Title: ACCU18 Trip Report

Author: Bob Schmidt

Date: 07 July 2018 17:16:22 +01:00 or Sat, 07 July 2018 17:16:22 +01:00

Summary: Balog Pal reports his experiences from the 2018 ACCU Conference.

Body: 

I had to skip last year, so was really eager to be there again and meet all the fine folks. This time I brought my young padawan and we decided to use a shared room (great deal as it’s practically half price) in the hotel instead of finding a bed outside.

We arrived on Tuesday evening and I hoped to find an extra-schedule presentation as in previous years. And there was a local group meeting too really, just it ended by the time of our arrival. Too bad. So fast check-in and off to the bar to see who is already there. And indeed the area around the counter was filled with our people. I waved hi to John Lakos before asking for some ale, then started to process internal confusion alerts. some parts of the picture didn’t match with my memory from 2 years ago. I even had to ask someone if that is John or not. And yeah, he was too, JL 2.0 (or maybe ½.0 would fit better). What is a great thing, last time I was actively worried for him.

Eventually I met most everyone already on site, but most people were pretty tired and time-lagged so it was off to sleep even before 2:00. There will be enough opportunity to catch up later.

I read the schedule around registration time and it, as usually showed an average over 3 ‘interested’ flags per slot. Though the keynotes did not look very impressive, based on title and speaker at least, so I was prepared to just one good of the four, leaving enough room for a pleasant surprise. And after the opening we headed into the first keynote, that I found without any substance. On the bright side it was not against expectations.

After that I headed for Anthony’s talk. (For sake of brevity I will not list my also-considereds as I could paste most of the schedule…) We learned about more ‘advanced’ tools in multithreading that are supposed to help in extreme scalability situations. Along with demonstrating problems that we encounter with the more basic tools. Especially without knowing details on how the hardware actually works. We also learned that hazard pointers and RCU proceeds well in the standardization.

After break I went for some nostalgia trip with Jez and Chris. In retrospect would probably better off with a different choice. Not even sure why, these days even nostalgia is not as it used to be? Then the last slot went to Nico, a choice that just can’t go wrong. And he showed a lot of new template-related material since C++11, traps with auto and decltype(auto), constexpr and much more. Followed by 12 great lightning talks filled with humor and energy.

As an extra event after that I crashed the SG14 teleconfed meeting where Herb Sutter presented his new ideas about lightweight exception handling. One that would allow code generation similar to using return codes without messing up the source code. Both the material and the meeting itself was really interesting. The idea would open up exception handling on gaming and other high-performance platforms where it is now usually on the forbidden list. Everyone in the room and on remote agreed it is a thing worth pursuing. Beyond that, on details I observed that the votes went almost exactly crossed: the number of likes in the room matched the dislikes on remote and vice versa. Go figure. But it’s okay, the pesky details can be figured out later to a better consensus.

It felt a long and exhausting day at that point but it seems the bar has really good regenerative environment, especially as almost everyone showed up. And the party went on to almost the sunrise as is supposed to happen.

Thursday started with keynote on Kotlin. A thing I knew like nothing about (except a few mentions as successor of java). And by the presentation I liked it a LOT really, both the language and its dev environment. The only down point being it is promised to be ready ‘by winter’ without a firm specification of a continent and a year OTOH different sources strongly suggest that Winter is actually coming, so I have my hopes up and will try this thing in the near future.

Next I decided (looking at the program not sure how) to see how not to lead a team of software professionals. And it turned out as a really lucky choice. While it went at half or even third pace compared to others, the content was really touching, and felt a really honest recollection of one’s mistakes and faults with a lot to learn from. On top of that I finally got the inspiration on how to make my next talk on ACCU that eluded me for good four years.

After the break I was really tempted to see John’s where the presentation was slimmed similar to the presenter to mere 260 slides – but I decided to go live with Dietmar for a double handful of good reasons. And learned some really interesting tech. That it’s even possible to get right on with simulated concepts before the real ones get implemented. It finished a little early so I could catch the tail of Hubert Matthews on reads and writes considered harmful. Turned out I completely misjudged the scope by the title and probably should have looked at this one from the start.

The last slot is a no-brainer choice with the pub quiz. (And who thinks it’s fair to run Kevlin in this slot too.) No matter if it is with Jon at the helm while Olve can’t make it to continue the series of yey-s. The format is a little different, instead of guessing what the code will produce we have to make up the shortest program that compiles and uses all the provided tokens. What could be better than a quiz with freebie drinks (kudos to Bloomberg folks) than to win it too. And it is not hard using my top talent of picking the proper team. That is not a hard guess either: the one with Richard Smith on. :) It turned out an interesting format in practice, though somewhat spoiled by abuse of templates where gcc maximizes opportunity in the ‘ill-formed, no diagnostic required’ realm, just make sure to have dependent expressions. Where Richard comes extremely handy. :) for the record the second team came up really close too.

After we had the lightnings squeezed in the small room as the big place was prepared for the dinner party. This year it was moved up from Friday which was a great thing. Another round of fun, and while the previous day Jon Kalb and Phil Nash lobbied for East const (the more agile people even could grab a bracelet beyond learning this cool name for the old phenomenon), now the conservatives pointed out how uncool it is to call the old-way ‘West const’. When it is correctly ‘const West’. Quite obviously – after the fact.

