Sean Edwards and Tom McSweeney - Informal Applied Mathematics Seminar
|Starts:||15:00 15 Nov 2019|
|Ends:||16:00 15 Nov 2019|
|What is it:||Seminar|
|Organiser:||Department of Mathematics|
|Who is it for:||Current University students|
The Informal Applied Mathematics seminar will continue with a double bill from Sean Edwards and Tom McSweeney. Talks will take place in Frank Adams 2 (next to the Alan Turing Building's kitchen), with complimentary hot drinks and biscuits from 2.45 pm on the atrium bridge. Title and abstracts for the talks can be found below.
Sean Edwards: Subsonic air flow over an aerofoil: the response due to heat transfer through a streamwise aligned narrow section.
Abstract: The high Reynolds number regime for fluid flow is common in everyday life, from air flow past an aerofoil to water flow past a ships hull. Often, when a solid cuts through a fluid, near to the leading edge (the tip of the wing for example), we observe smooth laminar flow, but downstream the flow becomes turbulent, leading to drag and increased fuel consumption. It is therefore important to understand the properties of such flows and how they transition from laminar to turbulent flow.
In this talk, we will model an aerofoil cutting through a viscous, incompressible fluid (eg air at subsonic speeds). Our aerofoil will be a semi-infinite flat plane and we first present the classical Blasius boundary layer, a 2D similarity solution to this problem. Then we will consider heat transfer through a streamwise aligned narrow section of the surface with a width comparable to the width of the boundary region. We will attempt to rationalise the resulting response using asymptotic analysis. Finally, time allowing, we will present the response to active injection rather than heat transfer, and compare and contrast with the thermally forced problem.
Tom McSweeney: Scheduling heuristics for CPU and GPU
Abstract: Using both GPUs and multicore CPUs for general computations is becoming more and more common, both in high-performance and everyday computing. Task-based programming is a paradigm that aims to exploit this heterogeneity. The idea is that computational jobs are divided into many smaller tasks which are then assigned to the available CPUs and GPUs according to how well-suited they are for tasks of that type. The difficult part is finding a good schedule: what tasks should each processor execute and in what order? In this talk I briefly outline a classic heuristic for finding good schedules in general heterogeneous computing environments, before describing a similar new heuristic I have developed which is optimized for CPU-GPU machines in particular.
FYI: The Informal Applied Seminar is a weekly seminar for all applied mathematics students in the department. We get together on a Friday afternoon for students to speak about what they're working on at the moment. It's not formal - lecturers and staff don't attend - so it can be a good place to get feedback on your talk before an upcoming conference or just good to practise for public speaking in general. We generally meet on the bridge (outside the Alan Turing building first-floor kitchen) 15mins before the start and head to Sandbar for a drink afterwards.
Travel and Contact Information
Frank Adams 2
Alan Turing Building