Greetings! "Warren A. Hunt Jr." <[EMAIL PROTECTED]> writes:
> Hi Bob, > > It looks to me as though Camm's si::p-let has an overhead of about 2 > milliseconds on elgin. That seems awfully fast to me! Great in fact. How > long do you think would be "long" for a fork operation to take? Of course, > I > probably am hopelessly confused. Just learning out loud. > > Two milliseconds sounds about right. But remember, "elgin" executes > something like one or two billion instructions a second, so two > milliseconds of computation is millions of instructions! > Just a note here that what we have now should in no wise be regarded as illustrating the floor of the overhead of fork, which in any case will be higher than that for threads, but by what fraction (to me) is yet unknown. There are significant areas remaining in which to tighten the existing implementation. It would be most helpful to have an anlagous figure for the thread overhead as a sort of bogey to shoot for. I undertand that the machiones may not be comparable, but perhaps we could express this in terms of fractions of the single process performance, etc. Take care, > Cheers, > > Warren > > > -- Camm Maguire [EMAIL PROTECTED] ========================================================================== "The earth is but one country, and mankind its citizens." -- Baha'u'llah _______________________________________________ Gcl-devel mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gcl-devel
