
fredrikj @ mathstodon Computer algebra & Arbitrary-precision arithmetic. Researcher at @Inria.
Bordeaux, France
Joined March 2016
- Tweets656
- Following129
- Followers615
- Likes990
fredrikj.net/blog/2023/04/fl…
FLINT furnished with faster FFT: we recently merged Dan Schultz's amazing new small-prime FFT code into FLINT, enabling much faster bignum arithmetic (in some cases 10x faster than GMP).
Why do furniture manufacturers even bother to give you these hex wrenches made of tin-playdough alloy?
Now my ANTS talk is also up on YouTube: youtube.com/watch?v=Un_Mqm2D…
Kind of fun. I created a minimal mpmath context object (github.com/fredrik-johansson…) around the floating-point arithmetic from FLINT/Arb fredrikj.net/blog/2022/07/ge… and it works:
Compare with this monstrosity of a flowchart that currently decorates the Wikipedia page en.wikipedia.org/wiki/Karats…:
Here is an attempt at a geometric explanation of the Karatsuba algorithm. The objective is to calculate the areas of the three distinctly-colored regions; Karatsuba does this with 3 multiplications instead of 4.
A key ingredient in hal.inria.fr/hal-03714660 is the ability to quickly approximate a real number to high precision by a smooth rational number. For example, here is a 131-smooth approximation of pi accurate to 131 digits, computed with Algorithm 1 in the paper.
New blog post: Parallelizing Arb for fun and profit.
fredrikj.net/blog/2022/05/pa…








