Hello Bruno,hcyrano wrote:Tord,
have you statistics (nodes, time, ... ) on Glaurung's code with 1, 2, 4, 8... cores?
Unfortunately, I have no such statistics. I have a dual-core CPU, but one of the cores is usually busy with some other task, so I very rarely have the chance to test my engine with more than one thread. Except for a single tournament two years ago, when I operated my engine remotely on a quad-core Linux machine, I have never even seen Glaurung run on more than 2 cores.
The only useful data I have are the results on the CEGT and CCRL rating lists. Based on the difference in strength between 1 CPU and 4 CPUs on these lists, it looks like the effective speedup with four threads is approximately 3.0.
Tord