Merge DDSS into DDS V3

[BSalita]

I am the author of DDSS, a significantly faster fork of DDS. DDSS benchmarks 50% faster than DDS when solving 200+ deals in a single call. Why not merge the improvements, mainly memory management and performance, into DDS V3? I offer my help to a DDS V3 maintainer although an AI can equally do the job.

https://github.com/bsalita/ddss

[zzcgumn]

My personal research needed better reuse of transposition tables and the capacity to parallelise both over processes and threads. I have hence not looked into performance for solving large batches of deals.

Could you give a high-level overview of your work?

[ThorvaldAagaard]

I did implement DDSS in BEN and did not gain any improvements, as the SolveAllTables in DDS in version 2.9 was faster. I assume it is because SolveAllTables reuse the transportation tables and not just the memory, where it is correct that DDSS is an improvement if you only solve one deal at a time and then switch to batc

BEN use DDS to find the best play based on DDS, so in a lot of situations, BEN find 200 samples mathcing the bidding, and then lets DDS solve these 200 deals

As all 200 deals will have the same hand for the play and the saaame hand for dummy, the boards are similar in some way, and the internal transportation tables seems to be very useful, and currently giving the best perfoamce.

Actually BEN combine this with some Neural networks, heurustics etc.