Skip to content

Performance of phasicFlow

Jump to bottom
GitHub Actions edited this page Apr 29, 2025 · 19 revisions

Benchmark

The benchmark for performance test is the simulation of a drum rotating at 12 rpm. Since particles are moving in this drum and some part of the drum is empty, so this benchmark can resemble most of the typical granular flows. The number of particles varied between 100K and 11M. Around 30-45 vol% of the drum was full of particles in all simulations. The time step for integration was 0.00001 s, so each second of simulation took 100K iteration to complete. In all evaluations, we excluded the time for writing the results to the disk.

Results

Scalability with respect to number of particles (CPU and GPU)

The execution time for one second of simulation was measured for various number of particles in the simulation. Both CPU and GPU builds showed almost a linear behavior with respect to number of particles, which is perfect for DEM simulation.

execution time vs. number of particles

Scalability with respect to number of CPU cores

For three different problem sizes, we tested the performance of PhasicFlow with respect to number of CPU cores. A linear relation was found in between the execution time and number of cores in the log-log plot.

execution time vs. number of cores

Execution time vs problem size

It is important for us to know at which problem size we are using the highest computational capacity of the system. The figure bellow shows the execution time per particle (for one second of simulation). For small size problems, this value is high, since each kernel lunch (entering parallel zone) has some overhead and the amount of computations in the parallel zone with respect to this overhead is low for small size problems. But for larger problems, these overheads become more negligible with respect to amount of computations, and we see almost constant execution time per particles for larger problems.

execution time per particle

Content from benchmarks/rotatingDrum/readme.md

Rotating Drum Benchmark (phasicFlow v-1.0)

Overview

This benchmark compares the performance of phasicFlow with a well-stablished commercial DEM software for simulating a rotating drum with varying particle counts (250k to 8M particles). The benchmark measures both computational efficiency and memory usage across different hardware configurations.

Simulation Setup

Figure 1. Commercial DEM simulation snapshot

Figure 2. phasicFlow simulation snapshot

Hardware Specifications

Table 1. Hardware specifications used for benchmarking.
System CPU GPU Operating System
Laptop Intel i9-13900HX 2.2 GHz NVIDIA GeForce RTX 4050Ti 6G Windows 11 24H2
Workstation Intel Xeon 4210 2.2 GHz NVIDIA RTX A4000 16G Ubuntu 22.04

Simulation Parameters

Table 2. Parameters for rotating drum simulations.
Case Particle Diameter Particle Count Drum Length Drum Radius
250k 6 mm 250,000 0.8 m 0.2 m
500k 5 mm 500,000 0.8 m 0.2 m
1M 4 mm 1,000,000 0.8 m 0.2 m
2M 3 mm 2,000,000 1.2 m 0.2 m
4M 3 mm 4,000,000 1.6 m 0.2 m
8M 2 mm 8,000,000 1.6 m 0.2 m

The time step for all simulations was set to 1.0e-5 seconds and the simulation ran for 4 seconds.

Performance Comparison

Execution Time

Table 3. Total calculation time (minutes) for different configurations.
Software 250k 500k 1M 2M 4M 8M
phasicFlow-4050Ti 54 min 111 min 216 min 432 min - -
Commercial DEM-4050Ti 68 min 136 min 275 min 570 min - -
phasicFlow-A4000 38 min 73 min 146 min 293 min 589 min 1188 min

The execution time scales linearly with particle count. phasicFlow demonstrates approximately:

  • 20% faster calculation than the well-established commercial DEM software on the same hardware
  • 30% performance improvement when using the NVIDIA RTX A4000 compared to the RTX 4050Ti

Figure 3. Calculation time comparison between phasicFlow and the well-established commercial DEM software.

