Closed lafita closed 6 years ago
The performance evaluation showed poor scaling when running independent CeSymm runs in different threads. The scaling for 8 threads is only of 2.5 and using more threads makes the calculation slower due to thread overhead.
The ultimate solution for running CE-Symm in large collections in parallel should be using MMTF format and Spark.
MMTF
Spark
The performance evaluation showed poor scaling when running independent CeSymm runs in different threads. The scaling for 8 threads is only of 2.5 and using more threads makes the calculation slower due to thread overhead.