N-Accumbens

That makes sense. What about Chris Rorden's [zlib optimizations](https://crnl.readthedocs.io/optimizing_spm_fsl/index.html) for the general performance of the container? Chris seems to get [huge uplifts](https://github.com/neurolabusc/pigz-bench-python#readme)

> I considered this, but the author of `pigz` (Mark Adler) says that there is limited improvement with decompression. The decompression apparently must be done in series, so it can't...

> Thanks for sharing. Many of the use-cases are related to compression, but given your benchmarks, we could expect some performance benefits with decompression. However, I think the bottleneck here...

> Dear @N-Accumbens, > > I had similar problems before and couldn't solve them in neurodocker, so we build an image with VNC integrated and lot's of neuroapplications build using...

Hi @stebo85, Sorry about the late reply. Thanks for the link, I'll take a look at it. Using NeuroDocker to add neuro-applications sounds ideal and is vaguely what I'm envisioning....

@stebo85 So I was browsing around and I was wondering if this solution was considered? https://www.geeksforgeeks.org/running-gui-applications-on-docker-in-linux/ In other words, instead of using a VNC/RDP into the container, hook it up...

I've attached the installation log here using `pip install --upgrade --force-reinstall nipype[all]` [nipype.install.log](https://github.com/nipy/nipype/files/6512496/nipype.install.log)

Related to this, is it possible to choose the GPU based on something other than the name? I have two of the same GPUs that have the same name, but...