| commit | b6b781d8d358099d5185b920b374e5ed46d5ed96 | |
| author | Petr Baudis <pasky@ucw.cz> | |
| Sat, 9 Jul 2011 12:34:18 +0000 (9 14:34 +0200) | ||
| committer | Petr Baudis <pasky@ucw.cz> | |
| Sat, 9 Jul 2011 12:34:24 +0000 (9 14:34 +0200) | ||
| tree | ceaa9a60608f6f4c1cef67b8988084ca25c7dcd1 | treesnapshot (tar.gz zip) |
| parent | 74c1c6495dafcbdc454a5aa88f19cbe7af3a7615 | commitdiff |
| parent | 6d8788da3b178c51dcb3e71279e1637ebafae0bd | commitdiff |
Merge branch 'ltreeowner'
Our general policy is not to merge experimental branches into master
until they show clear merit in play-testing compared to master.
local_tree should be slight improvement based on my tests, but
Jean-loup's results differ and we could not explain why yet.
However, in this case I'm merging in ltreeowner anyway, since the
local tree code is in master already anyway, and it is completely
dead. These are significant changes to the code and the whole
mechanism that should make the feature much more feasible; since
we already have local trees in master anyway, we might as well
finish up the development in master too.
Our general policy is not to merge experimental branches into master
until they show clear merit in play-testing compared to master.
local_tree should be slight improvement based on my tests, but
Jean-loup's results differ and we could not explain why yet.
However, in this case I'm merging in ltreeowner anyway, since the
local tree code is in master already anyway, and it is completely
dead. These are significant changes to the code and the whole
mechanism that should make the feature much more feasible; since
we already have local trees in master anyway, we might as well
finish up the development in master too.
| uct/internal.h | diff1 | diff2 | blobhistory |
| uct/policy/ucb1amaf.c | diff1 | diff2 | blobhistory |
| uct/uct.c | diff1 | diff2 | blobhistory |