Anyone have any trouble with the 15e ggnfs siever?
 

Had a weird experience tonight. I've been sieving this [URL="http://www.mersenneforum.org/showpost.php?p=227180&postcount=960"]poly[/URL] for a c170.

It's been going good until tonight. Twice tonight my system just powered itself right off. The first time I didn't think anything of it and restarted everything back up, including the sievers (two, on a dual-core). About 10 minutes in, the system shut itself down again. I didn't start anything else up except my local ECMnet server. So far after an hour or so, everything has been good. I'm about to start one of the sievers to see what happens, but I wonder if there could be any interaction between 15e, Vista64, and some strange corner case error with a sieve value; and whether anyone else has seen anything like this.

[QUOTE=schickel;229823]Had a weird experience tonight. I've been sieving this [URL="http://www.mersenneforum.org/showpost.php?p=227180&postcount=960"]poly[/URL] for a c170.

It's been going good until tonight. Twice tonight my system just powered itself right off. The first time I didn't think anything of it and restarted everything back up, including the sievers (two, on a dual-core). About 10 minutes in, the system shut itself down again. I didn't start anything else up except my local ECMnet server. So far after an hour or so, everything has been good. I'm about to start one of the sievers to see what happens, but I wonder if there could be any interaction between 15e, Vista64, and some strange corner case error with a sieve value; and whether anyone else has seen anything like this.[/QUOTE]

I don't think that this has to do with the 15e siever (at least not directly).

Possible causes of the problem which come into my mind:
* a brief power outage or current fluctuation - this might bring a PC to a reboot or switch off. (I have sometimes seen this happen a few times in a short period of time when e.g. the power supplier is experiencing a temporary problem and switching off (and later back on) one of the overland power lines.)
* a possible problem with the power supply of the PC
* a possible problem with the CPU fan: heavy loaded CPU is overheating and doing an emergency shutdown.

Have you tried to run other programs which are bringing heavy load to the PC (e.g. ECM or the 14e siever or prime95 torture test) to rule out a problem with your PC?

[QUOTE=Andi47;229826]Have you tried to run other programs which are bringing heavy load to the PC (e.g. ECM or the 14e siever or prime95 torture test) to rule out a problem with your PC?[/QUOTE]Well, I may have sussed it out. I think it might be a cooling issue. What threw me here was that after almost a year of perfect 24/7 100% load on both cores with nary a blip, I then had two shut downs tonight. I installed SpeedFan and based on the temp rise when I started one instance sieving, I opened the case and blew the dust out; there was actually very little, but I did raise a nice cloud.

Now, with two instances running, I get a max of 65-66 C, which is well below the 74C max on AMD's site.

I think I'll just chalk this up to a CPU cooler that got a little too much dust built-up and got overloaded....

[QUOTE=schickel;229828]Well, I may have sussed it out. I think it might be a cooling issue. What threw me here was that after almost a year of perfect 24/7 100% load on both cores with nary a blip, I then had two shut downs tonight. I installed SpeedFan and based on the temp rise when I started one instance sieving, I opened the case and blew the dust out; there was actually very little, but I did raise a nice cloud.

Now, with two instances running, I get a max of 65-66 C, which is well below the 74C max on AMD's site.

I think I'll just chalk this up to a CPU cooler that got a little too much dust built-up and got overloaded....[/QUOTE]
I've had this happen a few times on my own computer (Core 2 Duo E4500); it will spontaneously reboot, and keep doing so every 15-30 minutes until I actually shut it down, let it sit for a few minutes, and do a cold reboot. That always fixes it.

IIRC, this has only happened when the CPU's temperature is close to the limit (the real limit, 85 C in my case, not the lower thermal throttling limit that I expect is the 74 C figure in your case). But it's not an actual thermal shutdown; that would shut down and stay off.

In your case, did the computer shut down and stay off, or just reboot?

[quote=mdettweiler;229856]
In your case, did the computer shut down and stay off, or just reboot?[/quote]
BIOS settings can change this behavior I think.

[QUOTE=mdettweiler;229856]In your case, did the computer shut down and stay off, or just reboot?[/QUOTE]It was a solid shut down, requiring the power switch to restart. That's what threw me; if it had just rebooted, I might have thought of heat related trouble first.

[QUOTE=henryzz;229928]BIOS settings can change this behavior I think.[/QUOTE]Yeah, I'll have to spend a little time looking around in the BIOS to see what options there are. I just hate having the system offline not doing anything....

Sieve Region
 

Would someone please post the exact meaning of "14e", "15e", etc.
in terms of the sieve boundaries for each special q? What is the exact
sieve region for each of these instantiations of the code?

I believe the number is the power of two used for positive and negative c offsets in the (c,d) plane, and that d is always 2^12. So the 14e siever works on a +-2^14 x 2^12 grid.

[QUOTE=jasonp;229969]I believe the number is the power of two used for positive and negative c offsets in the (c,d) plane, and that d is always 2^12. So the 14e siever works on a +-2^14 x 2^12 grid.[/QUOTE]

the file INSTALL.and.USE in the ggnfs lasieve src (SVN 340) says this:

[QUOTE]
4) One of the parameters of the lattice siever has to be selected at compile
time. For a lattice sieving region of 4096x2048, (suitable for projects
with SNFS complexity 150, or GNFS for numbers with 110 digits),
type 'make gnfs-lasieve4I12e'. For 8192x4096 (suitable for SNFS complexity
around 180, or GNFS for numbers in the 130-140 digits range), type
'make gnfs-lasieve4I13e'. For 16394x8192 (for very large projects, like
GNFS on a number in the upper c150 or c160 digits ranke),
make gnfs-lasieve4I14e.

Note that the second dimension for the lattice siever is not fixed at compile
time. For instance, gnfs-lasieve4I13e can also be used with smaller sieving
areas like 8192x2048. However, it is best to use one of the parameters
indicated above.[/QUOTE]

Extrapolating, the 15e siever uses a region of size 2^15 x 2^14.

Well, I think I finally tracked it down.

After blowing the dust out, I had better temps for a couple of days. After that, though, the temps started creeping up. I backed off to running only one siever and stayed OK. Now, over the last couple of days, even the single instance (and surfing or doing anything else) have had the temps creeping back up. Checking the fans out this time, it looks like the fan on the CPU cooler is failing. In fact I think the fan is lying to the MB about its speed. I'm going to chalk the problems up to bad bearings on the fan causing a lot of.

Next step is to order at least a new fan. I think that I'm going to go ahead and buy a better heatsink/fan combo, though, especially since this system replaced my last one that died of a fried CPU (due to a dead fan, no less!)