SimWorks Finite Difference Solutions

Last release：v2.0.0 Release
Release Date: 2025-05-23

• FDTDSolver
  • Optimize data exchange for single-machine multi-NVIDIA GPU setups to improve multi-GPU parallel speed.
  • Fix the issue of abnormal magnitude in normalized field values output by the FDFD solver.
  • Fix the malfunction of the TFSF source when the boundary coincides with the structure boundary it crosses.
  • Fix the malfunction of the TFSF source in nonuniform mesh.
  • Fix the issue where the TFSF source crossed dispersive materials abnormally when using volume averaging or conformal variant VP-EP 1.
  • Fix a stability issue in GPU-accelerated simulations of the FDTD solver under the combination of dispersive materials and conformal variant V-EP.
• FDESolver
  • Add FDE accuracy selection in Resources, supporting single precision mode solving simulation.
  • Fix the issue of abnormal mode fields when using PML boundaries for multi-frequency mode solving.
• Mesh
  • Add snap pec to yee cell boundary feature to force PEC structures to align with simulation mesh boundaries.
  • Optimize mesh growth speed for 2D structures.
  • Fix the issue where refractive index creation for polygon structures failed in special cases.
  • Fix refractive index calculation errors at some conformal mesh points in ring structures.
  • Fix the issue of abnormal refractive index at certain conformal mesh points containing PEC materials under VEP.
  • Fix the issue of abnormal refractive index caused by mesh level errors when 2D RLC structures overlap PEC structures.
  • Fix the issue of incorrect refractive index assignment at boundary conformal mesh points in Sidewall bezier structures.
• Source
  • Add Resource Test feature.
  • Optimize preprocessing speed of multi-frequency light sources.
  • Fix the issue where the number of resource setting threads did not match the actual number of threads used in simulation.
• Scripts
  • Optimize the farfieldeql and farfieldeql2d functions to improve far-field analysis speed.
  • Optimize the speed of complex matrix dot product operations.
  • Fix the issue where the rfile command did not work properly.
  • Fix the issue of excessive memory demand in the grating series functions causing software crashes.
• UI
  • Optimize the instance resource table generation method, with the workbench customizing the instance resource table attributes.
  • Fix the issue where some properties in the frequency domain, time domain, and plotting were displayed inaccurately when the light source wave type was set to continuous wave.
  • Fix the issue where running scripts after simulation completion in analysis groups containing refractive index monitors caused the result data to be cleared.
• Sweep
  • Fix the issue of memory leaks during sweep execution that caused software and operating system crashes.

Learn more>>

SimWork Cloud Computing

Last release：v2.0.0 Release
Release Date: 2025-05-23

• Add software startup page.
  • Add license restrictions on software versions.
  • Add network proxy configuration tool.
  • Add 7-day automatic login feature: upgraded based on single sign-on functionality, users do not need to re-authenticate within 7 days after logging into the software.
  • Fix the issue where the software could no longer initiate cloud computing after being open for an extended period.

Learn more>>

Key Highlights

1. Optimize data exchange for single-machine multi-NVIDIA GPU setups to improve the multi-card parallel speed of the FDTD solver.

The software significantly optimizes the parallel efficiency of multiple GPUs on a single machine. Version v2.0 optimizes the communication mechanism between multiple GPUs, realizes the overlapping computation with data communication, and significantly improves the multi-card simulation efficiency.

After running simulations, the results demonstrate that the loss rate of Multi-Card GPUs in version v2.0 has been reduced by half compared to previous version, $loss\;rate= (speed\;of\;multi-GPUs - single\;GPU * number\;of\;GPUs)/(single\;GPU * number\;of\;GPUs).$ As the scale of the grid expands, the multi-card efficiency improvement in v2.0 becomes more obvious. With an 8-card configuration, the speed is increased by more than 50% in v2.0, greatly releasing the potential of the hardware.

2. Add snap pec to yee cell boundary

The Snap pec to Yee cell boundary option mandatorily requires all PEC structures to align with Yee cell boundary, ensuring that the electric field components at PEC interfaces maintain a tangential distribution. Enabled by default, this option should remain selected in antenna engineering applications to enhance the accuracy of radiation efficiency calculations.And this feature is equally essential for mode-solving at PEC boundaries. During simulation mesh generation, the software automatically calibrates specified PEC boundaries to achieve precise alignment, permitting a maximum offset of dx/2 (where dx denotes the Yee cell dimension).

As shown in the figure: In the horizontal direction, if the PEC structure extends less than dx/2 beyond the grid lines, the index distribution stays unchanged (PEC aligns with inner grid lines). In the vertical direction, if the PEC structure extends dx/2 beyond the grid lines, the index distribution expands outward to align the PEC with outer grid lines.