Mersenne prime exponent not randomly distributed?

[ewmayer]

B -smoothness: 24036583 is 1078 of 1171, percentile = 8.03
S_{L[sup]2[/sup]}-smoothness: 24036583 is 1078 of 1171, percentile = 8.03

[ewmayer]

B -smoothness: 25964951 is 316 of 1169, percentile = 73.05
S_{L[sup]2[/sup]}-smoothness: 25964951 is 265 of 1169, percentile = 77.42

[ewmayer]

B -smoothness: 30402457 is 16 of 1187, percentile = 98.74
S_{L[sup]2[/sup]}-smoothness: 30402457 is 67 of 1187, percentile = 94.44

[ewmayer]

B -smoothness: 32582657 is 148 of 1168, percentile = 87.41
S_{L[sup]2[/sup]}-smoothness: 32582657 is 9 of 1168, percentile = 99.32

[ewmayer]

B -smoothness: 37156667 is 501 of 1122, percentile = 55.44
S_{L[sup]2[/sup]}-smoothness: 37156667 is 675 of 1122, percentile = 39.93

[ewmayer]

B -smoothness: 42643801 is 36 of 1173, percentile = 97.02
S_{L[sup]2[/sup]}-smoothness: 42643801 is 17 of 1173, percentile = 98.64

[ewmayer]

B -smoothness: 43112609 is 512 of 1114, percentile = 54.13
S_{L[sup]2[/sup]}-smoothness: 43112609 is 165 of 1114, percentile = 85.28

[Dubslow]

May we please see the source?

Do you have a scatter plot of these percentiles, or shall I make one?

[ewmayer]

Originally Posted by Dubslow

May we please see the source?

Attached. I didn't take the time to un-bundle it from the rest of my comp-NT code, so standalone builders will need to replace the missing boilerplate header-includes with the few needed typedefs (e.g. uint32) and utility functions (e.g. modpow).

Do you have a scatter plot of these percentiles, or shall I make one?

Feel free - but consider waiting until tomorrow when one more datum will get added.

[firejuggler]

ewmayer, what about M48? hmm?

[Dubslow]

Originally Posted by ewmayer

Attached. I didn't take the time to un-bundle it from the rest of my comp-NT code, so standalone builders will need to replace the missing boilerplate header-includes with the few needed typedefs (e.g. uint32) and utility functions (e.g. modpow).


Feel free - but consider waiting until tomorrow when one more datum will get added.

Well I have little to do tonight, along with knowledge of M48, so...

Edit: I copied in Msieve's util.h, which looked like it fit the bill. But I'm not entirely sure how to go about replacing the rest.

mprime_exp_smooth.c:41:2: warning: implicit declaration of function ‘ncmp_uint32’ [-Wimplicit-function-declaration]
mprime_exp_smooth.c:49:2: warning: implicit declaration of function ‘ncmp_uint64’ [-Wimplicit-function-declaration]
mprime_exp_smooth.c:89:3: warning: implicit declaration of function ‘twopmodq32_x8’ [-Wimplicit-function-declaration]
mprime_exp_smooth.c:94:5: warning: implicit declaration of function ‘ASSERT’ [-Wimplicit-function-declaration]
mprime_exp_smooth.c:94:12: error: ‘HERE’ undeclared (first use in this function)
mprime_exp_smooth.c:132:3: warning: implicit declaration of function ‘twopmodq32’ [-Wimplicit-function-declaration]

The cmps should be easy to replace (assuming they don't do anything particularly fancy I'm unaware of), and it looks like twopmodq32(uint32 a, uint32 b) = 2^a mod b. It also looks like twopmodq32_x8(uint32 a-h) calculates each of 2^(a-1)%a, 2^(b-1)%b, ... presumably in some sort of optimized batch fashion?

I'm not sure what your ASSERT/HERE macros/thingies are though... sorry, that's probably a really newbie thing to say

Edit: A not-really-informative scatter plot.