Luke Gorrie

Yes. I am optimistic about table sealing. Generally I think that we need more optimisations that are mutually redundant with loop optimization, i.e. that optimize the head of the trace...

> Wouldn't hyperblocks and a better GC (Like the proposed luajit 3 gc) go a long ways towards this? Good question! #### Hyperblocks How do we implement that? My impression...

@capr Interesting! Can you tell me how to run your application in such a way that it will make the JIT misbehave? I'd be curious to profile it with RaptorJIT...

> also when new traces are formed around code that previously compiled fine and then the whole root trace is aborted because some obscure limit is reached. This is also...

@capr Just thinking about GUI domain a little more. On an intuitive level (= without experience :P) it seems like a fine fit for the tracing JIT. You will have...

@capr Thanks for sharing your positive experience with Terra. That's very interesting and I'll have to give it a proper look. I have heard other people expressing the same skeptical...

This has been talked about much but I don't think that anybody has done it. Amazon really has [three network interfaces](http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/enhanced-networking.html) the last time I checked: - Plain ol' presumably...

@tst2005 If you compile the VM with assertions enabled (`-DLUA_ASSERT=1`) then it will "balance the books" on shutdown. There is an [assertion](https://github.com/raptorjit/raptorjit/blob/master/src/lj_state.c#L176) to check that the VM freed exactly the...

@mwiget That's one more good reason to book a Snabb Lunch in Switzerland soon! cc @alexandergall @sleinen

Thanks @emmericp @dpino I updated above. The Ixy talk will be a real highlight! Great lead-in to my talk too :-)