Follow the process of creating a chess engine using NVIDIA's CUDA™.
Sunday, April 20, 2008
Parallel move generation
One of the advantages of using GPU's multiprocessors. Instead of generation all pseudo legal moves for a given chess position in sequence, I use 64 threads to generate all legal moves parallel!
1 comment:
Anonymous
said...
Heyho,
it's been some time, but i still try to figure out how you achieved 1Mnps on an Nv 8800 with 64 threads running during move generation on the same position.
My own approach with 64 bit based Kogge-Stone Vector move generation https://github.com/smatovic/Zeta/tree/v099a lacks far behind with max 150 Knps on 8800 GT, running 28 work-groups with 64 work-items each.
Please enlighten us and tell us what kind of move generator you used :)
1 comment:
Heyho,
it's been some time,
but i still try to figure out how you achieved 1Mnps on an Nv 8800
with 64 threads running during move generation on the same position.
My own approach with 64 bit based Kogge-Stone Vector move generation
https://github.com/smatovic/Zeta/tree/v099a
lacks far behind with max 150 Knps on 8800 GT,
running 28 work-groups with 64 work-items each.
Please enlighten us and tell us what kind of move generator you used :)
--
Srdja
Post a Comment