Re: Cycles per ASM instruction

<hancock4@[EMAIL PROTECTED]
> wrote in message 
news:1172005803.961733.21960@[EMAIL PROTECTED]
> On Feb 13, 9:46 am, "Roded" <rod...@[EMAIL PROTECTED]
> wrote:
>> Hello,
>> Anyone know where I can find the number of cycles per ASM instruction
>> for the z990 and the z9 computers?
>
> Hasn't that sort of thing always been model dependent ever since
> System/360 was introduced?  That is, don't different models have
> different data widths and memory access methods, so the slow Model 30
> would use more cycles than the faster Model 65?  Wouldn't this apply
> to present day "Z" series machines as well?
>
> Complicating the matter on present day machines, don't storage cache
> and "pipelining" also impact on cycles and thus instruction time?
>
> If I recall correctly, the pre-S/360 machines included detailed
> instruction time charts which I presume were regularly used to
> estimate processing time.
>
> But I don't recall such charts being readilly available for System/360
> and later machines.  I believe with some searching they could be found
> in more specialty manuals and utilized in certain applications, but I
> think this was discouraged since it caused hardware dependencies which
> made migration difficult.

Number of cycles per instruction was model dependent from the very first
S/360 model, and IBM issued charts for all (?) 360 models. However
as the 370 models were all microprogrammed such information became
not only model dependent but also dependent on the actual data values
operated upon. This meant that a specific machine instruction no longer
had a characteristic operation time so that such charts became more or
less meaningless and were no longer issued.

(Users noticed the difference as they got a new startup function: IMPL
in addition to IPL.)

regards Sven