COMP35112 - Chip Multiprocessors

Announcements

Course unit materials: The course unit materials (see below) are gradually being updated as the course unit proceeds. Please check the dates shown on the slides.

Access to lab multicore platform: We are using a multiprocessor machine called mcore48 for the lab work. Details appear in the labsheet for exercise 1 (see below). The list of people allowed to login to this machine will be updated in the third week of Semester 2. Once it has been updated, you should be able to log in using your usual university username and password and you should then see your normal filestore. If you have any problem logging in, please e-mail me and tell me what your username is.

Guest lecture: There may be a guest lecture later in the course. A previous guest lecturer was Dr Anton Lokhmotov from ARM. Anton described ARM’s Mali attached processor architecture and its associated OpenCL programming model and language. The slides are below, and Anton provided the following blogs for students to access giving supporting information about the things he showed during his talk, which you may find of interest: http://community.arm.com/groups/arm-mali-graphics/blog/2014/03/10/introduction, http://community.arm.com/groups/arm-mali-graphics/blog/2014/03/27/introducing-pyopencl, and http://community.arm.com/groups/arm-mali-graphics/blog/2014/04/23/arm-mali-compute-architecture-fundamentals .

Schedule

Lectures: Mondays 13:00-14:00 (Kilburn 2.15); Tuesdays 10:00-11:00 (LF15) – Starting from 30^th January, some of the Tuesday lecture slots will be used for supervised lab sessions in the 3^rd Year Lab (room 1.8 in the Kilburn Building).
Labs: Tuesdays 10:00-11:00 (Third Year Laboratory – Kilburn 1.8) – Starting from 21^st February. There will be a demonstrator in attendance at each scheduled lab session. There are three exercises, each scheduled for two one-hour sessions, with a deadline for submission a week after the final session is scheduled (taking into account the Easter break). Note: the lab exercise deadlines will be adhered to. Marks for reports submitted after the deadline will be subject to ARCADE’s usual rules.

1^stlab: Tuesday 21^st February and Tuesday 28^th February – deadline 10:00 Tuesday 7^th March - here is the labsheet. An example script for running a job using the OGE batch queue is available here; this example runs the Java class file called Demo using a range of numbers of cores ranging from 1 to 32 cores.
2^nd lab: Tuesday 14^th March and Tuesday 21^st March – deadline 10:00 Tuesday 28^th March - here is the labsheet.
3^rd lab: Tuesday 28^th March and Tuesday 25^th April – deadline 10:00 Tuesday 2^nd May - here is the labsheet.

Handouts

Lecture 1 – 13:00 Monday 30^th January: Introduction
Lecture 2 – 10:00 Tuesday 31^st January: The World of Parallelism
Lecture 3 – 13:00 Monday 6^th February: Parallel Programming
Lecture 4 – 10:00 Tuesday 7^th February: Shared Memory Multi-Processors
Lecture 5 – 13:00 Monday 13^th February: Other Protocols
Lecture 6 – 10:00 Tuesday 14^th February: Programming with Barriers and Locks
Lecture 7 – 13:00 Monday 20^th February : Hardware Support for Synchronisation
Lecture 8 – 13:00 Monday 27^th February: Lockfree Data Structures
Lecture 9 – 13:00 Monday 6^th March: OpenMP and MPI
Lecture 10 – 10:00 Tuesday 7^th March: Speculation
Lecture 11 – 13:00 Monday 13^th March: Transactional Memory
Lecture 12 – 13:00 Monday 20^th March: Hardware Support for Transactions
Lecture 13 – 13:00 Monday 27^th March: Memory Consistency
Lecture 14 – 13:00 Monday 24^th April: GPUs and CUDA Slides for Anton Lokhmotov’s old guest lecture are available here.
Lecture 15 – 13:00 Tuesday 2^nd May: Functional Programming and Dataflow Principles

There is also a reading list.

Course Lecturer: Graham Riley – room IT412.