Luke Gorrie

@jsnell The D-I-Y pitch could be "Writing drivers is much less crazy than running thinly-capitalized startup companies, and people do *that* all the time." :smile: It's really hard to assess...

> I was scared by the perceived complexity after I've found some documents on the proprietary version of DPDK. This is such a huge barrier, right? People see complex driver...

Here are some references into the code for the curious: - [memory.c](https://github.com/SnabbCo/snabbswitch/blob/master/src/core/memory.c) allocates huge pages and resolves their physical addresses. (This should probably be rewritten in Lua with ljsyscall.) -...

@RichardBarrell I believe this part of the address space is not used on Linux/x86-64 but this is a bit unscientific. It would be good to have a strong reference for...

Good question @proofit404 :) Our main application area is high-capacity networking applications. These usually do not involve passing packets through the Linux kernel on the host where Snabb Switch is...

Quite so :). I'm using a bit of artistic license too. The JIT would actually check the type while loading the slot from memory and you wouldn't really have other...

Please leave a comment if you know anything about LuaJIT internals :).

Could be that what I really want is better `perf` integration. For example, for `perf top` to be able to zoom in and show me machine code annotated with IR...

@darius maybe we can thrash our way through it together a bit. I'm looking now in [`lj_asm_trace`](https://github.com/SnabbCo/luajit/blob/v2.1/src/lj_asm.c#L2221-L2260) and it seems like the machine code is created by iterating backwards from...

This is really a follow-on to previous posts [Snabb Switch for Entrepreneurs](https://github.com/SnabbCo/snabbswitch/wiki/Snabb-Switch-for-Entrepreneurs) and [Snabb, my lab](http://lukego.github.io/blog/2013/04/03/snabb-my-lab/).