I think I understand that the next step in the analysis is something like "Nebula" which requires a huge dataset all available at the same time.

I wonder if some theoretician could come up with some processing after the step we usually do, that we could do "on the side" as an experiment to replicate other post-processing? Or some entirely new post-processing that we could do by cloning our results and shipping them to another server?

I am ignoring the issue of who "owns" the date and/or results. Don't ruin my thread with "hard reality" :)

Tom

---

A proud member of the OFA (Old Farts Association).

Greetings,

Bummer, cannot copy and paste from the Tasks tab in BOINC Manager. :(

WTF?

ap_09no18ab_B0_P0_00383_20181112_11668.wu_1

The above WU is estimated to take "1d 4:43:xx" time to complete. Really? Has anyone else come across anything like this while not running a 486 PC? ;)

Have a great day! :)

Siran

[edit]
Forgot to mention: AstroPulse v7 7.00
[/edit]

---

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath

The above WU is estimated to take "1d 4:43:xx" time to complete. Really? Has anyone else come across anything like this while not running a 486 PC? ;)

If the machine has not yet completed 10 APs, this is the server estimate, which will be modified to a more reasonable number when servers have received 10 completed APs. I HAVE seen such estimates before finishing my 10 APs before.

Is this being run on CPU or GPU?

---

If the machine has not yet completed 10 APs, this is the server estimate, which will be modified to a more reasonable number when servers have received 10 completed APs.

With (I think it is) less than 10% blanking. One of my systems had to do about 22 CPU WUs before the estimates came good.

---

Grant
Darwin NT

The above WU is estimated to take "1d 4:43:xx" time to complete. Really? Has anyone else come across anything like this while not running a 486 PC? ;)

If the machine has not yet completed 10 APs, this is the server estimate, which will be modified to a more reasonable number when servers have received 10 completed APs. I HAVE seen such estimates before finishing my 10 APs before.

Is this being run on CPU or GPU?

Hi,

It's running on the CPU. I just found another AstroPulse v7 7.00 with an estimate over 1 day. I have others with a higher version number one @ 7.03, one @ 7.10, that are not estimated over a day. The 7.03, a CPU WU, is just over 6 hours and the 7.10, a GPU WU, is under an hour. Only the 7.00s are estimated over a day. Seems to me like I have done at least 10 AP WUs. Although, I haven't kept count of how many. ;)

Thanks Cruncher and have a great day! :)

Siran

---

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath

You can read exactly where that machine is at from this website:

Application details for host 8603647

You can read exactly where that machine is at from this website:

Application details for host 8603647

Hi Richard,

Well, I guess I haven't finished 10 APs then, according to that page. I've only done one. I could have sworn I had seen more APs over the past couple weeks. But, I guess I was hallucinating. ;) Is there a link to that page other than in these posts? I'd sure like to keep an eye or 2 on it. ;)

Thanks and have a great day! :)

Siran

---

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath

There are standard links for all of that.

https://setiathome.berkeley.edu/home.php will show your account, it'll show me my account.

Computers on this account https://setiathome.berkeley.edu/hosts_user.php will show you your computers, will show me my computers.

Computer details will show, well, details. This is the one we started with.

Application details show is back to where we started.

There are standard links for all of that.

https://setiathome.berkeley.edu/home.php will show your account, it'll show me my account.

Computers on this account https://setiathome.berkeley.edu/hosts_user.php will show you your computers, will show me my computers.

Computer details will show, well, details. This is the one we started with.

Application details show is back to where we started.

Hi Richard,

I know where those first 3 links are at, outside of posts in these fora. What I cannot find and want to know is where outside of posts in these fora is the 4th link, the Application details page, located. I would hate to have to come here and go through page after page in this thread to get to the link in the future.

I have looked on my account page, on my hosts page, on my tasks page and even the menus on the website main page. I cannot find that link, except in these posts from you.

