I’ve been doing some teaching of students, specifically a discipline we’ve called “data engineering” and the documentation of Slurm is quite bad and obfuscates how to do a very simple thing - namely - how to set up a Slurm cluster.

A couple of weeks ago I got forwarded a chain of emails from a researcher who was having difficulty getting QLM to do what he expected when he multiplied a Qint by an integer.

If you have been around High Performance Computing for any serious period of time you will have heard about GPUs, usually in reverent tones, that you have to learn a new programming langauge (CUDA) and become a master of arcane and terrible magicks to apply this to your application. The good news is that for a long time, since the invention of OpenACC this has not been true, you could write code in Fortran like God intended and add directives and the PGI compilers would generate GPU code for you. The problem being of course that the PGI compilers were commercial and cost a reasonable amount of money. Some years ago, Nvidia bought PGI and last year made the rather amazing decision, after switching the backend to LLVM to make their compilers available for free.

Some time last year I decided that as part of my escape from Internet isolation due to my own literal madness I would start trying out streaming services and stop relying on physical media to watch stuff and so I signed up for Netflix by prepay (yay, avoiding giving them my card details) and opened an eye-watering array of content which for a short while quite made me lose my mind until I understood the game of Netflix which is to repeatedly present the same fairly limited set of knock-off content but with different cover images to confuse the user. Don’t get me wrong, Netflix is still much better than not having Netflix, but there are moments. Oh there are moments. Due largely to The Boys, I shortly thereafter signed up for Amazon Prime, therefore forever wedding myself to servitude to Bezos. Prime feels like a better value proposition than Netflix, because you also get free next day shipping you can share with your significant other and a whole host of other services, each not quite as good as Prime Video but good enough and comparatively free thus dissuading you from signing up for competitors.

One of the things that infuriates me about modern technology is planned obsolescence – that is the way that modern devices are designed to expire after a fixed period of time. In the “bad old days” this used to be that things mysteriously failed a few weeks after the warranty expired which in reality was at best sod’s law and at worst a sort of mad conspiracy theory. I have a PowerBook 150 (circa 1994) that works. OK, so it has lots of new parts in it but it works and is maintainable like and old car – I can take it apart, Apple published an excellent manual for servicing it, it uses standard tools and so on. Modern devices are sealed, batteries are non-user removable and companies like Apple deliberately use weirdly patterned screwdrivers to frustrate repair. Apple, incidentally work even harder to restrict the supply of spares and prevent even trained non-Apple technicians from repairing them (currently the subject of a strong campaign for right to repair legislation). Now with everything being Internet connected, this kind of behaviour is enforced in another way – in manufacturers limiting the period over which devices will get security updates.

So it’s been a long time since I touched this blog and a lot of stuff has happened.

For those not in the know, “Grace” is UCL’s primary High Performance Computing (“HPC”, as opposed to High Throughput Computing or “HTC”) resource and the first service I designed and commissioned. Grace is over three and a half years old now and we are in the process of tendering a replacement which will run alongside Grace for a number of years, before being expanded to replace it.

It’ll be no surprise to anyone who follows me on Twitter but I got quite into trains this year. This meant models (n gauge to fit the London flat lifestyle), taking photos of real trains and playing Train Simulator 2018/2019. Train Simulator (TS) is quite an old game on a very old engine which has a truly amazing amount of content available for it - hundreds of trains, lines, scenarios, and comes with the tools to make your own scenarios. The developers of Train Simulator, Dovetail games have, for some time been working on the “next generation” of the series with a new engine called Train Sim World (TSW) but it is way beyond the capabilities of the laptop I play games on.

I’m afraid this is one of those posts which are at least partly deliberately contrarian. I hope this will provoke people to think differently about how they do “open science”, about what they share and how. In a way this is an expansion of some comments I’ve made previously about open science on Twitter and I might seem flippant but I’m slightly concerned we (computational scientists) have taken a wrong turn, based on a confusion of a number of similar ideas. Luckily, it’s easily resolved.

