bupticybee / TexasSolver

🚀 A very efficient Texas Holdem GTO solver :spades::hearts::clubs::diamonds:
https://bupticybee.github.io/texassolver_page
GNU Affero General Public License v3.0
1.71k stars 304 forks source link

memory usage tree compression #33

Closed ghost closed 3 years ago

ghost commented 3 years ago

Are you using any tree or other memory compression tricks?

Compression is mentioned in the logs of jesolver (By the developer who did Cephesus)

https://jesolver.com/cmdref.html#set_compression https://jesolver.com/download/free/changelog.txt

He has some CFR compression work but it isn't clear if it applies - uses average strategy https://jeskola.net/cfr/

https://jeskola.net/cfr/dcfu/ https://jeskola.net/cfr/avgstr/

Simplepostflop mentions tree compression,

"Simple Postflop Solver has 3 different algorithms (A1, A2, A3) and SPF A1 algorithm could use different degrees of tree compression in the RAM."

https://simplepoker.com/en/News/Fair_Performance_Comparison_of_Simple_Postflop_and_PioSolver_56

". Also, algorithm can now use compression of strategy tree in RAM 2x, 3x and 4x times, which allows to run much more complex trees than previously your PCs allowed."

https://forumserver.twoplustwo.com/167/poker-software/simple-poker-best-gto-solvers-preflop-postflop-holdem-omaha-2-way-mw-amp-gto-tools-1518064/index8.html

bupticybee commented 3 years ago

I will look into that, thanks for the tips