-04 vmlinux vmlinux ext3_free_blocks_sb
428 1.8e-04 vmlinux vmlinux ext3_has_free_blocks
162 6.7e-05 vmlinux vmlinux ext3_new_inode
25 1.0e-05 vmlinux vmlinux ext3_free_inode

As we can using atomic_long_t is slower than percpu_counter so ...
forget my patch.

Regards,
Laurent

--
Laurent Vivier
Bull, Architect of an Open World (TM)
http://www.bullopensource.org/ext4

Attachment: signature.asc
Description: Ceci est une partie de message=?ISO-8859-1?Q?num=E9riquement?= =?ISO-8859-1?Q?_sign=E9e=2E?=