Leyland Primes (x^y+y^x primes)

[rogue]

Originally Posted by storm5510

Just for the record, I was able to get xyyxsievecl to run, like so:

xyyxsievecl -p3 -P5000 -x3 -X10000 -y3 -Y10000 -s- -M3200 -W2 -oresults.txt

The above produced an ABC file with over 3.3 million terms in 72 seconds. It didn't work my 2080 all that hard after I raised the -M parameter to 3,200 and added -W2. It would not start at the default of 800 for -M. I put the above in a batch file. It doesn't seem to want to run from an input file. I will learn more with some experimentation. I was able to do this by recalling my memories of sieving Riesel candidates.

Depending upon the speed of your GPU you can change -S.

[storm5510]

Originally Posted by rogue

Depending upon the speed of your GPU you can change -S.

-S Max steps iterated per call to GPU (default 100000)

The clock speed of this GPU (RTX 2080) when working at 100% capacity is around 1900 MHz. I don't know if this is what you're referring to.

If I remove the -W switch, it seems to work a lot harder, and warmer. If the parameters are large enough, it fails to start without a -W value set.

[rogue]

Originally Posted by storm5510

The clock speed of this GPU (RTX 2080) when working at 100% capacity is around 1900 MHz. I don't know if this is what you're referring to.

If I remove the -W switch, it seems to work a lot harder, and warmer. If the parameters are large enough, it fails to start without a -W value set.

-W sets cpu threads. -G sets gpu threads.

[storm5510]

Originally Posted by pxp

All of my data has had a table since April 2015 and I've posted the link here many times....

Your table has a column marked "Proven" but it is mostly blank beginning with entry 349. What is the story?

Originally Posted by rogue

W sets cpu threads. -G sets gpu threads

.

I would prefer not to get warned about being off-topic again. I will start my own topic in "Blogorrhea."

[pxp]

Originally Posted by pxp

All of my data has had a table since April 2015 ...

For the historically curious, my 2015 table looked like this. I had at this point checked all Leyland numbers smaller than 43413 digits, as indicated by the indexing going up to #954. The next number in that table turned out to be #955 but the one after that, #957. In January 2016 I found the missing #956, my fourth-ever Leyland prime find (having come across three somewhat larger PRPs in October 2015). By 4 July 2020 I had found 579 PRPs, all using Mathematica which (in case you didn't know) is slow as molasses. Since then (the last two years), I have found another 966.

[pxp]

Originally Posted by storm5510

Your table has a column marked "Proven" but it is mostly blank beginning with entry 349. What is the story?

The story is that these are entries that are P in factordb. The remaining exception is #715 (indicated by a K) which was proven prime a long time ago but has not (yet) had a Primo certificate submitted to factordb. I have never personally been concerned with the "proven" classification but many people are, so I added the column as a courtesy to them. In fact, quite a few of those P entries weren't there a couple of months ago!

[paulunderwood]

Congrats to frmky for the proof of 3^78296+78296^3 with 37,357 decimal digits

[storm5510]

Originally Posted by pxp

For the historically curious, my 2015 table looked like this...

I found your table very interesting and actually bookmarked it in my browser. I did an experiment based on it, but I won't go into it here. My results seemed to indicate 3-PRP does not mean prime. True?

[storm5510]

Originally Posted by kruoli

If you could dump the work that has been done I can host a website for that.

Off-topic: I haven't given up on it completely, yet. Some here say it is possible to sieve it. I haven't found anything that will. I didn't see anything in the mtsieve collection that would either. I looked at the parameters for each program using the -h switch.

[rogue]

Originally Posted by storm5510

Off-topic: I haven't given up on it completely, yet. Some here say it is possible to sieve it. I haven't found anything that will. I didn't see anything in the mtsieve collection that would either. I looked at the parameters for each program using the -h switch.

What do you mean by "it"? This clearly refers to a post from a few days ago, but there isn't enough context for me to know what you want to sieve.

[storm5510]

Attached is a snip from pxp's current table. There is a gap between the two highlighted areas. Has any of this been tested? I am looking for something to run.