Another new one to test for Linux, Mac and Windows

REMINDER TO ALL ALPHA TESTERS:
It's far easier for us to fix problems if you send message logs with the appropriate flags set. The main flags are:

<cpu_sched_debug>: problems involving the choice of applications to run.
<work_fetch_debug>: problems involving work fetch (which projects are asked for work, and how much).
<rr_simulation>: problems involving jobs being run in high-priority mode.

Use these flags from the cc_config.xml file.

Report any problems you get with it to the Alpha email list. This list needs registration.


The major change for the release and the reason for this release is the addition of the CPU Usage monitor configuration item in the advanced preferences dialog. We needed this to be available since most projects haven't upgraded their server software to support this preference item yet. Please try several different settings to see how it works. Please keep in mind it may take up to 20+ seconds before it engages.

----- Rom

Note from David

Specifically, the item is:
"Computing allowed
... While processor usage is less than ___ percent".

The default is 25%, meaning that BOINC will stop computing if non-BOINC CPU usage exceeds 25%. Either 100 or 0 remove the restriction and let BOINC compute whenever it otherwise would.

There have been reports that it has issues with cuda work fetch (ie it won't ask for cuda work).


Changes since 6.10.32 (6.10.33 for the Mac)

- WINSCR: cleanup configuration dialog text.

- Mac SS: change configuration dialog text to match new Windows wording.

- Mac installer: If installing an acct_mgr_url.xml file, don't install it directly into the BOINC Data directory. Instead, put it into the installer's resources and have the postinstall or postupgrade script copy it to the BOINC Data directory. This fixes a problem which would happen if the user later ran the standard BOINC installer without an acct_mgr_url.xml file: the Apple installer would then delete the file, causing BOINC to detach from the account manager.

- Mac Installer: add .txt extension to copyright files in GUI installer extras.

- Mac Installer: Fix receipt name for GridRepublic installer so we can remove it.

- graphics API: if font missing, find one that isn't (from Kevin)

- client: if GPU get available RAM fails, don't use the GPU

- client: detect CPU information via the cpuid instruction instead of the registry. This change is only for Windows but I haven't tested this with GCC.

- client: undo [17160]. <ncpus>0</ncpus> in cc_config.xml no longer means simulate zero CPUs. There are several places that divide by ncpus. Zero CPUs doesn't make any sense anyway.

- client: some more CPU detection cleanup.

- client: if a project is anonymous platform and it has no app versions that use a resource, don't request work from it for that resource.

- client: include CPU usage of client in BOINC total. Also include manager if its name includes "boinc" (can't figure out another way)

- manager: first whack at "suspend if CPU load exceeds x" in dialog

- screensaver: fix bug where sometimes no tasks are shown;

change font sizes for max of 2 sizes at a time

- client: don't accumulate LTD for projects w/ suspended jobs

- API: in txfRenderString, replace non-ASCII chars with ?. Otherwise it calls abort()!

- client: fix bug in "suspend if CPU load too high" feature. Forgot to convert between fraction and percentage

---

BOINC blog

There have been reports that it has issues with cuda work fetch (ie it won't ask for cuda work).

That was me. And only me, so far - we need confirmation checking and reporting to the list.

I have three CUDA hosts. None of them has downloaded any SETI 'work for CUDA' since I upgraded about 48 hours ago. I tested one with a downgrade to v6.10.32 - it fetched just fine (no change to app_info or any other settings). Two machines have successfully crunched work downloaded as CPU, but rebranded to CUDA. One machine is allowed to fetch SETI Beta work: it's been doing that, sporadically - I run Beta under app_info, too, so that's not the problem. All machines are downloading GPUGrid work without inhibition, even though it's grossly overworked after last week's outage here.

There have been reports that it has issues with cuda work fetch (ie it won't ask for cuda work).

That was me. And only me, so far - we need confirmation checking and reporting to the list.

