Stats

More Content

the parallel strategy of PointWise in high-order curved mesh generation

Studyhardhardhard

Hello, everyone. I have a question about the parallel strategy of PointWise in high-order curved mesh generation.

We all know that producing a high-order mesh can be a lengthy and computationally expensive process for larger grids. Based on my observations of PointWise in the Windows Task Manager, it seems that PointWise primarily runs on a single core when handling large-scale high-order curved mesh generation. However, for a very small fraction of the time, less than one-tenth of the total, PointWise appears to operate in multi-core parallel, as indicated by nearly 90%+ CPU usage.

I am particularly interested in this multi-core parallel phase. I have consulted the Grid, Elevate page in the Fidelity PointWise User Manual and related academic papers, learning that it consists of Elevation and WCN smoothing phases, with detailed explanations of the underlying principles provided in the papers.

However, I have not found any description of its parallel strategy. How does it allocate computational resources? Why does it run in parallel for such a brief period and in serial for the rest of the time? What is it doing during these brief periods of parallel operation? Where can I find more detailed information about this? I am genuinely interested in this, especially its short bursts of parallel activity.

Best Regards.

  • 0

    Hello, 

    Thank you very much for your question.

    Please note that Fidelity Pointwise has many algorithms used for different purposes. Some of these algorithms are performed serially while other are performed in a multi-threaded fashion. I would recommend to review this page of the User Manual for general details on this: Special Topics, Parallelism. 

    As for multi-threading and grid elevation, please note that with the 2022 release, we improved the performance of mesh curving by a large amount through parallel CPU threading. The computations for curved element quality improvement are nearly 100% parallel.  The element polynomial order elevation is performed serially, so if a mixed-order mesh is being created, there will be significant portions of the runtime which is serial.

    I hope this helps!

  • 0 in reply to Claudio M Pita

    Thank for your prompt reply.

    I'm glad to hear that in 2022 release the smoothing is running in parallel, which is helpful for users.

    I do read the User Manual for general details on Special Topics, Parallelism, but there is no more information about the high order curved mesh generation. Is there anywhere to learn more about the parallelism of higher-order curved mesh generation, or about its mixed-order mesh?  The User Manual may not be suitable for some academic detailing. Might there be an academic paper on it?

    By the way, after the high order curved mesh is generated, the Jacobian message is printed.But when I open the generated high order curved mesh with Gmsh and compute it using Gmsh’s mesh quality module, I find that the difference between Gmsh's minJ/maxJ and the Pointwise's Scaled Jacobian are not equal, please forgive my ignorance, I always thought Scaled Jacobian= minJ/maxJ, where did I misunderstand? How can I find out how the Jacobian, Scaled Jacobian, Normalized Jacobian is calculated after the high order curved mesh is generated and printed? I want to compare it with Gmsh's.

    Regards.

  • +1 in reply to Studyhardhardhard

    Hello, 

    Thank you very much for your additional question. 

    IMPORTANT: Since you have another ongoing thread dealing with the elevation process on your particular mesh and the Jacobian error you encountered, please add your last question on this post to that thread. This will help keep threads clean and easy to follow to other Fidelity Pointwise users. Thanks for your help with this!

    As for your first question, please refer to this material for additional information: https://link.springer.com/chapter/10.1007/978-3-030-92540-6_1

    I hope this helps! 

Cadence Guidelines

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.