[netperf-dev] netperf4 commit notice r88 - trunk/src

[Rick Jones]

burger at netperf.org wrote:
> Author: burger
> Date: 2006-03-20 13:48:31 -0800 (Mon, 20 Mar 2006)
> New Revision: 88
> 
> Modified:
>    trunk/src/disktest.c
>    trunk/src/disktest.h
>    trunk/src/netlib.c
>    trunk/src/netlib_hpux.c
>    trunk/src/netperf.h
>    trunk/src/netperf_docs.dtd
>    trunk/src/netperf_hist.h
>    trunk/src/nettest_bsd.c
>    trunk/src/nettest_dns.c
>    trunk/src/nettest_vst.c
> Log:
> Fix problem in netlib_hpux.c referencing tid because of glib thread_id change.
> 
> Added histogram routines to netlib.c
> Modified netperf_hist.h to make things work.
> Tested histogram code with disktest
> Modified nettest_bsd.c, nettest_dns.c and nettest_vst.c and made sure
> they would compile if WANT_HISTOGRAM was defined.
> Added a hist_stats element to netperf_docs.dtd so histogram stats can
> be passed back to netperf for reporting.  Currently only disktest has
> a complete report routine which will report histogram data.

Do multiple histograms get aggregated?

I'm cutting the rest of the diffs, but noticed there is some 
hundred_nano stuff when/if GETHRTIME is available.  I've not looked 
closely enough to see if the histogram report always includes the 
hundred nano field anyway, which I think it should for consistency of 
output.

also, there is the chance that the netserver side code was compiled on a 
system with HAVE_GETHRTIME, but the netperf side was not.

Sooo maybe it would be time to have all the histogram timestamps be in 
units of nanoseconds just across the board?

rick jones