I think you will find it very very difficult to come up with a real world application that can queue up a solid flow of ONLY FP instructions and at the same time do useful real world work..
This is one reason that theoretical situations are NOT a good measure of CMT, try and buy was created to allow customers to try their code and and see how CMT handles real world rather than theoretical corner cases... Peter Maciej Browarski wrote: > So, all FP instruction are queued from all virtual processors or > dropped ? > In theoretical situation, when all 8 virtual processor what to make > fsquare, what happen (In know that FP instruction need more than one > clock cycles) ? > > Regards > Maciej > > Octave Orgeron wrote: >> The FPU is separate from the crypto. It's shared in the core, so all the >> execution engines have access to it. >> >> *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-* >> Octave J. Orgeron >> Solaris Systems Engineer >> http://www.opensolaris.org/os/community/sysadmin/ >> http://unixconsole.blogspot.com >> unixconsole at yahoo.com >> *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-* >> >> >> ----- Original Message ---- >> From: Maciej Browarski <Maciej.Browarski at Sun.COM> >> To: ldoms-discuss at opensolaris.org >> Sent: Tuesday, June 24, 2008 4:03:21 AM >> Subject: [ldoms-discuss] guest domain and FPU >> >> Hello, >> What mean 1 FPU per core ? >> If I divide 1 core to 2 guest domain I also divide FPU (timeshared or >> else ?) or FPU goes with MAU ? >> If FPU goes with MAU how floating point instruction are executed ? >> >> Regards, >> >> > _______________________________________________ > ldoms-discuss mailing list > ldoms-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss
