GMP used during 3 and a half years to solve MIT's LCS35

Hans Åberg haberg-1 at
Tue Apr 30 22:37:29 UTC 2019

> On 30 Apr 2019, at 23:34, Bernard Fabrot <bfabrot at> wrote:
> > I got it to that you are using an average of about 4000 cycles per iteration, whereas the other guys are about 20 times faster.
> Yup close.
> Core i7-7600 running at 3.4 Ghz but turbo-boosting to 3.9 or 4.0 Ghz when only one or two cores are at 100%.  About 22 minutes for 1 billion iteration.
> So I was a bit slower than that: I'd say about 5150 cycles per iteration, something like that (unless my math is off).

One advantage with not being parallelizable is that one can run it on one core and still use the computer on the other.

> Funny development: now that the story is out about LCS35 being solved, a 3rd team came up mentioning they were at 75%. But these guys are building an actual ASIC and are using a FPGA to validate their ASIC design (from what I understand).
> It's all very exciting!

Since modular squarings are common in cryptography, it it curious what would happen if one develops very powerful specialized hardware: Then it may not be so safe anymore.

More information about the gmp-discuss mailing list