search

john-b-schneider / uFDTD

Content for "Understanding the FDTD Method" by John B. Schneider
212 stars 68 forks source link

GPU acceleration chapter/section? #13

Open 0xCoto opened 2 years ago

0xCoto commented 2 years ago

Hi, thank you very much for composing this fantastic book covering the FDTD technique in amazing detail and releasing it freely!

Are there any plans to include a chapter or section on GPU implementations for FDTD acceleration? There's also been some relatively recent advancements with CPU improvements as well, providing massive acceleration (even faster than GPUs):

Despite the significant speedup this technology achieves though, I couldn't find much more information on its implementation.

Many thanks,

Apostolos

john-b-schneider commented 2 years ago

Hi Apostolos,

Thanks for the kind words! I’m glad to know you found the work of some value.

I largely shut down my research program about 10 years ago and moved to the administrative side of university. More recently, I’ve moved away from administration and returned to teaching, but so far at the basic undergraduate level and haven’t returned to playing around with FDTD. I may yet do that, but have no definite plans at this time. It would be fun to explore the use of GPUs or other technologies that have come along. Thanks for the pointers to the XPU work!

Best, John Schneider, @.**@.>, https://eecs.wsu.edu/~schneidj/ School of Electrical Engineering and Computer Science P.O. Box 2752, EME Bldg. 221 Washington State University Pullman, WA 99164-2752 (509) 335-6457

On Jun 18, 2022, at 9:55 PM, Apostolos @.**@.>> wrote:

[EXTERNAL EMAIL] DO NOT CLICK links or attachments unless you recognize the sender and know the content is safe.

Hi, thank you very much for composing this fantastic book covering the FDTD technique in amazing detail and releasing it freely!

Are there any plans to include a chapter or section on GPU implementations for FDTD acceleration? There's also been some relatively recent advancements with CPU improvements as well, providing massive acceleration (even faster than GPUs):

Despite the significant speedup this technology achieves though, I couldn't find much more information on its implementation.

Many thanks,

Apostolos

— Reply to this email directly, view it on GitHubhttps://urldefense.com/v3/__https://github.com/john-b-schneider/uFDTD/issues/13__;!!JmPEgBY0HMszNaDT!rmg1da5paD3a35zV5Mn9pg4EfNMiBed3CIUtF-nToA1Y9E3c5aQyXfVTb9OgTdtO_juAcmmgMMsmeV4JjSM8Fyke1au80w$, or unsubscribehttps://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AAZ2SF5RNNIXALSWSQQQVCTVP2R4PANCNFSM5ZFY3OKA__;!!JmPEgBY0HMszNaDT!rmg1da5paD3a35zV5Mn9pg4EfNMiBed3CIUtF-nToA1Y9E3c5aQyXfVTb9OgTdtO_juAcmmgMMsmeV4JjSM8FylhFOtAHQ$. You are receiving this because you are subscribed to this thread.Message ID: @.***>

matsievskiysv commented 2 years ago

@0xCoto I've repeated some examples (not many) with PETSc. It supports GPU. You may look in that direction.

0xCoto commented 2 years ago

@matsievskiysv Thanks! Are the examples for 1D or have you also written code for full-fledged 3D antenna simulation examples?

matsievskiysv commented 2 years ago

Unfortunatelly, no. I've got sidetracked at the refactor chapter, when I tried to do a memory leak check (which is tricky for MPI). I'm mostle learning FEM now using DolfinX, but FDTD is in my backlog:)