From: Jérôme Hénin (
Date: Thu Nov 22 2018 - 11:30:53 CST

Update: in my previous build, I had compiled with the -no-pie flag to
override Ubuntu's default GCC settings, re-using some previously compiled
static libraries (VRPN, ACTC etc). This time I removed the -no-pie flag and
recompiled all the libraries - which was a more manual process than I
remembered, but I understand not everyone does that everyday.

The errors are gone. So this could have been a binary incompatibility
between GCC versions / settings.

As far as I can see now (using quicksurf, gofr), VMD & CUDA 10 seems to be
getting along.


On Thu, 22 Nov 2018 at 16:12, John Stone <> wrote:

> Jerome,
> I built and ran VMD with CUDA 10 without any errors or
> segfaults during August. The current source code should be
> CUDA 10-ready unless I accidentally broke something very recently.
> If you can send me both your compilation logs and/or describe
> how you're getting crashes that'd be great. If you can run VMD
> under 'gdb' and catch the crash with a traceback, that'd be even better.
> The only 'dangerous' issues I'm aware of with CUDA 10 are that there's
> a long standing bug that still isn't fixed in Thrust that VMD contains
> a workaround for (since CUDA 9.0), but I added #ifdefs for that a few
> months back and VMD should be using the workaround for the known problem.
> Best,
> John
> On Thu, Nov 22, 2018 at 01:47:27PM +0100, Jérôme Hénin wrote:
> > Hi,
> > Does anyone have experience with building VMD with CUDA 10? That's the
> > only supported release on Ubuntu 18, and building with that, I get
> either
> > erroneous display or segfaults when using QuickSurf.
> > If CUDA 10 is supposed to be supported, I'll send a more complete
> report.
> > If not, I'll try to use CUDA 9.2.
> > Thanks,
> > Jerome
> --
> NIH Center for Macromolecular Modeling and Bioinformatics
> Beckman Institute for Advanced Science and Technology
> University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> Phone: 217-244-3349