Advanced search

Forums : Wish list : Progress Indicator
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile ohiomike
Avatar

Send message
Joined: 17 Jul 07
Posts: 302
Credit: 5,006,319
RAC: 0
Message 1236 - Posted: 20 Jul 2007, 21:19:57 UTC

I know this has been discussed elsewhere, but I think it should be in the "wish list" as opposed to a "problem" forum. I know it is in the works, but is would nice to have some visual clue of what is happening, especially with the new longer WU's (for those of us who don't have lives and just sit around and watch the BoincView displays).


Boinc Button Abuser In Training >My Shrubbers<
ID: 1236 · Report as offensive     Reply Quote
Profile ohiomike
Avatar

Send message
Joined: 17 Jul 07
Posts: 302
Credit: 5,006,319
RAC: 0
Message 1246 - Posted: 20 Jul 2007, 22:45:33 UTC - in response to Message 1236.  

I know this has been discussed elsewhere, but I think it should be in the "wish list" as opposed to a "problem" forum. I know it is in the works, but is would nice to have some visual clue of what is happening, especially with the new longer WU's (for those of us who don't have lives and just sit around and watch the BoincView displays).


One more thought on the progress indicator. I just re-started Boinc on one of my machines (increased the size of the partition that had the Boinc directory), and could not tell if the checkpointing picked up where it should have of not, the work timer started over at 0.

Boinc Button Abuser In Training >My Shrubbers<
ID: 1246 · Report as offensive     Reply Quote
Profile Mr.Pernod
Avatar

Send message
Joined: 15 Jul 07
Posts: 60
Credit: 841,431
RAC: 0
Message 1295 - Posted: 21 Jul 2007, 10:36:12 UTC

I just posted the same question regarding the checkpointing in a thread called question on checkpoints.

Back to the progress indicator.
If the checkpointing is based on flop_count it shouldn't be too hard to integrate a rudimentary indicator for the progress bar, even if it only increments by 1 or 2 points per checkpoint and then jumps to 100% when the unit is finished.
But then again, I have no intimate knowledge of the inner workings of BOINC :-(
ID: 1295 · Report as offensive     Reply Quote
Profile Bryan Siegfried
Avatar

Send message
Joined: 8 Jul 07
Posts: 26
Credit: 3,740,940
RAC: 0
Message 1374 - Posted: 22 Jul 2007, 15:21:58 UTC - in response to Message 1295.  
Last modified: 22 Jul 2007, 15:24:24 UTC

I just posted the same question regarding the checkpointing in a thread called question on checkpoints.

Back to the progress indicator.
If the checkpointing is based on flop_count it shouldn't be too hard to integrate a rudimentary indicator for the progress bar, even if it only increments by 1 or 2 points per checkpoint and then jumps to 100% when the unit is finished.
But then again, I have no intimate knowledge of the inner workings of BOINC :-(


We have mixed environment of desktop linux and windows at the office, but if the screensaver is set to boinc, doesn't C@H show the little boinc icon and a statement "0% of project done" or somesuch? Does checkpointing make it show 33% of project done, and so on? If you can attach your boinc manager to other clients, wouldn't this be seen all in one place? I just cehcked my boinc manager on the linux machine and it shows 44min of 3.5 hours, but still 0%. I guess maybe the % is not updating frequently, but the total time appears to be estimated well.

Why the need for a progress bar?
ID: 1374 · Report as offensive     Reply Quote
Profile Mr.Pernod
Avatar

Send message
Joined: 15 Jul 07
Posts: 60
Credit: 841,431
RAC: 0
Message 1380 - Posted: 22 Jul 2007, 18:58:28 UTC - in response to Message 1374.  

<SNIP>
I guess maybe the % is not updating frequently, but the total time appears to be estimated well.

Why the need for a progress bar?

Even though the total processing time seems well-estimated on the whole, I currently have a result that has been running for over 9 hours on a host with an estimated time to completion of 6 hours and roughly 30 minutes.
It would be nice to have at least some sort of visual progress indication in the BOINC manager (in this case BOINCview) to check if it is still processing, instead of having to connect to the host and check the filesystem for the access/change-timestamp of the checkpoint-file.
ID: 1380 · Report as offensive     Reply Quote
Profile ohiomike
Avatar

Send message
Joined: 17 Jul 07
Posts: 302
Credit: 5,006,319
RAC: 0
Message 1484 - Posted: 24 Jul 2007, 18:38:24 UTC
Last modified: 24 Jul 2007, 18:39:04 UTC

Now that we have checkpointing. If the camb task does 20 checkpoints/WU, could it poke that checkpoint # out to the wrapper (through a shared memory area or such) and use that to drive the progress indicator. It would jump 0, 5, 10.. but would still be helpful.

Boinc Button Abuser In Training >My Shrubbers<
ID: 1484 · Report as offensive     Reply Quote
Profile ohiomike
Avatar

Send message
Joined: 17 Jul 07
Posts: 302
Credit: 5,006,319
RAC: 0
Message 1574 - Posted: 26 Jul 2007, 13:25:03 UTC
Last modified: 26 Jul 2007, 13:51:10 UTC

I believe it would be trivial to implement a low-level progress indicator.
You already have the MSG_CHANNEL setup for it based on what I see with <app_msg_send> turned on.

