Xeon phi

[Layosh]

Is there a way to use the intel xeon phi cards with labwindows?
These cards are becoming cheaper and support the I86 architecture.
Is ther a driver that could be used to run openMP commands on these cards?

[KeaneSenses]

Hi Layosh,

In general, if you have any general ANSI C APIs that you are interested in using, they can be used in CVI.  I found some extra CVI documentation that does mention using OpenMP commands:

Tutorial: Implementing Hello World in OpenMP

http://zone.ni.com/reference/en-XX/help/370051Y-01/cvi/libref/openmptutorialhelloworld/

Introduction to Open Multi-Processing (OpenMP) for LabWinsows/CVI

https://decibel.ni.com/content/docs/DOC-29830

Build on the Latest Technologies with LabWindows/CVI 2015

http://www.ni.com/white-paper/14617/en/

[gdargaud]

Hello,

I did an evaluation of the Xeon Phi a few years ago.

The compiler is different (LabWindows uses CLang, Phi uses the Intel compiler with specific options).

There is no problem in compiling separate programs, the Labwindows one running on the PC, the optimized one running on the Phi and both communicating through some method (TCP socket, direct file read/write to common shared directory or other).

There seems little point in compiling a CVI user interface for the xeon phi, but maybe it would be interesting to compile console progs that use the advanced analysis library or other NI libraries. If you have the source code, you can try replacing CLang with the intel compiler in CVI (there's a white paper about that), it would work in some cases. But if you need to link with some .dll or .lib, it's bound to fail dur to processor architecture differences.

[Layosh]

It sounds too complicated at this stage. CUDA is easier as one can generate Dll in visual C, and use that in CVi. The new NVIDIA cards are quite a bit more powerful than Xeon Phi.