P2P to GPUs

[Terry_ALE]

P2P is a very useful technology for sharing data between NI targets.

Could this be provided for GPUs?

[Nikn]

Hi Terry,

Looks like you might have already seen the post 

Are you looking for something beyond it? Can you describe your use case for me to understand how NI can support you better.

Regards,

Kiran

Kiran Nagaraj | Principal Product Owner | LabVIEW R&D

[Terry_ALE]

Hi Kiran,

The posted code is older and there is a concern for future support.  There is no specific need although customers do like the idea of offloading some processing on a GPU which has software level compile times (as opposed to the FPGA).

[nanocyte]

I need this right now. Basically, I need a low latency path between the GPU and the FPGA that doesn't bother the CPU. This also gives you more determinism and saves you CPU resources.

https://developer.nvidia.com/gpudirect

https://blog.zhaw.ch/high-performance/2019/10/03/direct-communication-between-fpga-and-gpu-using-fra...

[Terry_ALE]

Check out https://forums.ni.com/t5/NI-Labs-Toolkits/FlexRIO-Peer-to-Peer-GPU/ta-p/3660064

[nanocyte]

This warning scared me: Note: FlexRIO no longer officially supports x86 systems. See NI_FlexRIO-16.0.0_P2P_GPU_Driver/README.txt for more information.

[Terry_ALE]

I assume that x86 means 32-bit systems and x64 means 64-bit systems.

See http://net-informations.com/q/mis/x86.html