Yeah, I agree this is a valid concern.
From a performance point of view, we'd have to gather a few different big
test cases, but the previous changes lead to a factor 3 improvements. In
this very specific case (the BigResolutionTest), the improvement for
depth-first was an additional factory of
On 6/26/15 10:15 , David Bosschaert wrote:
Looks like a great piece of work, Guillaume. I left some minor
thoughts in a few places on github.
One potential concern is that the change to depth-first search will
produce different results than the previous approach.
I wouldn't worry too much
Hi Guillaume,
What a work you did !!
It looks really great !
I cloned your repo to test and take a deeper look.
Thanks !
Regards
JB
On 06/26/2015 03:06 PM, Guillaume Nodet wrote:
I've spent the last two weeks working on optimising the resolver.
The results are available in the following
I've spent the last two weeks working on optimising the resolver.
The results are available in the following github fork
https://github.com/gnodet/felix/commits/FELIX-4942
The commits are split so that they can be reviewed more easily.
Most of them do not really change the algorithm in any