From 9ac266c76f39620d8343e46ca41cb09d1534384c Mon Sep 17 00:00:00 2001 From: Sebastiano Tronto Date: Sat, 7 Dec 2024 16:38:40 +0100 Subject: Merge the "solver-experiments" branch that I have been working on for a few weeks. This include mainly three things: 1. Various tweaks for a total performance gain of around 30%. 2. Take into account symmetries and avoid repeated work. This required a re-work of the splitting into tasks before the solve. 3. Add a second fallback table (eoesep). This gives huge performance gains for particular scrambles (e.g. superflip). After merging this commit, remove and re-generate all pruning tables. Squashed commit of the following: commit 60f0705d2d69050e6a30581a2810f686d6f69b80 Author: Sebastiano Tronto Date: Sat Dec 7 16:06:48 2024 +0100 Fix indentation commit cc5d489a251812b6188c0ba264ac6cb2236f1afe Author: Sebastiano Tronto Date: Sat Dec 7 15:56:19 2024 +0100 Updated documentation commit a3f605dd628546e52564f82b139feb473b0725f3 Author: Sebastiano Tronto Date: Sat Dec 7 14:01:11 2024 +0100 use eoesep table as second fallback commit c75e43c9116c64f97e92b0fe038be8a032925a72 Author: Sebastiano Tronto Date: Fri Dec 6 16:13:23 2024 +0100 First commit for gendata_eoesep commit fea7688ab8bdc5ae0c3480622e2510a3bcd39248 Author: Sebastiano Tronto Date: Tue Dec 3 17:31:00 2024 +0100 Add scramble to tool commit 66866cb71dea4ca8278ecb9e90ff4295771feb42 Author: Sebastiano Tronto Date: Tue Dec 3 17:22:59 2024 +0100 Added tool to check multiple solutions commit ef65611c772c3996bddca8d181da3538e0af1674 Author: Sebastiano Tronto Date: Tue Dec 3 17:16:20 2024 +0100 Write all solutions for symmetric positions commit e3ded26db7d7d4ae7c0e2488151ed14f581f7b8e Author: Sebastiano Tronto Date: Tue Nov 26 09:14:54 2024 +0100 Added symmetry filter (TODO: print excluded solutions) commit 864c437a9751c58d58562650ca9eba4a9e6ad3eb Author: Sebastiano Tronto Date: Mon Nov 25 14:51:13 2024 +0100 Improved task split commit b88926d36d7ab0c64c5fe3bb954fd15d41267fba Author: Sebastiano Tronto Date: Fri Nov 22 19:06:41 2024 +0100 Reworked tasks for multi-threading in view of symmetry filter commit 26fa653f97df8cd601aecb80eaf89f0a00e9ba9f Author: Sebastiano Tronto Date: Thu Oct 31 15:37:43 2024 +0100 Added transform move commit 19f655ef94d658eaa2fefb5cea3c167a3ec58db6 Author: Sebastiano Tronto Date: Thu Oct 31 09:29:15 2024 +0100 Clarified doc commit 3b0fe1e5ef8b628854e30f0f0067300e2763c954 Author: Sebastiano Tronto Date: Thu Oct 31 08:36:58 2024 +0100 Handle solved cube correctly commit 57705cbc4982e3abe97a36ed64871738d4f721c0 Author: Sebastiano Tronto Date: Thu Oct 31 08:24:30 2024 +0100 Close file commit fc7d462b58bcf3d3a3fbf26c1f3bd04c640a4898 Author: Sebastiano Tronto Date: Tue Oct 29 15:05:48 2024 +0100 Removed stats tool commit 359bf7cb49ef405ee76ed662207d47cb2abcc5a9 Author: Sebastiano Tronto Date: Tue Oct 29 15:01:51 2024 +0100 Updated theory doc commit 39c315af562bc4ce896f41004388a4c34d475d37 Author: Sebastiano Tronto Date: Tue Oct 29 14:51:40 2024 +0100 Remove unused constants commit 57a5d24538aa59a4df9221dad2f99e9f0286bd9d Author: Sebastiano Tronto Date: Tue Oct 29 10:20:38 2024 +0100 Add tool to solve scrambles from file commit 07e2918c216636891b1fa6adecc9756086a901e9 Author: Sebastiano Tronto Date: Mon Oct 28 17:00:00 2024 +0100 Add make table to tool commit f5e5266c654eb027a5a35c57cc618555246f5e5e Author: Sebastiano Tronto Date: Mon Oct 28 09:35:49 2024 +0100 Remove old solver, other small things commit a1ec78025b7959dbb845213f7f4e6851ecebc204 Author: Sebastiano Tronto Date: Sun Oct 27 02:00:29 