Prof. Rachid Guerraoui visits LSD and teaches a Research Seminar on Robust Concurrent Computing

Because of physical limitations on clock speeds, major chip manufacturers have shifted their focus from trying to speed-up individual processes to putting several on the same ship.
We are now talking about potentially doubling efficiency on a 2x core, quadrupling on a 4x core and so forth. Some programs run 18 times faster on a 32 core machine.
This is x 18 - The concurrency gigs used to talk about 5% - 10% improvements by re-coding parts of the applications not 400% or 1800% - For a change, this is a change. Yet, multicore is useless without concurrent programming. For instance, a single-threaded application can exploit at most 1/100 of the potential throughput of a 100-core chip and Intel is indeed promising such a chip.
So the statement that "processors just keep getting faster and nothing needs to be done to get that benefit" is over. The constructors are calling for a new software revolution: the concurrency revolution.
This might look at first glance surprising for concurrency is almost as old as computing and tons of concurrent programming models and languages were invented. In fact, what the revolution is about is way more than concurrency alone: it is about concurrency for the masses. This is at odds with current practice in concurrent computing that are still based on low-level locks. These are very fragile.
They tend to be error prone and hard to scale unless programmed by experts. This course will cover non-blocking concurrent computing: a radically different approach that relies on solid foundations.