I noticed this is "feature" for all new Boinc managers version 7.
It is general very good idea, but in one instance it is "little tricky"
If you are hurry to be prime number finder, then one second can be enough to "be late"
So please, is there any option, to disable downloading new WU until old is 100% finished?

Thanks for answer

Perhaps set minimum and maximum additional buffer to 0.0
Or set the network to suspend after it downloaded one task.

However, even with a fast computer it takes quite a while to finish the large prime numbers, so the second that you're 'too late' doesn't matter, as even with a second late you can upload & report yours before any resend can have finished the same task.

Of course, if your computer is super slow, then all bets are off. ;-)

---

All buffers are set to 0.
And setting network off every time I downloaded new WU is way to bad idea :(
You are wrong: on PPS and SGS prime hunting WU-s are finished in few minutes,so your reason goes off.

Opposite, my CPU is very fast, but, there will always be little faster :)
So there is no option, command, or anything like that to solve my problem?

1) Client configuration

<fetch_minimal_work>0|1</fetch_minimal_work>
Fetch one job per device (see --fetch_minimal_work).

2) Backup project (v6.10.30)

If a project has zero resource share, treat it as a "backup project":
fetch work from it only if there is an idle instance and no other projects have work.

You are wrong: on PPS and SGS prime hunting WU-s are finished in few minutes,so your reason goes off.

You were stating that you're one second late, which means the server still has to state that the task is end-of-deadline and it has to be sent out again. That doesn't happen instantaneously, it can take a while. But let's say it takes a minute. In all this time you have the chance to report your task, so you're still number one or the second one to report the work.

Only at end-of-deadline will this be a problem. As with 2 tasks sent out to two computers and them needing to wait for each other to be able to validate the task, it'll wait until you sent in yours anyway. Until, that is, you're outside the deadline.

And even if it's a minute late, I doubt that there are computers that run any of the Primegrid tasks in under a minute. Not on CPU only at least.

So we're always talking several minutes late. And well, if the project has a GPU app for the same tasks, either it has to put GPUs against GPUs or you'll have to update and get a super-fast GPU.

or if the project runs with too short deadlines, complain that they should increase them. I doubt that there's any worldly need in finding the biggest prime now or next week.

Or use Richard's options. Do know that the first will affect all projects you run. The second will not just get one task, but get one task for as many CPU cores that you have.

---

You are totally wrong: allow me to explain
Let say I and my wingman compute PPS ( on my computer finished about 4:30 minutes)
In the first phase, after 80% Wu is finished my Boinc client request new work.
So now I got new WU that wait about 40 seconds before start. But someone else ( third wingman) that is first WU, so he start to compute immediately. In case we have same speed CPU, he will finish WU just that 40 seconds before me :)
So that is catch....

<fetch_minimal_work>0|1</fetch_minimal_work>
this doesnot work: if set to 1 then even there is no left WU, new WU doesnot bee requested by manager.

What catch? Unless your cache is so darn big that your computer is unable to run all tasks by the deadline, there is no 3rd wingman.

PPS has a deadline of 3 days. You want to tell me that your computer cannot run two tasks that run in 4m30s a piece on your computer in 3 days? No, not if you have a minimum work buffer of 0 and a maximum of 7 days. But according to your own post, you've got your buffers set to zero. Which means that you should get at least one task per CPU core. Unless, of course you only just set the buffers to zero, but had zero and seven before that. No way we can check, since you have your computer(s) hidden.

I see that PPS has a initial replication of 1, and a minimum quorum of 1. So unless you manage to miss the 3 day deadline on either of these two tasks that run for 4 minutes and 30 seconds, then the task is sent out to a second computer, which may indeed get the task uploaded and reported before you do.

But then that means your computer is doing something completely different than run PPS, and then there's not much BOINC can do about that.

As for the client_configuration option to work, you need to load it into BOINC.
Do so by either exiting & restarting BOINC (the safest way), or BOINC manager->Advanced view->Advanced->Read config file. (Needs a BOINC 6.12 at minimum to work)

---

I will try to explain last time.
first of all why are you always talking about deadline? Wu is finished in 4:30 minute, so who talk about deadline?
For the last time
For SGS there is double checker, so that means two results per WU, right?
Ok
I download WU and my wingman download unit: we start to process that WU and then when my progress bar reach 80% my BOINC manager download another WU.

