The New Sudoku Players' Forum

by **champagne** » Sun Apr 21, 2024 5:50 pm

coloin wrote:Pretty good work 1
On the generating side [ before the pruning] Its only necessary to use 44 gangster bands { rather than 416]
This of course doesnt reduce the grids to be pruned by minlexgriding.

Use of the gangsters is more in line with my opening statement to start the expansion using band1/stack1.
This is likely good to have the quickest catalog builder, but unless I miss a point, this does not deliver a catalog directly in minlex order.

In fact, as producing the catalog is a on shot problem, having the best code is not challenging.

I am not sure that there is clients for a direct link
solution grid <==> 5e9 rank

but if yes, then we need either a huge lookup file, or an efficient catalog builder in the right order as basis.

by **champagne** » Wed Apr 24, 2024 7:20 am

The builder of the catalog in the min lexical order produced the NED=5 472 730 538 solution grids matching mladen'index.
This is now a virtual storage of the catalog through the builder.

The builder can start from any of these existing starts of the solution grids

416 band 1
9992 band1 plus mini row 1 in the box 4
652408 band1 plus row4

the code (dirty and partial so far) is available here

The first test has been done using the band1 table of starts. Next step will be to use band1 plus row 4 as start to see how long it takes to cover the entire field.
The first run let's hope some hundreds of milliseconds per { band1,row4 ] to find the chunk of solution grids.

The final target remains to have quickly the link { solution grid <==> NED rank}
At the end, something I never done, a DLL should be created.

The final size of this virtual storage of the catalog should be around 4/5 MB

by **champagne** » Thu May 09, 2024 8:40 am

The draft is now ready,

The core of the package is clearly the catalog builder in the min lexical order of the solution grids.

Several entries are available

for a direct access
give the rank 1-5472730538 get the solution grid
give the solution grid get the rank (option known minimal or not)
if the solution grid is not minimal, then the minlex morph is built

This is the part should be proposed in a DLL. The response is expected to be in average in some tens of milliseconds.

Are also available
the possibility to print part or all the table of the rows 4 with the index 0-415 of the band 1
And some pieces of the catalog for a given range of the rows 4

Some tests and some cleaning are still to come, but the repository has the current status of the code.
Note: the code has not yet been checked with the gcc compilation

here below

the bottom part of the rows catalog
first 4 rows; band 1 index 0-415; rows 4 start 0-652407; solution grid rank start; number of solution grids attached

Hidden Text: Show

and the bottom part of the catalog with the rank (done in 86 ms using the rows 4 entries

Hidden Text: Show

EDIT: the .exe size with microsoft virtual studio is 3.5 MB so this is somehow a virtual catalog in min lexical order of that size.

by **dobrichev** » Thu May 09, 2024 10:48 am

Hi Champagne,

Congratulations for the achievements!
To my knowledge this will be the first independent verification of the gsf's catalogue.

I am waiting for timings for entire catalog generation.

Cheers,
MD

by **champagne** » Thu May 09, 2024 2:04 pm

Hi mladen,

dobrichev wrote:To my knowledge this will be the first independent verification of the gsf's catalogue.

This is not quite true, but the process is hidden in the search of the 17 clues.
In the last version of this scan, we have 3 builders of the catalog, each one with it's canonical form, not min lexical.
One basic test has been to check that we get the 5472730538 ED solution grids.

dobrichev wrote:I am waiting for timings for entire catalog generation.

I start the process on one of my workers. It will be an enumeration process, but it's not so important for the full generation. The process based on the rows 4 is expected to be very close to the generation from scratch for the first bands, but mush faster later as shown in the example above

Cheers
Gérard

The New Sudoku Players' Forum

High density files for solution grids and 18 clues puzzles

Re: High density files for solution grids and 18 clues puzzl

Re: High density files for solution grids and 18 clues puzzl

Re: High density files for solution grids and 18 clues puzzl

Re: High density files for solution grids and 18 clues puzzl

Re: High density files for solution grids and 18 clues puzzl