glamiorew.blogg.se - Danball stick ranger wiki

#Danball stick ranger wiki how to
#Danball stick ranger wiki update
#Danball stick ranger wiki crack

Maybe GPU acceleration? I think this is a problem not very suited for GPUs, but with 1500 SPs I could eliminate nearly two stones on complex levels by just calculating all possible combinations for them at once.

I’d expect a 10x boost from nice C code, proper ASM should help another 4x, although that might be hard to do. (Ironically my first prototype wasn’t recursive.) Early tests have shown, around 30% of the time is lost in just calling the function.

Recode the thing to not use a recursive function.

I do that by hand and not automatically and a bit more sophisticated.

I said it would be fastest to have the smallest piece last in the vector. I don’t yet sort the pieces automatically.But since many puzzles take longer with this approach, I wonder if I can find a better way to do this. It won’t be possible to always choose the correct position for a stone without trying the other stones. I didn’t yet solve this problem and probably won’t ever. On the other hand, hard 4 calls the function 67922 times without this approach and 589329 times with this approach. Also the execution time is around 30% lower for this approach. And then luck might just favor approach 3) without this optimization.įor instance, with medium 7, with only approach 3), the function gets called 15179 times, with this approach it only gets called 6703 times. It just might be, that the first stone I put on the field has to create several new 0s. The problem is, its not guaranteed that this reduces the number of combinations tried. That generates a little overhead, but has the potential to reduce the complexity further. What I did was, the first few stones I put on the field, I first try them on every position on the field and then try the paths that generate the most 1s first. Big stones, sometimes just seem to fit somewhere. I usually lay the stones in a way where they don’t create many new 0s. 4) Human approach optimizationĪs it still wasn’t fast enough to calculate the very complex puzzles, I tried to make it play a little more like a human does. The overall speedup to the naive approach is around 60 billion x. All other combinations could be early skipped. The deepest call happened only 17 times, so only 1360 different combinations were tried with all stones on the field. But after trying out the proposed combination, it indeed solved the puzzle! Awesome! I looked at how often the recursive function got called. After I implemented that, I was sure I had an error in the code, because I got a result in just a few seconds. That way I could skip many many more unnecessary paths. Then from the last to the first I summed them up, so I knew how many 1s the last x stones could cover. I calculated how many 1s all stones in the list had.

#Danball stick ranger wiki crack

That still wasn’t fast enough to crack medium 7.

So if you can skip the tests for a stone with many combinations it gives a higher speedup then when you skip a stone with few combinations. Small stones have more possibilities to move. Now when you reorder the list of stones that the smallest stone is the last one in the list, it increases the speed again. If they were the equal, then I tried all combinations. Instead of trying each combination, I first counted all the zeros in the remaining field. Now in the recursive function, I handled the last stone differently. I counted how many 1s the last stone in the list had and stored it.

What I did next was a first optimization. It means that for the 4th hard puzzle, you have to try 497,082,506,768,547,840 (thats 497*10^15) different combinations, my computer can calculate around 5 million combinations a second (on a single core), takes far too long.

increase the “next” position and go to 2.

if the stones list is empty check if the game is solved.

Then I have a recursive function that does the following: Now I have another array of matrixes that contain the stones you can move onto the field. My program has a simple matrix that consists of 1s and 0s that stores the base level data. Some of the hard puzzles are still way out of range, I’d need another speedup of 1000x to solve them in a day. So I sat down and started coding on my brute forcer again and got a huge speedup that made it feasible to crunch even harder problems. I think I solved the new hard puzzle by myself, but I still wasn’t able to solve Medium 7 and 8.

#Danball stick ranger wiki how to

So I took another look at Laminate Pieces ( how to play on a PC), since it also got updated.

#Danball stick ranger wiki update

New Rings, the white stone bug was finally fixed (after the first update it still appeared on items bought from the shop) and PG got that Pen Paint addition. The last weeks on Stick Ranger and Powder Game were very boring.