Folders and files

Name		Name	Last commit message	Last commit date
moderngpu-master		moderngpu-master
.gitattributes		.gitattributes
.gitignore		.gitignore
GlobalVars.cpp		GlobalVars.cpp
InfInt.h		InfInt.h
LICENSE		LICENSE
Magics.cpp		Magics.cpp
Makefile		Makefile
MoveGeneratorBitboard.h		MoveGeneratorBitboard.h
README		README
UciInterface.cpp		UciInterface.cpp
chess.h		chess.h
launcher.h		launcher.h
network.cpp		network.cpp
perft.cu		perft.cu
perft14.txt		perft14.txt
perft14_run4.txt		perft14_run4.txt
perft15_run3_1.7z		perft15_run3_1.7z
perft15_run3_2.7z		perft15_run3_2.7z
perft_bb.h		perft_bb.h
perft_gpu.sln		perft_gpu.sln
perft_gpu.vcxproj		perft_gpu.vcxproj
randoms.cpp		randoms.cpp
switches.h		switches.h
util.cpp		util.cpp
utils.h		utils.h

gpu chess move generation / perft experiments
- Move generation uses bitboards
-- position represented using a set of 6 bitboards
-- two variations for sliding piece move generation: kogge-stone and magics
--- three variations of magics: plain, fixed shift fancy, and byte-lookup fancy
--- table sizes: 2.3 MB, 760KB, 150KB respectively.
-- kogge stone faster in some cases and fixed shift fancy faster in some
- perft routine can work with or without transposition table support
- peak speeds on overclocked GTX 780:
-- without transposition tables:
--- 27 Billion moves per second (pos2 of cpw)
--- 16 Billion moves per second (starting pos)
--- 1.5 hours for perft(10) of start position
-- with transposition tables:
--- start position perft(11) in 1 hour!