Adding memory info to "usertest"

John Arbash Meinel john at arbash-meinel.com
Sun Oct 18 18:52:21 BST 2009


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hey Ian-

Since I've been focusing on memory consumption, I was thinking it would
be nice to have some sort of benchmark tracking of peak memory
consumption. I don't think it is appropriate to put it into the standard
test suite, because it is a bit too variable.

But I think it would be nice to have some sort of standard benchmarks
for memory consumption, and I think usertest would fit that.

So I'd like to get some ideas about how you think it would be good to
incorporate it. I'm wondering if peak memory measurements should be part
of every run, or if it should be a separate set of 'memory benchmarks'
that we run.

I think it could be used across tools without difficulty. For example,
on Linux, as long as you know the PID of the child process, you can
watch /proc/PID/status. You may not get the absolute last value, but it
will probably be close enough. Similarly, I believe the win32 api
functions can take the PID of the process that you want info about. (I
also think there might be an api for getting the value just-after the
process exited, but I'm not positive on that.)

I'm not going to be working on this right away, as I'm focusing more on
actually decreasing memory consumption. However, I would

1) Like to see this not regress in the future
2) Get a good feeling about how it helps (or even hurts) across the
   board, rather than just on the few commands I'm focusing on.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkrbVdUACgkQJdeBCYSNAANGPwCg2bO0EHibejKt/TxWe2fNv3ex
tuEAoIhY1fxJ+vFozF2teASac6t4Kgr7
=DYeh
-----END PGP SIGNATURE-----



More information about the bazaar mailing list