While the dinner commenced I went pubbing with Ralph, Anna and some more folks, and afterwards resumed the session in the bar till a good morning.

Friday started with Lisa’s keynote about ‘shape of the program’. Which turned out really enchanting for me. Possibly a big part was for her voice: it sounds extremely like Laurie Anderson and even matches the intonation. (I asked later, not deliberately, it just happens by chance). I only missed the music in the background, but I almost heard it internally. And the topology attached to the code was interesting too, providing a different section compared to what we encounter in our editors. Two out of three suspect keynotes with thumbs up, great outcome. After the note I was trying to query people on the mentioned Laurie Anderson similarity with not much luck: those I tried were not familiar. (Maybe we would need some talks on evolution of music?).

But that led to another lucky accident, my last victim was Michel Grootjans and I noticed his extremely artistic sketches of the keynote (really, look up his Facebook page!), and I missed the pick of the next session as it started in the room I appeared to be at the bell. So I learned about graphs from Dom Davis and an interesting new query facility based on those. Really insightful, I’m glad for this switch as I did not originally consider this to attend.

Also this day brought serious reinforcements: Francis Glassborow, Jonathan Wakeley and Mike Wong.

In the lunch break there was an ACCU meeting up in the lounge (this year it was dead space, no arcades, no booze and no people, hope it will return in the future). It was the preparation for the next-day’s AGM, and a successful one too as we found a willing candidate for secretary quite out of the blue.

Then I attended Jon Kalb’s summary of the past, present ad future of C++, that goes parallel to Uncle Bob’s talk on ‘future of programming’ (obviously 95% being about the past 70 years…:). followed by the event I had been waiting for for so long: a grilling of the committee. Which was pretty tuned back on actual grilling and more filled with all kinds of fun.

Another round of great lightning talks was followed by the Microbrewery event (more kudos to Bloomberg) with food, booze and super-short chess tournaments. No kidding, 2 and a half minutes on clock that is barely enough to just move a piece let alone think what to do. Alan won the event and deserves all the praise despite losing the extra play-off against the local champion. We also had demo of music played by running programs on a rPi based thing. That unfortunately turned boring after the first five minutes. Most people tried to talk with each other but it was really hard with the loud noise so I rather left for some night walk in Bristol and back to the bar as usual. And the event was fine and alive, unlike at similar time in the previous years with the ACCU dinner on Friday when it felt like the end of the world.

On the last day I started with another obvious choice: Marshall and Jon paired up like Penn and Teller and introduced us to the trials and tribulations of those who try to implement the standard library. (And with handicaps too: turned out they proposed plenty of things to make their life easier, and those were even adopted into the standard or are on the way – but to keep the codebase compatible with many different compilers most of those improvements are on the not-to-use list. DOH. And we learned some interesting tricks too.

After that came the only slot I was not really moved by either talk or by title. Decided to go with the Total War, but it was mostly a letdown. In the break we had the AGM, pretty effective, due to the preparation on the previous day.

For the last slot another (for me) obvious choice going with Mike Wong. He was fighting off some cold, but even with that radiates enough stamina to put a small space station into orbit. He told us about the progress in the many study groups he is involved in. And the framework created to finally harness the power in GPUs we have around that are quite underused. I really look forward to having this SYCL stuff put to use.

And the conference is heading into conclusion with Seb Rose’s keynote that I originally picked as the only one promising. And it was a great one too. Seb was out for hiking in the beautiful woods and mountains of South-France. And found some rally fitting parallel with our everyday practice of software development. From estimating to clashing our well-conceived plans with reality. Then embracing all the change the latter forces on us. This is really a talk everyone should see, packing much beauty, laughs along with wisdom.

At the end Seb gave the audience some exercise: grab a piece of paper and write down something one learned here at the conference and plans to put it into practical use as they return to work. And exchange the paper with someone around who will query for the progress a few weeks ahead. (Yeah, that wisdom from agile schools about making things actually happen by declaring them in public. Unlike all the promises we just make to ourselves and way too easy to drop out…)

I did not fill the paper as though I did learn plenty of interesting things, none of them fit as immediate change in my work. But I did think up some promises: one, to write and submit a trip report. That, if you read this, actually happened. Second, I want to hang on to the inspiration and prepare a talk for the next ACCU conf. (For what I also promised myself earlier, that I shall have the prez complete at submission time, not just a title, summary and hope to make it up in the last 3–4 months time…) And as a third thing I had an idea for a C++ proposal to be submitted to WG21 with some gathered feedback (Richard Smith said ‘it looks useful’ and ‘I see no immediate problems’ so it should have a chance). Let’s hope this public announcement can create enough starting momentum for them to not sink under boring problems of everyday struggles.

In the closing words Russel asked for some praise for Francis Glassborow who created this whole thing now called ACCU and he got a loud Hurrah. And also hinted that he has a preliminary promise for the venue same time next year, a good seed for ACCU2019.

As a summary, it was a great conference and I look forward to the next one already. In the meantime I started to watch the talks I missed – we have some great progress here, I recall in 2013 there was just one camera, then two and now almost all talks are recorded. Also it took 2–3 months for them to appear on YT, now the first ones came up just few days later and the rest is coming at steady pace. The one thing I miss are the photo galleries. If you have one, please send a link to our website so it is shared. Also waiting to read other people’s trip reports. If you were there, write up your experience! And see ya all on the next instances.

Notes: 

More fields may be available via dynamicdata ..