1 GTALUG 2020-06-09
- For the GTALUG June 9, 2020 meeting
- Drawn from the GTALUG June 2020 EtherPad
- Questions and answers and such like
1.1 Hugh on UEFI for ARM
Related to Hugh’s Comments about ARM Support https://www.yoctoproject.org/yocto-project-long-term-support-announced/
Not sure about uboot currently as related to upstreaming SOC https://www.bsdcan.org/events/bsdcan_2020/schedule/session/53-making-pi-serverready-time-for-cheap-arm-infrastructure-be-standards-compliant/ Video for the above talk should be published soon (by July)
- Original ARM boot loader was “uboot”
- Vendors have needed to customize it each time, and so the code has continually diverged and there are many versions, all different
- Talk about porting to UBoot and the issues with it (few years old now): https://www.youtube.com/watch?v=5E0sdYkvq-Q
Is the raspberry pi sufficiently open so to allow a single distro to be adaptable for multiple different SBCs?
- Raspbian is often used (forked to “armbian” for other systems)
- Different devices often need what is termed a “device tree” that is unique to a board or family of related boards
- Annoying thing about ARM is that many have GPUs, and unfortunately most of the GPUs have issues with NDAs so that it is troublesome to share low level code
What are your thoughts on using Anaconda in Ubuntu or other Linux distro vs installing all Python packages and using Python virtual environments instead? I’m coming from Windows, so Anaconda was a straight answer, but now that I’m becoming a Linux user, is not that clear that I should use Anaconda, and could better learn to use the Python venv. Just wanted to get some thoughts.
- pyenv
- Should we use distribution-managed packages? Or build a virtual environment to pull fresh code?
- These days, scripting languages have their own dependency systems which makes life difficult for distribution makers.
- Awesome Alex, thanks for your thorough explanation, I’ll certainly look into pyenv…. my name is Nestor Sanchez btw…
1.3 Python II – the Sunsetting of version2
Note that Python 2 is officially sunsetted… https://www.python.org/doc/sunset-python-2/
Some old Python code still lurks in distributions.
1.4 Powershell
Has anyone been playing with Powershell? What sorts of differences are there between that and our favorites?
- It grew up with Windows so has lots of Windows bits
- It operates on streams other than Unix “bags of bytes” which can be very different
1.5 WFH Learning
- I changed my commute time for walk/run around the neighbourhood time. To keep some routine.
- Some tendency for longer days because we aren’t consuming any time on commute but maintain broadly similar hours
- Daily “stand-up” meetings (common in SCRUM methodology) are useful to add a little bit of personal discipline
- Audio bandwidth limitations are an issue
- Microphones on laptops that are picking up fans and refrigerators are common problems
- Microphones intended as speaking headsets provide improvement
1.6 ROCm
Any experience among this group with using ROCm for high performance computing?
- Alex had tried this somewhat; installing components was troublesome, somewhat easier for the NVIDIA toolset
- https://www.amd.com/en/graphics/servers-solutions-rocm-hpc
- https://github.com/RadeonOpenCompute/ROCm
- https://developer.amd.com/resources/developer-guides-manuals/ At the bottom are AMD’s official manuals on open computing for GPUs, the page also has the reference docs for the GPU ISAs too
- I’ll be trying to give it a go, if I can get it to work fine I’ll come back to talk about it. But yeah, NVIDIA seems like the de-facto option. In my research, the reason seems that AMD has no GPUs that are able to compete on the high end. Intel seems to be making discrete GPUs as well.
- NVIDIA alternative, CUDA? https://en.wikipedia.org/wiki/CUDA
- Other standard, OpenCL https://en.wikipedia.org/wiki/OpenCL
- See also SYCL https://en.wikipedia.org/wiki/SYCL
1.7 systemd, 10 years later: a historical and technical retrospective
systemd, 10 years later: a historical and technical retrospective
- everything needs to be decrudded once in a while
1.8 JITSI results
- mixture of opinions, for sure
- a couple of people fell off the meeting, so not impressed
- several reporting that sound is quite good
- user interface is quite comparable to Google Meet and Zoom
- Can use custom backgrounds similar to what are commonly done on Zoom
1.8.1 Custom backgrounds for videoconferencing
Get a green board to go behind, and then compile a virtual camera Using OBS Studio for Google Hangouts/Meet
- implements a virtual camera that allows putting arbitrary backgrounds behind you
- GitHub: CatxFish/obs-v4l2sink
- I made a green screen for chroma key out of 4 pieces of green bristol board from Dollarama.
- This can work with Jitsi, Zoom, Google Meet