Luke Gorrie
Luke Gorrie
@eugeneia Awesome!!
@catwell Thanks for the links and the "real" Lua perspective! My plan with RaptorJIT is to start from LuaJIT, take away the sharp edges, add good tooling, and add a...
@jimblandy Sounds like a balanced take :). I do think that I have a couple of new things figured out, which can make a big improvement on LuaJIT in my...
@catwell Titan paper talks about LuaJIT having a 10x difference between JIT vs interpreted code. Interesting! I would ballpark it more like 100x. I see around 10x difference within good...
NB: Just a couple of weeks after this blog entry I chased a performance problem here: https://github.com/raptorjit/raptorjit/pull/150. This was an optimization that had introduced a separate fast-path and slow-path, which...
I added _hyperthreads_ which I think is a good fit. Interesting that people are commenting and also finding analogies to _more_ protocols than just TCP :). I also reckon there...
> Is there any "thing" like a Re-Order Buffer in TCP? Yep! The TCP packet transmission queue seems to be like a Re-Order Buffer. TCP always frees packet buffers in...
@hirenp I wonder if this whole model could be expressed in terms of HTTP instead of TCP too :).
Here is a dramatic conclusion from an excellent recent paper called [Array layouts for comparison based searching](https://arxiv.org/pdf/1509.05053.pdf) by @pkhuong and Pat Morin: > Our conclusion is contrary to our expectations...
@andywingo This is interesting... NUMMODE did not make a difference but I was able to eliminate one of those instructions by declaring the array as `long` instead of `int`. Here...