Andre Poenitz wrote: >> The above is missing and then we need a full rebreak on buffer switching > > Ah yes. > > Well, I still think we should better aim for 'speed from simplicity' > than for 'speed gained by caches'. But it might be a bit too early for > that. Ot too late...
If that is the case, then we can happily gain the speed by simplifying, and then remove the cache. Doing the other way around will buy us nothing but enemies. > Could you re-post the patch nevertheless just to give me an impression > on how bad it would be? Here it is, no idea if it applies anymore. (anyway is mostly a brainless chopping here and there) http://marc.theaimsgroup.com/?l=lyx-devel&m=106370263723174&w=2 > which involves a full rebreak and seems acceptable _for me_ even for > buffer switching. But that's an AMD 1700+. And there are people (boxes) for which it's unacceptable... I don't think the cache bothers much in terms of complexity right now. > I'd be happy to trade the TextCache for a 1.5s delay when switching > buffers of the size of the UserGuide. It's not completely free now, > either... rebreak time is not an absolute value... (I remember people complaining) > But I'd rather get rid all of this and use 'row break on demand...' But you need to cache y-positions anyway. Alfredo
