I have seen a couple of reports that Eugene's code (egtb.cpp) won't run if compiled with visual studio. Not having any windows boxes around, I have not tried to debug this yet. It works fine with any version of gcc I have tried, as well as with several versions of the Intel C compiler (all on linux). I think Mike or Tracy has compiled on a windows box and run into this issue as well...cdani wrote:Thanks! I was having trouble making it work in Visual Studio. I managed to compile a version without cpus, but with cpus was crashing.velmarin wrote:Hi, Dann and all, built with Intel compiler.Dann Corbit wrote:I will deliver one on Monday.
P.S.
It cannot be built with visual studio.
The EGTB code that gets created is invalid.
It has to be built with gcc.
The supplied makefile does not work, and it takes a lot of fiddling to make a good one.
I could do it again from here, but I am too lazy for that.
I am sure someone else can make one.
I am not familiar with the crafty code.
Defined CPUS how 8
POPCNT defined how on and off.
Nalimov EGTB I not have on my computer, I have not tried.
Define smpmt = x(número de cores) in crafty.rc (I think that it is the case)
Added Wb2Uci of crafty24.
Linked statically.
The three executables work on my system.
Waiting for some official relase, they have to play a little.
Merry Christmas.
http://www.mediafire.com/download/d4pkq ... ty-25.0.7z
I'm trying the 64 bits popcnt with cutechess-cli and works nicely.
One thing you absolutely must not use (for gcc, and probably other compilers using a different option name) is the -fno-alias. Eugene has multiple places where different pointers point to the same thing, and using the above option will make the compiler produce a broken executable that will crash on egtb probes. I use that option for my source as it is safe and produces faster code. But not for egtb.cpp