So that, another WU, have also wingmen, right?
And that wingman can immediately start to process WU, but I have to wait minute or so to start to process that WU.
And in that situation I "lost" minute, and if he has same CPu like me, he will return that result minute before me.
So noone speak about deadline....

And in that situation I "lost" minute, and if he has same CPu like me, he will return that result minute before me.
So noone speak about deadline....

So what? it doesn't matter if you return your result before or after his result, both are needed to validate the Workunit, and you'll both get credit,

Claggy

Yes... you know? The BOINC developers don't make just about anyone moderator on their forums, they want people who know a thing or two about how their software does its things. And where you explain things one last time, I don't mind explaining you again and again until you understand how BOINC does its things, such as how it validates work. Preferably before you continually dictate that others are wrong, when it's you yourself who is (making a fuss over nothing and an arse of himself).

So noone speak about deadline....

I did, all this time.

As when you do not go over the deadline of the task, in the case of SGS there won't be a third, who can theoretically take the credit away from you. Other than that, and as Claggy already said, it does not matter who comes first when returning the work, you or the SGS wingman.

In the case of PPS there won't be a wingman, as a minimum quorum and initial reproduction of 1 means that you're the only computer running it. Unless, of course, you go over the deadline of the task and a second is sent out. And then it depends on the cache settings of the other computer how long it takes before that one is returned...

In SGS' case the validator will wait for a second result before credit is given out, whether that second result comes from you, or --after the 3 day deadline has passed-- from a third wingman.
In PPS' case the validator will wait until you have returned the task, unless it's gone over the deadline, then it's sent to another computer, but even then you still have time to finish it before the other does.

BOINC can even cache a lot of tasks for both of these projects, it doesn't need to cache them one by one. So there's plenty of computers there which run with a 3 day cache, so it takes time for any of them to get this -over the deadline- task and eventually run it. More time than the first time second, more time than the now purported minute. More time even than the full run-time of the task.

---

And in that situation I "lost" minute, and if he has same CPu like me, he will return that result minute before me.
So noone speak about deadline....

So what? it doesn't matter if you return your result before or after his result, both are needed to validate the Workunit, and you'll both get credit,

Claggy

Claggy, did you ever make a single WU with BOINC?

Of course it does matter, and it does BIG metter,
If he return second result before me he will be prime finder, not I!
That is what i talking all the time!

Yes... you know? The BOINC developers don't make just about anyone moderator on their forums, they want people who know a thing or two about how their software does its things. And where you explain things one last time, I don't mind explaining you again and again until you understand how BOINC does its things, such as how it validates work. Preferably before you continually dictate that others are wrong, when it's you yourself who is (making a fuss over nothing and an arse of himself).

So noone speak about deadline....

I did, all this time.

As when you do not go over the deadline of the task, in the case of SGS there won't be a third, who can theoretically take the credit away from you. Other than that, and as Claggy already said, it does not matter who comes first when returning the work, you or the SGS wingman.

In the case of PPS there won't be a wingman, as a minimum quorum and initial reproduction of 1 means that you're the only computer running it. Unless, of course, you go over the deadline of the task and a second is sent out. And then it depends on the cache settings of the other computer how long it takes before that one is returned...

In SGS' case the validator will wait for a second result before credit is given out, whether that second result comes from you, or --after the 3 day deadline has passed-- from a third wingman.
In PPS' case the validator will wait until you have returned the task, unless it's gone over the deadline, then it's sent to another computer, but even then you still have time to finish it before the other does.

BOINC can even cache a lot of tasks for both of these projects, it doesn't need to cache them one by one. So there's plenty of computers there which run with a 3 day cache, so it takes time for any of them to get this -over the deadline- task and eventually run it. More time than the first time second, more time than the now purported minute. More time even than the full run-time of the task.

Why are you talking about deadline, if I dont talk about it?
I repeat question for first message in this topic:
Can or cannot BOINC manager stopped to download result BEFORE previous WU is finished. Version 6 have that as default settings, version 7 dont have it.
So answer is simple: yes or no?

Yes, with the setting that Richard showed you.

<cc_config>
<log_flags>
</log_flags>
<options>
<fetch_minimal_work>1</fetch_minimal_work>
</options>
</cc_config>

Make cc_config.xml with Notepad, save it as an All Files, format ANSI, in your BOINC Data directory. Make doubly sure it got the .xml extension, not an added .txt extension.
Next open BOINC Manager->Advanced view->Advanced->Read config file.

