search

lantonov / asmFish

A continuation of the nice project asmFish by Mohammed Li. Latest version: 07.08.2019
https://lantonov.github.io/asmFish/
Other
118 stars 49 forks source link

asmFishW_2018-05-06_popcnt caused an excpetion during 16min match #167

Closed bitsocket closed 6 years ago

bitsocket commented 6 years ago

During a game the engine suddenly threw an exception and hence the game was terminated. Fritz-GUI remained stable, engine could just be re-started afterwards, but of course the match was lost.

I guess it might have todo with the Large Pages option, which I had enabled with 6144MB of memory. The engine had already died once before, when I had increased LP from 4096 to 6144MB.

SYZYGY (5men) was active during the games.

For details see screenshot: asmfishw-2018-05-06_exception_this_is_a_problem_in_the_engine

CounterPly commented 6 years ago

Hi bitsocket,

From which source was this _popcnt executable assembled? Does the crash also occur with this pre-assembled binary?

https://github.com/lantonov/asmFish/raw/master/WindowsOS_binaries/asmFishW_2018-04-25_popcnt.exe

bitsocket commented 6 years ago

It was assembled on the same machine (Xeon X56) - on which it crashed - twice in a while on 16 min matches only.

I haven't tried yet to reproduce it with precompiled binaries, as it doesn't crash every now and then. The first crash appeared after I had increased hash memory from 4GB to 6GB, while LP option was still active and SYZYGY was configured in Fritz 16 GUI - not directly with the engine, because F16 GUI doesn't show the syzygy path option anymore.

Since I have reduced hash mem. size to 4GB again and disabled LP option, it did not crash again - so far.

I will give it a try with precompiled binary and same settings..

CounterPly commented 6 years ago

Sorry, I meant from which source code? I just wanted to make sure that it was after this commit: https://github.com/lantonov/asmFish/commit/f9b379372194f0c61177c19befebe769ffe92327.

This fixed the Large Pages issues encountered by many users.

bitsocket commented 6 years ago

It was built from this commit: f9b379372194f0c61177c19befebe769ffe92327, which, according to log messages, already contains the LP fix.

Now, I remember that I had built it from source, AFTER it had thrown the exception twice from the pre-build binary.

I thought it could be related to different hardware being used for bre-build binaries - compared to my older Xeon X56 LGA1366. Hence, I retried with binaries built on my X56 machine, which however made no difference with regard to the error.

CounterPly commented 6 years ago

Hi @bitsocket,

Just checking in with you re: this issue. I have not heard of any other reports of crashes within Fritz or otherwise. Are you still experiencing this LP issue?