#FluidX3D#CFD v2.17 is out! Some huge #GPU/#CPU hardware has been announced at #Computex, so I've made my code ready. Until now I've been using 32-bit indexing, which overflows for >2³² grid cells in a domain, equivalent to 225 GB VRAM. Now my #OpenCL code will at runtime automatically compile with 64-bit indexing when more cells are used. 🖖🧐
Also, I've added a new raytracing-based field visualization. Thank you @python for the idea! 💡
Release notes 👉 https://github.com/ProjectPhysX/FluidX3D/releases/tag/v2.17
#FluidX3D#CFD update alert! v2.15 speeds up framerate in interactive graphics by 20-70%. 🖖🥳💻
How? Turns out iterating over 2 million pixels with a single CPU core is... really slow. I did that 3 times more than necessary for every frame rendered on screen! 🖖😆
I've now eliminated a memory copy of the frame (in favor of pointer swap), and a clear frame/zbuffer operation on CPU since that's already done on #GPU.
One of my #PhD papers got selected for the 2022 Best Paper Award of MDPI Computation! 🖖🥳📃🏆
That was a very bold publication for multiple reasons:
I solo-authored it
I wrote that paper in only 2 weeks
the title contains "Esoteric" twice
I submitted it on April 1st
It's serious science though: I discovered a simple algorithm to cut memory demand of the #LBM in half, allowing huge simulations on cheap #GPUs. This is one of the key innovations in #FluidX3D#CFD.
How realistic can a #CFD simulation be? Here is a 1 billion cell #FluidX3D simulation of an impacting raindrop, fully raytraced in 8K. FluidX3D contains state-of-the-art volume-of-fluid and surface tension models for highly accurate free surface simulations. Combined with my own #OpenCL#raytracing engine, results are rendered on-the-fly at resolution as large as remaining #GPU VRAM can hold. 🖖😋💧📺 https://youtu.be/MmLNQIW_Sic
FluidX3D is on #GitHub: https://github.com/ProjectPhysX/FluidX3D
This is wild: #FluidX3D can "SLI" together 🔵 #Intel Arc A770 + 🟢 #Nvidia Titan Xp, pooling 12GB+12GB of their VRAM for one large 450M cell #CFD simulation. Top half on A770, bottom half on Titan Xp. They seamlessly communicate over PCIe. Performance is ~1.7x of what either #GPU could do on its own. 🖖😋🖥🔥 #OpenCL shows its true power here - one implementation works on literally all GPUs at full performance, even at the same time. Happy #SimulationFriday! https://youtu.be/PscbxGVs52o
So apparently there is almost no free software that does meshing for #CFD with good boundary layer support... I saw #engrid recommended a lot, which hasn't really been maintained since 2014 and was written for QT4. I managed to compile it! Let's see if I can make it work.
Oh wow this is weird. my #CFD list here on Mastodon suddenly has only one message. I checked the profile of some users I'm following that belong to the list (such as @ProjectPhysX ) and their messages aren't visible on their profile either —I had to load the most recent ones manually (copying the link from the original page), even though I could see them a few days ago. Did I miss some federation accident on this corner of the #Fediverse?
(cc @FrankSonntag)
@FrankSonntag I see some new messages, but most of the old ones are not there. Note that this isn't about the tag, though, but a List of mine. Maybe old messages get naturally expunged from lists after a certain amount of time? I don't know enough about Mastodon to know if this is expected behavior or not.
#FluidX3D has passed 2000 Stars! It is the most popular #CFD software on #GitHub now! 🖖😊⭐️ https://github.com/ProjectPhysX/FluidX3D
Feeling blessed that my work is useful to so many people across the globe, with users in 75 countries already! 🌍
42% EU, 30% Americas, 25% Asia, 3% Oceania+Africa
It's #SimulationFriday again! Here is a real time @FluidX3D#CFD simulation of a #Cessna 172, 20M cells, on my old Titan Xp Pascal #GPU. 33 lines of code for this setup.
For #FluidX3D v2.8 I'm improving composite geometries with moving parts, and I'm adding transparent rendering. https://youtu.be/Gd_MEHPecw0
@HalvarFlake
Here's an article that describes problems that Volkswagen had to implement adjoint methods. Vw (and aerospace industry before them) has the kind of R&D budget to work around such issues, for most companies its not worth the hassle.
Standard CFD packages got better support in the last years, but even then it forces you tot use a rigid toolchain. Geometry, meshing, solver, post, it all had tot work together in detail.