CUDA (Compiler)


The CUDA™ Toolkit is a C language development environment which greatly simplifies writing GPU-accelerated (General Purpose GPU, or GPGPU) code. This toolkit includes:

  • nvcc C compiler and profiler
  • CUDA FFT and BLAS libraries for the GPU
  • gdb debugger for the GPU
  • CUDA runtime driver
  • CUDA programming manual

Availability Table

System Versions
vdi-centos6 6.5.14, 7.5.18

Change Log

o Jan 15, 2016 - Cuda/7.5.18 module made default on copper and monk.
o Jan 14, 2016 - Removed cuda/5.5.22 from all non-cpu clusters including tembo.
o Jan 5, 2016 - Cuda/7.5.18 module made default on angel.
o Dec 18, 2015 - Cuda/7.5.18 installed into ang4 for testing.
o Oct 22, 2015 - Cuda/7.5.18 installed onto mos1 for testing.
o Oct 22, 2015 - Cuda/7.5.18 installed into cop1 for testing.
o Oct 16, 2015 - Cuda/7.5.18 module made default on monk.
o Sept 21, 2015 - Cuda/7.5.18 installed onto monk for testing.
o Oct 6, 2014 - Installed cuda/6.5.14 on vdi-centos6 and as default module on angel.
o Sept 18, 2014 - Installed cuda/6.0.37 as default module on monk.
o Dec 23, 2013 - Installed cuda/5.5.22 module on monk (not usable until driver updated).
o Feb 21, 2012 - Installed cuda/4.1 sharcnet module and driver on monk for testing.
o July 10, 2011 - Upgraded angel and sharcnet visualization workstations to cuda/4.0
o Feb 23, 2011 - Updated all systems with cuda installed from CUDA Toolkit 3.2 RC 2 (October 2010) to CUDA Toolkit 3.2 (January 2011). Some codes may need to be recompiled since the minor version of directories under /opt/sharcnet/cuda/3.2 have changed from 3.2.9 to 3.2.16 The cuda/3.2 module is still loaded by default on angel but NOT on the SHARCNET visualization workstations (where it must be manually loaded).