[rescue] RS/6000??

[Francisco Javier Mesa-Martinez]

On Wed, 17 Oct 2001, Joshua D Boyd wrote:
> Are the PPCs and POWERs binary compatible?

They have some ISA differences, mainly most PPC do not implement the whole
POWER ISA, so AIX has to emulate some of the instructions in software
(specially for the PPC 604 based RS6Ks). The POWER4 seems to be the first
CPU which is supposed to represent the convergence of both families

In theory the PowerPC was supposed to be a hybrid between POWER and the
M88K lines... not to be 100% binary compatible with POWER (different
markets at the time of introduction). I am wondering if the new G5 will be
based on the PWR4 architecture.....
  
> Yeah well, everyone claimed there was no more life to CISC, and yet the
> P4s are fairly fast.  Horribly painfull to work on, and tremendously
> wasteful, but still pretty stinking fast, especially for integer code.  I
> do most of my work in integers anyway.  I still think that if it can be
> done using rational numbers that rational numbers are the way to go.

Well, some of us need real Floating point, and P4 is more RISC than CISC,
it retains the CISC ISA for interface, the guts are pretty RISCy. Ever
since the P6 core came out, the x86 has been anything but CISC. So I would
consider CISC pretty much dead, I dunno of any modern desing that is CISC
(Maybe some embedded families, but that is just about it).

>   
> > Yup some of the POWER3 processors run at a "pathetic" 375MHz, yet they
> > kick the living shit out of a 1.4Ghz P4. And I am talking about real FP
> > code, not synthetic market-o-benchmarks. I am sick of people comparing
> > 32bit processors performance with 64bit processors (it is all fun and
> > games until someone asks them intel reps if those MFLOPS are DP!!!)
> 
> MFLOPS are DP?

Double Precission Floating point. It is funny to see a market-droid coming
to a scientific computing place and try to talk in terms of "winmarks",
"intelmarks", and all the rest of "whatevermakes us look good marks".

> And what the heck do you mean by a pathetic 375mhz.  My desktop box is a
> P2-350, and it has far from pathetic performance (although the performance
> to heat/noise ratio sucks, as does the stability at times).

I was being "sarcastic", nowadays a lot of people associate lotsa MHz with
actual performance, which is not true. The P4 is fast in paper, but there
are a lot of nasty things we have seen when running our codes, sometimes
(in fact in many times) it turned out to be slower than our P3.... Intel
is being saved by its actual strength: process technology.

A 500Mhz or less machine should be sufficient for most people. I was
disgusted when I was running win2K on a machine and I took the machine to
its knees just using a wordprocessor. That is just plain bad software
design... but I guess that is a whole new topic alltogether.

Cheers!