CUDAICA

What is CUDAICA?

Is an implementation of Infomax ICA made on CUDA.

What is Infomax ICA?

Infomax ICA is an algorithm to perform Independent Component Analysis

Requirements

Hardware

• A CUDA enabled GPU
• Enough RAM!

Software

• Cuda Toolkit: 4.0+ required
• GCC
• Autotools, Autoconf and M4
• GFORTRAN compiler
• BLAS library

Installation

1. Download the source code
2. Configure
3. Make

Advanced configurations options

Additionaly, when configuring the source code, the following options can be passed to the configure script:

• –-with-cuda: Use specified path as cuda base directory. Example: –with-cuda=/opt/cuda. Default: /usr/local/cuda.
• –-with-cuda-arch: Generate cudaica for the specified cuda compute capability. Options are:
  • 11 = Compute Capability 1.1
  • 12 = Compute Capability 1.2
  • 13 = Compute Capability 1.3
  • 20 = Compute Capability 2.0 (default)
  • 21 = Compute Capability 2.1
  • 30 = Compute Capability 3.0
  • 32 = Compute Capability 3.2
  • 35 = Compute Capability 3.5
  • 50 = Compute Capability 5.0
  • 52 = Compute Capability 5.2
  • 53 = Compute Capability 5.3
• –-with-double: Enables or disables double precision floating point. Options are:
  • yes (default)
  • no
• –-enable-debug: Enable debugging features and use specified level. Generates executable with debugging symbols and nvcc shows extra information. Example –enable-debug=2.
  1. Function calls.
  2. Memory alloc information.
  3. Function calls inside iterations.
• --enable-python: Enables python bindings (default=disabled)

More information LIAA website

Cite

Raimondo, Federico, Juan E. Kamienkowski, Mariano Sigman, and Diego Fernandez Slezak. “CUDAICA: GPU Optimization of Infomax-ICA EEG Analysis.” Computational Intelligence and Neuroscience 2012 (2012).