First the numbers from the "time" command on a small test program:
Sparc II or IPX, SunOS 4.1.3/ ICS Motif 1.2.3
5.7u 0.7s 0:12 53% 0+2304k 0+0io 0pf+0w
5.7u 1.1s 0:10 66% 0+2316k 6+0io 3pf+0w
5.7u 0.8s 0:08 74% 0+2324k 0+0io 0pf+0w
Sparcserver 1000, Solaris 2.3 / SunMotif 1.2.2 or CDE Snapshot 1.0 Motif
14.0u 1.0s 0:22 66% 0+0k 0+0io 0pf+0w
14.0u 1.0s 0:20 71% 0+0k 0+0io 0pf+0w
14.0u 1.0s 0:21 68% 0+0k 0+0io 0pf+0w
Sparc II, Solaris 2.3 / SunMotif 1.2.2
33.0u 2.0s 0:52 67% 0+0k 0+0io 0pf+0w
30.0u 1.0s 0:40 77% 0+0k 0+0io 0pf+0w
31.0u 1.0s 0:46 68% 0+0k 0+0io 0pf+0w
(yes, it does mean that the Sparc II/SunOS/ICS Motif uses 2.5 times less of
its lowly CPU than a 50MHz SuperSPARC running Soloaris / SunMotif, and
an incredible 5.5 times less CPU that the same CPU with Solaris / SunMotif.)
The test program does noting other than create and manage windows, a main
window, and 10 dialogs. The dialogs consists of about 30 label gadgets,
44 text fields, 4 buttons, managed by a form and 2 row columns.
The Sparcserver 1000 has 192 MB RAM, with all disks local. Solaris 2.3 is
patched with the recomended patch list as of July 16, plus a few more,
including the Motif Jumbo patch 101377-16. I get the same timing with
"libXm.so.3" pulled of the CDE snapshot CDROM, witch I think is 1.2.3.
I tend to think the problem is in the Motif lib, but my tests can't reveal
what component of performace is due to operating system vs. Motif. It is
quite apparent that a Sparc II is a much snappier system running SunOS
4.1.3 than Solaris 2.3, even when only OpenWindows applications are involved.
Can anyone shed some light on these results? I wonder if these numbers
reflect a general difference in IXI vs. ICS Motif (my understanding is that
IXI is Sun's Motif supplier), or if it is really a Solaris isssue.
DuPont Experimental Station