Closed SebastianM-C closed 4 years ago
This was a regression introduced in https://github.com/jgreener64/Molly.jl/commit/97ffa4c3dd3581b832d8cbcd17b2b94306cab5fd, thanks a lot for the fix.
Shows I need to do performance testing as part of CI. I'll look into a simple benchmark suite.
Totals | |
---|---|
Change from base Build 61: | -0.5% |
Covered Lines: | 219 |
Relevant Lines: | 276 |
Hi! I noticed that there some performance improvements that can be made. For example due to
atoms::Vector{<:Any}
a lot of dynamic dispatch was occurring inside force calculations. Adding another type parameter solved the problem and provided a ~40x speedup for Lenard-Jones interactions. I think further improvements can be made.