The GPU part of the code is way too complex to be efficient.
Consider simplifying it (a lot!)
that is the point, the binary transformations involved create unique values(no repeatence, no collisions) and hit areas of the range that random or sequential may never reach,
yes, its not that fast, there is trade-off, but its unique in its own way
so threoretically there are multiple targets (horizontally or vertically or inversion or reversed) - hitting which automatically reveals the main private key, and i believe it could be effective
and because all transformations are in binary format they are quite accurate, hitting those tiny places, that random or sequential may never reach
for a better understanding i made a pc version of the same algorithm on python
https://github.com/puzzleman22/Bitcoin-puzzle-transformations-CPUyes of course I realize that the range is still huge
I believe that with enough computational power this method can be very effective
Hello friend. Thank you for your project. I don't care how it works as long as it works

I'm already testing it on 45 bits to see how long it will take, but I believe it has great promise. I'm sorry I'm poor and can't contribute, but who knows, maybe something will happen, and I will definitely remember you. Thank you.
ps: can't really work with any interval other than starting with 1 to 1?

because full transformations are applied from 0 to F in hex
with "1" prefix is simplier to apply them, on prefix "2" or "3" we need to repeat them for both
yes you can use it on any range, but it will not be symmetrical for them
for now its just for simplicity
i recommend start from 21 puzzle, then increase to see how it goes...
GPUS are not meant for this kind of approaches, because it creates so much divergence, i made the infinite loop within a gpu, each thread, thousands of threads are looping the transformations until one of them finds the target, so no CPU at all, then it stops and show the private key on the console and returns to CPU to save to a file result.txt
i am still working on it speed optimization
Okay, I get it, friend. I narrowed it down to puzzle 33, and when I find the target, I'll see how long it took, since we're only talking about 9 characters in the private key. Once I find 33, I'll try it again the 33 and see how long it takes a second time, and I expect very different times if there's no set sequence. And I really hope there isn't