I think I have
misunderstood the references to "multi-processor" L-L tests.
Dr Nowak's M42 was
found in 50 days on his 2.4GHz P4, and verified in 5 days on an Itanium 16p
machine, but it looks like the Itanium app was single-threaded. I have searched
all of http://www.oxixares.com/glucas and
found no refernece to parallel options.
As mentioned many
times in this forum and the Gimps pages, the L-L test is a consecutive linear
algorithm, you can't cut it into parallel pieces.
The 50/5 ratio above
is only about what one would expect from the P4/IA64 chips and associated 8x
cache size improvement etc, esp. as Glucas apparently exploits an IA64
instruction set feature which doubles the performance of
L-L.
So it looks like I
can't offer to verify M43 in a few minutes on a 512p IA64 after
all.
Greg
E
Greg
Edwards
Technical Consultant
SGI
11-17 Khartoum Road
North Ryde NSW 2113
Australia
Ph: +61 (0)2 8875 9580 (diverts to mobile)
Fx: +61 (0)2 8875 9480
Technical Consultant
SGI
Ph: +61 (0)2 8875 9580 (diverts to mobile)
Fx: +61 (0)2 8875 9480
_______________________________________________ Prime mailing list [email protected] http://hogranch.com/mailman/listinfo/prime
