NVIDIA / cnmem

A simple memory manager for CUDA designed to help Deep Learning frameworks manage memory

223 Stars 66 Forks Last release: Not found Other 21 Commits 1 Releases

Available items

No Items, yet!

The developer of this repository has not created any items for sale yet. Need a bug fixed? Help with integration? A different license? Create a request here:

CNMeM Library

Simple library to help the Deep Learning frameworks manage CUDA memory.

CNMeM is not intended to be a general purpose memory management library. It was designed as a simple tool for applications which work on a limited number of large memory buffers.

CNMeM is mostly developed on Ubuntu Linux. It should support other operating systems as well. If you encounter an issue with the library on other operating systems, please submit a bug (or a fix).


CNMeM relies on the CUDA toolkit. It uses C++ STL and the Pthread library on Linux. On Windows, it uses the native Win32 threading library. The build system uses CMake. The unit tests are written using Google tests (but are not mandatory).


The CUDA toolkit is required. We recommend using CUDA >= 7.0 even if earlier versions will work. * Download from the CUDA website * Follow the installation instructions * Don't forget to set your path. For example: *


Build CNMeM

Grab the source

% cd $HOME
% git clone cnmem

Build CNMeM without the unit tests

% cd cnmem
% mkdir build
% cd build
% cmake ..
% make

Build CNMeM with the unit tests

To build the tests, you need to add an extra option to the cmake command.

% cd cnmem
% mkdir build
% cd build
% cmake -DWITH_TESTS=True ..
% make

Link with CNMeM

The source folder contains a header file 'include/cnmem.h' and the build directory contains the library '', 'cnmem.lib/cnmem.dll' or 'libcnmem.dylib', depending on your operating system.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.