TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Enzyme: Cross-language Automatic differentiation for LLVM IR

30 pointsby dhashover 4 years ago

2 comments

wsmosesover 4 years ago
Hi all, author here.<p>A couple of relevant links for the curious<p>Github: <a href="https:&#x2F;&#x2F;github.com&#x2F;wsmoses&#x2F;Enzyme" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;wsmoses&#x2F;Enzyme</a><p>Paper: <a href="https:&#x2F;&#x2F;proceedings.neurips.cc&#x2F;paper&#x2F;2020&#x2F;file&#x2F;9332c513ef44b682e9347822c2e457ac-Paper.pdf" rel="nofollow">https:&#x2F;&#x2F;proceedings.neurips.cc&#x2F;paper&#x2F;2020&#x2F;file&#x2F;9332c513ef44b...</a><p>Project: enzyme.mit.edu<p>Basically the long story short is that Enzyme has a couple of really interesting contributions:<p>1) Low-level AD IS possible and can be high performance<p>2) By working at LLVM we get cross-language and cross-platform AD<p>3) Working at the LLVM level actually can give more speedups (since it&#x27;s able to be performed after optimization)<p>4) We made a plugin for PyTorch&#x2F;TF that uses Enzyme to import foreign code into those frameworks with ease!
评论 #25358751 未加载
achernyaover 4 years ago
It looks like the links to the code have some junk characters at the end. It should be <a href="https:&#x2F;&#x2F;github.com&#x2F;wsmoses&#x2F;Enzyme" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;wsmoses&#x2F;Enzyme</a>.