Comment #6 on issue 436 by [email protected]: Numeric results are wrong on some systems due to x87 extended double precision http://code.google.com/p/v8/issues/detail?id=436
> The impact of extended-precision results [...] seems to be minimal, > and any intrusive changes to solve this problem would cause > performance hits. Should this be marked as "will not fix", or "fixed"? The Caja group is designing an enforced-deterministic subset of Cajita, for use for replicated computations -- both for high availability under partition as well as low latency query operations (as would benefit the planned Caja embeddings in Wave). Such deterministic replayability also benefits security <http://www.eecs.berkeley.edu/~finifter/pure-ccs08.pdf> and recreation (imagine <http://fantasticcontraption.net/> in Cajita). With the exception of this floating point issue, we believe that an enhanced Cajita rewriting strategy can enforce all the remaining determinism issues at reasonable cost. However, if the basic floating point operations are not of a deterministic precision, we have been unable to invent any other means of recovering affordable determinism by rewriting. -- You received this message because you are listed in the owner or CC fields of this issue, or because you starred this issue. You may adjust your issue notification preferences at: http://code.google.com/hosting/settings -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
