Closed jecisc closed 5 months ago
Did you see some performance improvement?
With Seba's profiler I see that we reduce by almost 10% the allocated size in memory during an import of a medium size model.
There is a little gain of speed but it's not huge because this represent just a small part of the import but also because we tweak the GC parameters to have a big young space (64MB instead of 2.8MB).
I guess that without this change of parameter we would have a bigger difference.
I was thinking to propose to reuse the same Parser in Number/Integer>>#readFrom:* methods but that would cause problems in case of concurrency since the class NumberParser has states.
This change caches the number parser. During the import of a model the number parser represented 5% en the objects created. With this change we only create 1.