Memory Usage

Table 4. Memory consumption for different configurations.
Software 250k 500k 1M 2M 4M 8M
phasicFlow-4050Ti 252 MB 412 MB 710 MB 1292 MB - -
Commercial DEM-4050Ti 485 MB 897 MB 1525 MB 2724 MB - -
phasicFlow-A4000 344 MB 480 MB 802 MB 1386 MB 2590 MB 4966 MB

Memory efficiency comparison:

  • phasicFlow uses approximately 0.7 GB of memory per million particles
  • Commercial DEM software uses approximately 1.2 GB of memory per million particles
  • phasicFlow shows ~42% lower memory consumption compared to the commercial alternative
  • The memory usage scales linearly with particle count in both software packages. But due to memory limitations on GPUs, it is possible to run larger simulation on GPUs with phasicFlow.

Run Your Own Benchmarks

The simulation case setup files are available in this folder for users interested in performing similar benchmarks on their own hardware. These files can be used to reproduce the tests and compare performance across different systems.

Content from benchmarks/rotatingDrum/readme.md

Rotating Drum Benchmark (phasicFlow v-1.0)

Overview

This benchmark compares the performance of phasicFlow with a well-stablished commercial DEM software for simulating a rotating drum with varying particle counts (250k to 8M particles). The benchmark measures both computational efficiency and memory usage across different hardware configurations.

Simulation Setup

Figure 1. Commercial DEM simulation snapshot

Figure 2. phasicFlow simulation snapshot

Hardware Specifications

Table 1. Hardware specifications used for benchmarking.
System CPU GPU Operating System
Laptop Intel i9-13900HX 2.2 GHz NVIDIA GeForce RTX 4050Ti 6G Windows 11 24H2
Workstation Intel Xeon 4210 2.2 GHz NVIDIA RTX A4000 16G Ubuntu 22.04

Simulation Parameters

Table 2. Parameters for rotating drum simulations.
Case Particle Diameter Particle Count Drum Length Drum Radius
250k 6 mm 250,000 0.8 m 0.2 m
500k 5 mm 500,000 0.8 m 0.2 m
1M 4 mm 1,000,000 0.8 m 0.2 m
2M 3 mm 2,000,000 1.2 m 0.2 m
4M 3 mm 4,000,000 1.6 m 0.2 m
8M 2 mm 8,000,000 1.6 m 0.2 m

The time step for all simulations was set to 1.0e-5 seconds and the simulation ran for 4 seconds.

Performance Comparison

Execution Time

Table 3. Total calculation time (minutes) for different configurations.
Software 250k 500k 1M 2M 4M 8M
phasicFlow-4050Ti 54 min 111 min 216 min 432 min - -
Commercial DEM-4050Ti 68 min 136 min 275 min 570 min - -
phasicFlow-A4000 38 min 73 min 146 min 293 min 589 min 1188 min

The execution time scales linearly with particle count. phasicFlow demonstrates approximately:

  • 20% faster calculation than the well-established commercial DEM software on the same hardware
  • 30% performance improvement when using the NVIDIA RTX A4000 compared to the RTX 4050Ti

Figure 3. Calculation time comparison between phasicFlow and the well-established commercial DEM software.

Memory Usage

Table 4. Memory consumption for different configurations.
Software 250k 500k 1M 2M 4M 8M
phasicFlow-4050Ti 252 MB 412 MB 710 MB 1292 MB - -
Commercial DEM-4050Ti 485 MB 897 MB 1525 MB 2724 MB - -
phasicFlow-A4000 344 MB 480 MB 802 MB 1386 MB 2590 MB 4966 MB

Memory efficiency comparison:

  • phasicFlow uses approximately 0.7 GB of memory per million particles
  • Commercial DEM software uses approximately 1.2 GB of memory per million particles
  • phasicFlow shows ~42% lower memory consumption compared to the commercial alternative
  • The memory usage scales linearly with particle count in both software packages. But due to memory limitations on GPUs, it is possible to run larger simulation on GPUs with phasicFlow.

Run Your Own Benchmarks

The simulation case setup files are available in this folder for users interested in performing similar benchmarks on their own hardware. These files can be used to reproduce the tests and compare performance across different systems.

Home

Clone this wiki locally