These are the Release Notes for the 2.0PL2 revision of netperf:

Things Changed in this release:

*) Netperf.ps is a dummy file in this patch release. If you want the
   manual, either consult the Web page, or grab an earlier 2.0
   release.

*) fixed the looper process CPU util for SGI Irix, thanks to pointers
   from folks on the net on how to do it. It should now be both
   accurate, and unobtrusive.

*) added similar looper process code for AIX - untested. Feedback
   appreciated.

*) fixed a bug in the CPU utilization code when the soaker gets zero
   CPU time. Exposed on Digital Unix where nice(39) is *really*
   effective...

*) fixed a bug where netperf -t REM_CPU would leave the soaker
   proceses behind after the test, and not cleanup the
   /tmp/netperf_cpu file. There is still a problem with trying to
   measure CPU on both sides of a loopback test.

*) started fixing bugs where sockets were left behind, precluding
   running more than ~57 iterations for a confidence interval test.
   initially pointed-out by Jim Yun. some of these may not have been
   found by the time PL2 is released. Of course, the code also limits
   the max iterations to 32, so it should be rather rare that someone
   hits this.

*) added progress indications to snapshot_script so people can know
   that the script is making forward progress. These indications go to
   stderr.

*) changed the permissions on the debug file

*) just saying "./netserver" should be enough to start the standalone
   daemon  

*) other things I've probably forgotten :)

Things Not Changed in this release:

*) USE_LOOPER and USE_PSTAT still require support for "long long"

*) Still seem to be problems with Solaris 2.[34] and gcc 2.6.3 and
   2.7.0. I am unsure if this is Solaris, gcc, or netperf. Netperf
   appears to work with the Sun compilers.