That will do it. Mind that it needs to run any remaining cache down first.
This option requires BOINC 6.12 or higher.

Oh, and next time you expect some help around here, make your intentions clear in the first post. There is absolutely no need to tell others off in the obnoxious way as you did. It'll only stop you from getting further help. I doubt that Claggy will ever take a second look at any of your posts, while others who enter this thread, or have been following it, will ever want to take you on again. Not with the way you think you can interact with people.

Now, as for knowing how BOINC works. Yep, I know how BOINC works.
Knowing as to how PG does its things, I must say I did not know that only the prime finder gets the credit as the prime finder, that the second person only gets named as verifier (although she/he will be named with computer system in the PDF as well). But luckily it doesn't say in my job-description that I must know everything about how all BOINC project do their things.

BOINC is more than just Primegrid. Not all projects follow the (strict) way that Primegrid seems to do about its discoveries.

So, now that credits are no longer making the man, but being first at being named the prime finder is, what's the next thing that people will need to fight about over BOINC?

---

I'm sure Claggy is too polite to say he's offended. We can see what he has crunched here. Quite a lot... He has also advised and helped many other BOINC crunchers for years.

<cc_config>
<log_flags>
</log_flags>
<options>
<fetch_minimal_work>1</fetch_minimal_work>
</options>
</cc_config>

I know how to make cc_config, already have one, and as I say before it doesnot not work.
Claggy is nice guy, just in this case he didnot see what I ask.

Oh, and next time you expect some help around here, make your intentions clear in the first post.

I think, reading my first post that I was perfectly clear for what I need help.
So please, is there any option, to disable downloading new WU until old is 100% finished?

Thanks for all reply and help, getting back to 6.12.34
Have a nice day

So please, is there any option, to disable downloading new WU until old is 100% finished?

Yes, and it is the fetch minimal work option.
This will fetch only 1 job per device (CPU, GPU).

If you don't get it to work, what --if any-- are the messages?
Does BOINC show after editing and re-reading of the config file that the <fetch_minimal_work> option is used? If not, check and recheck the extension and the place of the cc_config.xml file.

Look, you're making it quite difficult to get help by not offering one iota of information about what you see or don't see. Just stating that it doesn't work is insufficient information to get help on. Give a screen shot, show your messages, things like that. How much work does your computer still have in cache?

Unless you yourself give information, there's extremely little that any helper here can do to try to give help. And while I said earlier that I'll continue to try, even my attention span is waning fast, as I am getting fed up with having to pull information out of you.

---

If my english is so bad that you cannot understand what I ask, i is very simple to try yourself to see what will be happen when you use option you write...
Join to Primegrid, make cc_config as you statement, and choose PPS as project for finding prime numbers. Those WU will be finished in minutes, so you need to wait so long.

As I say before, when I use that option, and all buffers are set to 0, when all WU are finished, BOINC manager doesnot request any new work, until I remove that option from cc_ config

I think, reading my first post that I was perfectly clear for what I need help.

I have read each and every post of yours and I did NOT understand what you wanted to achieve and why, until you mentioned the prime finder for the first time waaaayy down the thread!

Gruß,
Gundolf

---

Computer sind nicht alles im Leben. (Kleiner Scherz)

I think, reading my first post that I was perfectly clear for what I need help.

I have read each and every post of yours and I did NOT understand what you wanted to achieve and why, until you mentioned the prime finder for the first time waaaayy down the thread!

Gruß,
Gundolf

From my first post, there is question: it is not important that only first get glory in prime finding, it is universal question

So please, is there any option, to disable downloading new WU until old is 100% finished?

This question is pretty clear, I hope?

So please, is there any option, to disable downloading new WU until old is 100% finished?

This question is pretty clear, I hope?

Yes, but you didn't say why you need such strange behaviour.
((almost) nobody can know there exists strange project which awards only the first returned task.
And is this true I still don't know (I have to check at the project's site but I'm not interested in a project which do something (award only the first) that no other BOINC project do).
)

If you ask "How can I cut my leg" everybody will say "But why? You can just wash it" ;)
Until you make it clear "I have cancer in the leg"

---

- ALF - "Find out what you don't do well ..... then don't do it!" :)