Optimal 3x3x3 solver for large set of scrambles?

[bcube]

Hi,

I have downloaded and used Cube Explorer once, so I am no expert in using it. I have never used any other optimal solver to solve scrambled 3x3x3 cube, therefore I don´t know what software is currently available.

I have a set of 10 000 3x3x3 scrambles in Singmaster notation (without cube rotations, slice moves and wide moves) in a .txt file, each scramble on new line.

I would like to ask if there is an optimal 3x3x3 solver which would take that file and output an optimal move count (optionally, however not necessarily, also solution in Singmaster notation) for each scramble in HTM?

I am willing to install Python (maybe Java) or use chatgpt if needed (e.g. if solver requires different input format), but it would be nice if the program could do all that stuff already.

 

[xyzzy]

Use Nissy.

It includes an optimal solver, alongside many other functions useful for FMC scramble analysis. Can be run completely CLI with no human interaction.

 

[UKCubeFan]

I want this prog.

 

[Herbert Kociemba]

I would like to ask if there is an optimal 3x3x3 solver which would take that file and output an optimal move count (optionally, however not necessarily, also solution in Singmaster notation) for each scramble in HTM?

Sorry for the late response, I do not look into the forum at a regular basis. But why do you not just use Cube Explorer? It can do what you want.

 

[kubesolver]

Sorry for the late response, I do not look into the forum at a regular basis. But why do you not just use Cube Explorer? It can do what you want.

I think the lack of command line interface makes it hard to automate and make part of a larger script

e.g. can cube explorer be used in a task that downloads all scrambles from a competition and runs analysis on them and then re-uploads the analysis results somewhere?

 

[Herbert Kociemba]

I will answer the question of bcube which is unambiguous. @kubesolver: "run analysis" is not specific enough to allow an answer.

1. Start Cube Explorer, go to Options||Huge Optimal Solver and load it.
2. Got to File||Load Maneuvers... to load the cubes.
3. Go to Run|Start Autorun for Optima Solver and wait until all cubes are solved automatically.
4. Goto Run|Stop Autorun for Optimal Solver
5. Go to Options|Generate statistics to generate informations about the length of the maneuvers etc.
6. Copy and paste this to any document you want.
7. Use File|Save Maneuvers... if you want to save the optimal solutions.

 

[Twisted101]

I would say to use batch solver, could u tell me what algorithm set u r genning?

 

[Herbert Kociemba]

could u tell me what algorithm set u r genning?

Could you speak English please so that I understand what you mean?

 

[Twisted101]

Could you speak English please so that I understand what you mean?

I am speaking English

 

[Herbert Kociemba]

Is "genning" an English word?

 

[Billabob]

It's shorthand for "generating". He's asking what algorithm set the user would like to mass-solve cases for.

Cube Explorer includes a feature to load in scrambles from a .txt file. Each scramble must be on a new line in the file. It's not as flexible as a CLI, but it should do for most use cases, albeit it requires some manual input to start solving.

 

[kubesolver]

I will answer the question of bcube which is unambiguous. @kubesolver: "run analysis" is not specific enough to allow an answer.

Any kind of analysis of a scramble.
E.g. find any solution, find optimal solution, find optimal solution in qtm.

I understand you might not want to change your program. But when a question is "can I have a program that takes this file as an input and produces output" in the context of using it within e.g. python, you respond with 7 steps of gui interactions which is a very convoluted way to say "no".

 

[Herbert Kociemba]

I understand you might not want to change your program. But when a question is "can I have a program that takes this file as an input and produces output" in the context of using it within e.g. python, you respond with 7 steps of gui interactions which is a very convoluted way to say "no".

You are right. Cube Explorer is a quite old standalone program and not designed to work together with any other programming language. It does what it does and I have not the intention to add any new features to it. If you think seven steps are too much feel free to use any other program which is better suited for your purpose.

 

[bcube]

why do you not just use Cube Explorer? It can do what you want.

Two reasons. When I started this thread, I didn´t know exactly how to use batch solver in Cube Explorer. Thank you for your tutorial. Also, I was looking for alternatives to Cube Explorer.

I am sure there are more alternatives out there than I found - feel free to add some if you know about them. The performance comparison below is kind of useless until executed with the same hardware/software configuration in order to mutually compare solver´s performance.

Online solvers:

• this solver by Michael Feather, average solve: 4.81 s on 1 000 random cubes (hardware/software specification: HP Pavilion desktop computer, AMD Ryzen 7: 5700G, 3.8 Ghz 8-Core/16-Thread, 16 GB RAM (DDR4 3200 MHz), Google Chrome on the Fedora operating system)

Offline solvers:

• vcube by Andrew Skalski, average solve: 0.16 s on 10 000 random cubes (hardware specification: 32 GB RAM, Intel Core i7-7700K (4.20 GHz, 4 core, 8 thread) desktop, using a 22 GB pruning table) (average solve 0.26 s for nxopt on the same hardware and data set)
• nxopt by Tomas Rokicki, average solve: about 0.03 s on unknown amount of random cubes (hardware specification: a two-node, 16-physical-core, 256 GB RAM, using a 176 GB prunning table)
• Cube Explorer by Herbert Kociemba, average solve: about 12.34 s on 100 000 random cubes (hardware specification: Intel Core i7 920 CPU machine (Vista 64 bit with 6 GB of RAM) with 8 cores (4 physical and 4 virtual cores))
• nissy by Sebastiano Tronto, average solve: about 60 s on unknown amount of random cubes (Eventually it will become faster than nissy-classic at optimal solving, but without all other features, hardware specification: 4 cores at 2.5 GHz and using about 3 GB of RAM)