worldstone on 12-cpu e4500 with various kernel options

Kris Kennaway kris at FreeBSD.org
Fri Mar 18 20:32:10 PST 2005


A few weeks ago I ran some buildworld benchmarks on the 12-cpu sun
E4500 loaned to me by hrs:

-j12, no mpsafevm, 4BSD, invariants
     6441.67 real     16670.23 user     26399.98 sys

-j12, mpsafevm, 4BSD, invariants
     6371.78 real     17428.00 user     24199.93 sys

-j12, mpsafevm, 4BSD, adaptive mutexes and giant, invariants
     4353.08 real     16103.17 user     17315.01 sys

-j12, mpsafevm, 4BSD, adaptive mutexes and giant, no invariants
     3905.40 real     16114.54 user     13510.60 sys

-j12, mpsafevm, ULE, adaptive mutexes and giant, invariants
     4359.15 real     15680.28 user     14409.26 sys

-j12, mpsafevm, ULE, adaptive mutexes and giant, no invariants
     3927.60 real     15672.65 user     11076.43 sys

-j16, mpsafevm, ULE, adaptive mutexes and giant, no invariants
     3880.31 real     15903.13 user     11243.04 sys

-j1, mpsafevm, ULE, adaptive mutexes and giant, no invariants
    14904.45 real     12461.48 user      2670.71 sys

All tests are with a RELENG_5 kernel.  System was rebooted between
each test and was otherwise idle.  Adding PREEMPTION (either
scheduler), or trying to run 6.0 on this machine causes spontaneous
reboots (i.e. not panics).

If someone can think of another interesting variant, I'd be happy to
try it.

Kris

--
In God we Trust -- all others must submit an X.509 certificate.
    -- Charles Forsythe <forsythe at alum.mit.edu>


More information about the freebsd-smp mailing list