Dear Roshan,

It depends on which version of PFC you are working, but I'm working with version 5.0, and the parallel computing is already implemented!

You have two ways to create functions in PFC, either using FISH or Python. The problem is that FISH is not multi-threaded in PFC 5.0 and the previous versions (it will be in the next version of PFC (6.0)).

So if the fish function is called 30000 times on your particles (for example 30000 particles), the code will run on a single thread and limits the calculation performance.So it depends how many time the FISH function is called.

In PFC5 there are 2 solutions to fix this problem:

- use Python with the Numpy module. This module allows performing operations on large arrays in a fast way.

- implement this function as a C++ FISH plugin, which would allow the multi-threaded

The most efficient computers today are the Intel Core I9 with a maximum number of processors(Depending on your budget). Then, I advise you a min 32Gb of RAM (it also depends on the number of particles used). For the GPU, a min 4GB of memory is often quite sufficient. Here too, the price is decisive because it can go up very quickly.

About doing the simulation using GPU cores, you can't; unless you implement it by yourself using another code. You can directly contact Itasca consulting who developed this software.

I hope it helps.

Good luck.

Regards,

Youssef

Youssef Fawaz , thank you very much for the clear answer. I am currently using only the FISH functions and some user defined contact models developed using C++. Now it makes sense why the simulations are bit slow. Thank you very much again for the answer :-)

Dear Roshan,

I Confirm response of Youssef Fawaz, with PFC version 4.0 and fish functions there are probably no parallel computing using PFC.

Best regards,

Tariq.

Here are the minimum requirements recommended by itasca

https://www.itasca.fr/software/pfc-system-requirements

Operating System: Windows 10 Pro (with the latest service packs installed)*

• Hard Disk: 2 TB, 7200 rpm or better
• Processor: Intel Core i7 CPU (3 GHz) or better
• Memory (RAM): 16 Gbytes or better (large models may require more memory)
• Video Card: HD graphics card, 1024 x 768 pixels, 32-bit color palette or better, OpenGL 3.1 or higher

Dear Tariq Ouahbi , thank you for the answer. However, I confirmed from Itasca that the PFC version 5.0 automatically supports multi threading and parallel processing. By default, programme selects the maximum number of processors available in the machine. We can also give the number of processors by "set processor" command.

Dear Abderrezak Bouziane, Thank you for the answer. The machine specs you have given are similar to the one I am using. However, I am struggling with slow processing speeds in my simulation (in PFC 3D with 30000 particles), I inquired about this from itasca, and they reccomended a machine with i9 processor and 64 GB RAM to have a considerable speed. However, the simulation time also can depend on the size of the critical time step and the time step calculation method used in the simulation.

Dear Roshan ,

The stiffness of the particles plays a major role in the time-step calculation.

The bigger the stiffness is, the smaller the time-step.

Having very small particles (small m) with high stiffness (k) results in a very small time-step.

Youssef Fawaz also damping plays a big role in time-step calculation.

Roshan Jayathilaka good luck for your simulations.

Roshan Jayathilaka Exactly, it depends on the sample size, the time step and especially on the contact parameters. The machine specs I have given are the minimum recommended by itasca.

Good luck