Komodo 12.3 x64 (MCTS) @ CEGT 3'+1" ponder=on

[ThatsIt]

Hi !

http://cegt.forumieren.com/t1064-testin ... 3-x64-mcts

Best wishes,
G.S.
(CEGT team)

[sainzlei]

will you test lc0 engine ?

[ThatsIt]

@ Larry !

Under the Shredder Classic GUI i get very often:



This causes crashes.

Best wishes,
G.S.

[lkaufman]

Thanks. This tells us that the crashes in CCC and TCEC were probably not due to any changes since 12.3. We're trying to find the problem. Have there been any crashes in CEGT non-ponder testing? Are there any common elements in the positions where it crashes, or any illegal moves displayed?

Larry

[ThatsIt]

Hi Larry,

Have there been any crashes in CEGT non-ponder testing?

I've made no tests with ponder=off so far.

Are there any common elements in the positions where it crashes, or any illegal moves displayed?

No pattern so far, but no illegal moves.
There must be a change, Komodo 12.2.1 (MCTS) does not show this behavior !

Best wishes,
G.S.

[ThatsIt]

Have there been any crashes in CEGT non-ponder testing?

I've started the test for our 40/4 + ... list (ponder=off).

http://cegt.forumieren.com/t1069-testin ... -mcts#2230

Best wishes,
G.S.
(CEGT team)

[Modern Times]

I had some crashes in CCRL 4040 4CPU testing. Not many, 4 in just over 300 games.

[pohl4711]

Thanks. This tells us that the crashes in CCC and TCEC were probably not due to any changes since 12.3. We're trying to find the problem. Have there been any crashes in CEGT non-ponder testing? Are there any common elements in the positions where it crashes, or any illegal moves displayed?

Larry

If this helps: I played 5000 games with Komodo 12.3 MCTS with LittleBlitzerGui, singlecore, ponder off, 3'+1“, without any problems. No crashes, no timelosses, no illegal moves.

https://www.sp-cc.de

[lkaufman]

It seems that with multiple threads (and perhaps also with ponder?) 12.3 MCTS is subject to occasional crashes and time forfeits, with the frequency increasing with thread count. We have made progress towards a fix, but we want to avoid multiple bugfixes so we want to be reasonably sure we have fixed the issue before releasing anything. Basically, the problem was that our tester didn't report crashes/forfeits so we weren't aware of the problem, but now it does so this should not be an issue in the future once we have a bugfix to release. As long as current 40/4 testing on one thread doesn't show crashes or forfeits it should be valid.

[ThatsIt]

Hi !

400 games played for our CEGT 40/4 + ... list.
No issues at all !

Best wishes,
G.S.
(CEGT team)