On Tue, Aug 07, 2012 at 02:16:35PM +0400, al...@yandex.ru wrote: > Well, going down to instruction encoding is a big deal to me. This is > technically doable but it's useless knowledge to most people > (especially if you multiply by number or arches). Code is pretty much > unreadable without this knowledge. What does a jump forward by 12 bytes > mean?
Yeah (but it is fun, sometimes). > This is the main reason why I use multi-platform jit library. I looked at that as well, and I would be realy interested to see direct performance comparisions. Especially for non x86 machines, where the LIR model does match less perfect - however, that is impossible with the current code state, unfortunately. I also think that this should have been evaluated/discussed before the import. > Perhaps this discussion should be moved to a different list. Yes, please! Martin
