NVIDIA / gpu-rest-engine

A REST API for Caffe using Docker and Go
BSD 3-Clause "New" or "Revised" License
421 stars 94 forks source link

requisitions #42

Closed jadsongmatos closed 4 years ago

jadsongmatos commented 4 years ago

Is it possible to use GPU to speed up requests? Would there be any simpler demonstration like just hello world?

flx42 commented 4 years ago

Sorry, I don't understand your question.

Is it possible to use GPU to speed up requests?

Which part of the request do you want to speed up?

jadsongmatos commented 4 years ago

all parts, from routes to assembly of responses

flx42 commented 4 years ago

Ok, I'm not sure there is any example of this to be honest.

You should ask your question (and provide more details) here: https://github.com/NVIDIA/triton-inference-server (if it relates to deep learning at all)

flx42 commented 4 years ago

The name of this repository might be a bit misleading by the way; this is still doing the HTTP processing on the CPU side; but offloading the deep learning inference to the GPU.

jadsongmatos commented 4 years ago

So I was wrong about the purpose of the project, I thought GPU REST was about speeding HTTP requests. Can I seriously speed up HTTP responses using GPU? If yes, there is a project about it.

flx42 commented 4 years ago

Can I seriously speed up HTTP responses using GPU? If yes, there is a project about it.

Not that I know of. But if you're interested in network processing offload with GPUs, there is the NVIDIA Aerial product for 5G packets: https://developer.nvidia.com/aerial-sdk