C++Now - trip report

As I’ve mentioned previously, I’m currently on a break between jobs. With my copious spare time I decided – rather late in the day – to head to C++Now, a C++ Conference held yearly in Aspen, Colorado. Several C++ folks had recommended that I come to it: it’s a smaller and more intimate conference than CppCon.

I was not disappointed! Firstly, Aspen is a beautiful place to hold a conference, although the thin air at 9,000 feet up takes a bit of getting used to.

The conference rooms used for C++Now nestle in a beautiful meadow.

After deciding to go, I thought it would be odd to attend a conference and not speak. Of course, submissions for talks had closed months earlier, but thankfully the conference has lightning talk sessions. I love these kinds of things, and so prepared a 5-minute talk: “Trip Report: 7 years in finance with C++”. I submitted the idea on the Thursday before the conference. To my surprise, one of the organisers – the amazing Jon Kalb – contacted me the next day to ask a favour. One of the scheduled presenters had dropped out: could I possibly fill in? With the over-confidence of someone who doesn’t do this enough to know better, I said yes! My talk was to be on Tuesday, so I had to cobble something together pretty quickly!

Aspen has a small airport but many people fly in to Denver and drive the remaining 4 hours. I’m blessed that Jason Turner (of C++ Weekly and CppCast fame) lives in Denver and offered to let me stay the night before (and the night of my return). I’m even more lucky that Charley Bay offered to drive me to Aspen…and also to let me share his room! As all the rooms were booked, this was a life-saver. Thank you so much Charley!

Sunday night was conference registration and a reception. It was great to meet so many great people. The conference is smaller than others I’ve been too: around 150 attendees. By the end of the week I knew pretty much everyone by name, which is a refreshing change from the anonymity of some larger conferences.

The sessions started early and finished late, with plenty of social activities in between. Most sessions were 90 minutes long: it seems a long time, but the small nature of the rooms and the informal setting lend themselves to a very interactive presentation style. That meant that even the best-planned talk would take a fair bit longer answer questions as the talk progressed. I love this style!

Here’s a list of all the talks I saw, along with a short description of them. Bash films has recorded them all, so they’ll be available on the C++Now YouTube channel once they’re postproduced.

Monday 7th June

• Lisa Lippincott’s opening keynote was a mind-blowing explanation of how she’s working on some theories of how imperative programs migth be modeled and perhaps even proven correct or otherwise. Beautiful slides, well presented, but mostly over my head! Although maybe it was the early start and lack of coffee…
• Charley Bay presented in his inimitable high-energy style a history of quantum computing, along with his thoughts and intuitions about how quantum phenomena might actually be easier for computer programmers to understand than physicists. Great fun, and thought-provoking.
• Arthur O’Dwyer gave a talk on std::pmr and why an allocator is more accurately described as a handle to a heap. (Although later he gave a lightning talk correcting this position!) I hadn’t heard of the new std::pmr allocators, so this added to my reading list!
• Mateusz Pusz gave a great overview of some of the features coming in C++20.
• The first night of lightning talks was great fun!

Tuesday 8th June

• Tony Van Eerd updated us on his quest for a lock-free queue. Tony’s an amazing presenter, and I enjoyed part 3/N of his ongoing talk series very much.
• Mateusz spoke again, giving some insight into how Conan is used.
• Jason Rice introduced us to cppdock, a novel container-based build system.
• Not a talk, but at lunch I spoke with Eberhard Gräther, and he showed me SourceTrail which looks like a fantastic way to start dipping into a novel codebase, or even to gain insight into your own existing code.
• I was very excited to meet Vittorio Romeo and hear his talk on std::future-like things which don’t need allocations. A very polished presentation, and a great “science report” of his findings along the way, which I think is worth pursuing further. Always good to see some allocation-free code too, alongside some expression template-type stuff. I’m a bit of an old fuddy-duddy about some of the new-fangled concepts in C++, but talks like Vittorio cause me to rethink my position in a very healthy way.
• My talk… “What else has my compiler done for me lately?”. I took all the cutting-room floor examples from my CppCon talk last year and made a (mostly) new talk showing yet more ways compilers can amaze us with their ability to optimize our code. The talk went about as well as I could possibly hope – mainly because of the many wise and learned folks in the room helping me out with great observations and questions as we went through the examples. I hope the video does this justice: I can imagine this kind of back-and-forth talk where the audience is more involved might not be captured well. We will see! Slides are on GitHub.

Wednesday 9th June

• Tony Van Eerd gave some sage “Words of Wisdom” in another triumph of well-reasoned observations on how to be a better programmer.
• My favourite technical talk of the whole conference was given by Malte Skarupke: a fast hash map. Malte has a series of blog posts on this, and will post his source code on his website soon. I can’t wait to see how it looks. Great talk, fantastic data-driven research and the kind of thing that gets me really excited about programming in C++!
• Jason Turner’s presentation was a great explanation of all the things wrong (and right) with initializer lists.
• Phil Nash’s talk on “Option(al) is Not a Failure” discussed many different strategies for error handling in programs. It introduced Herb Sutter’s early ideas on how static throw might be part of the solution.
• The second evening of lightning talks was an absolute blast too. Despite my giving a full-length talk, I still did my lightning talk too. Slides are on GitHub.

Thursday 10th June

• Ben Deane explained the use of the declarative style in C++. This was an illuminating introduction (to me at least!) to the style of putting more logic into declarations, using I+FE (Immediately Invoked Inline Function Expressions) to push setup code into discrete scopes, making more things const, and just generally making code even easier to reason about. This was a fantastic talk, and I believe it won the “talk of the conference” award. Well done, Ben!
• Peter Bindels gave a great overview of views, and how they can be implemented in a fairly straightforward way. Ranges is coming down the pipeline (it seems!) but Peter’s talk gave practical insight into how one can implement and benefit from views and lazy evaluation with existing codebases.
• Anastasiia Kazakova showed some of the cool things IDEs (not just her own, CLion, but others too!) can do to make our lives better.
• In the most interactive talk I went to, Gašper Ažman presented a novel paper on how to accept this in different ways, opening the door to less repetition of code in reference-qualified member functions and more excitingly lots of novel ways to do unified call syntax. An awful lot of back and forth with the audience made this an awful lot of fun. I suspect the video won’t work very well as a result. I guess you just had to be there!
• Titus Winters gave a polished and eloquent talk on modern C++ API design. As always, it has a skew towards Google’s way of doing things (which I mostly agree with), but others may differ. Either way, the talk was informative and worth a watch when the videos are out.

Friday 11th June

• The closing keynote was from one of my favourite presenters, John Regehr. His specialist subject is undefined behaviour, and in this talk he gave a great overview of why we have undefined behaviour and why it’s so useful. This is a pretty mind-mendingly complex subject, but John handled it well. I finally learnt the difference between undef and poison values in LLVM!

Conclusion

The notes above don’t do the conference justice. So much of the cool stuff happens during the breaks, or at dinners or lunches. I met some amazing people, and was inspired to read up on about 12 new papers, and to investigate more libraries. A special thanks to Michael Caisse for inviting me to dinner and giving me a great opportunity to speak with many of the boost developers. And again, huge thanks to Jason Turner for the pre- and post-conference accomodation and airport taxi service, Jon Kalb for giving me the opportunity to talk, and of course Charley Bay for letting me car-share and room-share with him.

I’m already looking forward to attending again next year!