Decimal arithmetic
Linas Vepstas
linas at austin.ibm.com
Wed Apr 4 19:31:26 CEST 2007
On Wed, Apr 04, 2007 at 01:46:13PM -0300, Décio Luiz Gazzoni Filho wrote:
>
> I've been asked off-list by a former member about digit extraction
> functions in GMP. I've given the usual lecture about how digit
> extraction is a slow operation if the library does binary arithmetic,
> since it requires a base-2 to base-10 conversion. With that said, a
> base-10 library would provide efficient digit extraction while not
> sacrificing much performance.
>
> So what are the choices of libraries that support base-10 arithmetic?
FWIW, some history of base-10:
-- IBM mainframes have supported base-10 arithmetic since the 1960's.
This was primarily to support accounting, in order to make life
easier for banks and utilities and anyone who sent out zillions
of invoices every month.
-- This base-10 arith has been backwards-compat supported all those
decades, and is now making its debut on a modern CPU, the Power6
due to ship "any day now". As a result, base-10 arithmetic support
was proposed to ANSI as a libc or libm math standard, and I beleive
the standard was recently (last year or two) accepted.
-- I don't beleive that it is widely supported in general. The only
hardware support will be on power6, so I don't see much point in
doing base-10 aritmetic on anything other than that.
> Décio
Unless, of course, this is another one of those "poisson d'avril"
emails.
More information about the gmp-discuss
mailing list