interesting, to me it eems "buddies" is the wrong term anyhow...
HN user
holg
Senior Developer and System Architect
Forgot to mention: it is all Open Source and on crates.io: github.com/holg (acadlisp, bimifc, gldf-rs)
As the title as well mentioned IFC, here how to create the Quonset Hut on bimifc.de (same engine as acadlisp.de)
(ifc-new) (ifc-set-project "Quonset Hut" "Airfield" "Hangar" "GF" "Engineer" "Corps") (ifc-storey "Ground Floor" 0.0) (ifc-add-slab 0 0 8 15 0.2 0 "Concrete Pad") (ifc-add-wall 0 0 8 0.15 4.0 0.2 "Front Wall") (ifc-add-wall 0 14.85 8 0.15 4.0 0.2 "Rear Wall") (ifc-add-door 3 0 2.0 0.15 3.0 0.2 "Hangar Door") (ifc-add-window 1 14.85 1.5 0.15 1.5 0.2 "Rear Window") (ifc-storey "Roof" 4.0) (ifc-add-roof-barrel 0 0 8 15 3.0 4.0 "Steel Arch")
Hi HN, author here. This builds on the very same AutoLISP Interpreter, i did for: https://news.ycombinator.com/item?id=46393271
The Lisp can be used to define the GLDF product in the LISP REPL, some files are in the wasm bundle, so for the demo you can simply: Click the [λ Script (Lisp)] Button and it will create the GLDF as defined in the LISP, and also start some Animation, tilting the light fixtures, changing Variants (and as such housing and light color) and some camera movement
Love it, bcs i am infested, i would love to see it in Rust (even embedded bcs of even more RTOS), but hey thanks for sharing, love it to repeat myself :)
it is the fantastic tool!
That's the neat idea!
Thanks, well i got more feedback, so i figured out it shall be done even better...
Thanks for your interest :D
i agree, sorry this is some edge-tech, so for safari it is a bit tricky, as it reports to have webgpu, but &force=webgl2 will do the trick. I am working on the fix to have webgpu behave properly on Safari
Yup, so actually my take as well is, the scene itself is imperfect, i took some known and easy to implement one. So some scene, where some high building is illuminated up, would be desirable...
That's the real good proposal. I must admit, my focus was on the street lights, which in general are already almost good, compared to what you mention Thanks a lot
https://iesna.eu/?wasm=skyglow_demo&force=webgl2, tested on wsl2 ubuntu, sorry linux machine in the office :D
Ok, even so the quality degrades a lot, i added the webgl2 version, it shall load as fallback, if webgpu is not enabled, but can as well be enforced, to see the difference:
I should have mentioned: WebGPU is needed, on Safari there is a bug in the Bevy Overlay, so you only see flickering (very annoying!) This is an upstream issue (on to it) So FF and Chrome works fine (if WebGPU is enabled!)
Thanks for the information, i should have mentioned, webgpu is explicitly used here, and this might be the issue! Bevy's usage of WebGL2 is not that mature, and i use Bindless rendering (from my own contribution)...
i miss the Apple GPUs, specially bcs for higher VRAM they compute quite well, not even talking about power efficiency. For my use-cases very often the low VRAM is the limiting factor, where the Apple GPUs often shine, the Apple Tax if seeing this way is nowadays quite low :D
interesting proof of concept, in php, who would have thought :D
Sounds interesting, will give it a try, thx for sharing the Code as well
the nice article, thanks for sharing... Oh yes winit everywhere even not expected :D
Quite the good idea and concept, thanks for sharing
Quite the nice idea, i will check it out and i'd be happy to contribute
Great idea and effort, thanks for sharing. It is even way more strict than my native chinese tryarounds :D
This all sounds impressive, but tbh there are no real provable numbers to evaluate...
Nice idea, how about even more badges? Might be some challenge for the clean looking Certificate...
OK, interesting, i am not into betting, but it seems the solid solution, the ui is fine to me, i only speak a very limited french, but it is enough to understand the pages...
First of all, nice idea, thanks for sharing! I ask myself, how you would tackle lost context with your App? And i am very comfortable with my terminals, i find it convenient to have it in different tabs/windows, so where is the benefit?
quite impressive for the tool, which does not upload anything, but does it in the local browser...
Great work. Making the Hessian calculation linear in depth is a solid intermediate step. Thanks for sharing this; I look forward to seeing the final results as this research matures.
Rust-only deep learning is the right direction. Backend-agnostic, no Python required, autodiff built in, runs on embedded - that's a stack that actually deploys cleanly. Curious how training performance compares to PyTorch on real workloads.