PRP3 segfault with big numbers.

[Washuu]

PRP3 v 3.1.0 crashes, when I try to process very big number, exceeding 10M in size (ie 20*8087^3453301+1).

What are current limits of PRP3?

[Prime95]

Is this a P4? IIRC prp3 only works on SSE2 machines. I need to build a new executable. In the meantime, you could use LLR 3.5.

[Washuu]

Originally Posted by Prime95

Is this a P4? IIRC prp3 only works on SSE2 machines. I need to build a new executable. In the meantime, you could use LLR 3.5.

Celeron 4, 1800 Mhz. It has SSE2 for sure.

[Washuu]

And the same for PRP.EXE and OpenPGFW. Hmmm, is PRIME95 only 10M-digit enabled program? ;-)

EDIT: checked also on "true" P4 2,8 Ghz. The same problem. :

[Prime95]

I missed the 8087 base. I'll build a new prp3 and test it. You should be aware that these PRP test will be very slow. About 3 times slower than a Mersenne Test.

The limit of prp3 on numbers of this type should be about 39 million bits.

[Washuu]

Originally Posted by Prime95

I missed the 8087 base. I'll build a new prp3 and test it. You should be aware that these PRP test will be very slow. About 3 times slower than a Mersenne Test.

The limit of prp3 on numbers of this type should be about 39 million bits.

I know. I was just learning PRP3 and OpenPGFW and testing on different numbers. I just haven't expected segfault if I put the wrong numbers.

But my question is still current: what are current expected limits (apart from computer resources)?

[Prime95]

Originally Posted by Washuu

But my question is still current: what are current expected limits (apart from computer resources)?

39 million bits when base not equal 2. Your numbers are 44.8 million bits.

Testing k*2^n+/-1 your limit for n will depend on k. For small k, n up to 70+ million. For large k, n up to 39 million.