Performance of different semirings #460
Replies: 1 comment 1 reply
|
Great question. Do you have specific examples to compare? In SuiteSpare:GraphBLAS, it's generally recommended to avoid "GENERIC" operations, which can occur for many reasons. To see whether an operation is "GENERIC", enable the "burble" and look for "generic" in the printed diagnostic info. For example: gb.ss.config["burble"] = TrueHowever, generic operations shouldn't be an issue once the JIT comes in SuiteSparse:GraphBLAS 8.0 (coming soon!). Also, UDFs and UDTs are generally slower too. Many of the operations used in I don't know how to answer this question in general other than "benchmark and compare". It may be nice to gether recommendations here for specific operations. |
|
Thank you, Erik! It would help if I had a specific example, but this was just a general thought I had when going through my notes. |
Uh oh!
There was an error while loading. Please reload this page.
When running specific calculations/algorithms, different semirings can yield the same answer. Is there any guideline/heuristic on which semirings might have better performance (memory footprint or speed)?
All reactions