NDAK14008U CHANGED: Programming Massively Parallel Hardware (PMPH)
MSc Programme in Computer Science
MSc Programme in Bioinformatics
In simple words, the aim of the course is to teach students how
to write programs that run fast on highly-parallel hardware, such
as general-purpose graphics processing units (GPGPUs), which are
now mainstream. Such architectures are however capricious;
unlocking their power requires understanding their design
principles and also specialized knowledge of code transformations,
for example aimed at optimising locality of reference, the degree
of parallelism, etc. As such, this course is organized on three
tracks: hardware, software, and lab.
The Software Track teaches how to think parallel. We introduce the map-reduce functional programming model, which builds programs naturally, like puzzles, from a nested composition of implicitly-parallel array operators. We reason about the asymptotic (work and depth) properties of such programs, and discuss the flattening transformation, which converts (all) arbitrarily-nested parallelism to a more-restricted form that can be directly mapped to the hardware. We then turn our attention to legacy-sequential code written in programming languages such as C. In this context we study dependence analysis, as a tools for reasoning about loop-based optimizations (e.g., Is it safe to execute a given loop in parallel, or to interchange two loops?). As time permits, we may cover more advanced topics, for example related to dynamic analysis for optimising locality of reference.
The Hardware Track studies the design space of the critical components of parallel hardware: processor, memory hierarchy and interconnect networks. We will find out that modern hardware design is governed by old ideas, which are merely adjusted or combined in different ways.
The Lab Track applies the theory learned in the other tracks. We will review the fundamental ideas that govern the GPGPU design and potential performance bottlenecks. We will quickly learn several parallel-programming models, and we will get our hands dirty by putting in practice the optimizations learned in the software track. We will use (the in-house developed) Futhark to write nested-parallel programs, to demonstrate flattening, and as a baseline. We will use OpenMP and CUDA to write "parallel-assembly" code for multi-core and GPGPU execution, respectively.
- the types and semantics of data-parallel operators.
- analyses for identifying and optimising parallelism and locality of reference, e.g., flattening, dependence analysis.
- the main hardware-design techniques for supporting parallelism at processor, memory hierarchy and interconnect levels.
- implementing parallel programs in high-level (Futhark) and lower-level programming models (OpenMP, CUDA).
- applying (by hand) the flattening transformation on specific instances of data-parallel programs.
- testing, measuring the impact of applied optimizations, and characterizing the performance of parallel programs.
- reasoning about the work-depth asymptotic behavior of specific instances of data-parallel programs.
- reasoning based on dependence analysis about the (in)correctness of specific instances of loop parallelization and related optimizations.
- identifying an effective parallelization solution for a given application.
The topics taught in the hardware track are selected from the book "Parallel Computer Organization and Design'', by Michel Dubois, Murali Annavaram and Per Stenstrom, Cambridge University Press, lates edition
Lecture notes covering the material on the software track will be provided on Absalon. Various other related material, such as scientific articles and tutorials (e.g., Futhark, CUDA) will be pointed out from the course pages.
- Project work
PhD’s can register for MSc-course by following the same procedure as credit-students, see link above.
- 7,5 ECTS
- Type of assessment
- Continuous assessmentFour individual assignments (40%), group project (report) with individual presentation and short oral examination (60%). No aids are allowed for the oral examination.
- All aids allowed
- Marking scale
- 7-point grading scale
- Censorship form
- No external censorship
Several internal examiners
Resubmission of the assignments (35%) and the project extended with additional tasks (40%), and a 30 minutes oral examination (25%) without preparation. No aids are allowed for the oral examination. Already passed assignments/report will be considered.
Project Work 68
Criteria for exam assesment
See Learning Outcome.