Hi Tomislav,
Regarding your grievance, Edit layers have been part of UE since UE4.24 now, and the burden of supporting both the non-Edit layers and the Edit layers part was becoming too great for the small team that we are, with no real benefit. After all, a non-edit layers landscape is identical to an edit layers one with a single edit layer, so the conversion is trivial and we made sure that existing landscapes remained 100% compatible on load. Therefore, from the user’s perspective, the switch to edit layers landscape should be transparent. The 2 systems used very different code paths (the former is CPU-only, the latter runs on the GPU and is read back on the CPU) and more and more features and optimizations we are making are only possible on the edit layers system.
What prevented us from doing so up until now was :
1) No support for the Retopologize tool (i.e. the XY offset texture system + convex collision mesh) when using edit layers
2) Edit layers are bound by VRAM and could lead to out of (video) memory situations on particularly large landscapes due to how the GPU merge of edit layers was previously implemented
The reasons we moved ahead with the removal of non-edit layers landscape in UE5.6 are:
1) The Retopologize tool was hardly understood by users, its impact, almost inexistent with the type of landscape resolutions that are used in games (including Mobile ones) nowadays, plus it didn’t allow for overhangs. It ended up being rarely used as a result (or it was, but by mistake). It also had a very bad impact on memory and performance both on the GPU (it requires an additional texture to store and sample the XY offsets) and on the CPU (it turned the landscape heighfield collision component into a full-blown convex mesh collision component, which requires more memory than heightfields and is less efficient). We deprecated the tool in UE5.5 for non-edit layers (it was already disabled for edit layers) with a CVar to let users know and prepare themselves for the full removal in UE5.6, should they ever need it. There was not a single complaint or comment to that effect, so we decided to move forward with the removal in UE5.6
2) For several UE versions we implemented a new way that edit layers get merged on the GPU with the so-called “batched merge” code path, which we finally enabled by default in UE5.6. The benefits of the new algorithm is mainly the VRAM consumption which is now capped (if you don’t user BP brushes), instead of being unbound in the past. This means that even for very large landscapes, the merge will now happen on smaller landscape regions, avoiding the problem of running the algorithm on the entire landscape (if all landscape actors are loaded in the editor), which could quickly lead to Out Of Memory issues. The new algorithm also opens the door for several optimizations in the future, since the notion of locality is now embedded in the system. The 2 previous versions of the algorithm were still ready to use in UE5.6 but the batched merge system is now solid enough that we’ve been able to remove the 2 old code paths in UE5.7
With those 2 issues solved, there was no benefit to keeping non-edit layers so this is why we went ahead with its removal in UE5.7. That also opened the door for implementing the new “advanced weight-blending” technique in UE5.7, which runs on the GPU and only as part of the batched merge algorithm. It solves the most nagging features of weight-blended layers when they’re being used on multiple edit layers, as described in the post above.
That being said, the removal of the “weight-time paint balancing” system was something else entirely and it was done to solve other recurring issues that users had when using weight-blended target layers on edit layers landscapes (with more than one edit layer). Once again, I apologize for the sudden removal, it was meant to be replaced by a better system that supports both weight-blended and advanced weight-blended layers and is user-togglable, which will be added to the engine in the coming days (available in UE5.8). We simply failed to meet the UE5.7 deadline and hoped that it would not feel too disruptive (based on the arguments in the post above) in the meantime, although we were probably wrong here.
As for UE5.7 users, we will submit a backout of that removal in early December, targeting UE5.7.2 (it’s ready but we cannot yet submit on this branch, because UE5.7.1 is being finalized ATM), so that users can find back the same behavior for weight-blended target layers than they had in UE5.6.
Hopefully you’ll be able to integrate either the new UE5.8 tool or the UE5.7.2 backout when it’s submitted.
Cheers,
Jonathan