Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Allow Multiple InputNode type for CUDA Kernels #120

Open
Tracked by #106
khosravipasha opened this issue Mar 17, 2022 · 0 comments
Open
Tracked by #106

Allow Multiple InputNode type for CUDA Kernels #120

khosravipasha opened this issue Mar 17, 2022 · 0 comments
Assignees

Comments

@khosravipasha
Copy link
Contributor

khosravipasha commented Mar 17, 2022

Having multiple distribution type for inputs fails gpu kernels due to Union split limiation. Currently, can have at most 3 types in the Union of a CuArray for cuda kernels. JuliaGPU/CUDA.jl#1385.

Another example here

For nodes, we already have Sum and Mul node types, so we can at most have one InputNode type for our cuda kernels.

Possible Workarounds:

  • Merge all input node types into one
  • Merge Sum and Mul nodes into one, (then we can have two input types at most)
@khosravipasha khosravipasha changed the title Having multiple distribution type for inputs fails gpu kernels due to Union split limiation Allow Multiple InputNode type for CUDA Kernels Mar 17, 2022
@khosravipasha khosravipasha self-assigned this Mar 17, 2022
# for free to join this conversation on GitHub. Already have an account? # to comment
Projects
None yet
Development

No branches or pull requests

1 participant