After I was struck by leonard (312 balls record)
I stopped any activity on this fornicating stupid record !
But the madness of the code recalled me.
Here is a point of my code advance :
Clearing code was nicely optimized
5 gained rasters lines on my previous (non published) 283 record !
My 36bytes unit clearing datas storing have even been reduced (I did not calculate but average rate should be 28 bytes ?!)
But that is not enough !
I've tested the OR and MOVE display code(leonard one and mine). They are quite similar and CPU time used is nearly the same (the difference time is unimportant)
Now, I have to work on my databuilder. It is very bad in finding OR incomplete sprite.
Leonard used 3 types of OR sprite (1 complete and 2 incomplete = 4 top lines sprite & 4 bottom lines sprite suppressed)
I used 5 differents types of OR sprite for my 283 record.
BTW, Leonard, I have noticed a strange thing in your OR display code.
The 7th shift of the complete sprite :
Your sprite is bugged : to be complete, 2 instructions are missing :
Did you do that on purpose ?
Strangely, I did not find any visible incidence on curve display... but it should ?!
What's up with this?
"Then I fastly code a brute force research for all different sprite routines"
Number of permutations should be 312 ! = 2,1.10^644 (or would it be 2^312 = 8,34.10^93 ... mathematic gaps)
Anyway brute force research is just impossible... especially if databuilder is running for less than a few minutes...
You should have reduce in a drastic way the number of sprite (intelligent or empirical way)
I am running out of idea for the moment...