Combined Sieve Guide Discussion

[Dagger]

Originally Posted by AMDave

Very much appreciated, Greenbank.

The chart reveals how much work has gone into recording the seiving runs so far and paints an interesting picture. Still a lot of white there for us to fill in.

I look forward to rating my little farm on the combined set in the next week.
I have managed 1350kp/sec on the non-combined. I have read of the decrease in kp/sec but everyone's combinations of CPUs is different so I wait to see.

Cheers.

I'm seeing a ~33% (PSP) and ~42% (SOB) decrease in sieving speeds. PSP alone was 520kp/s while SOB was 600 kp/s. I run a dualcore so I split my ranges in half. So would that be 1140 and 1200 kp/s respectfully? Right now i'm running 340 kp/s per core which is 680 total. Still 3 times faster than the last Reisel sieve i did.

AMDAthlon 64 X2 4200+
1GB RAM, 2.2GHz not OC'd

[Greenbank]

Originally Posted by Dagger

I'm seeing a ~33% (PSP) and ~42% (SOB) decrease in sieving speeds. PSP alone was 520kp/s while SOB was 600 kp/s. I run a dualcore so I split my ranges in half. So would that be 1140 and 1200 kp/s respectfully? Right now i'm running 340 kp/s per core which is 680 total. Still 3 times faster than the last Reisel sieve i did.

AMDAthlon 64 X2 4200+
1GB RAM, 2.2GHz not OC'd

Think of it this way:

You run PSP on a small range (lets say 100G) and it will take:-

100*10^9/520000 = 192307 seconds

You run SoB on the same 100G range and it will take:-

100*10^9/600000 = 166666 seconds.

So to do both you'll need 358973 seconds. But you've got a dual core so you run both at the same time (with a bit of fiddling as the SoB range will end first) you could do them in: 179486 seconds.

Or you could split the range in two, and do combined sieving:-

50*10^9/340000 = 147058 seconds.

So combined sieving is faster than doing both one after each other.

[VJS]

Originally Posted by Greenbank

k=265711 noskval=133752 mod 120: 0(45337) 24(46172) 48(42243)

k should be obvious.
noskval is the number of entries for that k in the dat file.

45337 of these numbers are n = 0 mod 120.
46172 of these numbers are n = 24 mod 120.
42243 of these numbers are n = 48 mod 120.

Hey Greenbank could you do your above analysis for the SoB k's?