Author malte
Recipients jendrik, malte, silvan
Date 2015-07-26.14:47:21
> If you think it's worth it, I can start a small experiment with a modified
> version of v3 that allows inlining again, so that we have some more data on
> whether or not this can explain the performance difference.

I did a tiny test locally with one configuration and three instances whose
runtime was adversely affected by v3: inlining the two method did not eliminated
the loss of performance, but it reduced it by a factor of three. If we can
extrapolate from this, we'd get a typical -0.01 on average speed, which perhaps
we might recover by getting rid of the additional indirection in the shrink
strategies (asking Distances for the distances directly rather than asking the
transition system to ask Distances). I suggest we let this be for now; it
doesn't seem large enough to worry about.
Date User Action Args
2015-07-26 14:47:22maltesetmessageid: <>
2015-07-26 14:47:22maltesetrecipients: + malte, jendrik, silvan
2015-07-26 14:47:22maltelinkissue561 messages
2015-07-26 14:47:21maltecreate