multi-threaded multiplication speed-up

Richard B. Kreckel kreckel at
Wed Jan 17 21:43:43 CET 2007

Pierrick Gaudry wrote:

>This kind of parallelization at the application level looks indeed vastly
>superior to having the parallelization at the level of the library. If an
>algorithm allows it (and this is often the case), this is the right thing
>to do. On the other hand, there are also algorithms for which the
>multiprecision operations are inherently sequential. Computing digits of
>Pi is one example.

I'm not disagreeing with your point, but let me just point out that 
computing digits of pi (or most other interesting constants) is quite 
amendable to parallelization.


Richard B. Kreckel

More information about the gmp-discuss mailing list