Week #127

(Date: 23.3.2014 – 27.3.2014)
GTC and me!Andi goes to USA.
Let me summarize stuff for, maybe you came in late.
At the beginning of this year I applied for a talk at the biggest GPU conference there is, the NVIDIA GPU Technology Conference (GTC). I filled out lots of forms and abstracts and motivations and stuff. And, finally, was accepted. 25 Minutes about »Enabling the Next Generation of Particle Physics Experiments: GPUs for Online Track Reconstruction«, in one of more particle physics-oriented science tracks of the conference.
I booked a hotel (quite a fancy one, as it’s one of the conference hotels) and a flight.
And now, well, it’s time to attend the flight. Saturday morning, 10:00, starting from Frankfurt Airport. Only 15 minutes or so from Frankfurt Main Station, where my hotel for the previous week was located next to.

Hotelzimmer. Ungemacht. Aber mit Fischauge.A good flight1 and short shared-shuttle ride through Silicon Valley later I arrive at The Fairmont San José. And am amazed by the size of my room — especially compared to the flophouse hotel2 in Frankfurt the last week.
San José itself, though, looks (probably) like any other functional American city. Large roads with large cars, surrounded by large3 houses, filled with drink cup-bearing people. Plus, very un-american: A tram, stretching through the city, and lots of buses.
Panorama Fairmont San José
Staying up late (20:00) on Saturday, to counteract jet lag, and getting up early on Sunday (same reason). Come noon, we drive to a shopping mall a little outside the center, to look for stuff. I buy a new pair of sport sneakers. Because they are cheap. And we eat something – finally Burrito again. How I missed you, my dear!

On Sunday afternoon we go to the convention center and register early to skip lines tomorrow. I get some presenter presents (a cheap, but usable backpack; a battery pack; a t-shirt) and we socialize with drinking a glass of wine.

Monday the actual conference starts. 20 rooms (I didn’t count, just my feeling) giving possibility for 20 parallel sessions. At every entrance you need to get your ID scanned and your passing confirmed. You had to favorite sessions beforehand online, so that you could be put on a white list if a room is too full. Didn’t happen for me, though.
Free drinks everywhere and, everyday, free food during lunch and something to eat in the evening. You got quite something for your 900 USD…4
The talks were mostly very computational-heavy. Stuff about optimizing compilers, getting more out of your GPU and squeezing out the last nanosecond of your program. Quite high level and for me, physicist newcomer, quite hard. But, sometimes there was quite an interesting thing to see.
Before First GTC14 KeynoteTuesday, Wednesday and Thursday each had big keynotes in the largest keynote hall I ever sat in. On Tuesday, the CEO of NVIDIA kicked off the convention. Showing current products and introducing next-generation ones. And giving a small present to every attendant.5 On Tuesday, Pixar showed how far they came from their first animated movie to the current one (Monster’s University), and what they now can do in real time to give their animators more interactiveness. Using GPUs, of course. On Thursday, for me the most impressive keynote took place. A medical professor showing how he’s using simple-looking games to train the brain and, hopefully, better Alzheimer’s and showing how they use EEG to look into the brain in real time. Using GPUs, of course. Very impressive, though.

Poster SessionThe whole conference had a very professional spirit, but also a bit playfulness surrounding it – it’s still a bunch of nerds coming together, I guess.

My talk on Wednesday was quite good. That’s my feeling at least. I was able to use my Mac, so I was able to show some animations. The slides, I transformed to 16:9 aspect ratio on Monday, looked also alright – although I didn’t put too much effort in the transformation. Only the audience was a bit small – only 15 people. Good for my nervousness, but I hoped that the GTC, being such a big conference, there would be some more people. But at least from the ones attending my talk, there were some good questions.
Afterwards, I was contacted by a person from Italy who works on data transport to GPUs for High Energy Physics. The same guy the professor at the Supercomputing Centre in Jülich also said we should contact but I never had the time to. Funny, how things work out sometimes.
He approached me after the talk because I put »data transfer to GPU« to the list as one of the points we still have to look into.
You find the slides to my talk in the dedicated posting. I only want to hear praise!

On the following Monday it’s the final deadline to finish our Triplet Finder paper (the one together with the Supercomputing Centre). The peer reviewer criticized a lack of literature review to show that we are actually doing something important (which he was quite sure about) and do not work in a scientific vacuum. Since this points to the science part of our paper I found it my duty to work on this.
So I read up on tracking and, during GTC, added a paragraph to our paper referring approximately ten papers on the topic. One paper (which I researched before GTC!) of the guy who had the talk after mine. Funny, how things…

San FranciscoOn Thursday at 17:00 the conference is over and I take my bags and go to San Francisco. I booked a hostel room until Sunday, when my flight leaves back to Germany.
As it’s not work-related, at this place not too much about SF (not »Frisco«). Just the essentials: Beautiful town – I like the hills; lot’s of homeless people, screaming at nights in front of my hostel room; taking a hostel was the right decision – it was easy to get to know some people and do something together; doing stuff alone < doing stuff together – seeing sights alone is possible and ok-ish, but seeing sights together, being able to talk to someone about it, is better; Germans are everywhere; you should check out Wild SF Walking Tours, it’s really great fun – even so if you get wet to your skin because you somehow managed to find one of the few whole-day raining days of Sf (not »Frisco«)


  1. Apart from being specially controlled by the hands-on manual version of the full body ›naked‹ scanner alternative. 

  2. I had to look that up 

  3. And tall. 

  4. I didn’t need to pay. I was speaker. Yay! 

  5. An NVIDIA SHIELD