The last message thread got a bit long and we now have a new Astropulse, so I thought i'd post updated instructions.


Setting up Seti@home to run cuda and non-cuda Optimised apps

Disclaimer:
If you are not comfortable with editing an app_info then this is not for you. If you run Windows you can use the Unified Installer for Windows available from the Lunatics web site. Click here

BOINC is very unforgiving of an incorrect app_info and usually will delete all tasks if you get it wrong.

Do NOT use Internet Explorer to edit the xml files, it will stuff up your app_info. Use Notepad or another text editor.

Notes:
Upgrade your BOINC client first and get it working before changing anything else. At the time of writing I am running the 6.6.37 client, although any client from 6.6.15 should be sufficient.

The app_info.xml below is based on a Windows XP platform (32 bit) and the cuda-capable card is a GTS250. If you are running on another platform you may need to add/amend the <platform_name> tags.

My computers support the SSSE3 instruction set. SSE2, SSE3 or SSE4.1 may be more appropiate for you. You will need to amend the program names as appropriate in the app_info.

I've assumed that you have your cuda-capable card up and running and have the necessary nvidia drivers (minimum version is 180.48). Use the 190.38 drivers (or later) if you want to use the cuda 2.3 dll files.


Programs needed:
a) Optimised multibeam and optimised Astropulse, available from the Lunatics web site. Click here
AK_v8_win_SSSE3x.exe
ap_5.03r112_SSE3.exe
ap_5.05r168_SSE3.exe

b) Cuda multibeam V12 and support libraries, available from the Lunatics web site. Click here
MB_6.08_CUDA_V12_noKill_FPLim2048.exe
cudart.dll
cufft.dll
libfftw3f-3-1-1a_upx.dll


Instructions:
1. Get BOINC working before changing anything else.

2. Download the optimised apps from the lunatics web site if you don't already have them.

3. Empty your cache of Seti@home work. This is best achieved by setting the project to No new work and letting it finish off its tasks. Make sure they are all uploaded and reported, there should be none on your tasks list. If you are feeling brave/confident then skip this step, but if all your tasks get deleted don’t say you weren’t warned.

4. Browse the BOINC log file to get the estimated speed of your GPU (or before you shut BOINC down, click on the messages tab). This is usually given at the top and is in Gflops. Some estimates from my testing are:
a) 9800GT = 60Gflops
b) GTS250 = 84Gflops
c) GTX260 (216 sp) = 96Gflops

5. Disable network communications in BOINC.

6. Shutdown BOINC. Make sure it and the science apps are shutdown.

7. Browse your client_state.xml file (its in the BOINC data directory) and look for the entry <p_fpops>. We need to use this number. Do NOT change this file.

8. For each of the apps multiply the p_fpops value by the factor below and put this into the appropiate flops entry in the app_info given below. For multibeam 608 you need the estimated Gflops. The app_info given below has the values for a GTS250.
Application Calculate
Astropulse 503 = p_fpops x 2.6
Astropulse 505 = P_fpops x 2.6
Multibeam 603 = p_fpops x 1.75
Multibeam 608 = Est.Gflops x 0.2

9. Make sure you have all the programs and DLL’s listed above in the projects\Setiathome.berkeley.edu folder. If not copy them there.

10. Save your app_info.xml in the projects\Setiathome.berkeley.edu folder.

11. Start up BOINC. Check the messages tab to see if it lists any [file error] messages. If there are shut BOINC down, check you have the correct program names referenced. Go back to step 9.

12. If okay then enable new work for the Seti@home project.

13. Enable network communications again.

14. BOINC should now download work of all types. If not check your Seti@home preferences on the Seti web site, that Astropulse_v5 and Use Graphics Processor (GPU) are all ticked. If you have a slower computer you may not get Astropulse work units anyway.

