MCODE: Fastest way to multiply

[PeterP]

Albert, this is great material! I wish I had been in contact with you when I wrote my MultiPrecision library in MCode a few years (we'll, more like many years) back. Given that MPL was designed for numbers with 100-1000 digits, this could have some meaningful impact. And in a later post you even show Fact as well. Thank you for sharing!

(04-15-2019 05:07 PM)Albert Chan Wrote:  

(04-15-2019 03:58 PM)PeterP Wrote:  I found this article on Quanta Magazine (my favorite guilty pleasure read).

Thanks for the link ! Many other nice articles

Too bad the article never touch any details about the fastest method, except O(n log(n))

If anyone want to see real code for other mentioned method, I have some code examples.
Note: MAPM number does based-100, and based-10000 for FFT calculations.

Grade school method: mapm_mul.c
Karatsuba’s method : mapm_fmul.c
Fast Fourier Transform: mapm_fft.c