Fargusson.Alan wrote:
Realize that MIPS actually means Misleading Indicator of Processor Speed.
I usually use a multiplier of .5 for CISC machines, which turns out about right
in most cases. For example someone recently posted that the clock is .8 GHz
for a z890 (which is 800 MHz), and that the MIPS is about 365. Most RISK
machines are about .7, and superscalar machines can be 2 to 3 instructions per
cycle.
Komputer?
There is a book by Henesy (sp?) and Peterson that talks about this in great
depth.
-----Original Message-----
From: Linux on 390 Port [mailto:[EMAIL PROTECTED] Behalf Of LJ
Mace
Sent: Tuesday, October 24, 2006 6:39 AM
To: LINUX-390@VM.MARIST.EDU
Subject: what is the conversion from mips to Ghz or back
I've been asked this question and have looked around
but can't seem to find an answer.
Consider the impact of optional features.
Floating-point and decimal arithmetic were optional features on early
S/360 models. I gather certain features (eg cryptographic?) are optional
on current zSeries.
We used a third-party firmware upgrade to make our S/370 model 145
functionally equivalent to a 148 for a 10%-30% improvement in performance.
VM works better with firmware assists.
In the 80s IBM reportedly implemented firmware assists for dfsort (and
maybe other software), improving its performance wrt syncsort, and
(probably) the S/370s wrt the PCM crowd.
To most people, the important measure is how well their C, Pl/1, COBOL
and Java programs run; the number of machine instructions run is only
one component of the equation; language and compiler are immensely
important too.
Programs compiled with Intel's icc (C compiler) reportedly run rings
around gcc on IA32 and IA64 computers. Different releases of gcc perform
differently, and probably relativities between hardware architectures
change from release to release: improved code generation for (say)
Pentium IV might not be matched by an equivalent improvement on zSeries,
Power* or *Sparc*.
Processor family is important; recent Intel IA32 processors clock way
below early Pentium IVs running at 3 Ghz or so, but have better
performance. One expects Xeons to outperform equivalently-clocked
Pentium* processors, and AMD processors, for a while at least,
outperformed Intel's processors.
Here's a crude test; I've published it here a few times in the past.
[EMAIL PROTECTED] ~]$ cat bin/bm.perl
#!/usr/bin/perl
#use integer;
$i = 0;
while ($i < 10000)
{
$j = 0;
while ($j < 10000)
{
++$j;
}
++$i;
}
[EMAIL PROTECTED] ~]$
Run it again without "use integer" commented out. Try it on different
computers.
I have here a Pentium IV 3.0, with hyperthreading enabled. It can run a
copy on each virtual CPU faster than my Sempron 2500+.
Consider too, the workload. I'd expect more "IPS" on z zSeries running
z/OS, COBOL (or Pl/1) applications against IMS than the same machine
running zlinux, PHP and DB2.
--
Cheers
John
-- spambait
[EMAIL PROTECTED] [EMAIL PROTECTED]
Tourist pics http://portgeographe.environmentaldisasters.cds.merseine.nu/
Please do not reply off-list
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
