Skip to content
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

Disable the vectorizer pipeline. #516

Merged
merged 1 commit into from
Sep 14, 2023
Merged

Disable the vectorizer pipeline. #516

merged 1 commit into from
Sep 14, 2023

Conversation

maleadt
Copy link
Member

@maleadt maleadt commented Sep 14, 2023

Although we may want some additional configurability here, let's quickly disable it now as our Legacy PM pipeline doesn't have vectorization passes, because we specifically disabled those. This results in compilation failures with Metal, but I'm sure not all GPU back-ends support arbitrary vector operations (or know how to legalize them).

@codecov
Copy link

codecov bot commented Sep 14, 2023

Codecov Report

Patch coverage: 100.00% and project coverage change: -0.34% ⚠️

Comparison is base (5fc8d91) 84.17% compared to head (0b4fe4c) 83.83%.

Additional details and impacted files
@@            Coverage Diff             @@
##           master     #516      +/-   ##
==========================================
- Coverage   84.17%   83.83%   -0.34%     
==========================================
  Files          24       24              
  Lines        3292     3292              
==========================================
- Hits         2771     2760      -11     
- Misses        521      532      +11     
Files Changed Coverage Δ
src/optim.jl 88.74% <100.00%> (-2.88%) ⬇️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@maleadt maleadt merged commit 276b134 into master Sep 14, 2023
@maleadt maleadt deleted the tb/no_vectorization branch September 14, 2023 12:13
@maleadt
Copy link
Member Author

maleadt commented Sep 14, 2023

This results in compilation failures with Meta

See JuliaGPU/Metal.jl#257

@maleadt
Copy link
Member Author

maleadt commented Sep 14, 2023

Apparently SPIR-V also can't handle this:

      From worker 5:	InvalidFunctionCall: Unexpected llvm intrinsic:
      From worker 5:	 llvm.vector.reduce.mul.v4i64

      From worker 3:	InvalidBitWidth: Invalid bit width in input: 4

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

Successfully merging this pull request may close these issues.

1 participant