sr2sieve client

[VJS]

Your numbers look about right.

As we decrease the number of k's we decrease the efficiency of the sieve but that's what it is all about. Less k's less factors... less k's more speed...

Considering this k wasn't that heavy and the number we have sieve is obviously still good to go.

BTW, thanks for checking and posting the speed increase.

[hhh]

BTW, how deep can we sieve with sr2sieve? I can put hilarious lines in the sr2work.txt, like
999999999-1000000000
and it still starts crunching (That's about 1000 times deeper than what I am crunching right now); but are the results still reliable up there?

H.

[geoff]

Originally Posted by hhh

BTW, how deep can we sieve with sr2sieve? I can put hilarious lines in the sr2work.txt, like
999999999-1000000000
and it still starts crunching (That's about 1000 times deeper than what I am crunching right now); but are the results still reliable up there?

H.

It should produce correct results for factors up to 2^62, or about 4.6e18.

It will not necessarily run efficiently at that depth though. There will be some slowdown of the modular multiplication routines between 2^61 and 2^62 due to branch mispredictions. Also I haven't tested the performance of the prime sieve at that depth (it might spend to much time sieving for prime candidate factors when it would be faster just to let some composite ones through).

[Lennart]

Originally Posted by dspdude2000

Oh, you're right. That seems needlessly confusing. I guess I'd have to run sr2sieve manually to make a comparison. Maybe I'll do that.

Here you have the WU info on PG.
sr2sieve -p3710472300e6 -P3710473075e6

/Lennart

[hhh]

Am I mistaken or should there be a speed advantage for 64bit?

I installed Ubuntu on a USB stick using the 64bit CD, and get 660 kp/s, instead of up to 700kp/s in Windows XP32, what am I doing wrong?

H.

[benjackson]

Originally Posted by hhh

Am I mistaken or should there be a speed advantage for 64bit?

I installed Ubuntu on a USB stick using the 64bit CD, and get 660 kp/s, instead of up to 700kp/s in Windows XP32, what am I doing wrong?

H.

Go to a console and type uname -m and make sure it's x86_64 to check that you're using 64bit linux. Otherwise, make sure that you're using the 64bit sr2sieve as well. You can do that by using readelf -a sr2sieve | grep Machine: and seeing if it says X86-64. What type of processor are you using? What version of Ubuntu, I'm assuming 8.04. There shouldn't be any other libraries you need installed so I'm not sure what to tell you. What type/speed processor do you have? Any C2D should be getting well above 660 per core in 64bit mode.

[hhh]

ja, ja, Hardy Heron, x86_64, Machine: Advanced Micro Devices X86-64. C2D T5600.

As I said, on pendrive, but persistent install. Installationwise, you can talk about a pain in the bottom. So many HowTo's out there, none works: you have to search your information at 5 different places and be imaginative. And of course, the keyboard layout is the only thing that isn't saved.

I'm a bit frustrated. I mean it's cool, and works, finally, but it's not what I imagined. But hey, right?

Back to my problem. I have not the slightest idea, but could it be that the processor is sleeping all the time at 1 GHz instead of 1.83, and that multiplied with the better performance, it gives me about the same thing as full throttle 32bit?

H.

[geoff]

Originally Posted by hhh

Back to my problem. I have not the slightest idea, but could it be that the processor is sleeping all the time at 1 GHz instead of 1.83, and that multiplied with the better performance, it gives me about the same thing as full throttle 32bit?

Try running sr2sieve with the -Z switch to prevent it lowering priority to idle. Newer Linux distributions run the CPU at low frequency if all running processes are at idle priority. You can disable this in the system configuration somewhere, or in the BIOS menus if that doesn't work.

I have thought about changing the default priority for sr2sieve, but I am not quite sure what the best default should be.

the 64-bit version should run at least 1.5x faster than the 32-bit version on Core 2 CPUs, closer to 2x faster when the n-range is very large as with PSP/SOB.

[geoff]

If you need a sanity check on the sieve speed, the last range I did was 12323060-12324060 which ran at about 1750 kp/s on a 2.66 GHz Core 2 Duo in 64-bit mode, or at about 3400 kp/s using two cores with the -t2 switch.

[hhh]

Thanks Lennart and geoff, the -Z swich did the trick. Now I'm getting 2Mp/s, instead of 1.3Mp/s.

But this pendrive install still doesn't convince me. Booting takes forever, for instance. One day, I will have to put a real one on the HD.

H.

[mdettweiler]

Originally Posted by geoff

Try running sr2sieve with the -Z switch to prevent it lowering priority to idle. Newer Linux distributions run the CPU at low frequency if all running processes are at idle priority. You can disable this in the system configuration somewhere, or in the BIOS menus if that doesn't work.

I have thought about changing the default priority for sr2sieve, but I am not quite sure what the best default should be.

the 64-bit version should run at least 1.5x faster than the 32-bit version on Core 2 CPUs, closer to 2x faster when the n-range is very large as with PSP/SOB.

Yeah, I've run into this same problem on Ubuntu 8.04 32-bit. My solution is to put one CPU Frequency Scaling Monitor on my taskbar for each core, and then use it to manually put the CPU into performance mode (i.e. no clocking down whatsoever) at each bootup. Through some research online I've traced this down to a flag in a certain file that's causing the problem--yes, you can fix it, but it will be back to "normal" at the next reboot. When I get around to it I'll report it on the Ubuntu message boards, hopefully they'll get it fixed in the next version (8.10).