How to use my gpu to work?

[dipta005]

I am running a problem where I have to work with 500u500u100u, the more resolution the better. But when running it is using almost 100% of my cpu, while my gpu sits idle. For this extensive work, how to allocate all my gpu and cpu?
-Thanks in advance,

[ProjectPhysX]

Hi @dipta005,

the (fastest available) GPU is selected automatically by default. When you start FluidX3D, a console window will open, displaying first a list of available devices and then the device that is selected.

• Please check that here your GPU shows up.
• Then the simulation initializes, which for a few seconds loads the CPU on all cores.
• Finally the simulation starts. If you're using the INTERACTIVE_GRAPHICS extension, a graphics window will also open in fullscreen, and after initialization the simulation is at first paused - to allow inspecting initial conditions visually. To unpause/start, press the key P.
• A 500x500x100 box will need only 2.2GB (FP32 memory storage) or 1.3GB (FP16S/C storage), and should be no problem even on very old toaster GPUs.

Kind regards,
Moritz

[dipta005]

I have NVDIA 1660s. I know that it runs. But I want to use more of gpus. If there’s any way to modify that. My task is to increase the grid resolution, the higher,the better.

[ProjectPhysX]

@dipta005 the GPU should be loaded close to 100% even at that smaller resolution. Check GPU load with nvidia-smi; Windows Task-Manager lies.

[dipta005]

| Device ID    0 | AMD Ryzen 7 4800H with Radeon Graphics                     |
|----------------'------------------------------------------------------------|
|----------------.------------------------------------------------------------|
| Device ID      | 0                                                          |
| Device Name    | AMD Ryzen 7 4800H with Radeon Graphics                     |
| Device Vendor  | Intel(R) Corporation                                       |
| Device Driver  | 2024.18.6.0.02_160000 (Linux)                              |
| OpenCL Version | OpenCL C 3.0                                               |
| Compute Units  | 16 at 0 MHz (8 cores, 0.000 TFLOPs/s)                      |
| Memory, Cache  | 23386 MB, 512 KB global / 32 KB local                      |
| Buffer Limits  | 23386 MB global, 128 KB constant                           |
|----------------'------------------------------------------------------------|
| Info: OpenCL C code successfully compiled.                                  |
| Info: Allocating memory. This may take a few seconds.                       |
| Info: Loading                                                               |
|       "/home/dipta/Desktop/FluidX3D-master/stl/KCS_hull_SVA_0.001.stl" with |
|       20463 triangles.                                                      |
|-----------------.-----------------------------------------------------------|
| Grid Resolution |                                200 x 500 x 100 = 10000000 |
| Grid Domains    |                                             1 x 1 x 1 = 1 |
| LBM Type        |                                    D3Q27 SRT (FP32/FP16S) |
| Memory Usage    |                                CPU 314 MB, GPU 1x 1045 MB |
| Max Alloc Size  |                                                    514 MB |
| Time Steps      |                                                         0 |
| Kin. Viscosity  |                                                0.00700000 |
| Relaxation Time |                                                0.52100004 |
| Reynolds Number |                                                Re < 45175 |
| Volume Force    |                0.00000000,     0.10000001,    -0.00050000 |
| Surface Tension |                                                0.00000000 |
|---------.-------'-----.-----------.-------------------.---------------------|

[ProjectPhysX]

Oh I now see you have the simulation indeed running on the CPU. Your Nvidia GTX 1660 is not showing up as OpenCL device. Please install Nvidia's Linux driver:

sudo apt update && sudo apt upgrade -y
sudo apt install -y g++ git make ocl-icd-libopencl1 ocl-icd-opencl-dev nvidia-driver-550
sudo shutdown -r now