Belay that, I found it. :)

Have a great day! :)

Siran

---

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath

The first three links have been there ever since we switched from Classic to BOINC. The fourth one was added later, about eight years ago (2010).

Greetings,

Ok, let me see if I'm getting the gist of this GPU processing and the use of the config file as a new GPU user and new config file user...

[1] To use a GPU for processing, a "app_info.xml" or "app_config.xml" file need to be created. Wait, the former needs creation, the latter is already there, just blank.

[2] To get the GPU to process WUs that are not being processed, one or the other of those files needs editing.

Would those statements be fairly accurate?

Here's the thing:

I noticed that my GPU was skipping over 2 AP WUs. I got to thinking about the 2 above mentioned files I've seen mentioned and looked into the info file I found in a different thread for parameters. All I could find was this:

    <app_version>
        <app_name>astropulse_v7</app_name>
        <version_num>705</version_num>
        <platform>windows_intelx86</platform>
        <avg_ncpus>0.04</avg_ncpus>
        <max_ncpus>0.2</max_ncpus>
        <plan_class>cuda_opencl_100</plan_class>
        	<cmdline></cmdline>	
        <coproc>
            <type>CUDA</type>
            <count>.5</count>
        </coproc>
        <file_ref>
            <file_name>AP7_win_x86_SSE2_OpenCL_NV_r2887.exe</file_name>
            <main_program/>
        </file_ref>
        <file_ref>
            <file_name>libfftw3f-3-3-4_x86.dll</file_name>
        </file_ref>
        <file_ref>
            <file_name>ap_cmdline_win_x86_SSE2_OpenCL_NV.txt</file_name>
            <open_name>ap_cmdline.txt</open_name>
        </file_ref>
    </app_version>

The pieces of that that I used was this:

	<app_version>
		<app_name>astropulse_v7</app_name>
		<plan_class>cuda_opencl_100</plan_class>
		<avg_ncpus>0.20</avg_ncpus>
		<ngpus>0.5</ngpus>
	</app_version>

When I read the config files into BOINC I got an error that an application was referenced but not found, or some such. Seeing in BOINCs tasks tab the app reference was "opencl_nvidia_100", so I modified my file above to:

	<app_version>
		<app_name>astropulse_v7</app_name>
		<plan_class>opencl_nvidia_100</plan_class>
		<avg_ncpus>0.20</avg_ncpus>
		<ngpus>0.5</ngpus>
	</app_version>

Now my whole "app_config.xml" file looks like this:

<app_config>
	<app_version>
		<app_name>setiathome_v8</app_name>
		<plan_class>opencl_nvidia_SoG</plan_class>
		<avg_ncpus>0.20</avg_ncpus>
		<ngpus>0.5</ngpus>
	</app_version>
	<app_version>
		<app_name>astropulse_v7</app_name>
		<plan_class>opencl_nvidia_100</plan_class>
		<avg_ncpus>0.20</avg_ncpus>
		<ngpus>0.5</ngpus>
	</app_version>
</app_config>

Now the GPU is processing the AP WUs it was skipping before. I'm sure that I'll end up getting more GPU WUs to be run by apps not configured in my "app_config.xml" file which means I will need to add more to it, correct?

Thanks to everyone here for inadvertently and deliberately helping me with this. I doubt I coulda done it withoutcha. :) ( Good thing I knew a little about xml files from working on my website. ;) )

Have a great day! :)

Siran

---

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath

It occurs to me that BOINC has a similarity to a paticular horse gait.


We are constantly "Racking".






Ouch, ouch, ouch...... please don't hit me with another awful pun.....





Sorry, I am so relieved to have my 2990WX up that I suddenly had an irresistible urge...

Tom

---

A proud member of the OFA (Old Farts Association).

Which mode of memory runs Seti faster? NUMA (Non-Uniform Memory Access) or UMA (Uniform Memory Access)?

I am asking for two reasons.

My Largest dual Intel has a NUMA toogle which is now on (I think). From what I am reading, games that can run/fit into 1 cpu thread's direct memory will run upto 5% faster.

