Fifth INFN International School on: "Architectures, tools and methodologies for developing efficient large scale scientific computing applications" ESC13 - Bertinoro (Forlì-Cesena) Italy 20-26 October 2013

20 Oct 2013, 16:00 → 26 Oct 2013, 15:00 Europe/Rome

Bertinoro

Exercises web page

Sunday 20 October

17:00 → 20:30 Registration and Welcome

20:30 → 22:00 Welcome Dinner

Monday 21 October

08:30 → 20:00 Session 1

08:30 Welcome and opening remarks

Speaker: Mauro Morandin (PD)

Slides welcome.pdf

09:00 Concepts of performance and efficiency

Speaker: Dr Peter Elmer (Princeton University)

Slides esc13-intro-perf.pdf

09:50 Modern processors and related optimisation topics - Part 1

Speaker: Mr Sverre Jarp (CERN)

Slides ESC_Lecture_Jarp_October2013_V2_presented.pdf

10:40 Coffee break

11:00 Modern processors and related optimisation topics - Part 2

Speaker: Mr Sverre Jarp (CERN)

11:50 Introduction to Performance tuning tools

Speaker: Dr Peter Elmer (Princeton University)

Exercises http://cern.ch/elmer/esc13

git repository for exercises https://github.com/pelmer/esc

12:40 Lunch break

14:15 Floating point computation: accuracy, optimization, vectorization (with exercises)

Speaker: Vincenzo Innocente (CERN)

Slides OptimalFloatingPoint2013.pdf

15:00 Floating point computation: accuracy, optimization, vectorization (with exercises)

Speaker: Vincenzo Innocente (CERN)

15:45 Coffee break

16:00 Floating point computation: accuracy, optimization, vectorization (with exercises)

Speaker: Vincenzo Innocente (CERN)

Vincenzo's git site https://github.com/VinInn/FPOptimization

16:45 Floating point computation: accuracy, optimization, vectorization (with exercises)

Speaker: Vincenzo Innocente (CERN)

17:30 Student lightning presentations

Slides MONDAY_21_OCTOBER.pdf

20:30 → 20:30 Dinner

Tuesday 22 October

08:30 → 20:00 Session 2

08:30 Efficient C++ coding (with exercises)

Speaker: Dr Francesco Giacomini (CNAF)

Slides efficient_c++.odp efficient_c++.pdf

09:20 Efficient C++ coding (with exercises)

Speaker: Dr Francesco Giacomini (CNAF)

10:10 Coffee break

10:30 Efficient C++ coding (with exercises)

Speaker: Dr Francesco Giacomini (CNAF)

11:15 Efficient C++ coding (with exercises)

Speaker: Dr Francesco Giacomini (CNAF)

12:40 Lunch break

14:15 The Memory Crisis

Speaker: Dr Peter Elmer (Princeton University)

Slides esc13-memory-lecture.pdf

15:00 How memory allocation works

Speaker: Dr Peter Elmer (Princeton University)

15:45 Coffee break

16:00 Exercises - Memory Allocations

Speaker: Dr Peter Elmer (Princeton University)

16:45 Exercises - Memory Allocations

Speaker: Dr Peter Elmer (Princeton University)

17:30 Student lightning presentations

Slides Tuesday_22_October.pdf

20:30 → 20:30 Dinner

Wednesday 23 October

08:30 → 14:00 Session 3

08:30 Exercises (Floating Point, Memory use, C++)

Speakers: Dr Francesco Giacomini (CNAF), Dr Peter Elmer (Princeton University), Vincenzo Innocente (CERN)

exercises (web site) http://elmer.web.cern.ch/elmer/esc13/

09:20 Exercises (Floating Point, Memory use, C++)

10:10 Coffee break

10:30 Exercises (Floating Point, Memory use, C++)

11:20 Exercises (Floating Point, Memory use, C++)

12:20 Lunch break

14:00 → 18:00 Session 4: Introduction to parallel computing

14:00 Motivation.... The power wall and the emergence of ubiquitous heterogeneous computing

Speaker: Dr Tim Mattson (Intel)

Slides 04_Intro_to_par_prog.pdf

14:45 Parallel Computing: basic concepts and vocabulary

Speaker: Dr Tim Mattson (Intel)

15:30 Coffee break

16:00 Parallel hardware: from SMP to GPU to clusters to massively parallel supercomputers

Speaker: Dr Tim Mattson (Intel)

16:45 Core design patterns of parallel algorithms

Speaker: Dr Tim Mattson (Intel)

18:00 → 20:00 Social Tour

20:30 → 22:30 Social dinner

Casa Artusi Restaurant

Thursday 24 October

08:30 → 12:50 Session 5: Hands on introduction to parallel programming with OpenMP

Convener: Dr Tim Mattson (Intel)

slides 05_hands_on_intro_openMP.pdf

08:30 Multithreaded programming with OpenMP: The SPMD pattern on the CPU

Speaker: Dr Tim Mattson (Intel)

09:30 Parallel loops with OpenMP

Speaker: Dr Tim Mattson (Intel)

10:30 Coffee break

11:00 The divide and conquer pattern with OpenMP tasks

Speaker: Dr Tim Mattson (Intel)

12:50 → 14:20 Lunch

14:20 → 19:30 Session 6: Hands on introduction to GPU programming with compiler directives

Convener: Dr Tim Mattson (Intel)

slides 06_hands_on_intro_gpu_programming_directives.pdf

14:20 GPU architectures

15:15 Core design patterns for the GPU programmer

Speaker: Dr Tim Mattson (Intel)

16:15 Coffee break

16:45 Programming GPUs with directives: OpenACC and OpenMP 4.0

Speaker: Dr Tim Mattson (Intel)

18:30 Evening lecture: Exploiting vector units

Speaker: Dr Tim Mattson (Intel)

20:30 → 20:30 Dinner

Friday 25 October

08:30 → 12:50 Session 7: Hands on introduction to GPU programming with CUDA and OpenCL

Convener: Dr Tim Mattson (Intel)

Introduction to CUDA 07b_hands_on_intro_ro_CUDA.pdf

Introduction to OpenCL 07_hands_on_intro_OpenCL.pdf

08:30 The kernel parallelism pattern

Speaker: Dr Tim Mattson (Intel)

09:30 Basics of kernel programming

Speaker: Dr Tim Mattson (Intel)

10:30 Coffee break

11:00 GPU memory hierarchy and reductions

Speaker: Dr Tim Mattson (Intel)

12:50 → 14:20 Lunch

14:20 → 18:30 Session 8: Hands on introduction to cluster computing

Convener: Dr Tim Mattson (Intel)

slides 08_hands_on_intro_MPI.pdf

14:30 MPI and the concept of message passing

Speaker: Dr Tim Mattson (Intel)

15:30 The SPMD pattern in MPI

Speaker: Dr Tim Mattson (Intel)

16:30 Coffee break

17:00 Programming highly scalable systems: "MPI+X"

Speaker: Dr Tim Mattson (Intel)

20:30 → 20:30 Dinner

Saturday 26 October

08:30 → 14:50 Session 9

Announcements annouoncements.pdf

08:30 Students feedback

09:00 Final examination

11:00 Coffee break

11:30 Delivery of certificates of attendance

12:00 Lunch

14:00 Shuttle departure (to Forli' railway station)