[lkml]   [2011]   [Mar]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH 10/15] ftrace: fix event alignment: jbd2:*
On Tue, Mar 08, 2011 at 07:03:25PM -0500, Steven Rostedt wrote:
> Now I know why I was so late at adding this. I was waiting for
> Acked-bys ;)
> On Fri, 2010-12-03 at 20:52 -0500, Steven Rostedt wrote:
> > [ Ccing Ted for Acked-by ]
> Ted, Ping?

A month or two ago I did an analysis using the pahole binary (part of
the 'dwarves' package in Debian/Ubuntu) and noticed many more
structures that were flagged as having 'holes' where there were
optimization opportunities.

In addition, (in an interesting example of synchronicity) just this
afternoon, I had talked to Jiaying about a further optimization of the
ext4 trace header files, since she pointed out to me that the 'perf'
binary can actually handle the MAJOR() and MINOR() macros. So we can
replace 8 bytes of "int major, minor", with a 4 byte dev_t, and save
another 4 bytes (assuming we can move the structure elements around so
we can save a full 8 bytes to avoid structure padding).

Jiaying was also going to send some patches to add some additional
tracepoints to the upstream kernel.

So probably the best thing to do at this point is to drop the ext4 and
jbd2 patches, since Jiaying is about to send me patches that to handle
these optimizations plus some further ones --- and add some additional
tracepoints. Getting one patch from her that I'll feed in via the
ext4 tree probably will be much simpler than coordinating patches to
the same file from two different trees.

- Ted

 \ /
  Last update: 2011-03-09 01:33    [W:0.064 / U:0.332 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site