View Single Post
Old 2011-11-12, 19:35   #8
Mar 2010

5110 Posts

Curious about the cuda11_test branch in svn. I have the regular msieve-149 version running on a
GTX 465 Nvidia card and has been running for about 24 hours. So far that has not really given me much.
About 20 poly candidates.

So I compiled the cuda11_test brach on linux 64-bit and am running that on a GT8800 Nvidia card.
That version gave me about 80 candidates in 2 hours with the best shown below. It's way faster.

Just curious why this branch is so much faster.

[root@msieve-linux]# cat number.poly-1
n: 8570085955425991975416682393322092672178022622762559843295498103055560639891369587
# norm 6.077880e-15 alpha -6.591518 e 3.228e-12 rroots 5
skew: 14813998.57
c0: 271830330224374546052564805914266259840
c1: -928034253578711160424890240906
c2: -26681915219460396213323881
c3: 680113183828316262
c4: 141225440666
c5: 84
Y0: -2521959333809067094313594330127
Y1: 53520081350873741
type: gnfs

Last fiddled with by sleigher on 2011-11-12 at 19:36
sleigher is offline   Reply With Quote