Cool VL Viewer forum

At last, and two weeks late on my initial schedule, due to a silly copy/paste bug that took me that much time to find (*), here is the first release of the new Cool VL Viewer v1.31.0 experimental branch, with a dual renderer: the usual Extended Environment renderer (with both ALM and forward rendering) and new Physically Based Rendering one. You can switch between the two on the fly, with a simple click in a check box (however, when in a place with PBR materials around, do start with PBR on, so that reflection probes loading won't be skipped) !

Of course, and as usual for an experimental branch, this viewer is not for everyday's use: it will likely find creative ways to break your SL experience (even though the EE renderer should not differ the least from the release branch), and maybe even with innovative crashes !

With this viewer you should be able, not only to render, but also to manage PBR inventory materials and build stuff using them. BEWARE however: the editing code has so far not been sufficiently tested to be deemed reliable enough, and LL clearly warned that anything using PBR materials in the few regions supporting it on the main grid should not be used out of those regions, and never put up for sale. I personally recommend to only experience with material uploads and building in Aditi (Rumpus room regions): at least you will not loose any L$ there, should anything break.

As for the bugs, in excess of the ones you will encounter in LL's current PBR material RC viewer, there are issues I still need to fix in my own backport, which itself was synced against LL's code base as it was 3 weeks ago or so (e.g. recent double-sided material fixes have not yet been backported) EDIT: done.
In particular, you will find out that the mesh upload preview is hosed for rigged meshes (they should yet still upload fine: it's just the preview avatar which is totally distorted). But you may simply switch (yes, in real time too) to the EE renderer to see the preview avatar rendered properly.
There are also times when the PBR materials refuse to load (or to reload after a relog): it is not as bad as in LL's current RC release (v7.0.0.581126 is horrible in this respect), but will still happen. The only "fix" is to clear the object cache (from the Advanced -> Caches menu) and relog. EDIT: fixed now.

I also have a few important things on my "to do" list, such as optionally allowing reflection probes to load and update even while in EE rendering mode (would avoid missing reflections when switching from EE to PBR mode), adding SMAA (and perhaps CAS) support (because FXAA is so lame and causes so many blurry textures that I cannot stand or even tolerate it), or adding more options for the color space selection. EDIT: all done now.

As for performances, well, judge by yourself... From my own testing, the Cool VL Viewer is still the world champion in EE mode (ALM or forward rendering alike), and equals LL's RC viewer for PBR in PBR-heavy regions (while still much faster than LL's elsewhere); I still have a few speed optimizations to do (in excess of the usual on-the-fly optimizing I already did while backporting LL's code), but I do not expect spectacular increases in fps rates at this point.
Yet, and as I anticipated months ago, PBR rendering does badly impact "weak" hardware, on which EE forward rendering provides +50% to +100% in fps rates, and does not fill up the VRAM to the point of degrading the textures LOD to fit all in.

Feel free to experiment and report here your findings (for bugs, please use the bug reporting forum, after duly checking the bug is mine and not LL's fault, by comparing with LL's RC viewer).

---

(*) LLShaderMgr::SPECULAR_COLOR used in place of LLShaderMgr::SPECULAR_MAP somewhere in lldrawpoolmaterials.cpp, a file that I did yet already verify at least 4 times, before I could spot the bug on the 5th full re-check; I really wish I still had the sharp eyes I had only 20 years ago... This was probably the most time-consuming and stupid bug I had to hunt so far in my whole life !!!