HIPERFIT

HIPERFIT Reading group

Description

This page documents the activities of the HIPERFIT reading group, which was initiated in April 2015. We meet on tuesdays from 13:00 to 15:00 (approx.) following the schedule below. The schedule will be updated continuously.

Time usage

The reading group will run for 16 times, with an expected time usage around 8 hours for each session (including preparation), which will total 128 hours, which should correspond to 5 ECTS.

Topics

We will be reading papers in several topics in the core area of HIPERFIT, namely functional programming languages for financial IT.

  • Programming language design for parallel programming
  • Type systems for array based languages (e.g. dependent types)
  • high-performance computing optimizations and algorithms.
  • Automatic/algorithmic differentiation
  • Finance and monte-carlo simulation

Participants

  • Troels Henriksen (Ph.D. student)
  • Danil Annenkov (Ph.D. student)
  • Frederik Meisner Madsen (Ph.D. student)
  • Martin Dybdal (Ph.D. student)
  • Andrzej Filinski (Supervisor)
  • Martin Elsman (Supervisor)
  • Cosmin Oancea (Supervisor)

Other HIPERFIT faculty and students are welcome to participate in the meetings on voluntary basis.

##Schedule (2015)

Suggestions for future sessions

If you add a topic, please include your name

Finance, automatic differentiation, Monte-Carlo simulations etc.

  • “Parallel random numbers: as easy as 1, 2, 3” - http://dx.doi.org/10.1145/2063384.2063405 - Martin Dybdal

  • Andreas Griewank, Andrea Walther, “Evaluating Derivatives: Principles and Techniques of Algorithmic Differentiation” – det er en bog; jeg har den på mit kontor (fra KUBIS). - Fritz Henglein

  • (“Specialising Simulator Generators for High-Performance Monte-Carlo Methods” - http://dx.doi.org/10.1007/978-3-540-77442-6_9 - Martin Dybdal)

High-performance computing

  • “Rodinia: A benchmark suite for heterogeneous computing” - http://dx.doi.org/10.1109/IISWC.2009.5306797 - Martin Dybdal

  • “A characterization of the Rodinia benchmark suite with comparison to contemporary CMP workloads” - http://dx.doi.org/10.1109/IISWC.2010.5650274 - Martin Dybdal

  • “Software Pipelined Execution of Stream Programs on GPUs” - http://dx.doi.org/10.1109/CGO.2009.20 - Martin Dybdal

  • Obsidian

Optimisation

Trying to stick to something that has general value, rather than one-off hacks - Troels

  • “Effective Automatic Data Allocation for Parallelization of Affine Loop Nests” - http://www.csa.iisc.ernet.in/TR/2014/2/tech-report.pdf - Troels Henriksen

  • “Metrics and Models for Reordering Transformations” - http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.83.6624 - Troels Henriksen

  • “Automatic Generation of Peephole Superoptimizers” - http://embedded.cse.iitd.ac.in/~sbansal/pubs/asplos06.pdf (maybe also Massalin, 1987) - Martin Dybdal

Cost models for data-parallel languages

  • J. Hoffmann and Z. Shao. Automatic static cost analysis for parallel programs. ESOP 2015. - http://www.cs.cmu.edu/~janh/papers/parallelcost2014.pdf - Danil Annenkov
  • D. B. Skillicorn , W Cai. A cost calculus for parallel functional programming, 1995. - http://www.sciencedirect.com/science/article/pii/S0743731585710891 - Danil Annenkov
  • C. Barry Jay. Costing Parallel Programs as a Function of Shapes, 1998. - http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.48.8225 - Danil Annenkov

Other topics:

  • Flattening, NESL, VCODE
  • CUDA PTX
  • Something on the architecture of Xeon Phi? - Martin Dybdal
  • Hierarchical cost-models - Martin Dybdal
  • “Fission” / Bohrium

Decided not to read

  • “pocl: A Performance-Portable OpenCL Implementation” - http://link.springer.com/article/10.1007/s10766-014-0320-y