From Friday, April 19th (11:00 PM CDT) through Saturday, April 20th (2:00 PM CDT), 2024, ni.com will undergo system upgrades that may result in temporary service interruption.
We appreciate your patience as we improve our online experience.
From Friday, April 19th (11:00 PM CDT) through Saturday, April 20th (2:00 PM CDT), 2024, ni.com will undergo system upgrades that may result in temporary service interruption.
We appreciate your patience as we improve our online experience.
03-10-2010 09:27 AM
Hi,
Can anyone explain me the relation between Labview and MKL (the Intel Math Kernel Library) ?
For example, the Labview Interp_1d VI, is this routine optimized using MKL or not? The same question
for the FFT VI, is it optimized with MKL or not?
I use Labview 2009 and want to know this because I am testing my own GPU-based DLL against native Labview VI's
and need to know if it is legitimate to make this comparison for high end computing.
Thank you,
03-11-2010 06:17 AM
anyone?
03-11-2010 03:43 PM
Hello Eluir,
Many of the matrix operations in LabVIEW use the Intel Math Kernel Library (MKL), and the LabVIEW 2009 upgrade used the latest version of the Intel MKL to accelerate linear algebra operations under Windows and Linux operating systems. Hope this helps.
03-11-2010 11:05 PM
Hi Eluri,
LabVIEW uses MKL to optimize many linear algebra VIs.
Interp_1d VI and FFT VI do not use MKL. Also please note that FFT VI does not support single precision. I am not sure if your GPU-based DLL support double precision or not.
Best Regards,
Michael
03-12-2010 03:16 AM
Hi Michael,
thank you for your response. Do you know what the interp_1d and FFT VI's are based on? I want to compare my
GPU algorithms to the native Labview ones, but then I need to know if the Labview VI's are optimally implemented.
For example, is the FFT based on FFTW? And the interpolation, is it optimized C++ code?
Thanks in advance,
Eluri
03-12-2010 03:44 AM
Hi Eluri,
FFT VI is optimally implemented by NI self, but is not based on FFTW.
LV FFT is a bit slower than MKL FFT.
In most cases, the execution time of LV FFT is about 1.5x than MKL FFT.
For power of 2 case, LV FFT might be 2x or more than MKL FFT.
For Interpolate 1D VI, the implementation is G. It is designed for general purpose usage, not high performance.
Best Regards,
Michael
03-12-2010 04:51 AM
Thanks again for your fast reply,
if I wanted to compare my GPU algorithm to optimized MKL C code or to the FFTW, am I right that
I then would have to purchase the MKL library license and implement it myself or has someone already investigated this?
Thanks,
Eluri
03-14-2010 08:08 PM
Hi Eluri,
Currently you would have to purchase the MKL and implement it yourself.
Best Regards,
Michael