Sorting feature exists, kind of half-finished. It was accidentally included on a release version and it worked quite bad, proving it's not ready yet.

Actually it was in an alpha build... but same difference.

Here's an example of one bug in it.




My guess is that they'll work on it for the 5.10.x release client (and it may be awhile since we haven't even started alpha testing for that release yet).

---

Kathryn :o)

On the boinc website page called 'Choosing boinc projects'

http://boinc.berkeley.edu/projects.php

the BBC Climate Change Experiment should be completely removed as it's no longer possible to attach to it.

---

On the boinc website page called 'Choosing boinc projects'

http://boinc.berkeley.edu/projects.php

the BBC Climate Change Experiment should be completely removed as it's no longer possible to attach to it.

Sent to David :)

---

Kathryn :o)

Thanks Kathryn.

---

I am running BOINC on a Macintosh. Here are a couple user interface suggestions:

1) In the menu bar there is a BOINC icon and when I click on it one of the options in the menu that pops up is "About BOINC Manager..." When I select that option, a dialog opens that reports the version number and offers a link to the main web page. I think the dialog should close when the user clicks on that link, in addition to closing when the user hits the "OK" button. Once you click on the link, the dialog disappears behind the web browser's window, and you can't get back to it by clicking on the menu bar icon again--that menu is inactive while the dialog is open. You have to hide the browser, or minify its window or something to bring the dialog back to where you can click on its "OK" button. That's a nuisance. I'm assuming that 99% of the time after jumping to the web site the user has no further use for the dialog. If the user wants to see the version number again while at the web site, they can always open the dialog again.

2) BOINC does not really need to be represented in the Dock. Dock space is precious, and BOINC is a "set it and forget it" application--it can be set to do its thing while the computer is otherwise inactive, and in that mode it gets out of the way immediately when the user starts to work. I can't speak for other users, but for my part I don't need to open BOINC Manager every three minutes to check whether the space aliens are sending me private messages. (Besides that, if they are not smart enough to post directly to my occipital cortex, I'm not interested in making their acquaintance. ;) So I think BOINC Manager should have a preference option to stay out of the Dock. The aforesaid icon in the menu bar serves perfectly well to open/switch to the application. If not appearing in the Dock torpedoes the application's menu bar (due to a quirk in OS X), the functionality of the menu bar should be transfered to buttons within the window.

Sometimes crunchers see the boinc manager message "Error on file upload: can't write file /home/boinc/data/*workunit number*: No space left on device".

Some crunchers worry that the 'device' may be their own computer's hard drive. Could the word be changed to 'server'?...

That would need modifications on the server kernel, as that's an error message returned by the system.

BOINC could search for that specific string and change it to 'server', but I'm wondering if it's really needed.

a more reasonable fix, IMHO, would be to change the front bit of the message, so the full message reads

Fileserver error on file upload: can't write file /home/boinc/data/*workunit number*: No space left on device.

I am suggesting that all server messages could usefully be disambiguated this way. At worst it is a waste of 13 bytes because the message is obviously from the server's insides, but at best it is worth those 13 bytes to reassure the newcomer that whatever the kernel is saying it's not their kernel and not their problem. And, in this case, it is easier to change all messages than screen for the really ambiguous ones.

River~~

---

It was already changed:

- file upload handler: say "out of space on server" instead of
"out of space on device"; users get confused by the latter,
think it refers to their disk

char* errmsg;
if (errno == ENOSPC) {
    errmsg = "No space left on server";
} else {
   errmsg = strerror(errno);
}

but yours is a quite interesting idea.

When our computers trickle to the server, we now get the message 'Scheduler RPC succeeded'. This isn't worrying, because what happened was clearly successful. But I'd guess that many crunchers don't know that RPC means 'remote procedure call'. Could the message please be made clearer for non-geeks eg

'Request to server succeeded' or

'Scheduler request succeeded' (as before)

---

I would like the Statistics and Disk displays to be changed to allow more than 35 projects. I presently run 37 different projects. It looks like the charts support more than 35 projects, the the adjoining ledgends don't.

---

Hi all.

If not been requested before...

I would like a checkbox saying: Make fixed assignment of BOINC jobs/tasks to one CPU/Core each.
That is, any BOINC computing process would run on one speciic CPU and only that, avoiding the processes to hop between CPUs, and thus being a bit more efficient on CPU cache use.

Thanks.

Hi all.

If not been requested before...

I would like a checkbox saying: Make fixed assignment of BOINC jobs/tasks to one CPU/Core each.
That is, any BOINC computing process would run on one speciic CPU and only that, avoiding the processes to hop between CPUs, and thus being a bit more efficient on CPU cache use.

Thanks.

This was discussed briefly on the developers list. Early this morning (my time) I saw some one had posted some numbers using Seti as a test bed. But at the moment, the results are escaping my brain. I'll try to find the email.

---

Kathryn :o)

Hi all.

If not been requested before...

I would like a checkbox saying: Make fixed assignment of BOINC jobs/tasks to one CPU/Core each.
That is, any BOINC computing process would run on one speciic CPU and only that, avoiding the processes to hop between CPUs, and thus being a bit more efficient on CPU cache use.

