I’m delighted to announce the launch of DoriFlow, a new addon aimed at elevating the fluid simulation experience within Blender. As a CFD engineer and Blender enthusiast, I’ve been observing the gradual improvements in Blender’s fluid simulation tools, and now, with DoriFlow, we’re introducing a fresh perspective.
DoriFlow is a GPU-accelerated fluid solver designed to utilises Blender’s ray tracing and 3D capabilities, with scalable scenes, and fluid-structure interaction (FSI). Our emphasis on realism extends to incorporating whitewater effects, ensuring simulations feel immersive and authentic.
Currently, we’re in the beta testing phase and are seeking feedback and bug reports from experienced users of fluid simulation in 3D graphics software. Your insights are invaluable as we refine DoriFlow to better meet the needs of the Blender community.
If you’re interested in participating in our beta test, we invite you to fill out the questionnaire linked below. We’re eager to share the beta version of the addon with you and hear your thoughts.
Looking forward to get all the feedbacks from you.
Thank you for your participation. We are nearing our target number of users and will soon distribute the beta test version. In the meantime, please visit our website at www.doriflow.com to learn more about our story and discover helpful guides on using the software. We look forward to engaging with you in the coming days!
Can you give any info about system requirements?
I assume the gpu-computations are likely CUDA-only. But some tangible information on such matters would be helpful (including does it fall back to CPU in absence of a supported GPU, etc.).
Hi Kologe,
You are correct. Currently, the solver can run only on CUDA-based GPU, as there is algorithm in the hashing procedure from the solver requires CUDA to run specifically.
Although it’s technically feasible for the solver to run on CPUs, our beta testing phase requires GPU usage. If a GPU isn’t detected, the system will correctly trigger an error message.
The answer to your 2nd question is no, it will not fall back to CPU automatically on no detection of GPU. As my aim was to design this solver for a real physical system, meaning it solves the full set of fluid governing equations (which can be simplified down a lot to reduce load when focusing on visualization only), CPU is rarely comparable to GPU for computing particles in my experience.
However, if we get many feedback demanding CPU capability, we can impose the choice of CPU or GPU when computing fluid.
