A while back I started working on a way to use The faster Prime95 or mprime P-1 first stage and proceed to the 2nd stage using GMP-Ecm running as P-1. Which is at least 20 times faster.

What is most interesting is that you can split the work using B2min-B2max. It doesn't require to save a B2 files. If you know the work as been done to a certain point. You can start from there.

GP2 was able to make it work as the B1 was saved backward, as shown in this

link
Here is an exemple of one run.

B1 is done on Prime95 using a faster PC using Windows10.

The second stage is done on a Linux 32bits with 1G Ram.

And I run two instances.

B1 is at 10 billion and took over 2 hours.

B2 is done in steps of 300 billion and takes about 5 minutes.

The factor was found during the first step.

GMP-ECM 7.0.4 [configured with GMP 6.1.2] [P-1]

Tuned for x86/params.h

Running on jocelyn-Dell-DV051

Resuming P-1 residue saved by jocelyn@jocelyn-Dell-DV051 with GMP-ECM 7.0.4 on Fri Feb 5 22:54:18 2021

Input number is ((2^1571-1))/1250961804686347854556649368571289367 (437 digits)

Using special division for factor of 2^1571-1

Using lmax = 524288 with NTT which takes about 294MB of memory

Using B1=10000000000-10000000000, B2=19999998798540-20364610926900, polynomial x^1

P = 2251799814330893, l = 241920, s_1 = 1, k = s_2 = 47833428732674048, m_1 = 15488384

Can't compute success probabilities for B1 <> B2min

Step 1 took 0ms

Computing F from factored S_1 took 56019ms

Computing h took 11125ms

Computing DCT-I of h took 7321ms

Multi-point evaluation 1 of 1:

Computing g_i took 44471ms

Computing g*h took 17020ms

Computing gcd of coefficients and N took 11921ms

Step 2 took 147985ms

Peak memory usage: 306MB

Joss