With the tapes containing AP work units appearing to being split at a decreasing rate, what is the best way to send v7 work to the GPUs only? I am trying to get some thing of a balance between AP tasks (CPU & GPU) and v7 GPU, and I don't want to stuff the queue with v7 CPU tasks. When the AP GPU tasks have run their course and none are available to d/l, I want to fill the queue with v7 GPU tasks only. There will be enough AP CPU tasks to carry me through any reasonable drought.

Also has anyone heard any word on when/if VLAR units will be allowed for GPU processing?

---

I don't buy computers, I build them!!

My SETI computing preferences are set at,

Run only the selected applications
SETI@home Enhanced: no
SETI@home v7: no
AstroPulse v6: yes
If no work for selected applications is available, accept work from other applications? yes

but my rigs are also not setup to run AP's on my video cards (no openCL apps installed).
This seems to give my a good balance of available work.

Cheers.

My SETI computing preferences are set at,

Run only the selected applications
SETI@home Enhanced: no
SETI@home v7: no
AstroPulse v6: yes
If no work for selected applications is available, accept work from other applications? yes

but my rigs are also not setup to run AP's on my video cards (no openCL apps installed).
This seems to give my a good balance of available work.

Cheers.

I'm trying very hard to keep this machine doing primarily AP work, so normally everything is no except for AstroPulse. Up to v7 there was never a problem getting AP work, keeping a steady 10/10 day limit. Even at the current limits it was extremely rare that I had to inject MB units to keep the machine full.

Now that we are at v7, this problem has become quite severe. In order to keep the GPUs busy during the AP droughts, I had to accept other work. Right now I have 47 v7 MB CPU tasks taking up space that normally would have been filled with AP units. Eventually they will be worked on to be sure. But, if I keep the accept other work on, then the opposite of what I am attempting to achieve will be in effect and the machine will turn into primarily a MB cruncher, getting AP units at a greatly reduced rate. If I have to accept other work, I much rather get only v7 GPU work then a mix of both.

---

I don't buy computers, I build them!!

'Twould be wonderful if one could set CPU preferences to AP only and GPU preferences to AP, but accept other work.
The ultimate solution...
Devs? What are the chances?

I wish.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

'Twould be wonderful if one could set CPU preferences to AP only and GPU preferences to AP, but accept other work.
The ultimate solution...
Devs? What are the chances?

I wish.

That is my big problem, when specifying accept other work you get everything. The only answer that I can think of is to edit the app_info.xml to remove everything but AP (openCL) and cuda50, but I'm not proficient enough to take the chance and completely screw things up. I guess I will just have to bite the bullet and create a new file to play with and hope I don't have everything that I currently have on the machine report back as being abandoned.

---

I don't buy computers, I build them!!

Ok, I've bitten the bullet, but I want to have someone to check this before I implement it. This is what I want to use as my app_info.xml after all v7 CPU work has completed on my machine. XMLMarker 2 states there are no errors, but another set of eyes doesn't hurt. This should allow me to process all AP tasks and only cuda50 will be allowed if I have to inject v7 into the machine.