In a way up until now, my life has been somewhat blessed; namely I have never had to understand Matlab for work. I started out on actual programming languages and stayed there. I have had to install and test it for other people though. And I did, in a fit of masochism, do the ubiquitous pi example for Matlab and its achingly slower Open Source semi-compatible cousin Octave.

In my previous post about FreeDOS tools I mentioned that I hadn’t yet tried to get Gambit Scheme to compile .scm code to a stand-alone .exe. There are a couple of things you need to do.

A significant part of the job of my team at UCL is to install and maintain user applications on UCL’s centrally-owned clusters (Legion, Grace) and the new Thomas national materials modelling service. A number of years ago, prodded by my colleague Dr Kirker, we embarked on a journey where the installation of all new applications is scripted (in bash, nothing fancy) so that subsequent deployments (either to other machines or upgraded packages) is as painless as possible. You can see our work over on Github, because transparancy is also something I think benefits university IT in general, not just “Research Software Development”.

We’ve had a number of tickets recently asking about running Jupyter Notebooks on Legion/Grace. Until the architecture of the Jupyter Notebook changes this will never be a good/safe idea. This sparked a discussion which descended into an argument between James and myself on the internal Slack about whether it is appropriate to encourage new researchers to use Jupyter notebooks.

I thought as I find useful things to add to FreeDOS, I’d maintain a list of them here for other people to find.

Bywater BASIC is very old fashioned, Open Source BASIC interpreter. If you have Ubuntu to hand, you can install it straight from apt.

Generally, I’m a tinkerer. I like playing with things and taking them apart and fixing them. I do this in my day job looking after UCL’s supercomputers, but I like to have some systems at home to play with. If I’m honest, I’m attracted to unusual, old pieces of technology - old programmable calculators, old computers - my most recent aquisition for example is an Amstrad NC200 - a Z80 based laptop computer which comes with BBC BASIC, a stunning 128K of memory and a DD floppy drive. I dream of having the room for DEC PDP-8s, 11s, and 10s in my flat. But that’s not an option today.

Some of you may be aware that my colleague Dr Kirker and I maintain a repository of implementations of a particular way of calculating PI in whichever language we are playing with at that moment in time here.

Julia is a new programming language for technical computing and shows a lot of promise when it comes to being a useful language to replace Python as the standard for scientific computing.

I mentioned previously that I recently bought a TI Nspire CX CAS. Today I thought I’d have a go at backing up the small programs I’d written using the tilp2 package in Ubuntu since I use Linux and there is no official TI support for that. This worked, sorta, but you end up with programs in a proprietary binary format that nothing can read (and you don’t really want to put into version control).

I’ve been frustrated for some time about the lack of programability of modern portable devices. “Smart” phones are anything but - sure, you can buy more apps for them, but when was the last time you opened an editor, wrote some code on one and ran it? It wasn’t always so - the coolest computer I own is a little Sharp PC-1251 - a pocket computer Sharp produced in the 80s. In computing terms it’s tiny - ~0.5mhz 8 bit processor, 4K of RAM and a one line display, but in one respect it is superior to every iPhone ever made - it comes, out the box, with a programming language. In this instance, a cut down version of Basic, called S’BASIC.

I tried editing posts in Libreoffice (and saving as plain text) so that I could use the spell checker and it broke the YAML front matter.

Since I’ve been put in the position of some responsibility here within Research IT Services at UCL, I’ve decided that it’s sensible to start having a public, semi-professional blog where I can talk about some of the things that we (or more specifically me) are doing here at UCL to enable our researchers to use computers to generate science, as well as posting long form about various things vaguely or not at all related. I’ve had a twitter account for a while, but UCL doesn’t really have a recommended blogging platform for individual staff, so a colleague suggested I try using Jekyll with Github pages and so here we are.