Banksia GUI released

[AdminX]

Can you run it with something like wine?

Good question, I have not tried it yet but I will put it on my todo list and see. Still, I would prefer not to use emulation but I may not have a choice as of today.

[AlexChess]

I am in the process of upgrading my Raspberry Pi 4 8GB model with a SSD and a new case. I just installed Ubuntu Mate on it. Then the thought hit me, how difficult would it be to port Banksia over so that it could run on a Arm processor? I am also in the process of installing Picochess on it, so in theory I should be able use all engines created to run on Arm processors in Banksia also if it were ported to Arm. Keep hope alive! This is just a thought, not a request yet ...

Welcome to Linux ARM64 Pi 4 It's perfectly compatible with Mac M1 and we will fight together to compile native engines. I know that BanksiaGUI will be ported asap, but some sdk aren't still available. I'm using Cutechess 1.20 compiled by cpeters on Linux ARM64.

Regards, Alexchess

[AdminX]

I am in the process of upgrading my Raspberry Pi 4 8GB model with a SSD and a new case. I just installed Ubuntu Mate on it. Then the thought hit me, how difficult would it be to port Banksia over so that it could run on a Arm processor? I am also in the process of installing Picochess on it, so in theory I should be able use all engines created to run on Arm processors in Banksia also if it were ported to Arm. Keep hope alive! This is just a thought, not a request yet ...

Welcome to Linux ARM64 Pi 4 It's perfectly compatible with Mac M1 and we will fight together to compile native engines. I know that BanksiaGUI will be ported asap, but some sdk aren't still available. I'm using Cutechess 1.20 compiled by cpeters on Linux ARM64.

Regards, Alexchess

Thanks, Good to know about Cutechess 1.20 as I was not aware of that Arm port.

[phhnguyen]

..

Look at this:




How is the ponderhit ratio calculated? (BTW can it be shown permanently? It's very difficult to see on short time control games, and it only shows when engines get a ponderhit. If they don't, it's not shown for that move. Also, it would be cool to show the move it's currently pondering to see if it will guess it or not depending on the rival's pv.)

In this case, Ethereal guessed Stockfish move 1 out of 3. It should be 33%. Even more, it's can't be 25% in any way you put it. 0/3, 1/3, 2/3 and 3/3 are then only possible combinations, and none of those give 25%

It should be number of ponderhits / (current move number - 1), because neither color can ponder on move 1, and to avoid "divide by zero" crap, it should only be shown/calculated starting on move 2.

Thanks for the feedback.

The ponderhit is calculated by the number of correct guesses / the number of commands “go” (or “ponderhit” in cases of being hit).

The above ponderhit is updated when that side in turn & starts computing. The white side made 3 moves but it has been thinking for 4th times, it means it has received already 4 “go” commands. That’s why it can be 25%.

I have noted down and will check further to improve.

[phhnguyen]

I am in the process of upgrading my Raspberry Pi 4 8GB model with a SSD and a new case. I just installed Ubuntu Mate on it. Then the thought hit me, how difficult would it be to port Banksia over so that it could run on a Arm processor? I am also in the process of installing Picochess on it, so in theory I should be able use all engines created to run on Arm processors in Banksia also if it were ported to Arm. Keep hope alive! This is just a thought, not a request yet ...

That’s doable. Even I was almost going to buy a Raspberry Pi just for testing. However, I postponed that plan later since needed to focus some other tasks, including real life ones. Sure sooner or later I will support directly some ARM hardware (Ras, M1)

[BrendanJNorman]

I am in the process of upgrading my Raspberry Pi 4 8GB model with a SSD and a new case. I just installed Ubuntu Mate on it. Then the thought hit me, how difficult would it be to port Banksia over so that it could run on a Arm processor? I am also in the process of installing Picochess on it, so in theory I should be able use all engines created to run on Arm processors in Banksia also if it were ported to Arm. Keep hope alive! This is just a thought, not a request yet ...

That’s doable. Even I was almost going to buy a Raspberry Pi just for testing. However, I postponed that plan later since needed to focus some other tasks, including real life ones. Sure sooner or later I will support directly some ARM hardware (Ras, M1)

Hey buddy, I hope your wife is doing well and feeling better...

I have a bug report:

I have noticed that there is a bug when 1. engines are asked to use a "seconds per move", especially in situations where 2. I am using the multi-time control and giving the human side unlimited time.

I think this bug is more related to the multi-time control feature.

Another feature that would be nice, but also easy to implement, is for the engine's rating to also show up in the exported pgn file. As of now it does not.

Thanks buddy, you're doing excellent work.

[AdminX]

Bug Report:

I have noticed that concurrency is not being respected. Today for example I had it set to 1, however when someone challenged my engine it tried to play anyway even when I was already playing and caused Banksia to crash over and over causing my account to lose the game I was already in the middle of. The only reason I know that this was the case was because every time I tried to relaunch Banksia it showed status Playing 2 before crashing again.