<app_info>
<app>
<name>astropulse_v6</name>
</app>
<file_info>
<name>AP6_win_x86_SSE2_OpenCL_NV_r1843.exe</name>
<executable/>
</file_info>
<file_info>
<name>libfftw3f-3.dll</name>
<executable/>
</file_info>
<file_info>
<name>ap_cmdline_win_x86_SSE2_OpenCL_NV.txt</name>
</file_info>
<file_info>
<name>AstroPulse_Kernels_r1843.cl</name>
</file_info>
<app_version>
<app_name>astropulse_v6</app_name>
<version_num>604</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>.5</avg_ncpus>
<max_ncpus>.5</max_ncpus>
<plan_class>cuda_opencl_100</plan_class>
<cmdline></cmdline>
<coproc>
<type>CUDA</type>
<count>.5</count>
</coproc>
<file_ref>
<file_name>AP6_win_x86_SSE2_OpenCL_NV_r1843.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>libfftw3f-3.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>
<file_ref>
<file_name>AstroPulse_Kernels_r1843.cl</file_name>
</file_ref>
</app_version>
<app_version>
<app_name>astropulse_v6</app_name>
<version_num>604</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>.5</avg_ncpus>
<max_ncpus>.5</max_ncpus>
<plan_class>opencl_nvidia_100</plan_class>
<cmdline></cmdline>
<coproc>
<type>CUDA</type>
<count>.5</count>
</coproc>
<file_ref>
<file_name>AP6_win_x86_SSE2_OpenCL_NV_r1843.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>libfftw3f-3.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>
<file_ref>
<file_name>AstroPulse_Kernels_r1843.cl</file_name>
</file_ref>
</app_version>
<app_version>
<app_name>astropulse_v6</app_name>
<version_num>604</version_num>
<platform>windows_x86_64</platform>
<avg_ncpus>.5</avg_ncpus>
<max_ncpus>.5</max_ncpus>
<plan_class>cuda_opencl_100</plan_class>
<cmdline></cmdline>
<coproc>
<type>CUDA</type>
<count>.5</count>
</coproc>
<file_ref>
<file_name>AP6_win_x86_SSE2_OpenCL_NV_r1843.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>libfftw3f-3.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>
<file_ref>
<file_name>AstroPulse_Kernels_r1843.cl</file_name>
</file_ref>
</app_version>
<app_version>
<app_name>astropulse_v6</app_name>
<version_num>604</version_num>
<platform>windows_x86_64</platform>
<avg_ncpus>.5</avg_ncpus>
<max_ncpus>.5</max_ncpus>
<plan_class>opencl_nvidia_100</plan_class>
<cmdline></cmdline>
<coproc>
<type>CUDA</type>
<count>.5</count>
</coproc>
<file_ref>
<file_name>AP6_win_x86_SSE2_OpenCL_NV_r1843.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>libfftw3f-3.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>
<file_ref>
<file_name>AstroPulse_Kernels_r1843.cl</file_name>
</file_ref>
</app_version>
<app>
<name>astropulse_v6</name>
</app>
<file_info>
<name>AP6_win_x86_SSE_CPU_r1797.exe</name>
<executable/>
</file_info>
<file_info>
<name>libfftw3f-3.dll</name>
<executable/>
</file_info>
<file_info>
<name>ap_cmdline_win_x86_SSE_CPU.txt</name>
</file_info>
<app_version>
<app_name>astropulse_v6</app_name>
<version_num>601</version_num>
<platform>windows_intelx86</platform>
<cmdline></cmdline>
<file_ref>
<file_name>AP6_win_x86_SSE_CPU_r1797.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>libfftw3f-3.dll</file_name>
</file_ref>
<file_ref>
<file_name>ap_cmdline_win_x86_SSE_CPU.txt</file_name>
<open_name>ap_cmdline.txt</open_name>
</file_ref>
</app_version>
<app_version>
<app_name>astropulse_v6</app_name>
<version_num>601</version_num>
<platform>windows_x86_64</platform>
<cmdline></cmdline>
<file_ref>
<file_name>AP6_win_x86_SSE_CPU_r1797.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>libfftw3f-3.dll</file_name>
</file_ref>
<file_ref>
<file_name>ap_cmdline_win_x86_SSE_CPU.txt</file_name>
<open_name>ap_cmdline.txt</open_name>
</file_ref>
</app_version>
<app>
<name>setiathome_v7</name>
</app>
<file_info>
<name>Lunatics_x41zc_win32_cuda50.exe</name>
<executable/>
</file_info>
<file_info>
<name>cudart32_50_35.dll</name>
<executable/>
</file_info>
<file_info>
<name>cufft32_50_35.dll</name>
<executable/>
</file_info>
<file_info>
<name>mbcuda.cfg</name>
</file_info>
<app_version>
<app_name>setiathome_v7</app_name>
<version_num>700</version_num>
<platform>windows_intelx86</platform>
<plan_class>cuda50</plan_class>
<avg_ncpus>0.5</avg_ncpus>
<max_ncpus>0.5</max_ncpus>
<coproc>
<type>CUDA</type>
<count>.25</count>
</coproc>
<file_ref>
<file_name>Lunatics_x41zc_win32_cuda50.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart32_50_35.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft32_50_35.dll</file_name>
</file_ref>
<file_ref>
<file_name>mbcuda.cfg</file_name>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_v7</app_name>
<version_num>700</version_num>
<platform>windows_x86_64</platform>
<plan_class>cuda50</plan_class>
<avg_ncpus>0.5</avg_ncpus>
<max_ncpus>0.5</max_ncpus>
<coproc>
<type>CUDA</type>
<count>.25</count>
</coproc>
<file_ref>
<file_name>Lunatics_x41zc_win32_cuda50.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart32_50_35.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft32_50_35.dll</file_name>
</file_ref>
<file_ref>
<file_name>mbcuda.cfg</file_name>
</file_ref>
</app_version>
</app_info>

