There is a new stockfish_151_x64_ja.exe dated 10/10/2009 compiled by Jim Ablett,
which is indicated as Update 10-10-09: Faster x64 compile.
Actually (looking at the infinite analysis of the start_position, using only 1 thread for reproducibility), it has the same speed (in kn/s) as the previous stockfish_151_x64_ja.exe dated 08/10/2009, but for the first time in a Stockfish version, the 64-bit version behaves exactly the same as the 32-bit version (and is 25% faster, at least on the start_position).
Bravo, Marco (and others who contributed to that result)!
There is a new stockfish_151_x64_ja.exe dated 10/10/2009 compiled by Jim Ablett,
which is indicated as Update 10-10-09: Faster x64 compile.
Actually (looking at the infinite analysis of the start_position, using only 1 thread for reproducibility), it has the same speed (in kn/s) as the previous stockfish_151_x64_ja.exe dated 08/10/2009, but for the first time in a Stockfish version, the 64-bit version behaves exactly the same as the 32-bit version (and is 25% faster, at least on the start_position).
Bravo, Marco (and others who contributed to that result)!
Bravo _only_ to Jim.
It is all up to him, we didn't touch a line of code in the released seurces. Not only it is faster (if Jim says it is faster then it means it is), but he has used the same STL library (probably MSVC one if result is the same of 32 bit) and that's the reason the functionality is the same between the two versions.
Thanks Jim !!!!!!!
We should inform testers to use the new build instead of the current one. Luckly testing is just starting now, this week was the Shredder week
Unfortunatly I cannot verify the speed of the 64 bit version because I have a 32 bit. But I _blindly_ trust Jim
I'll re-start my chess960 testing with the new exe.
I had played 2 opponents with the old exe, and I have to say results were not promising - a mere +15 ELO better than Glaurung 2.2 x64. Maybe the new exe will perform better
Spock wrote:I'll re-start my chess960 testing with the new exe.
I had played 2 opponents with the old exe, and I have to say results were not promising - a mere +15 ELO better than Glaurung 2.2 x64. Maybe the new exe will perform better
Fortunately I've been testing the 32-bit exe, but results aren't all they've been hyped up to be here either.
mcostalba wrote:Bravo _only_ to Jim.
Unfortunatly I cannot verify the speed of the 64 bit version because I have a 32 bit. But I _blindly_ trust Jim
Well then Bravo Jim!!!
As I said, since I could only compare in terms of kn/s with the previous 64-bit compile (the 1.5.1 of oct. 8), the question of compared speed is still open.
On the start_position, I found that speed was the same. I will see what happens in a few more positions (the ones I tested in http://www.talkchess.com/forum/viewtopi ... 13&start=0)
Maybe Jim can explain why he thinks his new compile is faster.
