DSP Math Accelerator Module Released for DADiSP

DADiSP/ProPac combines the new VectorXL module with the previously released FFTXL and MatrixXL modules to provide a single, optimized numeric processing engine. The solution accelerates common vector math computations by using the Intel Math Kernel Library (Intel MKL). Speed improvements of 20% to 60% are common, according to the company.

The MKL Library provides optimized vector routines based on the VML Vector Math Library. The algorithms are specifically tuned to Intel processors to provide enhanced performance.

The plug-in module automatically accelerates FFT computations and FFT based analysis routines. Also based on MKL, FFTXL takes advantage of the latest instruction sets, parallelism and algorithms to yield an optimized FFT function.

The DADiSP/MatrixXL Module builds upon the LAPACK Linear Algebra PACKage provided by Intel MKL to accelerate core matrix computations. LAPACK is an industry standard software library that provides a number of matrix routines including functions to solve linear equations, least squares systems, eigen value and singular value decomposition problems.

"DADiSP/ProPac delivers the combined benefits of VectorXL, FFTXL and MatrixXL in one simple package to provided highly optimized numeric analysis capabilities to DADiSP users in a transparent manner," said Randy Race, DSP's CTO. "Fast and highly tuned computing algorithms once limited to the domain of specialty programmers are now available to scientists and engineers across a broad spectrum of applications."

For more information, visit DSP.

Sources: Press materials received from the company and additional information gleaned from the company's website.