It would require adding 1 var to either the wrapper or the app (I don't know which signals Boinc) :
int checkpoint_number;

On task startup:
checkpoint_number = 0; // new var, count checkpoints
checkpoint_cpu_time = 0;

On check point:
checkpoint_number++;
checkpoint_cpu_time = current_cpu_time;
fraction_done = 0.05 * checkpoint_number;

This would also add the checkpoint_cpu_time that I think the Boinc scheduler uses.

(ps- I think the var names are correct, they are what Boinc/SETI uses,
and I think you use the same msg structure).

Hum- I just noticed that you don't notify Boinc when you checkpoint. ??
That should be a call akin to :
update_app_progress(current_cpu_time, checkpoint_cpu_time );


Boinc Button Abuser In Training >My Shrubbers<
ID: 1574 · Report as offensive     Reply Quote
Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 1577 - Posted: 26 Jul 2007, 14:28:13 UTC

The problem is communication between CAMB and the wrapper, of which I believe there is none; the only thing the wrapper cares about is whether or not CAMB has finished running yet.

In order to report progress, I will have to implement CAMB as a stand-alone boinc application and ditch the wrapper. However, the implementation is turning out to be difficult, since linking between the boinc API and CAMB (which is written in fortran 90) is complicated.

Once we figure that out, though, it should be smooth sailing.
Scott Kruger
Project Administrator, Cosmology@Home
ID: 1577 · Report as offensive     Reply Quote
Profile Jayargh
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 25 Jun 07
Posts: 508
Credit: 2,282,158
RAC: 0
Message 1583 - Posted: 26 Jul 2007, 15:02:58 UTC - in response to Message 1577.  

The problem is communication between CAMB and the wrapper, of which I believe there is none; the only thing the wrapper cares about is whether or not CAMB has finished running yet.

In order to report progress, I will have to implement CAMB as a stand-alone boinc application and ditch the wrapper. However, the implementation is turning out to be difficult, since linking between the boinc API and CAMB (which is written in fortran 90) is complicated.

Once we figure that out, though, it should be smooth sailing.


Scott I believe it took 2 developers about a year to get the sixtrack application to work with Boinc at LHC many moons ago for the same reason(the Garfield application seems to be taking the same route though I'm not sure how much they are working on it.)So don't feel bad if it takes a little longer than you 1st thought.
ID: 1583 · Report as offensive     Reply Quote
Profile ohiomike
Avatar

Send message
Joined: 17 Jul 07
Posts: 302
Credit: 5,006,319
RAC: 0
Message 1584 - Posted: 26 Jul 2007, 15:03:58 UTC - in response to Message 1577.  

The problem is communication between CAMB and the wrapper, of which I believe there is none; the only thing the wrapper cares about is whether or not CAMB has finished running yet.

In order to report progress, I will have to implement CAMB as a stand-alone boinc application and ditch the wrapper. However, the implementation is turning out to be difficult, since linking between the boinc API and CAMB (which is written in fortran 90) is complicated.

Once we figure that out, though, it should be smooth sailing.


Open a shared mem area between the wrapper and camb?

Boinc Button Abuser In Training >My Shrubbers<
ID: 1584 · Report as offensive     Reply Quote
rbpeake

Send message
Joined: 27 Jun 07
Posts: 118
Credit: 61,883
RAC: 0
Message 1723 - Posted: 3 Aug 2007, 18:43:37 UTC - in response to Message 1583.  

The problem is communication between CAMB and the wrapper, of which I believe there is none; the only thing the wrapper cares about is whether or not CAMB has finished running yet.

In order to report progress, I will have to implement CAMB as a stand-alone boinc application and ditch the wrapper. However, the implementation is turning out to be difficult, since linking between the boinc API and CAMB (which is written in fortran 90) is complicated.

Once we figure that out, though, it should be smooth sailing.


Scott I believe it took 2 developers about a year to get the sixtrack application to work with Boinc at LHC many moons ago for the same reason(the Garfield application seems to be taking the same route though I'm not sure how much they are working on it.)So don't feel bad if it takes a little longer than you 1st thought.


They were also able to get LHC to not have to use HR, which paper you have probably seen linked on the BOINC web site. Do you also have plans to make CAMB not dependent on HR? That would sure help process units more quickly! (But of course would not be ideal if it resulted in processing units incorrectly! ;D)

ID: 1723 · Report as offensive     Reply Quote
Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 1856 - Posted: 8 Aug 2007, 22:14:42 UTC - in response to Message 1723.  

[They were also able to get LHC to not have to use HR, which paper you have probably seen linked on the BOINC web site. Do you also have plans to make CAMB not dependent on HR? That would sure help process units more quickly! (But of course would not be ideal if it resulted in processing units incorrectly! ;D)

No, HR is more or less necessary at the moment.

I've been working on progress indication and it seems that it might be ready fairly soon.
Scott Kruger
Project Administrator, Cosmology@Home
ID: 1856 · Report as offensive     Reply Quote

Forums : Wish list : Progress Indicator