[netperf-dev] netperf2 commit notice r64 - in trunk: . src
raj at netperf.org
raj at netperf.org
Fri Mar 31 13:27:07 PST 2006
Author: raj
Date: 2006-03-31 13:27:06 -0800 (Fri, 31 Mar 2006)
New Revision: 64
Modified:
trunk/AUTHORS
trunk/Release_Notes
trunk/config.h.in
trunk/configure
trunk/configure.ac
trunk/src/netlib.c
Log:
Add processor affinity for AIX (5.3 at least, perhaps earlier) and some
housekeeping for AUTHORS and Release_Notes. Pretty soon we really aught
to cut a 2.4.2 "release"
Modified: trunk/AUTHORS
===================================================================
--- trunk/AUTHORS 2006-03-30 04:46:19 UTC (rev 63)
+++ trunk/AUTHORS 2006-03-31 21:27:06 UTC (rev 64)
@@ -167,4 +167,7 @@
Changes to deal with different floating-point formats.
Anonymous
-Changes to retrieve CPU util on MacOS X.
\ No newline at end of file
+Changes to retrieve CPU util on MacOS X.
+
+Dickon Reed
+Patches to attend to some windows in TCP_CRR and TCP_CC under Windows
\ No newline at end of file
Modified: trunk/Release_Notes
===================================================================
--- trunk/Release_Notes 2006-03-30 04:46:19 UTC (rev 63)
+++ trunk/Release_Notes 2006-03-31 21:27:06 UTC (rev 64)
@@ -7,6 +7,12 @@
*) Additions for CPU util support on MacOS X, courtesy of Anonymous.
+*) Processor affinity is now supported on AIX 5.3 (perhaps earlier)
+ via the bindprocessor system call.
+
+*) Fixes for test lockups with TCP_CRR and TCP_CC under Windows
+ courtesy of Dikon Reed.
+
These are the Release Notes for Revision 2.4.1 of netperf:
Things changed in this release:
Modified: trunk/config.h.in
===================================================================
--- trunk/config.h.in 2006-03-30 04:46:19 UTC (rev 63)
+++ trunk/config.h.in 2006-03-31 21:27:06 UTC (rev 64)
@@ -6,6 +6,9 @@
/* Define to 1 if you have the <arpa/inet.h> header file. */
#undef HAVE_ARPA_INET_H
+/* Define to 1 if you have the `bindprocessor' function. */
+#undef HAVE_BINDPROCESSOR
+
/* Define to 1 if you have the `bind_to_cpu_id' function. */
#undef HAVE_BIND_TO_CPU_ID
Modified: trunk/configure
===================================================================
--- trunk/configure 2006-03-30 04:46:19 UTC (rev 63)
+++ trunk/configure 2006-03-31 21:27:06 UTC (rev 64)
@@ -6407,11 +6407,13 @@
done
+# check for the various CPU binding calls
-for ac_func in mpctl processor_bind sched_setaffinity bind_to_cpu_id
+
+for ac_func in mpctl processor_bind sched_setaffinity bind_to_cpu_id bindprocessor
do
as_ac_var=`echo "ac_cv_func_$ac_func" | $as_tr_sh`
echo "$as_me:$LINENO: checking for $ac_func" >&5
Modified: trunk/configure.ac
===================================================================
--- trunk/configure.ac 2006-03-30 04:46:19 UTC (rev 63)
+++ trunk/configure.ac 2006-03-31 21:27:06 UTC (rev 64)
@@ -84,7 +84,8 @@
AC_CHECK_FUNCS(sendfile)
-AC_CHECK_FUNCS(mpctl processor_bind sched_setaffinity bind_to_cpu_id)
+# check for the various CPU binding calls
+AC_CHECK_FUNCS(mpctl processor_bind sched_setaffinity bind_to_cpu_id bindprocessor)
# see if we should be enabling histogram support
Modified: trunk/src/netlib.c
===================================================================
--- trunk/src/netlib.c 2006-03-30 04:46:19 UTC (rev 63)
+++ trunk/src/netlib.c 2006-03-31 21:27:06 UTC (rev 64)
@@ -1762,6 +1762,17 @@
#include <sys/processor.h>
#include <sys/procset.h>
processor_bind(P_PID,P_MYID,processor_affinity,NULL);
+#elif HAVE_BINDPROCESSOR
+#include <sys/processor.h>
+ /* this is the call on AIX. It takes a "what" of BINDPROCESS or
+ BINDTHRAD, then "who" and finally "where" which is a CPU number
+ or it seems PROCESSOR_CLASS_ANY there also seems to be a mycpu()
+ call to return the current CPU assignment. this is all based on
+ the sys/processor.h include file. from empirical testing, it
+ would seem that the my_cpu() call returns the current CPU on
+ which we are running rather than the CPU binding, so it's return
+ value will not tell you if you are bound vs unbound. */
+ bindprocessor(BINDPROCESS,getpid(),(cpu_t)processor_affinity);
#elif HAVE_SCHED_SETAFFINITY
#include <sched.h>
/* gee, I wonder what we would do on a system with > 32 or 64
More information about the netperf-dev
mailing list