Skip to content

[mlir] Consider moving away from linking LLVM IR for host and device code in SPIR-V CPU Runner flow #115348

New issue

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

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

Already on GitHub? Sign in to your account

Open
andfau-amd opened this issue Nov 7, 2024 · 2 comments

Comments

@andfau-amd
Copy link
Contributor

See this discussion for more details: #114563 (comment)

One possible outcome is we simply keep things as they are, but there may be a better option. In particular, if "device" CPU compilation/execution could be offloaded in a similar way as for GPU devices, there would be a better alignment between the SPIR-V CPU Runner and e.g. the Vulkan Runner.

@llvmbot
Copy link
Member

llvmbot commented Nov 7, 2024

@llvm/issue-subscribers-mlir-spirv

Author: Andrea Faulds (andfau-amd)

See this discussion for more details: https://github.com//pull/114563#discussion_r1831742718

One possible outcome is we simply keep things as they are, but there may be a better option. In particular, if "device" CPU compilation/execution could be offloaded in a similar way as for GPU devices, there would be a better alignment between the SPIR-V CPU Runner and e.g. the Vulkan Runner.

@EugeneZelenko EugeneZelenko removed the mlir label Nov 7, 2024
@kuhar
Copy link
Member

kuhar commented Nov 8, 2024

@fabianmcg explained how this could be implemented by serializing LLVM bitcode: #114563 (comment)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

4 participants