TalkChess.com

Looks like a patch submitted by Joona has generated a 50 point ELO gain for Stockfish running 16 threads. So perhaps a big step forward in Stockfish's multi-threaded searching. Seems very promising.

http://tests.stockfishchess.org/tests/v ... 3e9fac133b

zullil wrote:Looks like a patch submitted by Joona has generated a 50 point ELO gain for Stockfish running 16 threads. So perhaps a big step forward in Stockfish's multi-threaded searching. Seems very promising.

http://tests.stockfishchess.org/tests/v ... 3e9fac133b

Hope it will not pass, as otherwise I will not play your 16-core SF.

sounds very promising

Yes Komodo 9 with +20 ELO will be in very bad trouble. And there is already some +10 ELO patch after SF6 - unbelievable .

zullil wrote:Looks like a patch submitted by Joona has generated a 50 point ELO gain for Stockfish running 16 threads. So perhaps a big step forward in Stockfish's multi-threaded searching. Seems very promising.

http://tests.stockfishchess.org/tests/v ... 3e9fac133b

Yes, this is huge!

Note that the test was auto-purged (chi-square test failed) twice.

Initially we had:
* bravone: 0 elo
* glinscott: +20 elo
* fastgm, bking: +50 elo

The chi-square test failed, and gave bravone as worst offender:
* glinscott: +20 elo
* fastgm, bking: +50 elo

And the chi-square test failed again, giving glinscott as worst offender. So we're left with:
* fastgm, bking: +50 elo

I don't know what's wrong with these machines. Could be OS or hardware. The only thing we know is the OS:
* bravone: Windows 2008 (old)
* glinscott: Linux 2.6.32 (old)
* fastgm: Linux 3.16 (recent)
* bking: Windows 8 (recent)

In other words the machines with modern OS saw a massive gain, while the others not so much (no gain on Windows 2008, and "only" 20 elo on an old Linux). But maybe the OS is a red herring, and it's about the hardware. I don't know. We can only speculate at this point.

is the smp patch a regression for one core? or am I reading the test results wrong

is the smp patch a regression for one core? or am I reading the test results wrong

That's extremely unlikely given that it is a non-functional change on one core. It's probably just an unlucky run.

EDIT: Actually checking again I see that the test wasn't even finished. Never draw conclusions from an unfinished test (statistics 101).

styx wrote:is the smp patch a regression for one core? or am I reading the test results wrong

Yes, -1.8 after 32K games. That means -2% speed.
May be it's possible to reorganize "IF" conditions to not lose speed with nb threads = 1

Michel wrote:

is the smp patch a regression for one core? or am I reading the test results wrong

That's extremely unlikely given that it is a non-functional change on one core. It's probably just an unlucky run.

EDIT: Actually checking again I see that the test wasn't even finished. Never draw conclusions from an unfinished test (statistics 101).

Though it is a non-functional change on one core, I experience a small slowdown on my box.
But I guess even a small elo-loss on one core of 1 or 2 elo would be acceptable, given the huge gain for 16 cores.

Lyudmil Tsvetkov wrote:

zullil wrote:Looks like a patch submitted by Joona has generated a 50 point ELO gain for Stockfish running 16 threads. So perhaps a big step forward in Stockfish's multi-threaded searching. Seems very promising.

http://tests.stockfishchess.org/tests/v ... 3e9fac133b

Hope it will not pass, as otherwise I will not play your 16-core SF.