On Wed, 16 Feb 2022, Leah Neukirchen wrote:

> Apparently it was a popular benchmark back in the day:
> https://www.tuhs.org/Archive/Documentation/AUUGN/AUUGN-V05.1.pdf

Yep.

> > The benchmark was "echo 99k2vp8opq I /bin/time dc > /dev/null’. It
> > uses dc (the desk calculator) to calculate the square root of 2 to 99
> > decimal places,

Ugh; that "I" ought to be "|" (and I can see other typos), so someone
really needs to proofread those PDF scans if they're going to be regarded
as authoritative.  Yes, I can help...

> > This benchmark has been applied to a large number of machines° It has
> > (up until now) been useful because most manufacturers have not
> > optimised dc, so the results are not likely to have been distorted by
> > attempts to optimise for benchmarks.

Indeed, when a lot of compilers recognised the Sieve of Eratosthenes being
used and optimised for it...  Wasn't all that long ago that vehicle
manufacturers also started doing the same thing :-)

> > Wicat 150WS 27.3 sec
> > Unison 32.6 sec

The WICAT 150 was just a terminal with several serial ports, and was
p*ss-awful (I worked for Lionel Singer, and had to support the poxy
thing); I'm surprised that it beat another box, though...

> I looks like V7 dc used 100-limbs internally, so printing in decimal was
> fast, but printing in octal required conversion.

Yep; extra work for the CPU.  Not much of a benchmark these days, as CPUs
are really fast; when I was supporting Unify I used to use its automatic
tutorial as a benchmark, and I used to joke that I needed a calendar on
some boxes...

Trivia: Lionel Singer also sold the then-new Sun-3, and we had one set up
at an AUUG conference, giggling whenever someone tried that benchmark;
what they didn't know was that the prominent window was actually an
"rlogin" to the Gould that we also sold...  The looks on their faces were
priceless :-)

-- Dave