On Saturday, March 25, 2017 at 4:32:20 AM UTC+1, Brad wrote: > > Interested in any feedback about the idea of making a Go -> C++ > transpiler. [....] >
A very important point it to get a precise multi-threaded garbage collector in C++. This is painfully difficult, and if you succeed (perhaps by using Boehm's conservative Garbage Collector <https://www.hboehm.info/gc/> or Ravenbrook's Memory Pool System <http://www.ravenbrook.com/project/mps/>) you'll probably get a GC much less efficient that what Go has. (I have designed and coded GCC MELT <http://gcc-melt.org/> and I speak of experience: even debugging a single-threaded GC friendly to C++ is painful). Another point is the major difference between C++ containers and Go ones. IMHO, it is not very realistic. You might however code (after several years of hard work) some Go -> C compiler based above Boehm's GC. Similar stuff has been done for Scheme in Bigloo <https://www-sop.inria.fr/indes/fp/Bigloo/>. But the result will be slower than Go. Read the GC handbook <http://gchandbook.org/> to understand why. If you do things like this, you need to budget nearly ten years of work. Are you sure it is worth the trouble? Cheers. Basile Starynkevitch <http://starynkevitch.net/Basile/> (France). -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
