[SATLUG] comparison of processors (OT)

[Don Wright]

On Mon, 1 Aug 2005 09:04:43 -0500, Jeremy Mann
<jeremymann at gmail.com> wrote:

>The question was actually asked by an intern here in my office. I
>figured somebody on this list would have an answer. He has since wrote
>a program to figure out the timing. Thanks again.

You may find the timings are much less dependent on the numbers
being multiplied. Since transistor count is not as much of an
issue any more, many processors use a lookup table to multiply,
for example, 8 bits at a time using a 256x256 table of 16-bit
results, or 131,072 bytes. That's a small ROM compared to the
on-board L1/L2 cache.  --Don


>On 7/30/05, Travis H. <solinym at gmail.com> wrote:
>> > >> Back in the i386 days, multiplication took a variable amount of time,
>> > >> depending on the number of leading zeroes in the binary representation
>> > >> of multiplier.  Oddly, this was asymmetric; multiplying 2 by 100 took
>> > >> more time than multiplying 100 by 2.