I have three CUDA hosts. None of them has downloaded any SETI 'work for CUDA' since I upgraded about 48 hours ago. I tested one with a downgrade to v6.10.32 - it fetched just fine (no change to app_info or any other settings). Two machines have successfully crunched work downloaded as CPU, but rebranded to CUDA. One machine is allowed to fetch SETI Beta work: it's been doing that, sporadically - I run Beta under app_info, too, so that's not the problem. All machines are downloading GPUGrid work without inhibition, even though it's grossly overworked after last week's outage here.

Hi Richard, I will give it a shot and see if it lets mine download. Kinda full right now but with three cards in my machine it goes thru them pretty fast. Will watch my cache and let you know. 1st thing it did was put my existing work on hold and start some high priority.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

There have been reports that it has issues with cuda work fetch (ie it won't ask for cuda work).

That was me. And only me, so far - we need confirmation checking and reporting to the list.

I have three CUDA hosts. None of them has downloaded any SETI 'work for CUDA' since I upgraded about 48 hours ago. I tested one with a downgrade to v6.10.32 - it fetched just fine (no change to app_info or any other settings). Two machines have successfully crunched work downloaded as CPU, but rebranded to CUDA. One machine is allowed to fetch SETI Beta work: it's been doing that, sporadically - I run Beta under app_info, too, so that's not the problem. All machines are downloading GPUGrid work without inhibition, even though it's grossly overworked after last week's outage here.

Hi Richard, I will give it a shot and see if it lets mine download. Kinda full right now but with three cards in my machine it goes thru them pretty fast. Will watch my cache and let you know. 1st thing it did was put my existing work on hold and start some high priority.

When I did mine, the upload/download traffic was still absolutely maxxed out, and it was maintenance day as well. I had to be very careful to check that SETI wasn't requesting CUDA work, and then being backed off by repeated failures.

But

23-Feb-2010 13:26:39 [---] [wfd]: work fetch start
23-Feb-2010 13:26:39 [CPDN Beta] [wfd] FS: 0.500000 = 300.000000/600.000000
23-Feb-2010 13:26:39 [SETI@home] [wfd] FS: 0.500000 = 300.000000/600.000000
23-Feb-2010 13:26:39 [GPUGRID] chosen: major shortfall NVIDIA GPU: 0.00 inst, 10802.60 sec
23-Feb-2010 13:26:39 [---] [wfd] ------- start work fetch state -------
23-Feb-2010 13:26:39 [---] [wfd] target work buffer: 4320.00 + 43200.00 sec
23-Feb-2010 13:26:39 [---] [wfd] CPU: shortfall 0.00 nidle 0.00 saturated 175453.62 busy 0.00 RS fetchable 600.00 runnable 600.00
23-Feb-2010 13:26:39 [AQUA@home] [wfd] CPU: fetch share 0.00 LTD 0.00 backoff dt 0.00 int 0.00 (no new tasks)
23-Feb-2010 13:26:39 [climateprediction.net] [wfd] CPU: fetch share 0.00 LTD 0.00 backoff dt 0.00 int 0.00 (no new tasks)
23-Feb-2010 13:26:39 [CPDN Beta] [wfd] CPU: fetch share 0.50 LTD -374.51 backoff dt 0.00 int 0.00
23-Feb-2010 13:26:39 [Einstein@Home] [wfd] CPU: fetch share 0.00 LTD 0.00 backoff dt 0.00 int 0.00 (no new tasks)
23-Feb-2010 13:26:39 [SETI@home] [wfd] CPU: fetch share 0.50 LTD 0.00 backoff dt 0.00 int 0.00
23-Feb-2010 13:26:39 [SETI@home Beta Test] [wfd] CPU: fetch share 0.00 LTD 0.00 backoff dt 0.00 int 0.00 (no new tasks) (blocked by prefs)
23-Feb-2010 13:26:39 [GPUGRID] [wfd] CPU: fetch share 0.00 LTD 0.00 backoff dt 34804.91 int 86400.00
23-Feb-2010 13:26:39 [---] [wfd] NVIDIA GPU: shortfall 43210.40 nidle 0.00 saturated 4309.60 busy 0.00 RS fetchable 400.00 runnable 100.00
23-Feb-2010 13:26:39 [AQUA@home] [wfd] NVIDIA GPU: fetch share 0.00 LTD 0.00 backoff dt 0.00 int 0.00 (no new tasks)
23-Feb-2010 13:26:39 [climateprediction.net] [wfd] NVIDIA GPU: fetch share 0.00 LTD 0.00 backoff dt 0.00 int 86400.00 (no new tasks)
23-Feb-2010 13:26:39 [CPDN Beta] [wfd] NVIDIA GPU: fetch share 0.00 LTD 0.00 backoff dt 40587.36 int 86400.00
23-Feb-2010 13:26:39 [Einstein@Home] [wfd] NVIDIA GPU: fetch share 0.00 LTD 0.00 backoff dt 0.00 int 0.00 (no new tasks) (blocked by prefs)
23-Feb-2010 13:26:39 [SETI@home] [wfd] NVIDIA GPU: fetch share 0.75 LTD 0.00 backoff dt 0.00 int 0.00
23-Feb-2010 13:26:39 [SETI@home Beta Test] [wfd] NVIDIA GPU: fetch share 0.00 LTD 0.00 backoff dt 0.00 int 0.00 (no new tasks)
23-Feb-2010 13:26:39 [GPUGRID] [wfd] NVIDIA GPU: fetch share 0.25 LTD -500304.45 backoff dt 0.00 int 0.00 (overworked)
23-Feb-2010 13:26:39 [---] [wfd] ------- end work fetch state -------
23-Feb-2010 13:26:39 [GPUGRID] [wfd] request: 0.00 sec CPU (0.00 sec, 0.00) NVIDIA GPU (10802.60 sec, 0.00)

