M1061...

[cyrix]

Hi, "M1061 completed 100 ECM curves, B1=44000000, B2=4290000000"

Cyrix

[smh]

Originally Posted by ET_

Hmmm... I can only see version 4 for Win32 out there. Did you try version 5 under Cygwin?

Luigi

5.03, under windows, but it needs the cygwin dll to run

[ET_]

Originally Posted by smh

5.03, under windows, but it needs the cygwin dll to run

Thanks
I'll give it a try.

Luigi

[Xyzzy]

My friend Dommi did 60 curves with GMP-ECM using B1=44000000 and B2=184367799127...

[geoff]

Completed 300 curves, multiplier 1.32 (B1=44e6, B2=15e9).

I will construct a conversion table for multiplier values with various B2 and post it later.

[ET_]

I tried ecm5 (5.0.3) under Cygwin. It seems to work smoothly apart from a parameter. When I run

 ./ecm5 -c 23 44000000 <m1061

I obtain only one curve instead of 23.

What am I missing?

Luigi

[Axel Fox]

Yeah, it does that on my computer too. With the same version of gmp-ecm under cygwin.

But since I do stage one on Prime95 with the GmpEcmHook, I first do 10 curves stage 1 and then transfer the results file and resume them. Then it just does one curve for each input from the results file, so it continues all 10 curves.

However, at the moment I'm experimenting with different bounds and stuff, so I'm not posting my curves to count toward the total at this moment.

[geoff]

Here is a table with the multiplier to apply to gmp-ecm curves at B1=44 million and various B2 to convert to equivalent 50 digit Prime95 curves.

  B2         B2'       dF     P   k Expected  Multiplier
                                     curves
----------------------------------------------------
  4.29e9  4320585360   9216  -12  5  17831     1
  5e9     5049539880  10080  -12  5  17068     1.04
  6e9     6036383430  11040  -12  5  16482     1.08
  7e9     7142946720  12096  -12  5  15797     1.13
  8e9     7999948320  12672  -12  5  15529     1.15
  9e9     9046597560  13440  -12  5  15111     1.18
 11e9    11045144880  15120  -12  5  14434     1.24
 13e9    13495572090  16704  -12  5  13846     1.29
 15e9    15610434840  17280  -12  5  13553     1.32
 17e9    17151465240  18432  -12  5  13195     1.35
 18e9    18892404300  19200  -12  5  13002     1.37
 21e9    21231840630  20160  -12  5  12698     1.40
 24e9    24014815440  22080  -12  5  12406     1.44
 27e9    27718170480  23040  -12  5  12047     1.48
 30e9    30600888780  24960  -12  5  11799     1.51
 35e9    35069154120  25920  -30  5  11076     1.61
 40e9    43285242000  28800  -30  5  10658     1.67
 45e9    46495266510  30240  -30  5  10494     1.70
 50e9    52366223910  31680  -30  5  10281     1.73
 60e9    62312009760  34560  -30  5   9993     1.78
 70e9    73122389340  37440  -30  5   9677     1.84
 80e9    84797873160  40320  -30  5   9435     1.89
 90e9    91273101150  42240  -30  5   9300     1.92
100e9   100636912530  45600  -30  5   9112     1.96
120e9   122274758760  49920  -30  5   8815     2.02
140e9   147932254470  51840  -30  5   8616     2.07
160e9   161164286010  57024  -30  5   8418     2.12
180e9   180749221380  59904  -30  5   8256     2.16
184e9   190742241690  60480  -30  5   8198     2.17  (gmp-ecm 5.0.3 default)

Expected curves was calcualted as 1/ecmprob(44e6,B2',sqrt(10)*1e49,k*dF^2,P) using Alex Kruppa's rho.gp program, where B2', k, dF are the values reported by 'gmp-ecm -v 44e6 B2' and P is the polynomial used, -12 for dickson(12) etc.

Note that B2 is the value specified on the command line, B2' is the actual bound used by gmp-ecm.

The multiplier gives the number of equivalent curves at B2=4.29 billion. This may not be exactly equivalent to one curve with Prime95 because I don't know what parameters or polynomial Prime95 uses for its stage two step, but it should be close enough.

[Xyzzy]

68 curves using B1=44000000 and B2=184367799127...

[akruppa]

Originally Posted by ET_

 ./ecm5 -c 23 44000000 <m1061

I obtain only one curve instead of 23.

Strange. Right now I don't see how this could go wrong, it's a simple loop. I'll have to get Cygwin installed somewhere and try, may be a while though.

Alex

[thomasn]

[Sun Oct 24 22:44:24 2004]
M1061 completed 100 ECM curves, B1=44000000, B2=4290000000

Thomas