---

I don't buy computers, I build them!!

app_info looks fine, it starts and ends with the right tags, and has the correct tags between the different apps.

Claggy

app_info looks fine, it starts and ends with the right tags, and has the correct tags between the different apps.

Claggy

Thanks Claggy. I will suspend all AP work until the v7 CPU work is done, then implement the new app_info.xml. Hopefully it will resolve my problems.

---

I don't buy computers, I build them!!

My SETI computing preferences are set at,

Run only the selected applications
SETI@home Enhanced: no
SETI@home v7: no
AstroPulse v6: yes
If no work for selected applications is available, accept work from other applications? yes

but my rigs are also not setup to run AP's on my video cards (no openCL apps installed).
This seems to give my a good balance of available work.

Cheers.

I'm trying very hard to keep this machine doing primarily AP work, so normally everything is no except for AstroPulse. Up to v7 there was never a problem getting AP work, keeping a steady 10/10 day limit. Even at the current limits it was extremely rare that I had to inject MB units to keep the machine full.

Now that we are at v7, this problem has become quite severe. In order to keep the GPUs busy during the AP droughts, I had to accept other work. Right now I have 47 v7 MB CPU tasks taking up space that normally would have been filled with AP units. Eventually they will be worked on to be sure. But, if I keep the accept other work on, then the opposite of what I am attempting to achieve will be in effect and the machine will turn into primarily a MB cruncher, getting AP units at a greatly reduced rate. If I have to accept other work, I much rather get only v7 GPU work then a mix of both.

You can blame the credit imbalance for that and the few people who are now doing some....., look I can only call it cheating, to grab as many AP's as they can (there have been a couple of examples around here somewhere).

I was also surprised when I compared my 3570K rig to your 950 and seen the difference between the 2. Do you do anything else on it?

Cheers.

You can blame the credit imbalance for that and the few people who are now doing some....., look I can only call it cheating, to grab as many AP's as they can (there have been a couple of examples around here somewhere).

I was also surprised when I compared my 3570K rig to your 950 and seen the difference between the 2. Do you do anything else on it?

Cheers.

Remember that the 950 is primarily an AP machine and what v7 work that it is doing is only because there is no AP work available. I've had both of my machines set up this way for several years, one for AP and the other for MB. Since v7, it has kept my rac from completely dropping through the floor. My rac has dropped approx. 35 - 37k since v7 and I don't think I will ever recover unless the rules for new credit are changed to compensate for Autocorrelation or we get back to the point that there is more or less a steady supply of AP work. I wouldn't mind v7 so much if I can process VLARS on my GPUs.

The credit imbalance has nothing to do with what tapes are being split and what types of units are available. Yes, there are some of us that have multiple machines that run or attempt to run primarily AP on at least one of them, myself included, but I wouldn't call it cheating. It's a matter of preference, if I could I would run strictly AP on both machines.

---

I don't buy computers, I build them!!