was pretty conclusive bug evidence.

6.10.35 was just released. ;-)

---

6.10.35 was just released. ;-)

Where? Just checked download section and .34 was still there.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

6.10.35 was just released. ;-)

Don't you mean 'compiled', rather than 'released'? Rom hasn't sent the email yet....

6.10.35 was just released. ;-)

Don't you mean 'compiled', rather than 'released'? Rom hasn't sent the email yet....

he has now, but at the time I posted that it was available for download already. Thus released.

---

6.10.35 was just released. ;-)

Don't you mean 'compiled', rather than 'released'? Rom hasn't sent the email yet....

he has now, but at the time I posted that it was available for download already. Thus released.

My download copy is timestanped 17:29, so at the time I posted I'd already downloaded it. I still call that 'escaped', rather than 'released' ;-)

I thought you were one of the people that advocate allowing the pre-alpha testers to catch any gross finger-fumbles in the download package that might cause installation to fail, before advising people to give it a test drive for BOINC functionality? I usually assume that Rom only emails once he's happy about that most basic of quality controls.

opps

---

Official Abuser of Boinc Buttons...
And no good credit hound!

Just had to put a GPU unit on hold as it was running two on GPU 0.......using the 6.10.35

---

Official Abuser of Boinc Buttons...
And no good credit hound!

I thought you were one of the people that advocate allowing the pre-alpha testers to catch any gross finger-fumbles in the download package that might cause installation to fail, before advising people to give it a test drive for BOINC functionality? I usually assume that Rom only emails once he's happy about that most basic of quality controls.

It doesn't help with you... you get it as soon as you see it available on the list, 9 minutes after it was uploaded even. So what does it matter then that I tell about it? I'd logged it into the Alpha releases thread as well, which I shouldn't then do either.

The changes were minute anyway, just the CPU registers and a fix for the preferences window. Boohoo, big change, no one should get it. But fine, I'll go post about it until it's well after a month that they've been posted about on the Alpha list. If that much.

And if anyone then wants to know what the changes were, go harass Richard through PM on the BOINC Dev forums. He's downloading them within minutes of 'escape', so he should know what the changes are.

