-
Notifications
You must be signed in to change notification settings - Fork 7
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
use precompiled JLL #7
Conversation
Hi Dominique, thank you very much for this contribution. The failures on MacOS are likely because QPALM uses a relative and absolute tolerance (both default to 1e-4). I set the relative to 0 now in the tests. Could you please run them again and confirm whether it works, then I will integrate your pull request. (Sorry, I don't have a Mac, and travis made their service paid so I can't test it.) |
Yes that seems to fix it. Now I'm seeing output of the form
with a suspicious final residual, and I get a segmentation fault in the warm start tests:
that I can reproduce on linux. The link to the Bintray binaries that is in the documentation gives me a 404, which was the original motivation for building the JLL. Do the warm start tests pass for you? Here are the commands I used to compile QPALM: https://github.com/JuliaPackaging/Yggdrasil/blob/master/Q/QPALM/build_tarballs.jl#L31. Does anything seem off to you? |
The warm start function has been updated and needs to be one call now (since I set work->x/y to NULL in C and check for that later if the corresponding x/y_warm_start is NULL). Actually, that would have never worked, since with sequential warm start calls, the first variable to be set would have been overwritten by 0 in the second call. Anyways, I fixed the test, and this should run now. Can you confirm? |
All green lights on macOS and Linux! There's a segfault on Windows, but I don't know if anyone's tested the code there. |
Also it might make sense to move |
I don't understand your last comment. QPALM.jl is a submodule in QPALM_vLADEL, so qpalm_julia.c is already located in this folder, no? |
I'm saying that that's no longer necessary. With this PR, the C library and the Julia module can be separated since users no longer need to clone either repository to use the Julia interface. |
Would it make sense to merge this and sort out the issues on Windows separately? |
Sorry guys, I wanted to have a look at the CI since I have to fix it anyways, but got distracted since it was my public defense today. On the bright side, the issues on Windows should have been fixed. Turned out to be a different definition of the type long there. |
congratulations on your defense!
where was that fixed? |
Thanks. In the latest couple of commits of https://github.com/Benny44/LADEL and https://github.com/Benny44/QPALM_vLADEL/ |
woah, congrats! |
I prepared precompiled binaries for QPALM_vLADEL for a variety of flavors of linux, macOS, FreeBSD and Windows in the form of a JLL package: https://github.com/JuliaBinaryWrappers/QPALM_jll.j (you can see the available platforms here: https://github.com/JuliaBinaryWrappers/QPALM_jll.jl/releases/tag/QPALM-v0.1.1+0).
This PR makes it seamless for users to install QPALM.jl and have the appropriate binaries installed automatically as Julia artifacts.
On macOS, I observe two test failures:
The tolerances are nearly attained. I built QPALM_vLADEL against OpenBLAS. Perhaps that matters?!