Path: utzoo!utgpu!watmath!att!tut.cis.ohio-state.edu!ucbvax!pasteur!ames!vsi1!daver!yoda!gs
From: gs@yoda.chips.com (George Scolaro)
Newsgroups: comp.sys.nsc.32k
Subject: Re: NS 32532 information wanted
Message-ID: <20@yoda.chips.com>
Date: 9 Aug 89 21:53:06 GMT
References: <1711@garcon.cso.uiuc.edu> <13274@nsc.nsc.com>
Reply-To: gs@yoda.UUCP (George Scolaro)
Organization: Chips and Technologies, San Jose, CA
Lines: 25

In article <13274@nsc.nsc.com> roger@nsc.nsc.com (Roger Thompson) writes:
>In article <1711@garcon.cso.uiuc.edu>, blj@bach.csg.uiuc.edu (Bob Janssens) writes:
>> What I would like to have is a reference that goes into the specifics of
>> the 32532, especially instruction clock counts.  If anybody knows of any
>> references I would appreciate it if you would email me.
>> 
>
>tables etc really don't help.  To optimise the performance
>you are best off to trust the optimiser in the compiler.
>Trying to hand optimize is a unique skill.  To get a flavor of performance,
>write your test program in C or other supported HLL, turn on the
>Optimizer and let it run.
^^^^^^^^^^^^
This works great if you want to use a HLL for everything, but if you have
to write in assembly code to access special hardware etc (and we've had to
do it many times), then you must have access to instruction timing! Besides,
if the compiler (ie compiler writer) knows how to generate fast code then
the information is obviously available.

If you want a detailed description of how to calculate instruction timing,
refer to the 32GX32 data book (the newest one). It describes all instruction
timing including interactions with other instructions, the pipeline, cache
issues etc etc. Since the same timing refers to the 32GX32 and the 32532
you can use either new data book. I am not sure if the instruction timing
is in the 32532 data book, but it is definitely in the 32GX32 book.