<app_info>
    <app>
        <name>astropulse_v5</name>
    </app>
    <file_info>
        <name>ap_5.03r112_SSE3.exe</name>
        <executable/>
    </file_info>
    <app_version>
        <app_name>astropulse_v5</app_name>
        <version_num>503</version_num>
        <flops>6131559081</flops>
        <file_ref>
            <file_name>ap_5.03r112_SSE3.exe</file_name>
            <main_program/>
        </file_ref>
    </app_version>
    <app>
        <name>astropulse_v505</name>
    </app>
    <file_info>
        <name>ap_5.05r168_SSE3.exe</name>
        <executable/>
    </file_info>
    <app_version>
        <app_name>astropulse_v505</app_name>
        <version_num>505</version_num>
        <flops>6131559081</flops>
        <file_ref>
            <file_name>ap_5.05r168_SSE3.exe</file_name>
            <main_program/>
        </file_ref>
    </app_version>
    <app>
        <name>setiathome_enhanced</name>
    </app>
    <file_info>
        <name>AK_v8_win_SSSE3x.exe</name>
        <executable/>
    </file_info>
    <file_info>
        <name>MB_6.08_CUDA_V12_noKill_FPLim2048.exe</name>
        <executable/>
    </file_info>
    <file_info>
        <name>cudart.dll</name>
        <executable/>
    </file_info>
    <file_info>
        <name>cufft.dll</name>
        <executable/>
    </file_info>
    <file_info>
        <name>libfftw3f-3-1-1a_upx.dll</name>
        <executable/>
    </file_info>
    <app_version>
        <app_name>setiathome_enhanced</app_name>
        <version_num>603</version_num>
        <platform>windows_intelx86</platform>
        <flops>4127010920</flops>
        <file_ref>
            <file_name>AK_v8_win_SSSE3x.exe</file_name>
            <main_program/>
        </file_ref>
    </app_version>
    <app_version>
        <app_name>setiathome_enhanced</app_name>
        <version_num>608</version_num>
        <platform>windows_intelx86</platform>
        <avg_ncpus>0.127970</avg_ncpus>
        <max_ncpus>0.127970</max_ncpus>
        <flops>16800000000</flops>
        <plan_class>cuda</plan_class>
        <file_ref>
            <file_name>MB_6.08_CUDA_V12_noKill_FPLim2048.exe</file_name>
            <main_program/>
        </file_ref>
        <file_ref>
            <file_name>cudart.dll</file_name>
        </file_ref>
        <file_ref>
            <file_name>cufft.dll</file_name>
        </file_ref>
        <file_ref>
            <file_name>libfftw3f-3-1-1a_upx.dll</file_name>
        </file_ref>
        <coproc>
            <type>CUDA</type>
            <count>1</count>
        </coproc>
    </app_version>
</app_info>

---

BOINC blog

A question on xml.
Is there any way to "rem" out lines in an xml file ? Or do you have to delete them and then put them back when you've finished playing ?

Brodo

A question on xml.
Is there any way to "rem" out lines in an xml file ? Or do you have to delete them and then put them back when you've finished playing ?

Brodo

In true xml, yes. In the bowdlerised version used for Boinc, I haven't found one so I tend to save the old one in a Notepad .txt file with a suitable name and as much explanatory text as I will need to remember its special characteristics and then edit the xml file.

F.

---

...
8. For each of the apps multiply the p_fpops value by the factor below and put this into the appropiate flops entry in the app_info given below. For multibeam 608 you need the estimated Gflops. The app_info given below has the values for a GTS250.
Application Calculate
Astropulse 503 = p_fpops x 2.6
Astropulse 505 = P_fpops x 2.6
Multibeam 603 = p_fpops x 1.75
Multibeam 608 = Est.Gflops x 0.2
[...]

<app_info>
    [...]
        <flops>6131559081</flops>
        <file_ref>
            <file_name>ap_5.03r112_SSE3.exe</file_name>
            <main_program/>
        [...]
        <flops>6131559081</flops>
        <file_ref>
            <file_name>ap_5.05r168_SSE3.exe</file_name>
            <main_program/>
        [...]
        <flops>4127010920</flops>
        <file_ref>
            <file_name>AK_v8_win_SSSE3x.exe</file_name>
            <main_program/>
        [...]
        <flops>16800000000</flops>
        <plan_class>cuda</plan_class>
        <file_ref>
            <file_name>MB_6.08_CUDA_V12_noKill_FPLim2048.exe</file_name>
            <main_program/>
        [...]
    </app_version>
</app_info>

IIRC, in past Richard made some tests with BOINC V6.6.14 or .15 and this app_info.xml .

