Gennady Fedorov - Technical Consulting Engineer Intel Architecture, Graphics and Software (IAGS) LRZ workshop, June 2020
[email protected] Agenda • Introduction • MKL usage modes, tips • Known problems & Deprecations Copyright © 2020, Intel Corporation. All rights reserved. Intel IPP Overview 2 *Other names and brands may be claimed as the property of others. Intel® Math Kernel Library Linear Algebra FFTs Neural Networks Vector RNGs • BLAS • Multidimensional • Convolution • Congruential • LAPACK • FFTW interfaces • Pooling • Wichmann-Hill • ScaLAPACK • Cluster FFT • Normalization • Mersenne Twister • Sparse BLAS • Sobol • ReLU • Iterative sparse solvers • Neiderreiter • Inner Product • PARDISO* • Non-deterministic • Cluster Sparse Solver Summary Statistics Vector Math And More Benchmarks • Kurtosis • Trigonometric • Splines • Intel(R) Distribution for LINPACK* Benchmark • Variation coefficient • Hyperbolic • Interpolation • Exponential • High Performance • Order statistics • Trust Region • Log Computing Linpack • Min/max • Fast Poisson Solver • Power Benchmark • Variance-covariance • Root • High Performance Conjugate radient Benchmark Intel® Architecture Platforms Operating System: Windows*, Linux*, MacOS1* 1 Available only in Intel® Parallel Studio Composer Edition. Copyright © 2020, Intel Corporation. All rights reserved. 3 *Other names and brands may be claimed as the property of others. What’s New for Intel® MKL v.2019? • Just-In-Time Fast Small Matrix Multiplication : Improved speed of S/DGEMM for Intel® AVX2 and Intel® AVX-512 with JIT capabilities • CNR mode: independent of the number of threads (strict mode, BLAS) • New sparseQR Solvers : for sparse linear systems, sparse linear least squares problems, eigenvalue problems, rank and null-space determination, and others • Generate Random Numbers for Multinomial Experiments Highly optimized multinomial random number generator Great for finance, geological and biological applications Copyright © 2020, Intel Corporation.