scan solution grids for 17 clues as of blue

Programs which generate, solve, and analyze Sudoku puzzles

Re: scan solution grids for 17 clues as of blue

Postby champagne » Thu Dec 07, 2017 9:00 am

Search 17 milestone

The ongoing process is the search of 17 clues puzzles having >=7 clues in band 3 (blue’s pass 1).About 40 “band 1” have been fully scanned, and 20 more are in process. Each result is or will carefully be compared to the list of known such 17 clues.

So far, the current process finds all of them, but no new 17 showed up.

2 new bugs have been fixed, one in the process itself and one overflow in an interim table. I have still one failure to debug and I intend to release a new version of the pass 1 after the debugging of this case.

The current process should not be changed significantly for the bands 1 with a minimum of clues >=4.

I am still trying to improve the “bad cases” covering bands 1 with a minimum number of clues <=3. I drafted a search with and index 3X3Y, but I feel that it will be better to use a 2X2Y as main loop extended to 3X3Y only for the vectors.


The current version is faster for the bands with a high minimum of clues (5-6) due to the fact that the 2X2Y is not used when the number of ‘X’ ‘Y’ is small. The process itself is not changed out of bugs fix.

EDIT the last failure was due to another table overflow. The last code has been uploaded (see post 1 for the link)
champagne
2017 Supporter
 
Posts: 5741
Joined: 02 August 2007
Location: France Brittany

pass 1 last release

Postby champagne » Sun Dec 10, 2017 8:46 am

Search 17 pass 1 last release.

The frame for the search of 17 clues puzzles having >=7 clues in band 3 (blue’s pass 1) is now in a final status on my side. The code and the exec are available through the link given in the first post.

The comments will be updated ASAP.
Next step will be to release the final version of the 16 clues search.

Next changes in pass 1 will come from possible bugs unless some external contributions show possible improvements.

The work on the 3X3Y has been highly disappointing. Reversely, the variable mXpY frame with m;p<2 brought better run times.

Indirectly, the 3X3Y study was source of an unexpected huge improvement in the performance. In fact, moving pieces of the base tables in fix areas to process a chunk was a bad duplication of the cache work. The moves have been replaced by pointers to the main tables nearly halving the run time in my sample test.

I am now waiting for the results of the search on new ”bands 1” to verify that all known 17 show up. This should be done this week.

The current search is for bands in the range 40-90 far from the “best cases” where band had a minimum of 6 clues.


In parallel, I’ll make a deep analysis of closed bands and I’ll run new tests to evaluate the efficiency of the new code.
champagne
2017 Supporter
 
Posts: 5741
Joined: 02 August 2007
Location: France Brittany

Previous

Return to Software