Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Cullen/Woodall prime search :
New GCW sieve files generated
Author |
Message |
JimB Honorary cruncher Send message
Joined: 4 Aug 11 Posts: 920 ID: 107307 Credit: 989,666,286 RAC: 31,991
                     
|
We've just generated new GCW sieve files for all fourteen bases. What does that mean? It means that all candidates where a factor was found have been removed from the sieve file. Your computer then has fewer candidates to test, so the work will go faster. The sieve file is specified when we create the workunit, so it'll take an hour or two for these new sieves to come into use. Even though the work will go faster, your credit is unchanged.
As always with BOINC, these sieve files will be automatically downloaded as needed. As with the previous sieves, they're available as compressed files (.gz) for more recent BOINC clients. If your BOINC client supports that feature, it will automatically download the smaller files.
Don't expect a huge difference in sieving times. While the sieving will get faster, it won't be a dramatic decrease. Any future new sieves will be announced in this thread. Because it costs us bandwidth when new sieves are put into use, this won't happen very often - every few months at most. As we sieve higher and remove fewer candidates, the interval between new sieves will increase. | |
|
JimB Honorary cruncher Send message
Joined: 4 Aug 11 Posts: 920 ID: 107307 Credit: 989,666,286 RAC: 31,991
                     
|
New sieve files were again created since we're starting a GCW Sieve challenge in a few days. Everything said in the first post still applies. | |
|
|
Is it safe to delete older sieve files sitting in the projects directory, then?
____________
1 PPSE, 5 SGS, and 5 GFN-15 primes | |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 14037 ID: 53948 Credit: 477,954,885 RAC: 338,184
                               
|
Is it safe to delete older sieve files sitting in the projects directory, then?
Yes and no.
No is really the better answer here. You will still be getting a few "resend" tasks from older workunits that use the old sieve file, so you still actually need the old files for a while.
I think, however, it might technically be "safe" to delete them. I think BOINC will just re-download the sieve file if it needs it again for another WU. But I'm not 100% certain of this, and it's possible BOINC's behavior might vary depending on the version of BOINC that you have. So I won't guarantee that it's safe to delete them, but it probably is.
We pay for bandwidth, so I'm not crazy about people having to download a sieve file they already had, but after about 2 weeks the chance of getting an old task will be pretty low. So I'd wait a week or two before deleting them.
____________
My lucky number is 75898524288+1 | |
|
streamVolunteer moderator Project administrator Volunteer developer Volunteer tester Send message
Joined: 1 Mar 14 Posts: 1051 ID: 301928 Credit: 563,881,725 RAC: 1,034
                         
|
Is it safe to delete older sieve files sitting in the projects directory, then?
You must stop Boinc and edit client_state.xml too, carefully removing all <file>...</file> xml blocks referring these files. Otherwise Boinc will re-download "missing" files on next restart. Be sure that these filenames are not referenced anywhere else in client_state.
Mike already warned you about timeout necessary to avoid occasionally picking old task which still requires old files.
| |
|
|
Is it safe to delete older sieve files sitting in the projects directory, then?
You must stop Boinc and edit client_state.xml too, carefully removing all <file>...</file> xml blocks referring these files. Otherwise Boinc will re-download "missing" files on next restart. Be sure that these filenames are not referenced anywhere else in client_state.
Mike already warned you about timeout necessary to avoid occasionally picking old task which still requires old files.
Will the client automatically delete the old sieve files at some point?
If not, can you invoke the server's magickal powers to order the clients to delete the old sieve files please? Or do you already do this?
I know the RPC includes this facility. And the server knows better than any human when there is no more outstanding work for any given seive file.
R~~ | |
|
|
...We pay for bandwidth, so I'm not crazy about people having to download a sieve file they already had...
with up to 14 boxes on the same LAN, presumably it would have saved your bandwidth if I had installed a proxy so that the LAN only downloaded one copy of each file.
In fact it is worse than that, as I have re-initialised some of the boxes, some of them more than once. Maybe durint the challenge I downloaded 20x the files I could have done using a local proxy. With data files that are constant over a period of weeks, maybe I should look at doing the decent thing and caching the files locally.
A note about two other projects, both from about a decade ago.
Einstein@home used to use info about what files you already had when allocating tasks. If your prefs gave it a choice between A, B, and C but for historical reasons you only had the data files for B, you only ever got B work. Made sense for the project, not so popular with the users though, who wanted to see variety. Same might be true here, but if the costs are significant then it might save money to install file sensitive task allocation among the different GCW bases.
On the other hand, Rosetta explicitly ask their users NOT to cache, for some reason. You get sent a new copy of the data each time, even if it turns out to be identical. This used to irritate when I was on a metered connection. Not my problem now I don't pay by byte.
R~~ | |
|
streamVolunteer moderator Project administrator Volunteer developer Volunteer tester Send message
Joined: 1 Mar 14 Posts: 1051 ID: 301928 Credit: 563,881,725 RAC: 1,034
                         
|
Is it safe to delete older sieve files sitting in the projects directory, then?
You must stop Boinc and edit client_state.xml too, carefully removing all <file>...</file> xml blocks referring these files. Otherwise Boinc will re-download "missing" files on next restart. Be sure that these filenames are not referenced anywhere else in client_state.
Mike already warned you about timeout necessary to avoid occasionally picking old task which still requires old files.
Will the client automatically delete the old sieve files at some point?
No. They're are marked as "sticky", so client will never delete them. You have to delete them manually both from client_state.xml and from disk. Unlike "normal" files which are deleted automatically when no longer needed by any subproject (like old applications versions), sticky files will stay on disk forever.
Probably you may reset a project (completing all tasks first), so Boinc will delete whole directory (I didn't tried this myself).
If not, can you invoke the server's magickal powers to order the clients to delete the old sieve files please? Or do you already do this?
Last time I checked Boinc forums this problem was unsolved; there was no way to force deletion of old sticky files on all clients. Some project directories grown up to hundreds of megabytes...
| |
|
Message boards :
Generalized Cullen/Woodall prime search :
New GCW sieve files generated |