He asked also to AMD user for to compare the multiplier values for the CPU.

I'm not up-to-date.. ;-)
This multiplier values are now the same for AMD and Intel CPUs?


BTW.
Because of the GFLOPS of the GPU.
BOINC say for example 100 GFLOPS for the GPU, this are 100,000,000,000 flops and this value x 0.2 .

IIRC, this was asked very much.

---

...
8. For each of the apps multiply the p_fpops value by the factor below and put this into the appropiate flops entry in the app_info given below. For multibeam 608 you need the estimated Gflops. The app_info given below has the values for a GTS250.
Application Calculate
Astropulse 503 = p_fpops x 2.6
Astropulse 505 = P_fpops x 2.6
Multibeam 603 = p_fpops x 1.75
Multibeam 608 = Est.Gflops x 0.2

IIRC, in past Richard made some tests with BOINC V6.6.14 or .15 and this app_info.xml .

He asked also to AMD user for to compare the multiplier values for the CPU.

I'm not up-to-date.. ;-)
This multiplier values are now the same for AMD and Intel CPUs?


BTW.
Because of the GFLOPS of the GPU.
BOINC say for example 100 GFLOPS for the GPU, this are 100,000,000,000 flops and this value x 0.2 .

IIRC, this was asked very much.

I did ask Richard about the multiplier for AP505 and he mentioned that it was decided by the powers that be to leave it the same as AP503.

---

BOINC blog

Any news yet on a new multiplier for MB6.08 with CUDA 2.3. I believe there was a remark in a thread a few days ago about this (Richard?).

---

The last message thread got a bit long and we now have a new Astropulse, so I thought i'd post updated instructions.

.....

7. Browse your client_state.xml file (its in the BOINC data directory) and look for the entry <p_fpops>. We need to use this number. Do NOT change this file.

8. For each of the apps multiply the p_fpops value by the factor below and put this into the appropiate flops entry in the app_info given below. For multibeam 608 you need the estimated Gflops. The app_info given below has the values for a GTS250.
Application Calculate
Astropulse 503 = p_fpops x 2.6
Astropulse 505 = P_fpops x 2.6
Multibeam 603 = p_fpops x 1.75
Multibeam 608 = Est.Gflops x 0.2

[/code]

I used the earlier v0.1 Unified Installer from Lunatics a month ago. It did not, nor have I manually edited/ added, any <flops>number</flops> for any apps in apps_info.xml file. My system seems to be crunching along just fine.

What is the purpose of those flops values? Is there some speedup advantage by specifying them and/or some disadvantage if flops is not specified?

---

Sabertooth Z77, i7-3770K@4.2GHz, GTX680, W8.1Pro x64
P5N32-E SLI, C2D E8400@3Ghz, GTX580, Win7SP1Pro x64 & PCLinuxOS2015 x64

What is the purpose of those flops values? Is there some speedup advantage by specifying them and/or some disadvantage if flops is not specified?

No, the flops values are used in the calculation of the estimated time to completion (along with the DCF) so not having them doesn't stop anything working. However, if they are there and are close to correct, then the estimates will not bounce around so much when different apps (AP / MB / CUDA) finish crunching.

F.

---

Any news yet on a new multiplier for MB6.08 with CUDA 2.3. I believe there was a remark in a thread a few days ago about this (Richard?).

I did miss this thread but after looking into completion time I'm using multiplier factor of 14/9 now to increase my old value (nearly 9 minutes with version 2.3 versus 14 min at 2.2 - 8800 Ultra)

DCF now is 1.10

I'll look further...

If you are using the cuda 2.3 DLL's then the current suggestion is to take the Est. Gflops figure and multiply by 0.5 to calculate the MB608 flops figure.

So in the sample one given, based upon a GTS250, you take 84000000000 x 0.5 = 42000000000 flops for multibeam 608.

