Jump to content
The Dark Mod Forums

cabalistic

Development Role
  • Posts

    1545
  • Joined

  • Last visited

  • Days Won

    47

cabalistic last won the day on November 4 2021

cabalistic had the most liked content!

Reputation

1018 Deity

2 Followers

Profile Information

  • Gender
    Male
  • Location
    Germany, Munich

Recent Profile Visitors

3059 profile views
  1. I haven't really heard that laptops with discrete GPUs have a significantly higher failure rate. They do, however, affect battery life, weight and consequently mobility of the laptop. The coexistence of integrated and discrete GPUs can also be a painful on occasion, with apps sometimes launching on the wrong GPU and you having to figure out which port is connected to which GPU. If mobility is not your highest priority, I think a discrete GPU in a laptop is generally fine. Otherwise, the Ryzen 6000 chips definitely have the strongest integrated GPU right now (not counting Apple's M1), but they are also still rare to find in an actual laptop.
  2. I must confess, I'm a bit skeptical. While the 12700K is stronger in single-core performance than the 5800X, it shouldn't be that much stronger. There might be additional factors at play
  3. The problem with UI, I'm afraid, is first and foremost a technical issue. The existing UI system is an absolute pain to work with, and any effort to implement any reasonable modern UI concept in that system will drive every developer involved in the effort absolutely insane. I'm only half joking. Before any such effort could take place, I feel that the system needs to be replaced entirely. But that is also a massive pain because there's no good option to keep backwards-compatibility with existing missions. And the most obvious candidates for UI systems are rather large dependencies. So while I agree with the necessity of a UI overhaul, I'm not very optimistic for it to happen anytime soon.
  4. No, I do have a device that shows the issue, that's why I strongly suspect it's a GLSL compiler bug.
  5. No, it's definitely broken. The issue we have with bindless textures, for instance, is almost certainly a GLSL compiler bug. And it's not just NVIDIA that handles all of these things without issue, it's Intel, too. And Intel's driver isn't exactly stellar, either
  6. Given that TDM is based on the older non-BFG code base and has then been heavily modified, I would guess very hard. It's good for inspiration, no doubt, but a direct copy is out of the question.
  7. Maybe I need to start working on a Vulkan backend for real (Mostly kidding. There are a couple of improvements that should ideally be done before going to Vulkan, and I don't really have time, anyway. It is tempting, though...)
  8. At this point, just disable the new backend. It's not worth debugging it any further.
  9. If you can give me a particular map and coordinates (and any particular reproduction steps I may need beyond 64bits and Bloom), I can try and see if I can reproduce it on my Aya Neo.
  10. Please try the 2.10 beta. It has a different window and renderer creation path and might give better results.
  11. No clue, but to be honest, since neither of us can reproduce the issue, we'll just have to have both hypotheses tested. I imagine having a shader fix tested is slightly easier, so I would start with that, but if it's not that, then do try yours
  12. It's possible, but it sounds unlikely to me. I would expect any sensible driver to not actually sample the existing framebuffer if its blendweight is zero. Could be wrong, of course, wouldn't be the first time My personal hypothesis is that we may have negative values in the input color framebuffer. This has actually happened in another place where I already put a workaround for the resulting issues in place, so this isn't too far-fetched, even though I don't fully understand the cause. Now, if there is a negative value in the input to bloom, it would probably get smeared across the bloom stages and then added to the final image, which, if the negative value is high enough, would explain why it ends up as black. This is potentially simpler to test and workaround; all we'd need is a check in the bloom_downsample.frag.glsl brightpass function that returns 0 if the brightness is negative.
  13. Wrong shader, look at stages/frob/frob_apply.frag.glsl. Though honestly, returning to fresnel feels like we are moving in circles... If that had worked well and not be ugly as fuck, we wouldn't be where we are right now
  14. Well, obviously. The outline was always meant as an alternative, not an addition, to the existing highlight.
  15. Because the outline *is* the highlight in that case? I don't know what you are saying here.
×
×
  • Create New...