Heterogeneous Computing

By John Urbanic

Pittsburgh Supercomputing Center

Published on

Abstract


Bio

John Urbanic is Parallel Computing Specialist at the Pittsburgh Supercomputing Center where he spends as much time as possible implementing extremely scalable code on interesting machines. These days that means a lot of MPI, OpenMP and OpenACC. John also enjoys Big Data challenges which involve such things as graph analytics, SPARQL and novel architectures. John graduated with Physics degrees from Carnegie Mellon University (BS) and Pennsylvania State University (MS) and still appreciates working on applications that simulate physical phenomena.

John teaches workshops and classes on the above as much as possible and particularly likes meeting others with challenging new problems for HPC, so please introduce yourself and let him know what you are working on.

(Source)

Cite this work

Researchers should cite this work as follows:

  • John Urbanic (2013), "Heterogeneous Computing," http://nanohub.org/resources/19354.

    BibTex | EndNote

Time

Location

New York University, New York, NY

Submitter

NanoBio Node

University of Illinois at Urbana-Champaign

Tags

Heterogeneous Computing
by: John Urbanic
  • Heterogeous COmputing 1. Heterogeous COmputing 0
    00:00/00:00
  • WHat is Heterogeneous Computing? 2. WHat is Heterogeneous Computin… 20.578328588176184
    00:00/00:00
  • Objectives 3. Objectives 52.294885578252192
    00:00/00:00
  • Things started looking harder around 29004... 4. Things started looking harder … 121.64245123636923
    00:00/00:00
  • One popular culprit at the time: CPU vs. DRAM 5. One popular culprit at the tim… 179.11204629142557
    00:00/00:00
  • THe problem in 2004...didn't get better. 6. THe problem in 2004...didn't g… 215.68482397310535
    00:00/00:00
  • Moore's Law is not at all dead 7. Moore's Law is not at all dead 451.4539155958293
    00:00/00:00
  • Moore's Law at Intel 1970-2010 8. Moore's Law at Intel 1970-2010 540.75943275482518
    00:00/00:00
  • Not a new problem, just a new scale... 9. Not a new problem, just a new … 562.170019966211
    00:00/00:00
  • Basic Energy Consumption Parmeters 10. Basic Energy Consumption Parme… 634.86496356593113
    00:00/00:00
  • How to get same number of transistors to give us more performance without cranking up power? 11. How to get same number of tran… 676.675608798785
    00:00/00:00
  • Example: Dual core with voltage scaling 12. Example: Dual core with voltag… 746.40212290312127
    00:00/00:00
  • Top 10 Systems in November 2012 (Just got updated last week) 13. Top 10 Systems in November 201… 801.72859605112717
    00:00/00:00
  • Design Choices of the Leaders 14. Design Choices of the Leaders 834.38132050034983
    00:00/00:00
  • Titan System at Oak Ridge National Laboratory 15. Titan System at Oak Ridge Nati… 943.897421457704
    00:00/00:00
  • Most Power Efficient Architectures 16. Most Power Efficient Architect… 972.95013566784417
    00:00/00:00
  • Power Efficiency over Time 17. Power Efficiency over Time 1037.7503199713306
    00:00/00:00
  • Projected Performance Development 18. Projected Performance Developm… 1078.6135940886363
    00:00/00:00
  • Trends with ends. 19. Trends with ends. 1113.9821157357633
    00:00/00:00
  • Q.E.D. 20. Q.E.D. 1204.0455297871977
    00:00/00:00
  • Some truly interesting things we will deisregard for today... 21. Some truly interesting things … 1251.7930340108194
    00:00/00:00
  • to focus on the current pioritites of HPC 22. to focus on the current piorit… 1376.9723374119014
    00:00/00:00
  • First parallel Weather Modeling algorithm Richardson in 1917 23. First parallel Weather Modelin… 1411.014539497261
    00:00/00:00
  • Prototypical Application: Serial Weather Model 24. Prototypical Application: Seri… 1485.6673834024471
    00:00/00:00
  • Weather Model: Shared Memory 25. Weather Model: Shared Memory 1506.7621802419835
    00:00/00:00
  • Weather Model: Distributed Memory 26. Weather Model: Distributed Mem… 1545.2886056076043
    00:00/00:00
  • Weather Model: Accelerator 27. Weather Model: Accelerator 1605.036143961501
    00:00/00:00
  • GPU Architecture: Two Main Components 28. GPU Architecture: Two Main Com… 1722.2575299919795
    00:00/00:00
  • GPU Architecture - Kepler: Streaming Multiprocessor (SMX) 29. GPU Architecture - Kepler: Str… 1765.64712708408
    00:00/00:00
  • GPU ARchitectuer - Kepler: CUDA Core 30. GPU ARchitectuer - Kepler: CUD… 1803.7946040034813
    00:00/00:00
  • GPU Architecture - Kepler Streaming MUltiprocessor (SMX) 31. GPU Architecture - Kepler Stre… 1830.3555518338637
    00:00/00:00
  • GPU ARchitectuer - Kepler: CUDA Core 32. GPU ARchitectuer - Kepler: CUD… 1926.3844263553985
    00:00/00:00
  • Programming NVIDIA GPUs: OpenACC and CUDA 33. Programming NVIDIA GPUs: OpenA… 1933.7739210566735
    00:00/00:00
  • Anatomy of a CUDA Application 34. Anatomy of a CUDA Application 2131.5846232352442
    00:00/00:00
  • Threads in CUDA: fine control over the hardware 35. Threads in CUDA: fine control … 2195.6252472367655
    00:00/00:00
  • Threads in OpenACC 36. Threads in OpenACC 2244.5072575218687
    00:00/00:00
  • Intel's MIC Approach 37. Intel's MIC Approach 2413.3016866158869
    00:00/00:00
  • What is MIC? 38. What is MIC? 2468.9120658135284
    00:00/00:00
  • MIC Architecture 39. MIC Architecture 2594.0676928310231
    00:00/00:00
  • Kinghts Corner Core 40. Kinghts Corner Core 2673.5179845553043
    00:00/00:00
  • Vector Processing Unit 41. Vector Processing Unit 2697.3453950730859
    00:00/00:00
  • Current Implementation: Xeon Phi and Stampede 42. Current Implementation: Xeon P… 2757.4238630961813
    00:00/00:00
  • Programming MIC 43. Programming MIC 2812.4149184149187
    00:00/00:00
  • OpenMP 4.0 on Phi 44. OpenMP 4.0 on Phi 2934.84324956166
    00:00/00:00
  • Scaling up... 45. Scaling up... 3014.9557547715444
    00:00/00:00
  • Amdahl's Law 46. Amdahl's Law 3057.921951654625
    00:00/00:00
  • Going Hostless 47. Going Hostless 3255.076560554864
    00:00/00:00
  • Some things we did not mention 48. Some things we did not mention 3340.4369349005424
    00:00/00:00
  • Summary 49. Summary 3537.3525229755505
    00:00/00:00