[Python-Dev] tiny optimization in ceval mainloop

[Tim Peters]

[Jack Jansen]
> On vrijdag, augustus 30, 2002, at 06:12 , Tim Peters wrote:

Jack, I'm never on vrijdag -- vrijdag is illegal in the US <wink>.

> ...
> And it costs!
>
> Running pystone without another thread active I get 5500
> pystones out of my machine. Running it with another thread
> active (in a sleep(1000)) I get 4200.
> After setcheckinterval(100) I'm back up to 5200.
>
> For completeness' sake: with no other thread active raising
> setcheckinterval() doesn't make a difference (it's in the noise,
> in my measurement it was actually 0.5% slower).
>
> We could get a serious speedup for multithreaded programs if we
> could raise the check interval.

Guido already agreed to try boosting it to 100.

> Some wild ideas:
> - Use an exponential (or linear?) backoff. If you attempt to
> switch and nothing happens you double the check interval, up to
> a maximum. If you do switch you reset to the minimum.

On a pthreads system under 2.3, using semaphores, chances are good it will
always switch.  But unless you're trying to fake soft realtime, it's a real
drag on performance to switch so often  We can't out-guess this, because it
depends on what the *app* wants.  Most apps aren't trying to fake soft
realtime, so favoring less frequent switches is a good default.

> - Combine the above with resetting (to zero? to minimum value if
> currently >= minimum?) the check interval on anything we know
> could influence thread schedule (releasing a lock, etc).

You need a model for what it is you're trying to optimize here.  I'm just
trying to cut useless overheads <wink>.