[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Profiling of ext2fs
From: |
Michael Casadevall |
Subject: |
Profiling of ext2fs |
Date: |
Fri, 3 Aug 2007 21:46:54 -0400 (EDT) |
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
After much blood, sweat, and tears, I managed to build the translator with
profiling (not an easy task, and it flushed out a few bugs with GCC at the
same time that had to be patched around for the time being, and some more
which need to be resolved). I ran ext2fs.profiling (a static version of
the translator with profiling), and got the following results with
gmon.out (I put it as an attachment due to the sheer size of the output).
The first thing you'll notice is a distinct lack of times. Googling says
this is caused when the program spends virtually no time in user mode
(gprof can only track userspace times, not kernel or real). While this
might be the case, especially if the translator is using co-location
(which I think is disabled in Hurd though), it might also be a bug in
gprof itself (I get a function terminated message on console, so this
requires futher investagation), but we can decude some things. For the
test, I deleted a directory on the mounted HDD, and then copied that same
folder from the root drive to it. As you can see, there are loads of calls
to IO, and it also removes quite a few suspected laggers such as Mach IPC.
I'm not an expert in mach's guts, so I'll leave this for you all to pick
over while I see if I can figure out why timing is broken.
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU)
iD8DBQFGs9qPV/FBeJ6jsZQRAko4AJ9Ej0FXryBLt3YWc/Z79Sb6QVEngACfZn9Y
H4zF/edTBakKHm5KaFftdQ0=
=eIiq
-----END PGP SIGNATURE-----
ext2fs_profiling_data
Description: Text document
- Profiling of ext2fs,
Michael Casadevall <=