2024 +0200 Improvements commit 8eea23dbe888d923e662e24ae969130e2c67b999 Author: Sebastiano Tronto Date: Sat Oct 26 12:24:19 2024 +0200 Makefile fix commit 3fc3927beacc78971cefeb42da8d71fe6c015fc1 Author: Sebastiano Tronto Date: Fri Oct 25 18:25:09 2024 +0200 More performance gains commit 7b4efa1f9af9722de1ab9ccfc27899825a0d12c4 Author: Sebastiano Tronto Date: Fri Oct 25 15:53:20 2024 +0200 Alternative solver implementation, small performance gain --- doc/h48.md | 31 +++++++++++++++++++++++++++---- 1 file changed, 27 insertions(+), 4 deletions(-) (limited to 'doc/h48.md') diff --git a/doc/h48.md b/doc/h48.md index 323453a..678e160 100644 --- a/doc/h48.md +++ b/doc/h48.md @@ -230,9 +230,8 @@ can be one of three kinds: If the base value is `b`, a pruning value of 1, 2 or 3 can be used directly as a lower bound of b+1, b+2 and b+3 respectively. However, a value of 0 could mean that the actual lower bound is anything between 0 and b, so we - cannot take b as a lower bount. Instead we have to use a pruning value from - another table, for example the corner-only table mentioned in the previous - section, or a completely new one. + cannot take b as a lower bound. Instead we have to use a pruning value from + another table - see the section "Fallback tables" below. * 1 bit per entry, or `k1`: With one bit per entry, the only information we can get from the pruning table is wether or not the current position requires more or fewer moves than a fixed base value b. This can still be @@ -240,6 +239,24 @@ can be one of three kinds: pruning values. (Work in progress - `k1` tables not available in the code yet) +### Fallback tables + +When a pruning table does not store the exact lower bound value, for +example in the case of `k2` table as described above, we need to refine +our estimate using another table, which we call *fallback table*. +In the current implementation, we actually use two different tables: + +* **h0** table: for larger `k2` tables we get a fallback value from the full + table for the **h0** coordinate. +* edges-only table: to improve the pruning value for certain specific + scrambles, namely whose with solved corners such as the superflip, + we employ a second table that takes into account the edges of a full + **h11** coordinate. This table is small (around 1MB). + +More tables could be used to refine the fallback estimate, but each +additional table leads to longer lookup times, especially if it is +too large to fit in cache. + ### Estimation refinements After computing the pruning value, there are a number of different tricks @@ -285,6 +302,12 @@ as described above. For doing this, we need to replace the cube with its inverse, and keep track of the moves done from now on so that we can invert them at the end to construct the final solution. +When this technique is used, it is also possible to avoid some table +lookups: if the last move applied to the cube is a 180° move *on the +inverse position*, then the coordinate on the normal position has not +changed. Thus if we keep track of the last computed pruning value, +we can reuse it and avoid an expensive table lookup. + ### Other optimizations Other possible (low-level) optimizations include: @@ -344,7 +367,7 @@ don't have this much memory. ### 2 bits tables with for h0 and h11 -(Work in progress - currently I there is no specialized routine for +(Work in progress - currently there is no specialized routine for computing 2 bits table for "real" coordinates; instead, an optimized version of the generic method explained below is used) -- cgit v1.3