[CMCanavessi]

I've found another issue with pondering, but I don't know if it's a Minic thing or a Banksia thing

Code: Select all

Minic 3.09 > bestmove e6d5 ponder f1e1
Minic 3.09 < position fen rnbkrqnb/pppppppp/8/8/8/8/PPPPPPPP/RNBKRQNB w EAea - 0 1 moves f2f4 d7d5 g1f3 b8c6 b1c3 g8f6 d2d3 c8e6 h2h3 g7g6 f1f2 f6d7 d1e1 f7f5 e2e4 d5e4 d3e4 h8c3 b2c3 f5e4 f3g5 f8g8 g5e4 e6d5 f1e1
Minic 3.09 < go ponder wtime 102100 btime 143677 winc 2000 binc 2000
DanaSah 8.8 < ponderhit
.
.
.
DanaSah 8.8 > bestmove e4d2 ponder d8a8
DanaSah 8.8 < position fen rnbkrqnb/pppppppp/8/8/8/8/PPPPPPPP/RNBKRQNB w EAea - 0 1 moves f2f4 d7d5 g1f3 b8c6 b1c3 g8f6 d2d3 c8e6 h2h3 g7g6 f1f2 f6d7 d1e1 f7f5 e2e4 d5e4 d3e4 h8c3 b2c3 f5e4 f3g5 f8g8 g5e4 e6d5 e4d2 d8a8
DanaSah 8.8 < go ponder wtime 97681 btime 145677 winc 2000 binc 2000
Minic 3.09 < stop
Minic 3.09 > bestmove O-O-O ponder c1d2
.
.
.
//MINIC LOSES ON TIME//

After DanaSah 8.8 moves (and Minic does NOT guess the correct move), Minic is told to stop, returns the bestmove, but then Banksia doesn't send the new position and go command...

Edit: tried some more games, looks like sometimes Banksia ignores Minic's reply... Here's the full log of one of those matches:

Edit2: in both games, Banksia just ignored Minic's bestmove O-O-O. Can it be that Banksia is filtering the "O-O-O" string out?

[amanjpro]

I've found another issue with pondering, but I don't know if it's a Minic thing or a Banksia thing

Code: Select all

Minic 3.09 > bestmove e6d5 ponder f1e1
Minic 3.09 < position fen rnbkrqnb/pppppppp/8/8/8/8/PPPPPPPP/RNBKRQNB w EAea - 0 1 moves f2f4 d7d5 g1f3 b8c6 b1c3 g8f6 d2d3 c8e6 h2h3 g7g6 f1f2 f6d7 d1e1 f7f5 e2e4 d5e4 d3e4 h8c3 b2c3 f5e4 f3g5 f8g8 g5e4 e6d5 f1e1
Minic 3.09 < go ponder wtime 102100 btime 143677 winc 2000 binc 2000
DanaSah 8.8 < ponderhit
.
.
.
DanaSah 8.8 > bestmove e4d2 ponder d8a8
DanaSah 8.8 < position fen rnbkrqnb/pppppppp/8/8/8/8/PPPPPPPP/RNBKRQNB w EAea - 0 1 moves f2f4 d7d5 g1f3 b8c6 b1c3 g8f6 d2d3 c8e6 h2h3 g7g6 f1f2 f6d7 d1e1 f7f5 e2e4 d5e4 d3e4 h8c3 b2c3 f5e4 f3g5 f8g8 g5e4 e6d5 e4d2 d8a8
DanaSah 8.8 < go ponder wtime 97681 btime 145677 winc 2000 binc 2000
Minic 3.09 < stop
Minic 3.09 > bestmove O-O-O ponder c1d2
.
.
.
//MINIC LOSES ON TIME//

After DanaSah 8.8 moves (and Minic does NOT guess the correct move), Minic is told to stop, returns the bestmove, but then Banksia doesn't send the new position and go command...

Edit: tried some more games, looks like sometimes Banksia ignores Minic's reply... Here's the full log of one of those matches:

Edit2: in both games, Banksia just ignored Minic's bestmove O-O-O. Can it be that Banksia is filtering the "O-O-O" string out?

Banksia correctly ignores the response. Stop is sent to the pondering engine when the pondering move is wrong... The engine responds with a "best"move only to comply with the protocol. And then banksia MUST send another go request to the engine with the start position and the moves, and a go command. If this is not done, then the bug is in the GUI else it is in the engine.

You can test Banksia's implementation against my engine Zahak, if Zahak also had issues then I'm sure it is a GUI bug

[CMCanavessi]

Banksia correctly ignores the response. Stop is sent to the pondering engine when the pondering move is wrong... The engine responds with a "best"move only to comply with the protocol. And then banksia MUST send another go request to the engine with the start position and the moves, and a go command. If this is not done, then the bug is in the GUI else it is in the engine.

You can test Banksia's implementation against my engine Zahak, if Zahak also had issues then I'm sure it is a GUI bug

Yes, I know how it should work. What I mean is that maybe banksia is filtering Minic's bestmove after the stop command and then waits forever to receive the bestmove that it just filtered out (for whatever reason). Still, if you look at the full log that I posted from the other game, Banksia again ignores Minic's bestmove and that one wasn't while pondering, it was just a regular move.

As Banksia works ok with all other engines, there must be something that Minic is doing different (wrong or right I don't know, but definitely there's something different). I suspect that castling string.