🚀 go-pugleaf

In article <1992Nov18.180357.3105@sparky.imd.sterling.com>, yankus@IMD.Sterling.COM (Mike Yankus) writes:
>
>In article <1992Nov16.231949.1214@ryn.mro4.dec.com> Blair.Phillips@cao.mts.dec.com writes:
>>
>> ...
>>Most real programs will use the runtime libraries for string manipulation, and
>>anyone serious about speed won't access single bytes very often, even on a VAX,
>>so trading off the byte access against cycle time is beneficial in almost all
>>cases except artificial benchmarks (e.g. Dhrystone).
>
>You are forgetting about a large volume of programs which mainly do
>nothing but character manipulation.  These include editors, compilers,
>mail handling, and the very programs which do the posting and reading
>of News.  I would consider a machine which does not have efficient
>character handling capability a step backwards of about 30 years.
>Is this really the case with the Alpha?
>--
The SPECmark benchmarks include the GNU C compiler compiling itself, and other
programs which do a lot of character manipulation. Alpha systems do very well
on these benchmarks.

The Dhrystone "benchmark" is almost entirely character manipulation, and is
small enough to fit in cache. It is close to the worst case for the Alpha
design decisions.

The Dhrystone MIPS numbers for various Alpha systems range from 134 to 202.2,
scaling very closely with the clock speed. I'm not sure how this compares with
HP and IBM.

However, if you write your own character manipulation code in C, and compile it
without optimization, you will probably get slower results than on another
architecture with similar performance and with byte load/store instructions.
Of course, right now, you can't buy another (microprocessor) machine with
similar performance...(HP is not shipping yet, I believe)
--
----------
Blair Phillips					Blair.Phillips@cao.mts.dec.com
Digital Equipment Corp (Aust) P/L		Phone: (+61 6) 2754874
Canberra, Australia				FAX  : (+61 6) 2473654
{Any opinions expressed are my own, not those of Digital Equipment Corporation}