View Single Post
Old 2009-04-09, 07:09   #7
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

186916 Posts
Default

Quote:
Originally Posted by gamer007 View Post
I just started the search since I had some work queued. Then I noticed this in the console. Is this normal?
Oh, it looks like you're running these through the LLRnet client. Since this is non-LLRNet work, you should run it through the manual LLR program instead--and you'll pick up about a 5% speed boost to boot (which is due to the LLRnet client being based on a slightly older version of LLR).

You can get LLR at the following links:

Windows (GUI): http://jpenne.free.fr/llr3/llr371c.zip
Windows (command-line): http://jpenne.free.fr/llr3/cllr371c.zip
Linux (dynamically linked): http://jpenne.free.fr/llr3/llr371clinux.zip
Linux (statically linked): http://jpenne.free.fr/llr3/llr371cslinux.zip
Mac OS X (Intel): http://jpenne.free.fr/llr3/llr371cmac.zip

If you're running Windows, you probably want the GUI version. The command-line Windows version, as well as the Linux and Mac versions, run only from the command line, and thus can be slightly less user-friendly but are useful if you're using LLR as part of a script. (Of course, if you're running Linux or Mac you'll need to use the command-line version regardless since there isn't a GUI version available for those OSes.)

Assuming you're using the Windows GUI version, you can proceed as follows to run your range:

1. Download the program and extract it to a new folder.
2. Plop your file you downloaded from this thread in the same folder.
3. Run LLR.exe.
4. Choose "Input Data" from the Test menu.
5. In the dialog box that appears, enter the following information:
Input file (from NewPGEN): (the name of the file you got from this thread)
Output file: primes.txt
Starting line number: 1
6. Now click OK. You should see LLR's console output some messages indicating that it's started the first test. If it doesn't, click Test>Continue. (If it's already running, "Continue" will show up as "Stop" instead, so if you're not sure if it's actually running, that will tell you.)

Minimizing the window will send it down to the system tray. The tray icon will show green when LLR is running, or red when it's idle (such as after it's finished its current range). Double-click on the icon to make it come back up again.

When your range is complete, zip up lresults.txt and either attach it to a post here reporting completion of your range, or, if it's too big to attach here, then simply post completion here and email it to Gary at the email address provided in the first post of this thread. After doing this, you'll want to either delete or rename lresults.txt so as to not get it mixed up with results from any further ranges.

If you need to stop the program, simply close it (using the close button, or Text>Exit, or right-clicking on the tray icon and choosing Exit). It will automatically save its current state and exit. It automatically saves its state every 30 minutes in case it's interrupted unexpectedly. (You can change how often it saves its state in Options>Preferences. I'm pretty sure the minimum is 10 minutes, which is what I have mine set to.)

There are a couple other handy little options you can configure, such as Options>Start at Bootup (which installs, or uninstalls if it's already installed, LLR as a system service set to run at bootup). Also, the Options>Preferences box contains some other options such as "Iterations between screen output"--for example, you may want to set that to 1000 iterations if you want more frequent status updates on screen.

Hope this helps!

Max

P.S.: Oh, I almost forgot to mention about how to run the Linux, Mac, and Windows CLI clients. Essentially it's the same as with the GUI, except that you need to run the program as "./llr -m" to get a menu which shows all of the GUI's options in a phone-menu like display. (That is, "press 1 for Test>Input Data"...you get the picture. ) From the menu you can run Test>Continue to run the program, or exit the menu and run the program again as "./llr -d" to simply set it running on its current workload without displaying the menu. Simply running it as "./llr" is similar to with the -d option, except that its console output is silent (i.e. no onscreen progress updates).

P.P.S.: For a multicore system (regardless of OS), simply set up multiple directories containing separate copies of LLR, extract separate sieve files to each, and run them all individually.

P.P.P.S.: If you want to concatenate more than one file to run together on the same core, you can join two files together. However, make sure that they both have the same header line (the first line of the sieve file that looks like 4000000000000:M:1:2:258), and be sure to remove any additional such lines except the first one.

Last fiddled with by mdettweiler on 2009-04-09 at 07:15
mdettweiler is offline   Reply With Quote