>> It could be the procps thing? `tiobench --threads 256' shows
>> up as a single process in top and ps due to the new thread
>> consolidation feature. If you run `ps auxm' or hit 'H' in top,
>> all is revealed. Not my fave feature that.
The feature is both buggy (both false consolidation and failure
to consolidate) and slow. While I do eventually need to add the
feature, I'm not doing so until it can be implemented properly.
So go ahead and enjoy procps-3.1.0 without it:
Future-compatibility exists. The "m" and "-m" options are currently
ignored so that you may use them in scripts without causing errors.
Um, BTW, don't we need /proc/42/threads/123/stat and friends
before the Linux 2.6.xx release? Without them, the new clone
flags supposedly let a user hide a task. That's really bad.
> Turns out monitoring things via /proc/ slowed it down by
> some ridiculous factor while it was trying to spawn threads.
> 9 hours became less than 1s when I stopped looking.
Not that procps will ever be fast, but 3.1.0 is a lot faster
than 2.x.xx is. Five out of the last seven releases have
included performance improvements, and more are coming.
You also get IO-wait info in vmstat. Upgrade today! :-)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/