On Fri, 2005-10-21 at 16:47 +0200, Nicolas Cannasse wrote: > > Nice. I will have a look sometime, but I'm up to my > > ears upgrading Felix from a simple event queue model > > to what seems to be more or less a complete user space OS .. > > > > I want to have an 'in depth' look at the VM -- the other > > stuff isn't as interesting in the sense that the VM > > what really defines the architectural properties. > > You should then look at the Neko Specifications, since Neko is specified by > its language and not by its vm bytecode.
What I'm interested in is this: > Neko : One VM to run them all Neko language is much too high level for me, and seems to make inappropriate architectural assumptions. It isn't clear to what extent these are in Neko language or the VM. For example: I do not need any symbolic names (all the symbols are already uniquely identified by integers). Any construction creating a local context is useless because I need to yield at any point, and later resume, without losing any local variables. Much of my interest is in (a) wrapping the C FFI, and (b) custom extensions. Felix does (a) almost automatically (if the API is simple C) the wrapper generator will do it in 1 second. Extensions are more work but doable. The idea is embedding. Executing Neko compiler and running bytecode is not very useful: how do I get data in and out? -- John Skaller <skaller at users dot sf dot net> Felix, successor to C++: http://felix.sf.net --- Neko : One VM to run them all
