Replies: 2 comments 1 reply
-
|
Hi @het1803, I'm afraid we're not participating in GSoC this year. |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Of course! We did a lot of work on ISPC last year in #710 but the reason it hasn't been merged is because it ends up copy-pasting most of the single threaded CPU BACKEND resulting in a maintainance burden. The solution is probably to split the single threaded CPU backend into a BackendCPUBase class and then derive ISPC and the single threaded CPU off that (similar to the relationship between backend SIMT and CUDAHIP) |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Hi GeNN Team,
I am Het Patel, a 1st year EEE student from BITS Pilani, Goa Campus. I want to work on the ISPC Backend project for GSoC 2026.
I have already:
1)Set up GeNN on WSL and ran the Potjans-Diesmann model.
2)Written a manual "Hello World" ISPC kernel and linked it to a C++ driver to verify parallel execution.
3)Explored the source code and found where the loops are generated, specifically in backendSIMT.cc and initGroupMerged.cc.
I can see how groupEnv.getStream() is used to print C++ loops, and I want to learn how to adapt this to generate ISPC foreach blocks instead. I am very motivated to help make GeNN faster for CPU users.
Looking forward to hearing from you!
Best, Het Patel
Beta Was this translation helpful? Give feedback.
All reactions