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.

Sign up for GitHub

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

Jump to bottom

[amdgpu] Enable llvm FpOpFusion option on AMDGPU backend #7398

Merged
merged 2 commits into from
Feb 20, 2023
Merged

[amdgpu] Enable llvm FpOpFusion option on AMDGPU backend #7398

merged 2 commits into from
Feb 20, 2023

Conversation

galeselee
Copy link
Contributor

Brief Summary

  1. Enable the FpOpFusion option. If not, it will cut saxpy performance in half

  • a. Turn off

    performance:
    image

  • b. Turn on

    performance:
    image

  1. Disable AsmVerbose in machine option. Thus we will get pure LLVM-IR after module and function passes. Or there will be some instructions for target in the module.

@netlify
Copy link

netlify bot commented Feb 19, 2023 •
edited

✅ Deploy Preview for docsite-preview canceled.

Name Link
🔨 Latest commit dfc72d8
🔍 Latest deploy log https://app.netlify.com/sites/docsite-preview/deploys/63f2c4dde515b4000880a90b

@galeselee galeselee mentioned this pull request Feb 19, 2023
Open
@galeselee galeselee merged commit 1d8ccc9 into taichi-dev:master Feb 20, 2023
@galeselee galeselee deleted the amdgpu_enable_fma branch February 20, 2023 07:20
quadpixels pushed a commit to quadpixels/taichi that referenced this pull request May 13, 2023
@galeselee @quadpixels
[amdgpu] Enable llvm FpOpFusion option on AMDGPU backend (taichi-dev#…
d6e10bd 
…7398)

### Brief Summary
1. Enable the FpOpFusion option. If not, it will cut `saxpy` performance
in half

- a. Turn off

  performance:
<img width="634" alt="image"
src="https://user-images.githubusercontent.com/47965866/219954146-1e54a0e7-ffa7-441e-ad09-20aeba0f7954.png">

- b. Turn on

  performance:
<img width="604" alt="image"
src="https://user-images.githubusercontent.com/47965866/219954180-0730fc4a-7711-4f26-b752-ef441fc87aab.png">
2. Disable AsmVerbose in machine option. Thus we will get pure LLVM-IR
after module and function passes. Or there will be some instructions for
target in the module.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@jim19930609 jim19930609 jim19930609 approved these changes

@turbo0628 turbo0628 Awaiting requested review from turbo0628

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
2 participants
@galeselee @jim19930609