Hacker News Reader

Comment by dnautics | Hacker News Reader

Wow, thanks for the detail.

> Re-wrote entire stack from python to C++ threadpool async grpc

Incredible. Presumably this is for latency/performance on the inference side?

Yes accuracy, latency & throughput are the 3 poles we try to achieve, c++ helps with latency & throughput and helps keep the cost low.

Why would c++ help with latency in comparison to say Python with numpy / numba / Cython? All the production critical “this needs to be as fast as possible stuff” I’ve ever worked on has been all Python, achieving complete speed parity with C, at a much faster development speed and with way way less boilerplate code.

If you have hard constraints at inference time, then it can be much easier to tune to a time budget with C++.

Like, it's normally not worth it, but when you need it, you really need it.

I definitely agree that could be a case where you want a statically compiled module that avoid any interpreted language overheads or high cost abstractions. But what would make C++ easier to write, tune, integrate or deploy in that case than using Cython to create the C++ extension for you?

I dunno man, I was always against running stuff in C++ if I didn't have to, but I got over-ruled. I guess that the high availability of C++ developers helped swing the decision.

I personally find C++ + pybind11 vastly easier to work with, also transitioning completely to c++ from there was a pretty small leap.

Interesting, I’ve never heard anyone who frequently uses Python and C++ together express this preference, it’s always the other direction that Cython is easier.

pytorch is pybind11 + c++

True, but that one project is just a drop in the bucket of scientific computing and C++ interop in Python, even despite the success and popularity of PyTorch - so it doesn’t really say much in favor of pybind that this or that project got good mileage out of it, it’s still such a deep minority compared to Cython.

Reply on news.ycombinator.com