In the end someone has to crunch those MB's so I'm happy to let my cards chew as many as they can, except VLAR's, but then I think I remember that we agreed to disagree on this point before (I thought that someone was going to make that an option for those that wanted to do them anyway). :-)

The only reason that I was interested in the difference bit between the 2 rigs is because that rig of mine just doing AP's on all its CPU cores (when it can get them) and MB's only on my 2 GTX660's giving it a RAC of around 36-37K.

Cheers.

Wiggo -

Why do you use BOINC Version 6.10.60?

Because I like it and living out in this part of the country (out in the mountain bush miles from anywhere) poses problems that make the later versions totally unsuitable for me (get those ridiculous back off times fixed and then I may consider upgrading). ;-)

Cheers.

Wiggo -

Why do you use BOINC Version 6.10.60?

Because I like it and living out in this part of the country (out in the mountain bush miles from anywhere) poses problems that make the later versions totally unsuitable for me (get those ridiculous back off times fixed and then I may consider upgrading). ;-)

Cheers.

That was fixed with the move to the co-lo.

---

Good point.

---

(some snippage)
You can blame the credit imbalance for that and the few people who are now doing some....., look I can only call it cheating, to grab as many AP's as they can (there have been a couple of examples around here somewhere).

I was also surprised when I compared my 3570K rig to your 950 and seen the difference between the 2. Do you do anything else on it?

Cheers.

A while back, it was discovered that APs didn't 'pay' as well as MBs. There was a great shift, of those who placed emphasis on a large RAC, to crunch
only MBs. So I and others that are more interested in seeing the science get done than having a high RAC saw that the APs needed to be crunched, decided to
crunch APs primarily. Some to the point of exclusivity.

When Seti@home V7 became the new MB it was found out that it didn't 'pay' as well as the old V6 or the APs. To make their flagging RACs proud again they about faced and began to primarily crunch APs.

Now this is legal under the present rules, though some consider this to be cheating, it isn't. If the powers that be think it is, they can change the rules to fix the 'problem'. I find it a bit sad that those who try to make as many points as possible are labeled cheaters when they are just using the program as it's written. It's not cheating if the rules permit it. I don't remember anyone being labeled a cheater when the RAC chasers exclusively crunched MBs and I and a few others crunched APs only.

It should be noted that burning through the APs as quickly as we presently are can create a bit of a log jam while AP crunchers have to wait for the MBs to get sent out so new tapes can be loaded and new APs split. This gets some people upset for some reason. Although why is hard to fathom. It's not as if the project has some sort of deadline to produce a BEM from outer space.

I can even imagine that some people are crunching APs only for the purpose of punishing the project for the paucity of credits being awarded to MB V7 work units.

It's a funny old world.

Wiggo -

Why do you use BOINC Version 6.10.60?

Because I like it and living out in this part of the country (out in the mountain bush miles from anywhere) poses problems that make the later versions totally unsuitable for me (get those ridiculous back off times fixed and then I may consider upgrading). ;-)

Cheers.

That was fixed with the move to the co-lo.

Not much good when the main problem here is my local exchange. ;-)

Cheers.

(some snippage)
You can blame the credit imbalance for that and the few people who are now doing some....., look I can only call it cheating, to grab as many AP's as they can (there have been a couple of examples around here somewhere).

I was also surprised when I compared my 3570K rig to your 950 and seen the difference between the 2. Do you do anything else on it?

Cheers.

A while back, it was discovered that APs didn't 'pay' as well as MBs. There was a great shift, of those who placed emphasis on a large RAC, to crunch
only MBs. So I and others that are more interested in seeing the science get done than having a high RAC saw that the APs needed to be crunched, decided to
crunch APs primarily. Some to the point of exclusivity.

When Seti@home V7 became the new MB it was found out that it didn't 'pay' as well as the old V6 or the APs. To make their flagging RACs proud again they about faced and began to primarily crunch APs.

