Am 28.11.2018 um 23:05 schrieb jr:

> btw, how much parse time difference do you expect between this and the
> x.tokenizer (compared to the previous)?

I haven't found time to work on the tokenizer recently, so there's not 
much difference between x.tokenizer.9945666 and x.tokenizer.9844488, and 
I expect no measurable difference at all between those.

As for the difference between x.tokenizer.9945666 and alpha.9945627, 
that totally depends on the structure of the source code, and the 
architectural changes run too deep to make any quantitative predictions 
for any given scene. I had originally hoped for roughly similar parsing 
times at this stage while fearing slightly worse. So far it seems that 
the implementation is performing a tad beyond my hopes in most cases, 
while matching my fears (but for other reasons than I had anticipated) 
in a few other cases.

In other words, from experience so far I would expect most scenes to 
parse slightly faster, while some special cases may still parse slower.

I'd like to mention that I'm interested in hearing about any of the latter.

Post a reply to this message