-
Notifications
You must be signed in to change notification settings - Fork 61
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Newest update crashes on Android #198
Comments
I didn't had a much time but i did 3 different tests, with short time controls (no increments), and single and multi core usage, didn't had a single crash, time forfeit or engine flag. I can ensure you that the engine is working flawlessly. :) |
Try playing a game manually with Droidfish and see if it makes it through a whole one. |
The "Remove Coordination" patch is unlikely to be the cause of any crashes. If it has to do with queen moves (which also seems unlikely), it should happen in single-threaded mode. In that case it should be possible to find a position on which Cfish crashes reliably. If you can find such a position, that would be very helpful. |
Ok, Im sure I can. Ill get back with you as soon as I do. |
It 1st crashed in this game at 3r1r2/1pk1b1p1/nq2p1n1/2ppPb2/Q4P1p/1NP1BN1P/1R2B1P1/5R1K w - - 5 27 So I backed it up a move to 1k1r1r2/1p2b1p1/nq2p1n1/2ppPb2/Q4P1p/1NP1BN1P/R3B1P1/5R1K w - - 3 26 ... And it crashed again. Start at move 26 above and it goes 26. Rb2 Kc7 27. (Q wants to move to A1 and it crashes) (this was on DroidFish) |
@SwiftSnips |
2N4r/5p1p/3pk1pb/4p1B1/6P1/1P6/3N3q/R4KR1 w - - 2 29
If you need the entire game of either let me know. |
I was using 6(of 8) threads and let it sit for 10-15 seconds. Just so ya know the conditions. |
Heres the entire game.
|
Oddly enough the one you posted isnt crashing. |
It is compiled with Clang (as usual), not GCC that you use. ;-) P.s. i think this issue can be closed. |
@SwiftSnips make sure you use original SRCs. |
Game-play crashes with filled hash tables and multiple threads aren't reproducible. A reproducible crash needs a position searched with an empty TT table (of a known size) and 1 thread. |
And of course anything special about how the binary was produced has to be mentioned. |
I have been using clang to compile my engines for the past few weeks also. Incredibly they are all faster than gcc builds now. An example of my command line would be "make build -j ARCH=armv8 COMP=clang COMPCXX=aarch64-linux-android-clang++ lto=yes native=no popcnt=yes sparse=yes" |
I can input that FEN I pasted above, put my Cfish on 6threads, and begin to crash time and time again. It doesnt take a full hash to do it. I'll grab a few screenshots later this afternoon. Something I did do I forgot to mention, was use a makefile from an older Cfish. Because I kept getting a linker error with the makefile it came with and at the time was having trouble figuring out why. The older Cfish that older makefile came with, works just fine. |
0
I have many, many SRCs I have changed. But for this build I downloaded a fresh Cfish, and never opened any file outside of just the Makefile. |
I just reopened DroidFish to this very same position after probably 45 mins to an hour of being off the app, and the engine crashed within 10 seconds of analyzing it. Just for the record, I have tried it on 1 thread, and just analyze til the hash table fills up... hasnt crashed the 2 times I tried it at this same position. |
Mr. Syzygy1, the newest update of Cfish crashes on Android during certain moments in games. In the past few weeks I have spent time updating the Cfish source also, for myself, only for this crash to happen to me too. Its only on Android, not PC, but I spent a lot of time narrowing down the problem -- and it has something to do with the patch "Remove Coordination between Searching Threads". Its usually mid-to-end game this crash occurs. I thought this could be something you can check out. Ive found you can remove parts of this patch, but not all or the crash occurs.
(The crash usually involves moving the friendly Queen)
The text was updated successfully, but these errors were encountered: