@ProjectPhysX Good choice! We were just in the morning discussing with Jan and wondering what happened to the #fluidx3d patches you made in Spring to improve its multi-node scaling with PoCL-R. Any plans with that you can share?

Over the weekend I got to test #FluidX3D on the world's largest #HPC #GPU server, #GigaIOSuperNODE. Here is one of the largest #CFD simulations ever, the Concorde for 1 second at 300km/h landing speed. 40 *Billion* cells resolution. 33 hours runtime on 32 AMD Instinct #MI210 with a total 2TB VRAM.
https://youtu.be/clAqgNtySow

#LBM compute was 29 hours for 67k timesteps at 2976×8936×1489 (12.4mm)³ cells, plus 4h for rendering 5×600 4K frames.
🧵1/3

#FluidX3D v2.9 is out! The long startup time for very large (multi-)#GPU simulations annoyed me, so I made it a lot quicker! Geometry initialization and sanity checks are now multithreaded, with my own cross-plattform parallel_for implementation. Make the cores go brrrr! 🖖😎🖥️🔥

The Pioneers used to ride these for miles! 🖖🧐💺
Aerodynamics of an arm chair at 50km/h. 1 second real-time, 3 billion cells, 66k timesteps, 3 hours on 4x
@NVIDIADC
#A100 40GB #GPU​s. Happy #SimulationFriday
https://youtu.be/vD7Tj7H17jA
#FluidX3D #CFD: https://github.com/ProjectPhysX/FluidX3D

Here is the most detailed #CFD #simulation of a #quadcopter ever, at 3 Billion microscopic (0.2mm)³ grid cells. Compute time was 8 hours on 4 Nvidia #A100 40GB #GPU​s, for 86k timesteps + 5400 videoframes. Happy #SimulationFriday!
https://youtu.be/QvpNUdX8LNY

#FluidX3D on #GitHub: https://github.com/ProjectPhysX/FluidX3D

Kudos to Musa Maluegha for the awesome #CAD model: https://www.thingiverse.com/thing:2540161/files

#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

Proud to announce the release of #FluidX3D #CFD v2.8! This update finally adds documentation, and also loads of refactoring, bug fixes and significant usability improvements, especially for beginners. 🔎 📃

Full release notes 👉 https://lnkd.in/exHmwHbD
Have fun with the software!

PS: Thanks for ⭐2k #GitHub Stargazers across 74 countries! 🖖🤩🗺

I've always wanted to do a helicopter! 🖖😎🚁
Here it is, a Bell 222 in #CFD. With massive 10 billion cells. 71 TeraByte data visualized. Just for #SimulationFriday fun, because I can! Took 6.4 hours on 8x AMD Instinct #MI200 #GPU​s.
https://youtu.be/BStzTRmLW7Q
#FluidX3D on GitHub: https://github.com/ProjectPhysX/FluidX3D

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

#FluidX3D v2.7 is out! I added slice and surface pressure visualization modes, both much requested. This was 1 day of programming for me! 🖖😁
The video is a 25M cell #CFD #simulation in real time, featuring the new #datavisualization modes. Have fun! 🖖😎
https://youtu.be/uL8usTb0Czg
https://github.com/ProjectPhysX/FluidX3D/releases/tag/v2.7

#OpenRadioss と #FluidX3D に興味をもったものの，そこまで手を広げると何も進まないからしばらくは様子見のつもり．

5 years ago I had this wild idea to write my own #CFD software from scratch in #OpenCL. I wanted to know how fluid simulations work, and to make them ridiculously fast on any #GPU. Today, #FluidX3D has ⭐1.4k stars on #GitHub: https://github.com/ProjectPhysX/FluidX3D

how it started: how it's going:

In update #FluidX3D v2.6 I patched the remaining #OpenCL driver issues on Intel Arc #GPU​s. >4GB VRAM allocations now work and VRAM capacity reporting is corrected. FluidX3D is now fully operational on #IntelArc! 🖖🥳
Release: https://github.com/ProjectPhysX/FluidX3D/releases/tag/v2.6

Gave the @FluidX3D #GitHub page an update with some colerful markdown illustrations: https://github.com/ProjectPhysX/FluidX3D
Did you know? #FluidX3D on a cheap 24GB #GPU does the same 460 million cells resolution for which other #LBM #CFD codes need 2x >$11k #A100 80GB.

In the latest #FluidX3D v2.5 update I revised the #raytracing graphics and implemented light absorption by the fluid. 🖖😎🌊
Demo: https://youtu.be/XOfXHgP4jnQ
I also fixed a bug where moving objects during re-voxelization would leave a trail of solid grid cells behind. Full release notes: https://github.com/ProjectPhysX/FluidX3D/releases/tag/v2.5

Our new 8x #GPU servers at BiofluidUBT lab are up and running! Temperatures under load are just fine, despite the dense packing. And yes, "SLI" (via PCIe) between 7x #RTX 2080 Ti + 1x #A100 is totally a thing with #OpenCL! #FluidX3D benchmarks are here: https://github.com/ProjectPhysX/FluidX3D#multi-gpu-benchmarks

For now, the way I initially did #FluidX3D multi-GPU communication over PCIe to CPU memory & back is still the best (&only) option. The major #GPU vendors have yet to implement/fix P2P communication in their compilers to claim that performance advantage on their platform.⚠
🧵6/9

When running #FluidX3D with the #CUDA backend of #PoCL + P2P cudaMemcpy, performance is 40% faster compared to #OpenCL PCIe copy over CPU memory. PoCL's P2P backend is >3x faster than Nvidias own runtime here. This is the perf delta #Nvidia are giving up on.
🧵3/9

Update #FluidX3D v2.4 is out! This contains several UI improvements (thanks @fclc for the feedback!) like a help menu to show keyboard/mouse controls. Also: minor optimizations, refactoring, under-the-hood upgrades, bug fixes. 🖖🧐
Full release notes: https://github.com/ProjectPhysX/FluidX3D/releases/tag/v2.4

Was playing around with #fluidx3d from @ProjectPhysX.

Specifically looking at a model of the RedBull RB-18 #F1 car.

A feature that would have been great to see is a way to "snap" the lowest part of a mesh to a bounding box.

As of now it's tweaking coordinates by fractional offsets within the example suite (Using the Ferrari example) which is a PITA to put it mildly