Intel i5-7400 CPU Benchmark

[danaj]

Presumably given the forum you're looking at testing Mersenne numbers, not general purpose testing. It isn't clear from your post, and I hate guessing at the question.

First check out the simple GMP program here: Rosetta-Code Lucas-Lehmer in GMP.

That should give you a decent idea of what GMP can do. There are certainly more optimizations, but they start getting more complicated. If it's worth doing, prime95 probably does it regardless of how "complicated" it is to write.

gwnum is more efficient than GMP once past 2000-4000 digits, and it keeps getting faster. This is pretty well known -- the underlying math library is just better tuned for large numbers and the particular operations. GMP is more portable and easier to write custom programs with, which is great, but these aren't a big problem for prime95, which would far rather have the speed.

[Wetitpig]

Actually I am just worried because it is not completely open-source.

[Wetitpig]

Originally Posted by danaj

Presumably given the forum you're looking at testing Mersenne numbers, not general purpose testing. It isn't clear from your post, and I hate guessing at the question.

First check out the simple GMP program here: Rosetta-Code Lucas-Lehmer in GMP.

That should give you a decent idea of what GMP can do. There are certainly more optimizations, but they start getting more complicated. If it's worth doing, prime95 probably does it regardless of how "complicated" it is to write.

gwnum is more efficient than GMP once past 2000-4000 digits, and it keeps getting faster. This is pretty well known -- the underlying math library is just better tuned for large numbers and the particular operations. GMP is more portable and easier to write custom programs with, which is great, but these aren't a big problem for prime95, which would far rather have the speed.

I have seen it and the program done will be very slowly. I know.

[Dubslow]

Originally Posted by Wetitpig

Actually I am just worried because it is not completely open-source.

Mlucas may interest you then. Even Mlucas will leave a simple GMP program in the dust.

[LaurV]

Originally Posted by Wetitpig

Actually I am just worried because it is not completely open-source.

What do you mean? Only the "crc" part (that does the server reports in a "trusted" form) is not open source. The rest, you can download the sources and compile it by yourself. What you will get, will be an "untrusted" client, but that you have anyhow, if you write your own code which will be much slower, and still untrusted.

[Dubslow]

Originally Posted by LaurV

What do you mean? Only the "crc" part (that does the server reports in a "trusted" form) is not open source. The rest, you can download the sources and compile it by yourself. What you will get, will be an "untrusted" client, but that you have anyhow, if you write your own code which will be much slower, and still untrusted.

Forget not the usage restrictions. Certain definitions of "open source" can be construed so that unrestricted usage is a requirement. (Certainly "libre software" definitely requires that.)

[Wetitpig]

Originally Posted by LaurV

What do you mean? Only the "crc" part (that does the server reports in a "trusted" form) is not open source. The rest, you can download the sources and compile it by yourself. What you will get, will be an "untrusted" client, but that you have anyhow, if you write your own code which will be much slower, and still untrusted.

So I just do manual reporting

[Wetitpig]

Originally Posted by Dubslow

Mlucas may interest you then. Even Mlucas will leave a simple GMP program in the dust.

Thanks. I will build it and try.

[Wetitpig]

Originally Posted by Batalov

We didn't want to give away the answer to your quest too early!
After all, "life is journey - not a destination!"

But if you want to see the destination, then you can mouse over this hidden text:
..."a custom program for testing prime number using libgmp" has no chances against Prime95 for Mersenne primes/composites.
It will be beat with Prime95's both hands tied behind his back.

After some trials I now found that the break even point is at around 2^400000 - 1.
Now I will optimize wven morw to see how it goes.