On Monday 23 March 2009 14:19:00 Xavier Leroy wrote: > "But shadow stacks are the only way to go for GC interface!" > No, it's probably the worst approach performance-wise; even a > conservative GC should work better.
I blogged a quick analysis of the performance of HLVM's current shadow stack code: http://flyingfrogblog.blogspot.com/2009/03/current-shadow-stack-overheads-in-hlvm.html There is a lot of scope for optimization but these results were enlightening to show where the effort should be put. In particular, shadow stack updates by the mutator (and not the collector, as I had incorrectly assumed) account for the entire performance difference between OCaml and HLVM on the 10-queens benchmark. -- Dr Jon Harrop, Flying Frog Consultancy Ltd. http://www.ffconsultancy.com/?e _______________________________________________ Caml-list mailing list. Subscription management: http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list Archives: http://caml.inria.fr Beginner's list: http://groups.yahoo.com/group/ocaml_beginners Bug reports: http://caml.inria.fr/bin/caml-bugs
