-
Notifications
You must be signed in to change notification settings - Fork 8
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
Cannot compile on MacOS: library not found for -lEnzyme-19 #175
Comments
I revisited this with the latest commit, 2a23a80, but failed with the same error: I poked around and found a folder of built libraries, and running
allowed the
command to succeed. I then ran (as per book instructions)
but failed to successfully compile the rustbook samples (as of commit 0ffed11799d9ee4be1db84e01f88bdba2405be10):
|
Hi, and thanks for investigating! The code for building and linking Enzyme is currently in the process of upstreaming and should land in upstream Rust within the next few hours, so I don't want to touch it right now. Once that happens I'll update the master branch of this fork to be built on the newest rustc. During upstreaming I got a few suggestions to improve the code quality, so maybe that will already be enough to fix your linker issue. Otherwise I'll make a follow up PR against rustc to fix this and let you know. One of the tests failing also ins't surprising, I haven't checked for them in a while since I'm currently busy with upstreaming and since the Rust frontend isn't tested that well yet rustc/llvm updates might break rust-enzyme test. That should quickly get better once I have everything upstream and time to fix some of these bugs. |
This is very, very cool-I'm so glad I decided to revisit this today! Just wanted to chime in here as well to say that while the build seems to now work, the examples fail. I have created a (horrible) container to test things locally in isolation, the setup can be found here. Attempting to run the examples as instructed in the docs fails:
|
I just rebased the master branch on the latest upstream rustc (which now includes the first autodiff code). Please feel free to test again. If it's still failing without your manual export, I'll look into fixing it in rustc @lynaghk |
Thanks @ZuseZ4! I just did a clean checkout and build (enzyme_rust_log.txt), but ran into the same linker error. The same
allowed the build to succeed. Unfortunately, there are still errors when trying to build the tests for the book (book_log.txt). |
@lynaghk Just as a small note, I did not forget this issue, my lab ordered me a MacMini to debug Apple issues, but apparently my config wasn't available directly from the company my lab works with. So it might take some more weeks till it finally arrives, but I'll look at it afterwards. |
Thanks for following up and no worries at all about the timing. I actually just purchased one of the new mac minis myself and am waiting for delivery in a few weeks too =D They look like incredible lil' machines. |
(Hi @lynaghk!) So I was trying to compile this too on my ARM Macbook and the LIBRARY_PATH setting here fixed the missing export LIBRARY_PATH="$(pwd)/build/aarch64-apple-darwin/enzyme/lib" By the way, this is super cool, I'm going to try to differentiate through an svg rendering library (not sure if it will work but we'll see how close I get!). |
First, let me say that I'm extremely excited about having automatic differentiation like this in Rust --- thank you for your hard work on everything!
I'm trying to compile from a clean checkout of e47e0ba on my M1 Macbook Air (MacOS 12.7.2) and the build fails with
library not found for -lEnzyme-19
.I did everything as per the book's instructions:
I've attached the full log as a txt file.
I see some references to
AppleClang 14.0.0.14000029
, which is interesting since on my PATH I have a much newer clang:I'm not super familiar with the dark magic of all these linker flags and environment variable stuff, so I figured I'd run it by y'all and document publicly in case anyone else runs into the same issue. Please let me know if you need any more info from my side. I also saw on some other issue threads that you may not have access to Mac hardware --- happy to hop on a call to pair if you prefer that to async discussion.
The text was updated successfully, but these errors were encountered: