ArdaxT

Agreed. I only did Previs for current cell, and made sure to put myself on the cluster's center. It's trivial to figure out in Excel if you want to do that. Also, don't forget to clean up your plug in in xEdit after. PreVis will add extra cells to your mod that you didn't modify. Since the PreVis timestamp is updated, you can't just clean ITMs either.

Oh, and to reply to your actual question Zorkaz: It seems to be required to do both PreVis and PrecombinedVis.

See, I did up my Excel sheet a different way, it was just a running list of cells. Added 2 columns to calculate the nearest 3x3 center. I wish I knew how xEdit sorted cell lists, because I'd resort my lists depending on whether I was working in xEdit or Creation Kit and matching sort orders helps a bit. That said, having the data of what to build would have been cool too. I was thiiiis close --><-- to having a largely touch-free system for rebuilding precombines and previs. xEdit isn't really designed for unattended operation, so you had to run scripts against plugins by hand, but that was it. Used the CK's command line params to generate precombines and previs data, xEdit scripts to handle putting it all into a plugin for the user, and a PowerShell script to guide the user through everything, bundle the results into a BA2 archive, and clean up afterward. I had script variants that theoretically could have handled an entire load order and rebuilt precombines for every interior and exterior cell. Think FO4FI, only less disk space by using shared precombines, better coverage, and customized for your load order. It would have taken a really, really, really long time to do. My mid-range 2019 laptop (9th gen quad core i5, 16 GB RAM, NVMe SSD) was churning for 4 days straight and was -- by my estimate -- about 40% of the way through the Commonwealth before I killed it. A high-end desktop with gobs of RAM may have been able to finish in about that time. After getting burnt out trying to solve the "green diamond glitch" in my precombines, I kinda put things off to the side for a bit. I'll go back and rebuild the plugins I made in my FO4FI patches mod (the long way :cry:), but until I solve that bug I'm done for the time being. I'm happy to share my work with anyone that's interested and wants to take a crack at it. I suck at documenting things, but I'm willing to email/chat anyone through what I've created provided they're comfortable with xEdit and the Creation Kit.

Well, that shouldn't happen at all really. Do you run anything that might use CPU in the background? Distributed computing like Folding @ Home or anything like that? Long running video transcodes?

There's almost certainly a list search or sort or something similar that scales poorly. It's fine for smaller numbers of items, but really quickly goes from "acceptable" to "I'm not sure this is ever going to finish!" (Anyone who's done enough CompSci study to understand big-O knows what I'm talking about.)

You could also write an xEdit script to copy all the REFRs from every parent into your cell and apply the matswaps to them. It'd look a little something like in the spoiler. I have something I could adapt out of some scripts I was working on trying to build an automated precombine building pipeline. if you're interested.

This is a known issue with precombines. You need to apply the matswaps to the object reference instead of the base record. Zilav posted an xEdit script in the performance megathread, link is below. https://forums.nexusmods.com/index.php?/topic/5522717-fallout-4-optimization-and-performance-systems-explained/?view=findpost&p=56626651

Delete the plugins.txt and start Fallout 4 to create a new one. Zilav wins the thread.

So maybe there's something wrong with the contents of the file itself? What's in your MO2 profile's plugins.txt? And does xEdit work outside of MO2?

Yeah, Windows will occasionally give you a hard time with accented characters in your username. It shouldn't be a thing in 2020, yet here we are. I think Steam still has some bugs around that too. Have you updated xEdit recently? Maybe try rolling forward or back through some different versions and see if that changes things.

Looking at xEdit's source, it does all the right things to find the path to plugins.txt. Something as simple as a space in the path shouldn't be causing problems. Do you have any non-ASCII characters in your user path, like accented characters or something? They shouldn't be causing a problem, but Windows can get weird at times.

Yeah, this is a known issue with material swaps. Swaps on the base record don't show up in precombines, you have to do it on the placed object. There's more discussion about it in the performance megathread.

Unfortunately it's all over that area, not just that building. Not every cell, but it seems to happen in clusters. The only mods installed are UF4P and Beantown Interiors, then the plugin generated to hold the precombines. Maybe it's something to do with how I'm merging the precombine data back into my patch plugin? I was never able to get the Version Control stuff working properly, so I just wrote an xEdit script that copies the PCMB and XCRI data from CombinedObjects.esp back into the target plugin. It's in the spoiler below. I have a 2nd one I run after generating the PreVis data.

Yeah, it does the same thing up close. I can walk right into them. There's no collision on them. So I noticed that if I generate them by running "-GeneratePreCombined:<plugin> clean all" then I get the .psg file, and yes I generated the .csg and .cdx files. If I generate by running "-GeneratePreCombined:<plugin> filtered all" then I don't get the .psg shared geometry file and the generated .NIF files are a lot larger, as if I'd done the generation from within the UI. But I get the green diamonds in both cases. No. I've only seen this by doing the generation from the command line. Setting bUseCombinedObjects=0 does make them go away. Toggling tpc from the console doesn't change anything. So that points to it being precombines in some way. I tried looking at the NIFs in Nifskope, but there wasn't really anything there. I made the hex edits. If I didn't, the game crashed as soon as it tried to load any of the generated precombines. I am using Nukem's CK patches now too, but this was happening before I tried using them. I'm not going to have any problems using both, am I? I figured that Nukem's fixes weren't doing the same thing as the hex edits.

Good eye. I changed the UI color and the diamonds stayed green though, so that didn't gain me anything. I also tried regenerating the precombines/previs with and without shared geometry (i.e. using -GeneratePreCombined:<plugin> clean all and filtered all) and the same thing happens.