Skip to content

Instantly share code, notes, and snippets.

@pozorvlak
Last active December 14, 2015 20:59
Show Gist options
  • Save pozorvlak/5148078 to your computer and use it in GitHub Desktop.
Save pozorvlak/5148078 to your computer and use it in GitHub Desktop.
HiPEAC '13 Berlin conference report

What I saw on my holidays

by Miles, age 32¾

The giant fishtank in the Radisson Blue

General themes

Statue of a woman standing on her head

  • Big computations aren't just array-munging any more
  • Profile-driven compilation: still really hard
  • Heterogeneous multicore programming: kill me now
  • FPGAs: buy our toolchain and we can make it less hateful, honest

Keynote: Wen-Mei Hwu

The Reichstag debating chamber from above

  • GPUs rock for HPC
  • high-level tools = high overhead
  • we badly need scalable GPU libraries
  • must be done by hand - nothing is affine any more!

GPUVerify - Alastair Donaldson, Imperial

The fishtank from underneath

  • prove CUDA/OpenCL code free of data races
  • compiles CUDA to Boogie IR
  • Boogie invokes an SMT solver
  • doesn't work if results used as indices

Understanding Fundamental Design Choices in Single-ISA Heterogeneous Multi-Core Architectures - Van Craeynest et al, U. Ghent

Accuracy of STP prediction

  • Exhaustive design-space exploration with analytical model (MPPM)

Understanding Fundamental Design Choices in Single-ISA Heterogeneous Multi-Core Architectures - Van Craeynest et al, U. Ghent

Pareto frontier

  • Identification of Pareto frontier
  • It's mostly made up of heterogeneous designs

Profile-guided Floating- to Fixed-point Conversion for Hybrid FPGA-Processor Applications - Doris Chen, Altera Corporation

Big ball of string sculpture

  • LLVM plugin
  • decimal point alignment = shift = just wires in FPGA
  • 2x-3x reduction in logic on some benchmarks

Keynote: The Golden Age of Architecture May Be Now - Doug Burger, Microsoft

The cathedral of St Hedwig

  • Go big or go home
  • power limits mean multicore is over
  • specialisation is the new hotness
  • "direct" architecture work, eg Kinect, vs "indirect" (traditional)
  • we won't know if we're irrelevant - ten-year pipeline!

Keynote: Chip-level Reliability - Michel Dubois, USC

A scuba-diver cleaning a fishtank

  • to simulate error correction, must inject errors at many times true rate
  • hence our results are full of lies
  • by 2016, all single event upsets will flip >1 bit
  • block-level SECDED is overkill
  • single-bit parity checks are not-enough kill

Workshop: the Sniper multicore simulator

A frozen washbasin

  • world's most trivial JIT
  • otherwise much like ArcSim
  • totally awesome benchmarking/visualisation toolchain
  • we should steal it

Thanks for listening!

You Are Leaving The American Sector

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment