Re: Cycles per ASM instruction

John W. Kennedy wrote:

(snip regarding instruction timing)

> It was available for most of the original S/360 line, in the "Functional

> Characteristics" manuals for each model, but not for the 91/92/195, and 
> not, I think, for the 85. Caches, pipelines, and other factors made them

> meaningless.

The timings for math library routines for Fortran and PL/I were 
available including for the 91.  (The 92 was never built.)

Also, the 360/91 and 370/168-1 were, for typical use, reasonably
comparable in speed.  I knew of them being run under ASP where jobs
could go to either machine.  In most cases where they weren't equal
the 91 was slower (most likely when floating point was not used much).

-- glen

-- glen