I did a search for ‘long parsing times’ the corona forum gave me 1 topic and it was said to be solved with more RAM. im running 32gb DDR4 but its only using 16gb at the moment and my parsing times are taking about 10 minutes. problem is they are really unstable and can cause crashes or moments of not responding where it sometimes recovers.
the scene is using forest pack and xrefs for housing types. is there much i can do to optimise this?
if my RAM is only 50% used and my cpu is going from 40-90% ish during the process of parsing. is the bottleneck because its loading files from a network drive?
Hi, please answer on the questions below. This might help us find the culprit.
You mentioned that you are getting freezes and crashes. Please provide us with minidumps created during those freezes and crashes. It is very simple, here is how to do it:
https://coronarenderer.freshdesk.com/support/solutions/articles/5000524006
corona 1.4
and its mostly “not responding” errors on max and render window during parsing times. just trying to find a way to optimise this part. can be frustrating when trying to check stuff if i cant even get a low res render off quickly.
I’ve had this happen before to some of my scenes too; I solved it by merging my scene into a fresh new file.
Since you are using xref’s this might not apply, but might worth a try.
Also I would try a few things:
i’ll try the xrefs one by one n see if i can pinpoint a culprit thats slowing down the scene.
I realised there is also railclone objects within the xrefs, does railclone have any issues with corona? i think i remember a while ago there was a particular material feature that didnt work because of the way corona and RC worked together, i dont remember the specific problem but it was to do with using multiple material IDs within the RC object and corona rendering the geometry differently.
randomizing UVW mapping in RC does not work with Corona. It is also possible to disable instancing in RC, maybe that is the issue?
Instancing engine is turned on but on top of the RC there is a material by element. Does this stop the instancing? I’m pretty sure I can just use a random element within the RC editor rather than the max modifier but this is where i believe i ran into problems with Corona a long time ago, which might be why theres a material by element modifier instead of the RC material node.
maybe im talking nonsense and i can just use the material node to get random IDs without the modifier?
I dont know details how RC works, but you can determine it from the number of unique/instanced primitives - unique should be fairly low and isntanced high, if both are ± same, then there is no instancing going on
a quick test with material by element on gives me 545k unique and instanced. remove the modifier and the unique drops down. but when i add the material node to my segments and tell it to randomise my IDs like the modifier was doing, nothing gets randomised. they all use the same material maps. this is where i think i got stuck last time.
maybe im using RC wrong? anyone else use multiple IDs and RC with Corona?
ok, so it seems RC decides not to instance it in the first case, that is the cause of the problem
without materialbyelement the unique dropped and instanced stayed up, which means it is correctly instancing? but my problem is, without that materialbyelement im not getting randomised maps to add variation.
so if i remove the materialbyelement then i should get faster parsing times because it will be working correctly but without that modifier i lose my variation. fix one problem, end up with a different problem 
Yep, you can have instances or random material IDs, but not both. Only Vray supports some black woodoo that RC does and lets you have instances and random material IDs at the same time.
Maybe this will be helpful:
http://docs.itoosoft.com/display/RAILCLONE/Requirements
Although I do not see a specific point for “randomizing UVW”.
so because RC doesnt support the random material in Corona would it be better to collapse the tiles into editable polys and attach them together per roof as such. would that be faster than it rendering each tile as unique? or at this point does it make no difference if they are attached in groups or not?
we need the variation but also need speed and stability
Hi. I don’t use RC, but I was wondering if you have tried using a Corona multi map set to instances?
This would give you the variation + the “instance” speed (if it works with RC).
Regards
I did try setting it to instance but it didn’t seem to do anything. I am using corona multi map in the material but it required the MBE modifier to work with the RC object.
Hi. The MBE modifier doesn’t work with instances, and if I understood your writing correct, the MBE modifier “breaks” the RC instances into unique objects. Therefore the Corona multi map set to “instances” won’t work, since it needs instances in this mode.
If RC instances doesn’t work with the Corona multi map directly, then maybe you could convert the RC object into intanciated polys/meshes - then the Corona multi map will work in instance mode.
Hope it helps
sorry i didnt type that very well. I took off the MBE to try RC with instance multi map and it didnt work but changing it to IDs and using MBE did work, but it only works because it breaks the instancing. i cant seem to get instancing and multi maps to work together.
with what Romullus said earlier and the link that Maru posted, it seems itoo only supports VRay with this magic at the moment. Which is a massive pain for me because of how unstable my machines are being 
now im just wondering how to collapse it all to be best optimised for Corona to render it without overloading it like i already am.
its a largish site of houses but only for the aerial will everything be seen. so i wonder what the best way to render the smaller shots is? i obviously want everything around to be in window reflections etc but not everything is seen. do i need to start hiding as much as i can?
Hey OP, I had a similar problem last week. I was using multiple xrefs as well, that contained RailClone and Forest objects. After going though each of the xrefs, I discovered none of them had any issue rendering on their own.
It really just came down to some sort of mystery corruption in the final render file that just had cameras and an HDRI in it. Starting from a blank file, importing everything from the corrupt scene (even the render preset), solved the problem for me.