And I am wondering how/if NUMA/UMA effects the processing speed when half the cpu cores need an extra jump to get to memory (like a 2990WX). I haven't actually found a NUMA/UMA toggle in my Amd MB bios though.

Thank you.

Tom

---

A proud member of the OFA (Old Farts Association).

NUMA(UMA) access through the BIOS is in the memory channel interleaving configuration. I can't remember which board you have. Depends on the board where it is the BIOS. For ASUS it is in the NBIO >>Common Options >

Setting the memory access to Channel mode is setting for UMA memory access. Setting for Auto is for NUMA access.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

NUMA(UMA) access through the BIOS is in the memory channel interleaving configuration. I can't remember which board you have. Depends on the board where it is the BIOS. For ASUS it is in the NBIO >>Common Options >

Setting the memory access to Channel mode is setting for UMA memory access. Setting for Auto is for NUMA access.

Thank you, I did find the setting. Under the dram stuff but apparently they got it backwards in the google lookup. They said "channel" turned on NUMA :)

Tom

---

A proud member of the OFA (Old Farts Association).

Which mode of memory runs Seti faster? NUMA (Non-Uniform Memory Access) or UMA (Uniform Memory Access)?

I am asking for two reasons.

My Largest dual Intel has a NUMA toogle which is now on (I think). From what I am reading, games that can run/fit into 1 cpu thread's direct memory will run upto 5% faster.

And I am wondering how/if NUMA/UMA effects the processing speed when half the cpu cores need an extra jump to get to memory (like a 2990WX). I haven't actually found a NUMA/UMA toggle in my Amd MB bios though.

Thank you.

Tom

Some results on turning off NUMA on my dual Intel box are in. Instead of pretty much running 1 hour and 12 minutes the cpu tasks started becoming more like the current results of my 2990WX. Some tasks started taking 2 odd hours. So I rebooted and turned NUMA back on :)

Tom

---

A proud member of the OFA (Old Farts Association).

I thought you were asking for the TR. It could very well be backwards for Intel. It all depends on how the BIOS writers define memory interlacing. So dependent on motherboard.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

I thought you were asking for the TR. It could very well be backwards for Intel. It all depends on how the BIOS writers define memory interlacing. So dependent on motherboard.

I was asking about both. Your response was clearly aimed at the 2990WX (at least clear to me).

In addition to seeing early results on the Intel MB I now have early results with memory interweaving = channel.

I now have 4 tasks that are taking above 4 hours at this moment. Since I haven't seen that level of outlier since I got a MB with a fully working memory system I believe I have empirical support for =channel turning off NUMA.

What is fascinating is it as a setting for "die" (I think this refers to cpu die) which sounds like it would "force" NUMA on cpus with direct local memory. The default is "auto" which means I don't know which it is actually doing :(

I guess I need to re-boot and byte the bullet and see if there is a visible difference between "auto" and "die".

Re-boot. re-boot...

Tom

---

A proud member of the OFA (Old Farts Association).

Anyone running Win 7 & got this weeks updates from MS?

Reason I ask is because of November's Malicious tool (KB890830). On starting its installation, a license agreement box popped up - Pre-release software for service pack 2 for Win XP? WTH? Clicked on cancel & then unchecked update.

Think someone within MS has lost the plot!

Strange. I can not find any recent issues reported on (KB890830). Most recent find was April 2017.

I now wait one or two weeks to install my M$ updates as I have been burned by M$'s "fixes".

I wait to get a Defcon rating that says it is OK to do updates from ComputerWorld's WOODY ON WINDOWS.

His blog Ask Woody is where the Defcon ratings and user feedback are located.

I choose the updates I want to install and NOT do the "Quality Preview Roll-up's" that M$ sends out usually on week 4 of the month. Did it once and had lots of trouble with very long boot times or the "BSOD" - Blue Screen Of Death at boot up. Never again.

---