---

I thought you were one of the people that advocate allowing the pre-alpha testers to catch any gross finger-fumbles in the download package that might cause installation to fail, before advising people to give it a test drive for BOINC functionality? I usually assume that Rom only emails once he's happy about that most basic of quality controls.

It doesn't help with you... you get it as soon as you see it available on the list, 9 minutes after it was uploaded even. So what does it matter then that I tell about it? I'd logged it into the Alpha releases thread as well, which I shouldn't then do either.

The changes were minute anyway, just the CPU registers and a fix for the preferences window. Boohoo, big change, no one should get it. But fine, I'll go post about it until it's well after a month that they've been posted about on the Alpha list. If that much.

And if anyone then wants to know what the changes were, go harass Richard through PM on the BOINC Dev forums. He's downloading them within minutes of 'escape', so he should know what the changes are.

And if you do that, you won't get a reply. You're welcome to PM me about a bug, or a feature, that you're worried about: if I can reproduce it, I'll post about it here, or at BOINC, or on the BOINC Alpha mailing list - sometimes more than one, depending on circumstances.

I watch the BOINC source code changesets (which I how I generally know there's a new compilation in the offing: it helps that Rom usually does them at a convenient time of day for me). But I'm not interested in a full list of all the social bells and whistles - Jord does an admirable job itemising those. I'm just concerned with whether they work or not.

And I'm selective in my posting. I won't mention a new release on a public board until I've tested it myself (which as it happens I haven't yet for v6.10.35) - unless it happens to be relevant to another discussion that I can't test, like CUDA on Mac OS X. Then my post will be full of 'allegedly' and other disclaimers.

And sometimes this approach is useful. Anyone remember v6.6.34? That had a red warning flag (high priority email) from me on boinc_alpha within 40 minutes of 'escape', because they'd introduced a catastrophic bug which trashed all CUDA work. But you didn't hear from me on this board for another four hours, by which time the bug had been confirmed, found and fixed.

And in case you were wondering about this 6.10.35 version mentioned above.

Changes since 6.10.34

This release contains the following three fixes:

- MGR: Fix a long-standing bug that made sizers fail on Mac for Advanced Preferences dialog: I added bSizer1->Fit(this) to end of constructor. To work around this bug, the dialog had been made resizeable and the height and width were saved in the Manager settings. But this meant that when we add an item to the dialog, it is too small until the user expands it. And Modal dialogs like this should not be resizeable. I also made the dialog no longer resizeable by the user. Remove second "between" in "Switch between applications between every".

- WIN: Add some missing CPU features to the feature list

- WIN: Make sure the Brand string for a CPU is printable. Older Intel processors have some control characters.

---

BOINC blog

Last night the 6.10.35 went into download heaven but only asked for CPU units. It got up to 2200 CPU units but only had 400 GPU units which I obviously use more of. I finally shut off new work and did a reschedule.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

Just had to put a GPU unit on hold as it was running two on GPU 0.......using the 6.10.35

If you change your cc_config.xml to the following, you'll be able to see which GPU Boinc assigns a particular Wu to:

<cc_config>
<log_flags>
<coproc_debug>1</coproc_debug>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

Then post your startup messages, and any Wu url's where this happens.

Claggy

Just had to put a GPU unit on hold as it was running two on GPU 0.......using the 6.10.35

If you change your cc_config.xml to the following, you'll be able to see which GPU Boinc assigns a particular Wu to:

<cc_config>
<log_flags>
<coproc_debug>1</coproc_debug>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

Then post your startup messages, and any Wu url's where this happens.

Claggy

If I see it happen again I will get the info to you.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

If you change your cc_config.xml to the following, you'll be able to see which GPU Boinc assigns a particular Wu to:

<cc_config>
<log_flags>
<coproc_debug>1</coproc_debug>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

Then post your startup messages, and any Wu url's where this happens.

