Excerpts from evanphx's message of Sun Dec 27 19:11:39 +0200 2009: > On Dec 27, 2009, at 6:27 AM, Eero Saynatkari wrote: > > The inline cache and method inlining are implemented in the > > interpreter as well. Actually JITting will use both of these > > features much more efficiently (along with exploiting other > > type information it is given or deducts itself.) > > Only the inline caches are used in the interpreter. No method inlining is done > in the interpreter. So to get the full effects of what Roger is talking about, > the JIT, which uses LLVM, is required.
I stand corrected - I thought the early inlining was at the bytecode level. On the other hand, it would be possible to implement inlining in the interpreter, though there seems to be little reason to go through the effort. -- Magic is insufficiently advanced technology. -- --- !ruby/object:MailingList name: rubinius-dev view: http://groups.google.com/group/rubinius-dev?hl=en post: [email protected] unsubscribe: [email protected]
