dungpa / parallelalgorithms Goto Github PK

Results on an 8-core (hyperthreading) machine:
1. Primes:
	4x speedup for n = 10000000

2. PI:
	perform better than C# versions
	8x speedup with Parallel.ForEach, 5.5x speedup with PLINQ

3. NQueens:
	perform better than C# versions.
	8x speedup when pre-placing the first two queens (N = 14).

4. Sudoku:
	Failed to parallelize its depth-first search algorithm.

5. Strassens:
	perform better than C# version, but scalability seems to be strange.
	7x speedup for the standard algorithm, 6x speedup for Strassen's algorithm.

6. BootyDivision:
	perform better than C# versions.
	Some parallelism consideration with Booty Division Problem.
	Speeup is 3.5x on a 4-core machine.
	It's possible to further optimize it.

7. Geonames:
	Speedups are marginal due to sequential IO operation.
	Two versions using PLINQ and Parallel.ForEach are comparable for seq.

8. Mersenne:
	Speedups are comparable between PSeq version and Parallel.For version
	They are 3.5x on a quadcore machine.

9. EditDistance:
	Use parallel wavefront algorithm.
	3x faster than C# version.
	Speedup is around 3x.