Slime Finder

A Slime Chunk finder implementation in CUDA to find the 16x16 area with the most slime chunks around.

Commands

# compile
nvcc src/main.cu -o build/slime_finder

# run
./build/slime_finder

Algorithms

• v1:
  • Each device thread computes 1 full row of chunks at a given Z coord.
  • The host finds the best row from the results of the previous step.
  • Duration: ~1:50 secs.
• v2:
  • Same algorithm as v1.
  • A queue is used to cache previous computed columns.
  • Duration: ~0:45 secs.
• v3 (TODO):
  • Takaoka's parallel algorithm (check sources).

Sources

• Cubiome's implementation of java random in C.
• Tadao Takaoka's paper on Efficient Parallel Algorithms for the Maximum Subarray Problem.