Performance Problem Detected and Fixed in the Mullocator
The Mullocator was written a bit lazily in a certain area. This
has come to bite me now. So I spent the evening rewriting this
portion of the routine to get good speed there too. I used the
malloc-test.c
from Wolfram Gloger. The pleasant result
now is:
Apple:
real 0m0.455s user 0m0.320s sys 0m0.040s
Mullocator:
real 0m0.460s user 0m0.280s sys 0m0.050s
(The user time is of interest). So in that benchmark the win isn't that big. Still I was losing here before.
[Playing right now: By The Pain I See In Others from the album "Deliverance" by Opeth]