Hi,
it seems that the following message that came before the two patches
didn't make it to the list, probably because of an eps graphics
attachment. So I try it again, without the graphics and I'll try to
send the graphic as a pdf in a separate mail.
My apologies
Jens

+++++++++++++++++++++++

Hi everybody,

with my current testing of the <stdatomic.h> implementation I came to
test and modify some part of the lock and wait primitives in
musl. Here is one stress test that I think shows interesting
performance differences. At the bottom it just uses only lock-free
atomics, so the result itself is independent of <stdatomic.h>.

The program is an implementation of a list structure that is shared
between threads and used as LIFO (stack). Threads draw a random number
and then randomly decide to insert (and malloc) or remove (and free)
as many members from the list. The probabilities are such that the
list is not growing too large on expectation.

All of this is narrowly intertwined between threads, so this results
in a lot of allocations that are then usually freed by another thread,
and it puts the malloc system under quiet a stress.

The question that I asked myself is how much spinning to we need in
such lock/wait primitives. In the first attached graphic, the three
bottom curves show what difference three spinning strategies can make
for this test on a 2x2 hyperthreaded core machine. The bottom is no
spinning at all, the next is the current strategy implemented in musl,
and the third is a modification of that. As you can see, for a high
load of threads they can make a substantial difference, but you also
see that musl's actually strategy is not very different from doing no
spinning at all.

The "new" strategy is simply to avoid to take the shortcut that the
actual code takes when spinning. Here, currently when we detect that
are already waiters, we stop spinning and try to go into futex_wait
immediately. Where this sounds nice at a first glance, the figures
seem to indicate that this is not a good idea and that we would be
better off without. I'll send a patch for that in a next mail, and
also one that lets you modify the number of spins easily.

Now the only situation I thought of where this could be important is
monoprocessors where actually spinning might not be so good and
aborting it early good be necessary. So I rand the same test with
taskset to nail the process to just one core. The result of that are
the top two curves. As you can see, here the spinning strategy has
almost no influence so I think we are safe to apply this patch.

Now all of this can also be read as a performance test of the malloc
subsystem, and here my feeling goes in the direction that Rich
recently indicated. The performance of the "application" is much
better if I eliminate all parallelism. As an additional indication
there are to additional curves that fix the process to one core and
its hyperthreaded sibling.

So maybe we might be better off with a simpler malloc strategy that
serializes all requests.

Jens



-- 
:: INRIA Nancy Grand Est ::: Camus ::::::: ICube/ICPS :::
:: ::::::::::::::: office Strasbourg : +33 368854536   ::
:: :::::::::::::::::::::: gsm France : +33 651400183   ::
:: ::::::::::::::: gsm international : +49 15737185122 ::
:: http://icube-icps.unistra.fr/index.php/Jens_Gustedt ::