Vladimir wrote: Yuri Censor wrote: Vladimir wrote:
Your test must have been flawed.

Are you kidding me? the test is 100% accurate/correct!
Yuri Censor wrote: There are several possibilities:
(a) You used Jose Velasco's no-frills Firenzina compile capable of running on my Intel Core 2 x64 Quad at work. (It doesn't support SSE4.2.)
Your compile does not work, so i used his compile.
That explains it, thank you. What equipment are you using? If your computer has a modern Intel processor that supports SSE4.2 and hardware popcnt, then you should be able to run my /MT compile that I released yesterday to remedy the problem. It's available at
http://dgusev.cs.edinboro.edu/GameDev/Chess/Firenzina_2-2-2_xTreme_popcnt_sse4-2_intel13_MT_x64.zipThe initial /MD compile required DLLs from Microsoft Visual C++ 2012 Redistributable Package (x64). I overlooked it, because, as a developer, I have the DLLs, and I didn't realize that they're uncommon.

I ran a 500-game 1/1 match to quickly make sure that the /MT compile is no weaker than the initial /MD. The match ended in a tie today, +75-75=350, the kNodes/sec numbers were similar. The /MT executable is noticeably larger, due to static linking.
Compiler capabilities vary quite a bit, see some performance test results at
http://software.intel.com/en-us/intel-parallel-studio-xe . Jose Velasco's is an MSVC compile.