Claggy

Surely the wrong thread but search didn't find another:

I'm trying to get two GPU running. Main: 275GTX. Secondary: 8500GT. Both CUDA capable, both seem installed OK, both using 190.38. cc_config:
<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

Messages show cc_config is read but only 1 GPU shows up (highlit with >>>>):

2/26/2010 3:28:25 PM Starting BOINC client version 6.10.18 for windows_intelx86
2/26/2010 3:28:25 PM Config: use all coprocessors
2/26/2010 3:28:25 PM log flags: file_xfer, sched_ops, task
2/26/2010 3:28:25 PM Libraries: libcurl/7.19.4 OpenSSL/0.9.8l zlib/1.2.3
2/26/2010 3:28:25 PM Data directory: D:\Boinc
2/26/2010 3:28:25 PM Running under account Art
2/26/2010 3:28:25 PM Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q9650 @ 3.00GHz [x86 Family 6 Model 23 Stepping 10]
2/26/2010 3:28:25 PM Processor: 6.00 MB cache
2/26/2010 3:28:25 PM Processor features: fpu tsc pae nx sse sse2 pni mmx
2/26/2010 3:28:25 PM OS: Microsoft Windows 7: Ultimate x86 Edition, (06.01.7600.00)
2/26/2010 3:28:25 PM Memory: 2.00 GB physical, 4.00 GB virtual
2/26/2010 3:28:25 PM Disk: 238.24 GB total, 232.24 GB free
2/26/2010 3:28:25 PM Local time is UTC -7 hours
>>>>2/26/2010 3:28:25 PM NVIDIA GPU 0: GeForce GTX 275 (driver version 19038, CUDA version 2030, compute capability 1.3, 896MB, 708 GFLOPS peak)
2/26/2010 3:28:25 PM Milkyway@home Found app_info.xml; using anonymous platform
2/26/2010 3:28:25 PM SETI@home Found app_info.xml; using anonymous platform

Any help from the experts?
Thanks
Art

---

Surely the wrong thread but search didn't find another:

I'm trying to get two GPU running. Main: 275GTX. Secondary: 8500GT. Both CUDA capable, both seem installed OK, both using 190.38. cc_config:
<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

Messages show cc_config is read but only 1 GPU shows up (highlit with >>>>):

2/26/2010 3:28:25 PM Starting BOINC client version 6.10.18 for windows_intelx86
2/26/2010 3:28:25 PM Config: use all coprocessors
2/26/2010 3:28:25 PM log flags: file_xfer, sched_ops, task
2/26/2010 3:28:25 PM Libraries: libcurl/7.19.4 OpenSSL/0.9.8l zlib/1.2.3
2/26/2010 3:28:25 PM Data directory: D:\Boinc
2/26/2010 3:28:25 PM Running under account Art
2/26/2010 3:28:25 PM Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q9650 @ 3.00GHz [x86 Family 6 Model 23 Stepping 10]
2/26/2010 3:28:25 PM Processor: 6.00 MB cache
2/26/2010 3:28:25 PM Processor features: fpu tsc pae nx sse sse2 pni mmx
2/26/2010 3:28:25 PM OS: Microsoft Windows 7: Ultimate x86 Edition, (06.01.7600.00)
2/26/2010 3:28:25 PM Memory: 2.00 GB physical, 4.00 GB virtual
2/26/2010 3:28:25 PM Disk: 238.24 GB total, 232.24 GB free
2/26/2010 3:28:25 PM Local time is UTC -7 hours
>>>>2/26/2010 3:28:25 PM NVIDIA GPU 0: GeForce GTX 275 (driver version 19038, CUDA version 2030, compute capability 1.3, 896MB, 708 GFLOPS peak)
2/26/2010 3:28:25 PM Milkyway@home Found app_info.xml; using anonymous platform
2/26/2010 3:28:25 PM SETI@home Found app_info.xml; using anonymous platform

Any help from the experts?
Thanks
Art

Have you got a monitor or a "dummy plug" on your 8500?

F.

---

Fred - No monitor on the 8500. I don't know what a dummy plug is so probably not.
THanks and regards
Art

---