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

Fix compilation issues with CUDA 9.1 #2212

Merged
merged 2 commits into from
Feb 25, 2018

Conversation

SergioRAgostinho
Copy link
Member

@SergioRAgostinho SergioRAgostinho commented Feb 13, 2018

Superseeds and closes #2208, fixes #2207.

This allows compilation for me in CUDA 9.1 and CUDA 8.0.

@taketwo @cmcmurrough @Robzz please give it a try.

Edit: Not ready yet. Found more errors of the same type.

@SergioRAgostinho SergioRAgostinho added the needs: testing Specify why not closed/merged yet label Feb 13, 2018
@SergioRAgostinho SergioRAgostinho added this to the pcl-1.9.0 milestone Feb 13, 2018
@Robzz
Copy link
Contributor

Robzz commented Feb 13, 2018

Just saw your edit, I'll wait until you think this is ready, but FYI #2208 works for me (Windows) and for colleagues running both Linux and OS X. We had some other breakage but that's related to another recent merge (#1421), nothing to do with CUDA.

@SergioRAgostinho SergioRAgostinho added the needs: code review Specify why not closed/merged yet label Feb 13, 2018
@SergioRAgostinho
Copy link
Member Author

SergioRAgostinho commented Feb 13, 2018

I haven't built all the stuff in the GPU module but the fix is similar for all. The operator needs to be defined with a const argument.

I also added some modifications to suppress the deprecations messages that started popping up. Inspired from here facebookarchive/caffe2@78fbcaa

@jasjuang
Copy link
Contributor

I can also confirm this PR fixes the compilation problem with CUDA 9.1 on Ubuntu 16.04 for me. Look forward to see this getting merged!

@taketwo
Copy link
Member

taketwo commented Feb 14, 2018

Same here, compiles fine 16.04 + 9.1.

Edit: turned out I have CUDA 9.0.

@Robzz
Copy link
Contributor

Robzz commented Feb 14, 2018

This doesn't break anything for me on Windows but I wasn't able to get anyone on Linux or OS X to try it so far.

@SergioRAgostinho
Copy link
Member Author

I'm not sure we're gonna get many OSX users on this. Macbooks stopped shipping Nvidias a while back.

I did hear recently that some support was added for running external cards but there's been no report whatsoever from our users in this configuration.

Just out of completeness @fran6co you wouldn't happen to be running an OSX with an Nvidia card to give it a try to this thing would you?

I'll ask around the mailing list as well.

@fran6co
Copy link
Contributor

fran6co commented Feb 15, 2018

@SergioRAgostinho sorry, I don't have a Nvidia card to give it a test.

@Robzz
Copy link
Contributor

Robzz commented Feb 15, 2018

I have a coworker working on a Mac with a nVidia card, I'll ask her to try it out if she has the time.

@paulsoning
Copy link

paulsoning commented Feb 15, 2018

@SergioRAgostinho, MacOS 10.13.4 Beta with CUDA 9.1 build was okay

@taketwo taketwo added status: ready to merge and removed needs: code review Specify why not closed/merged yet needs: testing Specify why not closed/merged yet labels Feb 23, 2018
@cmcmurrough
Copy link
Contributor

I tested on Ubuntu 16.04 LTS x64 with CUDA 9.1. It works.

@SergioRAgostinho
Copy link
Member Author

@PointCloudLibrary/maintainers need a merge on this please.

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

Successfully merging this pull request may close these issues.

PCL build fails with CUDA 9.1 on Ubuntu
7 participants