64 in fact. there are negative depths in QS. so the patch is already bogus with depths above 64. plus using 16 bits will surely backfire one day. besides, the testing was not done correctly mixing new stuff with hacky optimization, in an attempt to push the whole branch across the finish line.syzygy wrote: An example you should be interested in: only using 8 bits of the hash entry to store depth makes it impossible to store depths > 128 ply (ONE_PLY being 2)
I also agree with Marco. He has made it clear to Ron that he was not going to commit his mess of crap, which Ron ignored. Pure waste of time and fishtest resources. The guy is obsessed with his TT packing, and never listened to any comment from Marco. I dont think we can work with people like that.
The amount of resources that have been wasted by Ron and his idee-fix of packing TT entries is phenomenal and sets a new record. He has beaten Uri Blass in fishtest trolling, which is no small achievement
The hardest job of a maintainer, besides the actual coding and testing, is to say NO. Marco has done a very good job, and, if anything, he is guilty of having caved under pressure and accepted crappy patches to please people who insist too much, rather than the opposite.