Please visit donation page to help the project cover running costs for this month
1) Message boards : General discussion : Intel KNL to advance Number Theory (PrimeGrid too) (Message 98527)
Posted 2216 days ago by Carlos Pinho
Members of the mathematics, research, and development community at are working to purchase an Intel Xeon Phi -Knights Landing shared development machine.

We have several projects in Number Theory research incuding mlucas, prime95 ( and others that can be significantly advanced by developing for upcoming High Performance Computing technology. Intel's Xeon Phi processors offer these volunteer developers early access to several key features, including 6 channel DDR4 bandwidth, HBM on-die, large shared caches, AVX-512 instructions, and 64 cores with 256 threads.

We strongly believe that future consumer and enterprise processors will continue to move towards similar architectures. Allowing benchmarking, research, and development to occur today will ensure we can continue to see the ever improving performance of CPUs reflected in the work output of these projects.

More information on this project and a wealth of other on-going efforts can be found at the thread here:

Logistics of colocation and power for this system have already been committed to by members of the community, but we need your help to purchase the system and push this work forward. Ideally we will be able to achieve our funding goals and proceed with this purchase by the end of September so that we can begin work as soon as possible.

This system will also remain fully powered and online to provide contributions to when not being actively used for development, so you can be assured that the hardware will be fully utilized for its entire lifetime.

We are immensely thankful for any and all donations towards this project. We also welcome any requests for shared access to this system from other researchers and developers within the community.

If we are able to exceed our goal we may be able to purchase more than one development system.

Help spread the word!
2) Message boards : General discussion : Great Internet Mersenne Prime Search New Prime? (Message 91059)
Posted 2455 days ago by Carlos Pinho
Great Internet Mersenne Prime Search New Prime!?
Take a look at:
3) Message boards : Number crunching : LLR Version 3.8.14 released (Message 84633)
Posted 2733 days ago by Carlos Pinho
Batalov is running some 14,000 random double-check tests and no RES64 mismatches so far in (from 0.25 to 4 hours run time range). 7 known primes re-confirmed.
Keep watching that thread for updates. If anything is missed here I will chase and post.

Iain Bethune, I was also aware but more worried that this bug would be very likely to happen but Batalov said to me it is a very improbable bug. I was afraid my search would be compromised even now that I restrict to only 4 cores when in the past I had more than 30 cores on my search.

4) Message boards : Number crunching : LLR Version 3.8.14 released (Message 84622)
Posted 2734 days ago by Carlos Pinho
Please be aware of this thread:
5) Message boards : Number crunching : LLR Version 3.8.14 released (Message 84301)
Posted 2749 days ago by Carlos Pinho
Hi All,

I uploaded today the version 3.8.14 of the LLR program.
You can find it now on my personal site :

The 32bit Windows and Linux compressed binaries are available as usual.
The Linux 64bit binaries are also released here.
I uploaded also the complete source in a compressed file ; it may be used to
build the Mac-Intel executable and also the 64bit Windows binary.
This LLR version is linked with the Version 28.5 of George Woltman's gwnum library.
I released also a cllrd binary, which is linked with the debug version of this
gwnum library.
You can see in the Readme.text file, what is new in this release.
As usual, I need help to build the 32bit Mac Intel binary, and also the 64bit Mac Intel and Windows ones.
Please, inform me if you encountered any problem while using this new version.
Best Regards,
6) Message boards : The Riesel Problem : Hans Ivar Riesel passed away last year (Message 84254)
Posted 2750 days ago by Carlos Pinho
Just to let you guys know that Hans Ivar Riesel passed away last year.
Maybe an update of this thread is needed.


7) Message boards : Sierpinski/Riesel Base 5 Problem : SR5 Double Check (Message 76598)
Posted 3052 days ago by Carlos Pinho

You were with Joe O and ltd in front of RS5 in mersenneforum for quite some time. If I remember there were some faulty clients running this, right?

8) Message boards : Sierpinski/Riesel Base 5 Problem : SR5 Double Check (Message 76266)
Posted 3067 days ago by Carlos Pinho

The odds should actually be somewhat better than that because the chance of finding a prime is better with smaller numbers.

Can you prove your statement?
9) Message boards : Number crunching : Haswell AVX2 version of LLR available (Message 74091)
Posted 3130 days ago by Carlos Pinho

I don't measure power consumption and take note of temperatures just for two simple reasons, the first one is that electricity is very cheap here and second I live in a tropical country so ambient temps are always very high at the room I have my laptop. Basically since the start of the year I've turned on the AC 85 % of the time during afternoons and nights.

By CoreTemp software I know the laptop is consuming something like 29 Wh but I don't know if this number is real, I should use a energy meter instead. Also I don't know if the latter has a good accuracy for small energy consumptions. This consumption is only for the use of 2 threads out of 8, I can't switch off HT ( nothing on BIOS) to increase speed and decrease temperatures of the cores. Also I use only 2 threads due to memory bottleneck.

When I first tested this new version I was getting a speed up of 2.5 %. In the last 2-3 days the ambient temperature decreased something like 5-10 ºC and so now I am getting a speed up of 3.8%. Also it depends a lot if I have the AC turned on or turned off.

The other cores are in controlled atmospheres (lab) with AC and therefore I don't care about how much they are consuming.


10) Message boards : Number crunching : Haswell AVX2 version of LLR available (Message 74084)
Posted 3130 days ago by Carlos Pinho
If you don't have a Haswell CPU there's no need to switch to this version of LLR.

The new gwnum library (that this release is based on) is supposed to be about 1-2% faster on Sandy/Ivy as well.

Yes you need to switch.
On an Ivy Bridge processor I get a 3.5 % improvement.
I always learned with Prime95 that better a small improvement than nothing. Even at least 1 % is even better than 0% speed up.

BTW, someone said (Michael Goetz)here the following:

What I was saying was that AVX2 is not very useful. And probably never will be. The AVX2 instruction set, for the most part, is of no use to us, and the one instruction that is useful doesn't provide a significant benefit. It's not a game changer. It's barely noticeable above the background noise, to be honest. That's unlikely to change. Anyone who was expecting a big improvement from an "AVX2 optimized app" wasn't well informed on the nature of AVX2.

There are other improvements in Haswell, but AVX2 isn't a big deal, and never will be.

Now it is a funny comment...

Next 10 posts
[Return to PrimeGrid main page]
Copyright © 2005 - 2022 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 0.08, 0.02, 0.01
Generated 30 Sep 2022 | 11:46:40 UTC