MCNLA

Monte Carlo Numerical Linear Algebra Package

Information

Git

• https://github.com/wlab-pro/mcnla

Documentation

• http://wlab-pro.github.io/mcnla

Author

• Mu Yang <[email protected]>

Tutorial

• @ref tutorial_main

Requirements

• CMake 2.8+ (CMake 3+ recommended).
• C++ compiler with C++11 support (GCC 5+ recommended).
• BLAS & LAPACK library.
• OpenMP Library.
• MPI Library (MPICH or OpenMPI).

Optional

• Intel® Math Kernel Library (Recommended for better performance, used for BLAS & LAPACK).

• DOxygen (Used for documentation).
• Google Test Library (Used for code testing).

Installation

Please use the following commands to create Makefiles

mkdir build
cd build
cmake <path-to-source>

Options

Use the following command to set options

ccmake <path-to-source>

• Set 'MCNLA_BLAS' to select BLAS/LAPACK library. [BLAS/MKL]

• Set 'MCNLA_BUILD_DEMO' to build demo codes.

• Set 'MCNLA_BUILD_DOC' to build documentation.

• Set 'MCNLA_BUILD_DRIVER' to build driver codes.

• Set 'MCNLA_BUILD_REPORT' to build report codes.

• Set 'MCNLA_BUILD_TEST' to build test codes.

• Set 'MCNLA_INSTALL_DEMO' to install demos.

• Set 'MCNLA_INT_SIZE' to select system integer size. [32/64]

• Set 'MCNLA_OMP' to select OpenMP library. [OFF/GOMP/IOMP] (Require 'MCNLA_BLAS = MKL')

• Set MCNLA_USE_GPU to enable GPU support.

• Set 'GTEST_ROOT' for the root path of Google Test. (Require 'MCNLA_BUILD_TEST')

• Set 'INTEL_ROOT' for the root path of Intel libraries. (Require 'MCNLA_BLAS = MKL')

• Set 'MKL_ROOT' for the root path of Intel MKL. (Require 'MCNLA_BLAS = MKL')

• Set 'MPI_PROCS' to set the number of MPI processes used in demo codes.

Makefile

Command	Detail
make all	build all binaries
make install	install package
make run	run demo code
make check	build and run test codes
make doc	build documentation
make help	display make-rules

Definitions

• Define MCNLA_USE_ILP64 to use 64-bit integer.
• Define MCNLA_USE_MKL to enable Intel MKL.
• Define MCNLA_USE_GPU to enable GPU support.
• Define MCNLA_DISABLE_LITERAL to disable literal definition in global namespace.

Q&A

How to set CMake options?

• Use ccmake .. or ccmake <path-to-source> in the build folder.
• Press <Enter> on the option you want to change, change it, and press <Enter> again.
• After changing the options, press <c> to configure, and press <g> to generate Makefiles if configured successfully.
• To quit without saving, press <q>.

Why Intel MKL is not found?

• Source mklvars.sh in the bin folder of your Intel MKL to set the environment variables.

Why Google Test is not found?

• Set GTEST_ROOT to a folder containing include and lib of Google Test.

How to use 64-bit integer?

• Set MCNLA_USE_ILP64 with ccmake.
• If you want to compile directly, add -DMCNLA_USE_ILP64 to compiler flag.
• Make sure your LAPACK&BLAS / Intel MKL uses 64bit integer. Make sure you uses the correct library and flags of Intel MKL.

std::bad_alloc occurs. What should I do?

• Possible issues:
  • Not enough memory. You might request a vector/matrix which size exceed your memory size.
  • Index overflow. Use 64-bit integer instead of 32-bit integer. For example, if you request a 1e5 x 1e5 matrix, there will be 1e10 elements, which is larger than the maximum number in 32-bit integer (about 2.1e9). Using 64-bit integer may solve this problem.

Reference

• Ting-Li Chen, Dawei D. Chang, Su-Yun Huang, Hung Chen, Chienyao Lin, Weichung Wang, “Integrating Multiple Random Sketches for Singular Value Decomposition”

License

@include LICENSE.md