You didn't read carefully. Making a program search far more nodes for a given depth does NOT improve it. It just makes it look better at fixed depth since the extra nodes do not penalize the depth, and since time is not being measured, so what if it takes 5x longer than the opponent to reach the same depth? This is nothing more than a time-handicap match, which doesn't say a thing about actual Elo. I doubt Larry/Don would claim that parallel komodo is stronger at a fixed TIME limit... ignoring the extra nodes searched...Uri Blass wrote:"poor parallel implementation?"Daniel Shawul wrote:This is misleading because komodo probably compensates for poor parallel implementation by searching wider, otherwise it shouldn't be getting any elos for fixed depth test. You should do the test with time and see how much each gain.
The target of parallel implementation is not to get bigger depth but to play better.
If the implementation is good in helping komodo to play better than I think that it is wrong to call it poor.
Maybe it is the opposite and komodo compensates for poor pruning of lines that it should not prune by parallel implementation that prevent it to prune good moves in the relevant lines.
Comparing T(1cpu) to depth X to T(4cpu) to depth X will give a good estimate for the Elo gain for the SMP search. To my eye, it will be lower than expected because the tree grows too much.