jingchangshi/m2-genial-hpc

m2-genial-hpc

Resources pour le cours d'introduction au HPC / calcul haute performance du master Génie Informatique en Alternance GENIAL

biblio

Scientific Programming and Computer Architecture

• Scientific Programming and Computer Architecture book by Divakar Viswanath
• https://github.com/divakarvi/bk-spca : source associated to the book
• Definition of latency oriented architecture
• Seven Dwarfs of HPC
• CSCI-5576/4576: High Performance Scientific Computing
• Mark Horowitz talk at ISSCC_2014: Computing's energy problem
• Introduction to High-Performance Scientific Computing, book and slides by Victor Eijkhout
• San Diego Summer institute
• Finnish CSC summer school
• Computational Physics book by K. N. Anagnostopoulos
• Modern computer architecture slides, see e.g. slides Intro_Architecture.pdf
• Structure and Interpretation of Computer Programs

CUDA / GPU training

• NVIDIA's latest CUDA programming guide
• Julich training on CUDA
• Oxford training on CUDA
• Swiss CSCS summer school
• Amina Guermouche (Telecom Paris)
• EPCC, Univ Edinburgh, GPU training
• ARCHER GPU course
• Univ Luxembourg HPC
• SC19 Introduction to GPU programming with CUDA
• https://codingbyexample.com/category/cuda/
• http://turing.une.edu.au/~cosc330/lectures/display_notes.php?lecture=18
• https://www.nersc.gov/users/training/gpus-for-science/
• https://dl.acm.org/citation.cfm?id=3318192
• git@bitbucket.org:hwuligans/gputeachingkit-labs.git
• http://syllabus.gputeachingkit.com/
• udemy/cuda-programming-masterclass
• SDL2 Graphics User Interface : https://github.com/rogerallen/smandelbrotr
• mgbench : a multi-GPU benchmark
• performance analysis : parallelforall blog on Nsight
• misc : convert CUDA to portable C++ for AMD GPU
• List of Nvidia GPUs
• https://github.com/ashokyannam/GPU_Acceleration_Using_CUDA_C_CPP
• https://github.com/karlrupp/cpu-gpu-mic-comparison
• https://perso.centrale-marseille.fr/~gchiavassa/visible/HPC/01%20-%20GR%20%20Intro%20to%20GPU%20programming%20V2%20OpenACC%20.pdf

CUDA / performance analysis

• https://devblogs.nvidia.com/using-nsight-compute-to-inspect-your-kernels/
• https://www.olcf.ornl.gov/wp-content/uploads/2019/08/NVIDIA-Profilers.pdf
• http://on-demand.gputechconf.com/gtc/2017/presentation/s7445-jakob-progsch-what-the-profiler-is-telling-you.pdf
• monitoring performance : https://github.com/NERSC/timemory
• roofline model

Other CUDA resources

• C++ wrapper library
• template CMake project for CUDA

CUDA / python

• Numba // recommended numba tutorial for GPU programming
• CuPy
• pycuda
• python / C++ CUDA interface (SWIG and Cython)
• python / C++ CUDA interface with pybind11
• legate
• PythonHPC
• HPC Python video's
• Hands-On GPU Programming with Python and CUDA and examples
• 2020-geilo-gpu-python
• Numba introduction

Machine learning and Deep Learning

• https://towardsdatascience.com/fast-data-augmentation-in-pytorch-using-nvidia-dali-68f5432e1f5f
• https://ep2019.europython.eu/media/conference/slides/fX8dJsD-distributed-multi-gpu-computing-with-dask-cupy-and-rapids.pdf
• https://github.com/NVIDIA/DeepLearningExamples
• https://github.com/chagaz/hpc-ai-ml-2019
• tensorflow tutorial
• AI cheatsheet
• m2dsupsdlclass

Graphics / GPU

• https://raytracing.github.io/
• https://github.com/RayTracing/raytracing.github.io
• https://github.com/rogerallen/raytracinginoneweekendincuda : code très clean, super

