S/360-Z series timings, then and now? 
Author Message
 S/360-Z series timings, then and now?

I found (on the aek/pdf internet site) the IBM System/360
model 40 functional characteristics.  Included are detailed
machine timings for all sorts of functions, such as
external interrupts, supervisor call, program interrupts.

I tried looking for equivalent manuals for today's Z series
machines in Bookmanager, but I couldn't find anything.

I'm just curious how the timings compare today to back
then.  Would anyone know how to find S/390 or Z series
machine timings, or better still, the equivalent
"functional characteristics" manual?

Some 360-40 timings:

Arithmetic logic:  cycle time 0.625 microsecs.
Local storage:  two bytes 1.25 microsecs cycle time (split cycle 0.625)
General registers: 1.25 microsecs per 2 bytes.
Main storage r/w cycle 2 bytes is 2.5 microsecs.

BTW, the amount of detail they go into in describing the
design of the CPU and the I/O channels is substantial.



Wed, 08 Feb 2006 00:41:30 GMT  
 S/360-Z series timings, then and now?

just curious how the timings compare today to back

Quote:
> then.  Would anyone know how to find S/390 or Z series
> machine timings, or better still, the equivalent
> "functional characteristics" manual?

> Some 360-40 timings:

> Arithmetic logic:  cycle time 0.625 microsecs.
> Local storage:  two bytes 1.25 microsecs cycle time (split
cycle 0.625)
> General registers: 1.25 microsecs per 2 bytes.
> Main storage r/w cycle 2 bytes is 2.5 microsecs.

I recall that on earlier generations, such as the 7000-series
machines, instruction timings were given directly in the
Principles of Operation manual. I had one such for the 7094, and
it told how many cycles it took to execute each instruction, or a
range in the case of variable ones, like most of the floating
point ops.

Since the 360s came in many different sizes and designs, each one
had its own "functional Characteristics" manual to provide the
details. I don't remember seeing one for any 370s past the first
generation of them, such as the 145, 155, etc.

As CPU's became more sophisticated, with cacheing, pipelining,
speculative execution and branch prediction, the whole idea of
writing down the speed info became more problematical, and
therefore pointless. The older tendency to do that probably
reflects the general tendency in the electronics industry for
providing detailed technical specs for anything that cost a lot
of money ... just to prove you're getting your money's worth. As
devices become more complex and sophisticated, the "spec sheets"
get less informative.

I remember using one for the 370/145 once, to answer a broadside
criticism I'd gotten from someone who was looking at my code, on
microfiche, for the INASIRMI program, which was used to download
fixes. The guy made the wild claim that my text compression
routines would take longer to execute than the amount of time
they saved in transmission. Mind you, this was for transmitting
data over a primitive start-stop link operating at 600bps!  I was
able to show that each packet of information (some 700-odd bytes)
could be compressed and translated in about the time it took to
transmit a single character on the slowest CPU I was required to
support. The "translation" by the way, was a technique for
sending 8-bit binary data over a communications adaptor designed
for a 6-bit character code. I did this in 1973, and I think it
may have been the first instance of using a form of base-64
encoding on a communications link. At least, I've never heard of
an earlier one.

The communications controller in this case was the 2955, AKA "the
electric table".

<emoticon depicting gauntlet being thrown>

--
Russ Holsclaw



Thu, 09 Feb 2006 01:19:39 GMT  
 S/360-Z series timings, then and now?


<snip>
Quote:
> I'm just curious how the timings compare today to back
> then.  Would anyone know how to find S/390 or Z series
> machine timings, or better still, the equivalent
> "functional characteristics" manual?

<snip>

I know of no such table of timings. I believe that the reason is
that all processors these days are pipe-lined. As a result, the
amount of time an instruction spends in the pipe is directly
related to what the previous instruction is as well as the
"local of reference". Add to this the cache and type of cache as
well as the caching algorithm(s) and you have so many variables
in determining an instruction/interrupt timing that such tables
are no longer available. Then there is the idea that MIPS has
come to mean, Meaningless Indicator of Performance instead of
Millions of Instructions / Per Second.

--
Steve Thompson
www.vsstrategies.com
330/335-7228 office
Notice: By sending UCE/BCE (a/k/a, spam) to this address,
you are accepting and agreeing to our charging a $1000 fee,
per spam item, for handling and processing, and you agree to
pay any and all costs incurred (including court costs and/or
attorney's fees) for collecting this fee.



Thu, 09 Feb 2006 12:35:08 GMT  
 S/360-Z series timings, then and now?
On Sun, 24 Aug 2003 00:35:08 -0400, Steve Thompson hath writ:

Quote:


><snip>
>> I'm just curious how the timings compare today to back
>> then.  Would anyone know how to find S/390 or Z series
>> machine timings, or better still, the equivalent
>> "functional characteristics" manual?
><snip>

> I know of no such table of timings. I believe that the reason is
> that all processors these days are pipe-lined. As a result, the
> amount of time an instruction spends in the pipe is directly
> related to what the previous instruction is as well as the
> "local of reference". Add to this the cache and type of cache as
> well as the caching algorithm(s) and you have so many variables
> in determining an instruction/interrupt timing that such tables
> are no longer available. Then there is the idea that MIPS has
> come to mean, Meaningless Indicator of Performance instead of
> Millions of Instructions / Per Second.

IIRC, the last one I had was for the SYS/370-Mod155.  When we
swapped it out for a 158 I got a "Functional Characteristics"
manual for the new model.  It did not contain the "Appendix"
I was seeking!  The explanations I received at the time were
pretty much in line with Steve's comments above.  

(Some boot-leg, model-specific, multiply-copied sheets circulated
after a time with instruction timings.  But, these formulas had
SO MANY factors and footnotes for misc. Fiddle Factors that we set
such inquiries aside and moved on with our lives. :-)

Jonesy -- IBM MainFrame in 1966
--
  | Marvin L Jones       | jonz         |  W3DHJ   |  OS/2

  |   7,703' -- 2,345m   |   config.com |  DM68mn             SK



Fri, 10 Feb 2006 00:41:06 GMT  
 
 [ 4 post ] 

 Relevant Pages 

1. IBM 360 Emulator

2. IBM 360 Emulator

3. Assembly Source for APL\360

4. APL\360 Computers

5. 360/40 APL RPQ

6. APL/360 Computers

7. SPITBOL 360, a SNOBOL4 Compiler, Re-released Under GPL

8. ASM/360 tutorial

9. S/360-390 64 bit?

10. Complex Numbers from APL\360

11. FS: IBM System/360 Reference Card

12. AS/400 Assembler like S/360-390

 

 
Powered by phpBB® Forum Software