|
|
Jacket - The GPU Acceleration Engine for Matlab
|
|
Harmonic Software has partnered with AccelerEyes to provide distribution,
support, and discount sales pricing for the new AccelerEyes product, Jacket,
The GPU Acceleration Engine for Matlab.
Jacket Overview
Jacket uses the CUDA technology from Nvidia to utilize commodity
video cards to dramatically increase the speed of many Matlab applications.
Jacket enables standard Matlab code to run on the GPU, (graphics processing unit)
with minimal code modifications or learning required of the Matlab developer.
Many Matlab applications can be accelerated by up to a factor of 50, providing
application throughput otherwise not available in a workstation environment.
GPU Computing Simplified
Jacket GPU-enables standard Matlab. It greatly simplifies GPU computing
for engineers, scientists, and technical computing professionals. Users do
not need to learn CUDA, SIMD, HPC, and other complicated parallel programming
technologies. Within minutes of installing Jacket, most Matlab users
are using the GPU and obtaining significant performance gains for the
typical Matlab application.
Jacket Eliminates the Port. Many development projects and
even prototyping efforts will start with an interpreted environment
but grow to demand more performance than available within most tools.
Jacket enables many projects to continue with Matlab without the need
of rewriting to C, FORTRAN, or other low-level languages.
A Transparent and Optimized System
Jacket is not a GPU API, nor is it simply a collection of GPU MEX functions.
Rather, it is a complete and transparent system, automatically making memory transfer and
execution optimization decisions. Jacket uses a compile on-the-fly system to allow
GPU functions to run in MATLAB's interpretive style. Currently,
Jacket is built to use Nvidia graphics cards with support for Tesla
and other cuda capable gpus.
Example MATLAB code:
>> G = gones( 3 ); % Create a GPU matrix
>> G = fft( G ); % Perform a GPU FFT
>> G = G * G; % GPU Matrix Multiply
>> C = double( G ); % Bring back to CPU
By simply changing the data type, the example above taps into the power of the GPU to
accelerate the calculation of an fft and matrix multiply.
Run many iterations of a for loop simultaneously on the GPU ...
>> A = gones( n,n,m ); % Create a GPU matrix
>> B = gones(n );
>> gfor i = 1:m
>> A(:,:,i) = A(:,:,i) * B
>> gend
In the example above, all m matrix multiplications are performed simultaneously.
The Graphics Toolbox for Matlab
For a limited time,
the Graphics Toolbox for MATLAB is being bundled free of charge with Jacket.
The Graphics Toolbox integrates the Jacket computational engine with the
full OpenGL capabilities of your GPU. This coupling of computation and graphics
allows you to develop true Visual Computing applications.
The Graphics Toolbox includes a port of the entire OpenGL API,
including the newest extensions. You can now view your computations
in realtime as they evolve on the GPU by simply adding one line (e.g. gsurf)
to your MATLAB scripts. Jacket scripts, such as gsurf, are simply MATLAB
scripts containing OpenGL code and are completely load balanced for maximum speed.
Best of all, these scripts are totally open-source.
Don't like a visualization? Just change it.
The Jacket Graphics Toolbox for MATLAB provides truly
open, easy to use, blazingly fast visual computing.
Discount Pricing from Harmonic Software Inc.
For a limited time, Harmonic Software is providing a 10% discount
on the typical Jacket pricing. Note that your order must be
placed with Harmonic to obtain these discounts.
Please see our
Pricing and Ordering Information
page for details.
|
