runtime error when using redc

[ltd]

I have a problem when I use ecm under windows (both win32 and win64 Bit version).

I have started ecm with the following commandline:

ecm.exe -v -pp1 5000 <test.txt >>out.txt

test.txt contains only the following line:

168451*2^6730668+1


The output at the time of the crash is:

GMP-ECM 6.2.3 [powered by GMP 4.2.1_MPIR_1.1.1] [P+1]
Input number is 168451*2^6730668+1 (2026139 digits)
Using REDC
Using B1=5000, B2=287928, polynomial x^1, x0=982738599
P = 525, l = 512, s_1 = 240, k = s_2 = 1, m_1 = 3

When I run ecm with the following line everything runs without a problem:

>ecm.exe -v -pp1 -mpzmod 5000 <test.txt

Is there a known problem with the "redc" code or what did I do wrong?

[ltd]

Correction: -mpzmod did not work also. Output here is:

GMP-ECM 6.2.2 [configured with GMP 4.2.1 and --enable-asm-redc] [P+1]
Input number is 79817*2^1191+1 (364 digits)
Using B1=5000, B2=287928, polynomial x^1, x0=158395725
Step 1 took 46ms
Step 2 took 78ms
Input number is 168451*2^6730668+1 (2026139 digits)
Using B1=5000, B2=287928, polynomial x^1, x0=1798276254
Step 1 took 5574695ms
Could not initialise ntt_context, presumably out of memory
Please report internal errors at <ecm-discuss@lists.gforge.inria.fr>.

[henryzz]

the problem is your number:)
it is way too large

[ltd]

This is a little bit unexpected as ecm and p-1 also crash and they work with prime95 without any problems.

[henryzz]

Originally Posted by ltd

This is a little bit unexpected as ecm and p-1 also crash and they work with prime95 without any problems.

prime95 uses its own code that is designed to cope with huge numbers
gmp-ecm isnt designed for more than a few hundred(it will do several thousand digits slowly)

[ltd]

Thanks for the info. I knew that ecm uses its own implementation but I thought till now that both implementations would be able to handle numbers of the same size.