OpenMP

• https://www.openmp.org/wp-content/uploads/openmp-examples-4.5.0.pdf
• https://github.com/OpenMP/Examples/tree/v4.5.0/sources
• https://ukopenmpusers.co.uk/wp-content/uploads/uk-openmp-users-2018-OpenMP45Tutorial_new.pdf
• https://www.nas.nasa.gov/hecc/assets/pdf/training/OpenMP4.5_3-20-19.pdf
• http://www.admin-magazine.com/HPC/Articles/OpenMP-Coding-Habits-and-GPUs?utm_source=AMEP

OpenMP target

• How to build yourself clang with OpenMP target support for Nvidia GPUs
  • https://hpc-wiki.info/hpc/Building_LLVM/Clang_with_OpenMP_Offloading_to_NVIDIA_GPUs
  • //devmesh.intel.com/blog/724749/how-to-build-and-run-your-modern-parallel-code-in-c-17-and-openmp-4-5-library-on-nvidia-gpus
• https://www.openmp.org/wp-content/uploads/SC17-OpenMPBooth_jlarkin.pdf
• OpenMP 5.0 for accelerators at GTC 2019
• LLVM/Clang based compiler for both AMD/NVidia GPUs
• OpenMP target examples

How to build clang++ with openmp target (off-loading) support ?

• https://devmesh.intel.com/blog/724749/how-to-build-and-run-your-modern-parallel-code-in-c-17-and-openmp-4-5-library-on-nvidia-gpus
• https://hpc-wiki.info/hpc/Building_LLVM/Clang_with_OpenMP_Offloading_to_NVIDIA_GPUs

OpenACC

• OpenACC Programming and Best Practices Guide
• PGI compiler - OpenACC getting started guide
• https://www.fz-juelich.de/SharedDocs/Downloads/IAS/JSC/EN/slides/openacc/2-openacc-introduction.pdf?__blob=publicationFile
• Introduction to GPU programming using OpenACC
• https://github.com/eth-cscs/SummerSchool2019/tree/master/topics/openacc
• https://developer.nvidia.com/openacc-overview-course
• https://perso.centrale-marseille.fr/~gchiavassa/visible/HPC/01%20-%20GR%20%20Intro%20to%20GPU%20programming%20V2%20OpenACC%20.pdf
• Jeff Larkin (Nvidia) Introduction to OpenACC
• Jeff Larkin (Nvidia) OpenACC data management
• Jeff Larkin (Nvidia) OpenACC optimizations
• OpenAcc training material as notebooks
• https://www.pgroup.com/resources/docs/19.10/pdf/pgi19proftut.pdf
• https://github.com/OpenACCUserGroup/openacc_concept_strategies_book
• https://developer.nvidia.com/blog/solar-storm-modeling-gpu-openacc/

Which compiler with OpenAcc support ?

• Nvidia/PGI compiler is the oldest and probably more mature OpenACC compiler.
• GNU/gcc provided by Spack is the easiest way to get started for OpenMP/OpenACC offload with the GNU compiler.

C++17 and parallel STL for CPU/GPU

• accelerating-standard-c-with-gpus-using-stdpar/ for Nivia GPUs
• a real life example in CFD: LULESH
• another reference in CFD stdpar for Lattice Boltzmann simulation and its companion code
• https://github.com/shwina/stdpar-cython/
• https://software.intel.com/content/www/us/en/develop/articles/get-started-with-parallel-stl.html

Which compiler ?