Thanks.

Somebody posted a message on the mailing list showing how making it run on a single CPU gives really tiny extra performance, that is, the difference is quite small compared to leaving it on its own.

If your operating system lets processes hop between CPUs, then it's a problem with your operating system. It has been proved Linux kernel 2.6 is good at NOT doing that.

And it's not just adding a checkbox, it's QUITE hard to make the client actually do what you want when the checkbox is enabled :)

Somebody posted a message on the mailing list showing how making it run on a single CPU gives really tiny extra performance, that is, the difference is quite small compared to leaving it on its own.

If your operating system lets processes hop between CPUs, then it's a problem with your operating system. It has been proved Linux kernel 2.6 is good at NOT doing that.

And it's not just adding a checkbox, it's QUITE hard to make the client actually do what you want when the checkbox is enabled :)

Yes, i know about the situation on Linux. The greater concern is Windows and the typical workload there. I would be interested in the results measured and also interested if the measurement has been done leaving the System alone crunching away and just few processes or in a situation of multiple apps opened, that is office usage pattern. As well as the number of cores in the system.

My expectation is that the gain increases with both, more cores and more active apps. IIRC there is an M$ recommendation out there for SQL admins on larger machines to bind SQL processes to CPUs for better performance.

Even a small gain multiplied but a lot of active BOINC clients in the world will be a bit of difference probably.

But if it's hard to do and gain is really tiny or it's plain not wanted, i'm certainly with you. I would just not like it being dropped for weak review.

Is there a way i could measure the difference myself using identical test WUs ?

I would like the Statistics and Disk displays to be changed to allow more than 35 projects. I presently run 37 different projects. It looks like the charts support more than 35 projects, the the adjoining ledgends don't.

I see that the ledgend can be scrolled using the next project button on the left. This is only true for the Statistics display.

---

Don't know if this is the place to suggest this, but for what it is worth, the next time changes are made to the design of the BOINC Manager, under the TASK descriptions, I'd like to see a column available that would show time remaining for the current project running. I'm not talking about time remaining until completion of the current project, but rather the amount of time that remains until the project that is currently running will switch to the next consecutive item on the list which will begin to run.

Don't know if this is the place to suggest this, but for what it is worth, the next time changes are made to the design of the BOINC Manager, under the TASK descriptions, I'd like to see a column available that would show time remaining for the current project running. I'm not talking about time remaining until completion of the current project, but rather the amount of time that remains until the project that is currently running will switch to the next consecutive item on the list which will begin to run.

I think the client can't predict when it will switch, it depends on way too many factors.

Don't know if this is the place to suggest this, but for what it is worth, the next time changes are made to the design of the BOINC Manager, under the TASK descriptions, I'd like to see a column available that would show time remaining for the current project running. I'm not talking about time remaining until completion of the current project, but rather the amount of time that remains until the project that is currently running will switch to the next consecutive item on the list which will begin to run.

I think the client can't predict when it will switch, it depends on way too many factors.

That is correct. The closest possible display would be something along the lines of "next projected reevaluation of running project". Even that would not always be accurate since there are events that trigger reevaluation immediately when they occur.

note: At any reevaluation the client may decide to continue running the same task.

---

BOINC WIKI

BOINCing since 2002/12/8

A button to Clear the messages would be nice.
I have enormous buffers that use up ram and I have to restart BOINC to clear them.
This has caused issues when running sensitive projects like Rosetta and would be nice to avoid.

---

The only reason some people get lost in thought is because it's unfamiliar territory.
--Paul Fix

A button to Clear the messages would be nice.
I have enormous buffers that use up ram and I have to restart BOINC to clear them.
This has caused issues when running sensitive projects like Rosetta and would be nice to avoid.

Even if the manager had a button to clear messages, the core client would still be keeping them all in memory. The way the manager asks for them is "give me all messages with ID > X", that way if it stops asking for some seconds, it won't miss any. For example, when you first start it, manager asks for messages with ID > 0, and client sends, say, 10 lines of messages. Next request (one second later) would be for ID > 10. If you restart the manager alone (or connect remotely, or use another 3rd party program to control BOINC, whatever) it would fetch all of them (by asking > 0). In order to provide that, the client is presumably keeping all messages in memory. Clearing those wouldn't be as simple as it sounds to program, and it could cause other problems.

Still, a way to clear messages on the manager alone would be nice to have, especially to make scrolling a bit easier :-

A button to Clear the messages would be nice.

Even if the manager had a button to clear messages, the core client would still be keeping them all in memory. The way the manager asks for them is "give me all messages with ID > X", that way if it stops asking for some seconds, it won't miss any. [...] Clearing those wouldn't be as simple as it sounds to program, and it could cause other problems.

As the Manager knows, that the most recent message's ID was e.g. 3765, it would be no problem at all to use this number for subsequent request like "give me all messages with ID > 3765", even after discarding them fully from memory ;-)

BTW, BoincView (which is doing it the same way) maintains only last few (currently 1000 per host IIRC, sometimes a bit low number for me) messages in memory and lets the user specify, how many of them to display.

Peter