TalkChess.com

[Daniel Shawul]

I guess it depends also on whether you probe in qsearch as Ricardo pointed out. I forgot about that thinking I have an 85% tt miss on all nodes when it actually is on small subset. I see what you mean with not having enough time after prefecthing. There is really not much code for me after my do_move and a tt probe. I have static pruning testing, clearing node and testing for draws (probably biggest consumer). I maybe able to add legality testing in there too.
Multiple probes for the eval cache and using a smaller pawn tt that could fit in L3 is something I need to test. Eval cache takes more time than the main cache in my case but I didn't use buckets.
_________________
https://sites.google.com/site/dshawul/
https://github.com/dshawul