• Nvidia/PGI compiler for Nvidia GPUs
• GNU g++ version >= 9.1 (+ TBB) for multicore CPUs
• clang >= 10.0.1 for multicore CPUs
• Intel OneApi HPC Toolkit

SYCL

• Khronos
• syclacademy
• oneAPI-samples
• more oneAPI / SYCL samples
• a short tutorial
• Compilers / toolchain
  • codeplay
  • Intel OneAPI. If you want Nvidia GPU support, you'll have to rebuild llvm/clang from the source code, see instructions; OneAPI DPC++ actually is a SYCL implementation + extensions (Unified Shared Memory, Explicit SIMD, ...)
  • triSYCL for Xilinx FPGA target
• Comparison Kokkos/SYCL (early 2020)

Books on GPU programming / recommended reading

• The CUDA Handbook: A Comprehensive Guide to GPU Programming, by Nicholas Wilt, Pearson Education.
• CUDA by example, by Sanders and Kandrot, Addison-Wesley, 2010. Also available in pdf
• Learn CUDA programming by B. Sharma and J. Han, Packt Publishing, 2019
• Python + CUDA : https://github.com/PacktPublishing/Hands-On-GPU-Programming-with-Python-and-CUDA
• https://www.oreilly.com/library/view/hands-on-gpu-programming/9781788993913/ by Brian Tuomanen

C++ resources

• Discovering Modern C++: An Intensive Course for Scientists, Engineers, and Programmers, and companion github website
• https://github.com/changkun/modern-cpp-tutorial
• https://github.com/eth-cscs/examples_cpp
• https://github.com/mandliya/algorithms_and_data_structures
• https://www.fz-juelich.de/SharedDocs/Downloads/IAS/JSC/EN/slides/cplusplus/cplusplus.pdf?__blob=publicationFile
• https://gitlab.maisondelasimulation.fr/tpadiole/hpcpp
• http://www.cppstdlib.com/
• http://101.lv/learn/C++/
• https://github.com/caveofprogramming/advanced-cplusplus
• https://en.cppreference.com/w/
• list of Lists of C++ related resources: https://github.com/fffaraz/awesome-cpp
• list of books on C++ : https://github.com/fffaraz/awesome-cpp/blob/master/books.md
• C++ idioms
• Design Patterns and Book on design patterns for modern c++
• Julich training on C++
• CSCS computing center training on C++ videos
• CppCon and videos on YouTube
• Bo Qiang YouTube channel on C++11
• https://github.com/TheAlgorithms/C-Plus-Plus
• cours de C++ de l'université de Strasbourg

high-level C++ libraries for programming GPUs

Alternate programming models for programming modern computing architectures in a performance portable way:

• introduction to performance portability
• https://github.com/arrayfire/arrayfire
• https://docs.nvidia.com/cuda/thrust/index.html
• https://github.com/kokkos/kokkos
• https://github.com/LLNL/RAJA et https://github.com/LLNL/RAJA-tutorials
• https://github.com/triSYCL/triSYCL
• https://github.com/codeplaysoftware/computecpp-sdk

CMake

• cmake-cookbook and the book
• Modern CMake tutorial
• template CMake project for CUDA
• GPUs for science day

Git

• Git cheatsheet

Misc

• Udacity CS344 video archive
• cuda related : https://gist.github.com/allanmac/f91b67c112bcba98649d - cuda_assert
• FPGA, loop transformation, matrix multiplication
• Cycle du hype
• https://press3.mcs.anl.gov/atpesc/files/2019/08/ATPESC_2019_Dinner_Talk_8_8-7_Foster-Coding_the_Continuum.pdf

Shell and command line skills

• Learn/improve your skill on Linux’s command line/Bash
  e.g. http://swcarpentry.github.io/shell-novice/
• http://www.tldp.org/LDP/abs/html/
• http://www.epons.org/commandes-base-linux.php
• The art of command line

Blogs or newsletters on HPC

• https://www.nextplatform.com/
• subscribe blog/news letters on HPC; e.g. Admin-magazine / HPC
• (En anglais) Intel Parallel Universe Magazine

MOOC

• Amazon
• udemy

Projet

• Portage d'un code C++ de simulation des équations de Navier-Stokes par la méthode de Boltzmann sur réseau.