Now this is legal under the present rules, though some consider this to be cheating, it isn't. If the powers that be think it is, they can change the rules to fix the 'problem'. I find it a bit sad that those who try to make as many points as possible are labeled cheaters when they are just using the program as it's written. It's not cheating if the rules permit it. I don't remember anyone being labeled a cheater when the RAC chasers exclusively crunched MBs and I and a few others crunched APs only.

It should be noted that burning through the APs as quickly as we presently are can create a bit of a log jam while AP crunchers have to wait for the MBs to get sent out so new tapes can be loaded and new APs split. This gets some people upset for some reason. Although why is hard to fathom. It's not as if the project has some sort of deadline to produce a BEM from outer space.

I can even imagine that some people are crunching APs only for the purpose of punishing the project for the paucity of credits being awarded to MB V7 work units.

It's a funny old world.

Do you abort V7 work just to get more AP's or move assigned CPU AP work to your video cards using a rescheduler to get more?

Unless you do, then I (and others) won't count you as a cheater. ;-)

Cheers.

The kitties crunch everything sent their way.
My preferences are set to prefer AP work, but when it is not available, they gladly work on MB until some more AP is split.

My interest in being able to limit the CPUs to AP only is that they are obviously MUCH slower at it than my GPUs, and a little AP will keep the CPUs busy for a long time.

No matter......
I personally am not going to go to the lengths of manually editing app_info on 9 computers to try to achieve this and then go through it again when the next Lunatics installer is released with new apps.

But I also have no problem if somebody else wishes to do so to process the work they choose to do.

If my wish came true and separate preferences were available for CPU and GPU, I would limit my CPUs to AP and they would rarely run out of it, but the GPUs which do most of my work would always still do both AP and MB.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

Switched over to the new app_info this morning and received 64 new cuda50 tasks as expected with no v7 CPU units. Now if only a way can be found to pre-split AP tasks at their origin so that when they get to Berkeley they are on their own tapes and we can have a better supply of them.

On another matter, after the switch I recycled the machine in my normal manner and one of the openCL tasks (3154423812) aborted 194 (0xc2) EXIT_ABORTED_BY_CLIENT. The first couple of lines in the stderr.txt is

<core_client_version>7.2.11</core_client_version>
<![CDATA[
<message>
finish file present too long
</message>
<stderr_txt>

Is there anything I can do so this doesn't happen again, or is it the client?

On still another matter - those of you that consider those of us that desire to primarily process AP tasks as "AP HOGS", I contend as being wrong and you do not understand the nature of the beast; especially this beast. I would rather spend 13 hrs. working on one type of task, then a couple of hours working on x of another type. Its not how many tasks you process in any 24 hour period, unless you are a "credit king/queen", it's how much science that you do in the same period. If you look at my machines you will see that the other primarily process MB and has one AP task that is stalled and probably sit there till &^(&^$ freezes over. As far as rac is concerned, as I stated in a previous post to this thread, mine has been cut in half since v7 and I'm just trying to keep my head above water and maintain a reasonable rac for my efforts. As the way things stand at this particular moment the 950 machine will probably spend more time on MB than AP, but that is the nature of the splitting beast and I will have no choice but to process everything.

---

I don't buy computers, I build them!!

On still another matter - those of you that consider those of us that desire to primarily process AP tasks as "AP HOGS", I contend as being wrong and you do not understand the nature of the beast; especially this beast. I would rather spend 13 hrs. working on one type of task, then a couple of hours working on x of another type. Its not how many tasks you process in any 24 hour period, unless you are a "credit king/queen", it's how much science that you do in the same period.

No worries, mate.
It's not like you are cherry picking amongst work sent to you and aborting what you don't like, as some have done in the past. You are simply choosing what you wish to process on your computers with your donated resources.
As do I, and many others on the project.
It's all good.

I know what runs best on my computers......believe me, I do...LOL.
And if I or you or anybody else wishes to set things up to get the most work that runs best on their computers.....go for it! When AP work is available, it's available to all. When it runs out, you can either choose to idle or do some MB work in the interim. No shame or foul in either choice.

Hope you get some answers to your other question.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster