reading-group.md

title	layout
HIPERFIT Reading group	page

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)

• April 7th: Type systems for array languages (Discussion lead by Martin Elsman and Martin Dybdal)
  • "QUBE - Dependently typed array programs don't go wrong" (JLAP, 2009) - http://www.sciencedirect.com/science/article/pii/S1567832609000411
  • "QUBE - Array Programming With Dependent Types" (Ph.D. dissertation) - http://www.students.informatik.uni-luebeck.de/zhb/ediss1099.pdf
• April 14th: Type systems for array languages (Discussion lead by Danil Annenkov)
  • "From Contracts Towards Dependent Types: Proofs by Partial Evaluation" - http://link.springer.com/chapter/10.1007%2F978-3-540-85373-2_15
• April 21st: Type systems for array languages (Discussion lead by Martin Dybdal)
  • "An Array-oriented language with Static Rank Polymorphism" - http://dx.doi.org/10.1007/978-3-642-54833-8_3
• April 28th: Type systems for array languages
  • "Dependent Types in Practical Programming" - http://www.cs.bu.edu/~hwxi/academic/papers/popl99.pdf
• May 5th: Program optimisation (Discussion lead by Troels Henriksen)
  • "A framework for enhancing data reuse via associative reordering" - http://dl.acm.org/citation.cfm?id=2594342
• May 12th: Program optimisation (Discussion lead by Frederik)
  • "Optimising Purely Functional GPU Programs" (skip Section 3) - http://www.cse.unsw.edu.au/~tmcdonell/papers/acc-optim-icfp2013.pdf
  • "Embedding Foreign Code" - https://www.cse.unsw.edu.au/~chak/papers/acc-ffi.pdf
  • Skim: "High-Performance Code Generation for Stencil Computations on GPU Architectures" http://www.cs.ucla.edu/~pouchet/doc/ics-article.12.pdf
• May 19th
  • Cancelled
• May 26th: Program optimisation
  • "Automatic Parallelization in the Polytope Model" http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.39.3581
• June 2nd: Program optimisation (Discussion lead by Martin Elsman)
  • "Region-based memory management for GPU programming languages: enabling rich data structures on a spartan host" http://dl.acm.org/citation.cfm?id=2660244
• August 25th: Programming language implementation
  • SaC JFP paper - http://dl.acm.org/citation.cfm?id=967501
• September 15th: Finance / Automatic Differentiation
  • Esben Bistrup Halvorsen. "Calculating Key Ratios for Financial Products using Automatic Differentiation and Monte Carlo Simulation" - http://hiperfit.dk/pdf/ad_esben.pdf - Martin Elsman
  • White-paper: Structured products desks join the AAD revolution - http://www.fincad.com/sites/default/files/wysiwyg/Structured%20products%20desks%20join%20the%20AAD%20revolution.pdf
• September 22th: Programming language implementation
  • Draft-paper: "A Haskell EDSL for Nested Data-parallel Design-space exploration on GPUs" http://www.cse.chalmers.se/edu/year/2014/course/DAT280_Parallel_Functional_Programming/exploration-draft-Obsidian.pdf
  • And their own reflections: http://www.cse.chalmers.se/edu/course/DAT280_Parallel_Functional_Programming/LectureGPU115.pdf
• October 13th: Programming language implementation
  • A provable time and space efficient implementation of NESL
• October 20th: Program transformation
  • Automatically Improving Accuracy for Floating Point Expressions
• October 27th: Finance / Automatic Differentiation
  • "Valuing American Options by Simulation: A Simple Least Approach" (Longstaff and Schwartz): https://people.math.ethz.ch/~hjfurrer/teaching/LongstaffSchwartzAmericanOptionsLeastSquareMonteCarlo.pdf
• November 10th: Programming language implementation (kind of)
  • Supporting Extended Precision on Graphics Processors
• November 24th:
  • Cancelled
• December 1st:
  • Cancelled
• December 17th: Cost models
  • D. B. Skillicorn , W Cai. A cost calculus for parallel functional programming

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