Levelling up TF on 100M digit exponents with LL/PRP for future P-1

[Mark Rose]

Inspired by the effort to do better P-1 on exponents with one LL result, I've decided to prepare for doing the same on the 100M digit exponents.

Over the past couple weeks I've been poking at some exponents. In a couple of days I'll have every unassigned exponent over 333M with an LL result fully TF'ed (as well as between 130M and 332M).

There are a little over 500 unassigned exponents between 332M and 333M with an LL result without full TF. I've queued them all to be fully TF'ed by the end of the month (then I'll be back to whatever GPU72 feeds my machines).

Sadly the majority of exponents between 332.2M and 333M with P-1 have very poor P-1. Out of 3304 exponents with P-1, 1338 (40%) have P-1 with B2 under or equal to 1M, and a further 525 (16%) with B2 under 10M. Only 237 have P-1 with B2 over 100M.

At the time I'm posting this, there are 188 fully TF'ed unassigned exponents between 332.2M and 333M ready for P-1 with no LL or PRP. There are about 600 unassigned exponents with very poor P-1.

[Uncwilly]

Bless you. Proceed. Glad you are doing this.

[Mark Rose]

I hope there are people with hardware available to do P-1 with greater than 5% chance of success (~120 GB to give to mprime).

I've only got one such system and it has the capacity for 1 exponent every 24 hours or so. I've got it queued up into May already.

We'll have to get George to run his P-1 assignment generator script for batches of 332M exponents. It would be worth getting those machines capable started.

[Mark Rose]

I'm pleased to announce I've finished this DCTF effort.

All LL-tested exponents have been fully TF to current bit levels.

There may be a few more from Alpha2378 who has turned in a few higher level LL results in recent months.

I'm moving on to PRPed exponents. These should take about a month.

[Mark Rose]

All exponents below 332M with PRP have been fully TF'ed.

Working on the exponents above 332M now, 669 remaining (was 692). Should take less than a month.

I know some have verified certification, but a lot don't, so I'm just doing them all as I'm lazy.

[chalsall]

Originally Posted by Mark Rose

I know some have verified certification, but a lot don't, so I'm just doing them all as I'm lazy.

Thanks for doing this.

I've currently got my Colab TF instances configured to bring the few candidates in 332.1 up to 81. This is just to get them "off the books".

Yes, I'm OCD. I don't see a huge downside (beyond time).

[Mark Rose]

Originally Posted by chalsall

Thanks for doing this.

I've currently got my Colab TF instances configured to bring the few candidates in 332.1 up to 81. This is just to get them "off the books".

Yes, I'm OCD. I don't see a huge downside (beyond time).

I was wondering who was picking those up :-)

I do hope to get everything below 332.4 to 81 before the summer heat arrives. After the PRP exponent effort, it's currently another ~45 days of my throughput, which would finish up about mid May.

There are currently 1,582 unassigned 332M exponents without LL or PRP at 81 bits or higher, of which 99% or so could use initial/better P-1.

[Uncwilly]

I am also working in that realm, such as GPU72 will feed my colabs. I wish GPU72 would just feed me the lowest bit assignments that stay below 332.4 I am also willing to just bit level exponents up that are done or idle.