A view of the tea fields in the Kennan Devan hills near Munnar, Kerala.
After having experienced the hustle and bustle of Bangalore, the next plan was
to explore Kerala, a state which is located on the tropical south-western tip of
the subcontinent. On its coast, it is home to vast stretches of palm tree
adorned beaches, which are flanked by rice fields criss-crossed by small canals,
just as in Venice–though maybe a bit less flamboyant. Going further inland, one
meets the mighty Western Ghats mountain range, a high-altitude rainforest,
home to elephants and tigers, which is ideal for escaping the oppressive
humidity on the coast. Politically, it is also well-known within India for its
high degree of development: it has the highest literacy rates, life expectancy
and lowest impoverishment rates in the country. All great things. And they also
love a bit of communism too it turns out.
A streetscape in a slightly edgy neighbourhood in Bangalore.
Bangalore, which is known as India’s “silicon valley”, was my first port of call
for my trip to India. It was the winter semester break and I had just been
travelling around Venice and Triest, so my mind was hitherto occupied with the
calm and cultured sights of northern Italy. However, this collected state
disappeared as soon as my first night in India, when my
long-neglected sense of culture shock arose again and brought me back into the
present with the overwhelming feeling of newness that all travelers crave.
Bells and whistles: unless you have one of these in your cupboard, you might have to settle for classical simulation.
I was recently playing around with a research artefact from the paper
Grafeyn which implemented some
interesting techniques for circuit simulation. Also, it was written in Rust, so
that doubly piqued my interest! Here’s what the paper has to say about itself…
I was returning to Triest on the coach, when overhead a troupe of fighter jets
flew right over us, emanating smoke trails in the colours of the Italian flag
behind them. Needless to say this was quite a surprise! It was because there
turned out to be a major event occurring in Triest on that weekend: the return of
the Amerigo Vespucci to Italy after a two-year long tour around the world.
The gloomy weather in February keeps all but the most keen tourists away.
I’ve long been putting off visiting Venice as it has a reputation of being one
of the most touristy places on Earth. It’s known as a place where the streets
are clogged with Chinese and American group tours and everything is absurdly
expensive to match. Then there’s the government’s response to the
overtourism–fleece the tourists even more. Nope, neither of those sound
particularly appealing. However, now that I live in Munich, it’s pretty easy to
get a coach there and it’s not only cheap, but traverses some of the most
beautiful countryside in Europe on the way there as it goes directly through the
Alps on the Brenner Pass. So clearly it was finally time to make the plunge and
visit.
I previously (probably like many “old-skool” C++ devs) thought of cloud computing
primarily as renting other peoples’ machines at a high mark-up. However, I’ve
been reading up on papers in the databases research
field and so thought I would share some of my findings. The use-case I will be talking about here is how
modern OLTP databases are designed for the cloud and their corresponding
advantages.
Over Christmas, I was browsing eBay for mandolins and came across this beauty
and my better senses did not prevail–so I bought it for €45. It’s a flat-top mandolin
that was made in Leningrad in the mid 1980s.
Small angle neutron scattering for materials research uses a high-energy particle accelerator to produce scattering amplitude functions. However, the physics behind the neutron scattering is well-known and can be simulated on supercomputer clusters, provided an initial seed trajectory and molecular (MD) information is given.
I recently completed a project where I improved the performance of a program for neutron
scattering physics simulation by using CUDA to get some quite large
performance gains. Despite the existing application already being
written for super-computer clusters with MPI, writing a CUDA scattering
implementation, offloading some nodes to the GPU, and
allowing for a “hybrid” CPU/GPU computation model that still supports CPU clusters
worked surprisingly well.
