W_2_736 factors as:
[CODE]prp51 factor: 557407254155247166712738656005711709410906698630783
prp52 factor: 5560881101474937445745768799788053987340369105823599
prp58 factor: 1280564727679448478678574877582570543716410369024706830467[/CODE]

VictordeHolland - You should also check out Yafu. It has it's own suforum. It ties together msieve, ggnfs, gmp-ecm as well as many innate functions.

For post processing the NFS@Home jobs, you only need to use msieve. PM frmky for the url and credentials if you don't have them already. Once you reserve and download/extract/rename the appropriate files to the msieve directory, just run a single command (-nc) in msieve to run filter, LA and sqrt routines. The msieve readme has details.

One hard lesson - LA efficiency drops off a cliff with hyper threading. In your case use 3-4 threads. e.g.

[code]
msieve -v -nc -t 3
[/code]

Man, I just learned something myself. I've been using -t 8 for a quadcore with hyperthreading. I'll have to try the -t 4 and see how it looks.

I've contacted frmky and he provided me with an login for the data files.
So I'll take GC_8_246 .

[QUOTE=wombatman;357321]Man, I just learned something myself. I've been using -t 8 for a quadcore with hyperthreading. I'll have to try the -t 4 and see how it looks.[/QUOTE]

You will see a major improve.

I don't know about major, but the expected time went from ~20.5 hours to ~18 hours at the start. Once it finishes, I'll try 3 threads and see how that looks.

[QUOTE=wombatman;357406]I don't know about major, but the expected time went from ~20.5 hours to ~18 hours at the start. Once it finishes, I'll try 3 threads and see how that looks.[/QUOTE]

Turn off HT.

Ah, ok. I'll try that out tonight. Why does turning off hyperthreading give such an assist?

msieve has to do a lot of work to communicate between threads during linear algebra. With hyperthreading the overhead from this exceeds the extra CPU power.

Sieving on the other hand runs independently. So in my experience running 8 sievers in parallel on a system with 4 cores + hyperthreading gets about 25% more releations per minute than running 4 sievers.

Chris

Good to know. Unfortunately, I don't see an option in my BIOS to turn off hyperthreading. If I use 4 threads instead of 8, does hyperthreading play a significant role there?

Also, for what it's worth, I tried -t 4 and -t 8 just now to see what difference there might be in the time. With -t 4, the expected time was ~9h 15 min (and slightly climbing). With -t 8, it's ~8h and dropping. So maybe some of the issues with hyperthreading have been improved upon?

HT is automatically engaged - not sure there is a way to turn it off other than don't use more than 4 threads.

You're working with 30 bit lpb there. Try it with a 31 bit. LA can take a week on factorizations of that size. I'd be curious to see how much the ETA changes.

I also note you are running the latest msieve, which includes a huge improvement in LA performance. Perhaps this mitigates much of the HT effect. But I'm just guessing.

Tried to run your compiled binary for msieve but had problems. Will direct my questions to your other thread, if you are willing.