---

BOINC blog

for 608 cuda 2.3 .dll's and MB_6.08_CUDA_V12_noKill_FPLim2048.exe for a GTX275/I7 920 i use x 0.012

CPU <p_fpops> = 3002658944.320127

608 <flops> = 236000000000.000000

which gives the approximate mid point between slow and fast WU and a duration_correction_factor in the range of >0.895 ~<1.1

---

My gtx275 is at 123 Gflops and my i7 920 is at 2568143784.797343 pflops.
your calcs are a bit out.

---

My gtx275 is at 123 Gflops and my i7 920 is at 2568143784.797343 pflops.
your calcs are a bit out.

my calcs are fine to get a duration_correction_factor of 1.0 (+ - ~11%) for 608 WU with cuda 2.3 DLL's & MB_6.08_CUDA_V12_noKill_FPLim2048.exe

My I7 920 = <p_fpops>3002658944.320127</p_fpops>

My gtx275 / 608 = <flops>236000000000.000000</flops> in the app_info.xml which is the average point between the fastest & normal slowest WU, VLAR I manually change to 603 CPU and my average for 603 flops is <flops>30800000000.000000</flops> (AK_v8_win_SSE41.exe) also matching a duration_correction_factor of 1.0 (+ - ~6%)

---

Could someone please post a sample MB/Cuda/AP app_info for Linux ?

TIA
Brodo

For my GTX 275 could I find the correct flops by reading this thread.

But how have I to translate my CPU ?

17/09/2009 23:48:21 Benchmark results:
17/09/2009 23:48:21 Number of CPUs: 4
17/09/2009 23:48:21 3593 floating point MIPS (Whetstone) per CPU
17/09/2009 23:48:21 7506 integer MIPS (Dhrystone) per CPU

Could someone convert this "MIPS" into (G)(P)flops that I need for app_info.xml ?

Thanks a lot !

---

Supporting BOINC, a great concept !

Unfortunately, there is no simple way to convert MIPS to FLOPS (or MFLOPS, or GFLOPS). In fact, MIPS is pretty much useless anyway and not directly comparable to FLOPS.

Here is an old writeup about the subject back when processors were still rated in MegaFLOPS instead of GigaFLOPS, but the information still holds true.

---

For my GTX 275 could I find the correct flops by reading this thread.

But how have I to translate my CPU ?

17/09/2009 23:48:21 Benchmark results:
17/09/2009 23:48:21 Number of CPUs: 4
17/09/2009 23:48:21 3593 floating point MIPS (Whetstone) per CPU
17/09/2009 23:48:21 7506 integer MIPS (Dhrystone) per CPU

Could someone convert this "MIPS" into (G)(P)flops that I need for app_info.xml ?

Thanks a lot !

First hit on Google, here, suggests a value of 1010 GFlops for the GTX 275.

F.

---

For my GTX 275 could I find the correct flops by reading this thread.

But how have I to translate my CPU ?

17/09/2009 23:48:21 Benchmark results:
17/09/2009 23:48:21 Number of CPUs: 4
17/09/2009 23:48:21 3593 floating point MIPS (Whetstone) per CPU
17/09/2009 23:48:21 7506 integer MIPS (Dhrystone) per CPU

Could someone convert this "MIPS" into (G)(P)flops that I need for app_info.xml ?

Thanks a lot !

You don't get a message like this when boinc starts up?
ATI GPU 0: RV770 (CAL version 1.4.317, 512MB, 1200GFLOPS)

I don't recall at the moment, but i thought my machien at work with a cuda card showed something like that as well.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the BP6/VP6 User Group today!

I think what Yeti wants to do is to convert the CPU MIPS values (drystone and whetstone) to GFLOPS so as to compare his CPU to his GPU.

---

For my GTX 275 could I find the correct flops by reading this thread.

But how have I to translate my CPU ?

17/09/2009 23:48:21 Benchmark results:
17/09/2009 23:48:21 Number of CPUs: 4
17/09/2009 23:48:21 3593 floating point MIPS (Whetstone) per CPU
17/09/2009 23:48:21 7506 integer MIPS (Dhrystone) per CPU

Could someone convert this "MIPS" into (G)(P)flops that I need for app_info.xml ?

Thanks a lot !

For 3593 floating point MIPS (Whetstone) per CPU, <flops>3593e+6</flops> is the simplest. 3.593e+9 or 3593000000 are other forms you could use.

                                                                Joe