Hi again,

lost my bet they all validated perfectly. I don't mind the little ones they are as important as the big ones and they don't take long.

will we get the GPU-App for N-body automatically if it comes out ? :)

---

I'm getting all n-body V.84 and all are failing. What's up???

Hi, new in Milkyway

This is normal? WU:

http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=32886191

Less than 11 secs to compleate in a Q8300!!

Thanks

I'm using milkyway_nbody_0.84_i686-pc-linux-gnu__mt.
The problem is that my libc is not 2.11, but only 2.9 so:
milkyway_nbody_0.84_i686-pc-linux-gnu__mt: /lib/libc.so.6: version `GLIBC_2.11' not found (required by milkyway_nbody_0.84_i686-pc-linux-gnu__mt)

and all NBody simulation fails.
Is it possible have binaries compiled against an older version of libc?

Thanks in advance

Luca

Is it possible have binaries compiled against an older version of libc?

I thought I did? Apparently I did for the 64bit one and screwed up the 32 bit. I'll rebuild it in a bit.

Is it possible have binaries compiled against an older version of libc?

I thought I did? Apparently I did for the 64bit one and screwed up the 32 bit. I'll rebuild it in a bit.

Done

thanks Matt!
Goshy, I have another error and I don't know if you could help as before.
I've a missing library (libgomp), 'cause my linux environment is on Fedora10 based :(
you'd statically link the executable, but I guess it's not what you want to do, so
I'll ask linuxulator people!

thanks again
Luca

thanks Matt!
Goshy, I have another error and I don't know if you could help as before.
I've a missing library (libgomp), 'cause my linux environment is on Fedora10 based :(
you'd statically link the executable, but I guess it's not what you want to do, so
I'll ask linuxulator people!

thanks again
Luca

libgomp ships with it? It shouldn't be a problem

thanks Matt!
Goshy, I have another error and I don't know if you could help as before.
I've a missing library (libgomp), 'cause my linux environment is on Fedora10 based :(
you'd statically link the executable, but I guess it's not what you want to do, so
I'll ask linuxulator people!

thanks again
Luca

libgomp ships with it? It shouldn't be a problem

Apparently it didn't in the last one. See if it redownloads for you with it.

My machine (E7600 + GTX260) got a few N-Body Simulation v0.80 (mt) WUs, all finished after ~ 10 seconds. Got credits for it. So I guess they were well.

http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=43607717

Online they are named: MilkyWay@Home N-Body Simulation v0.80 (mt)

The Stderr output:
<core_client_version>6.12.33</core_client_version>
<![CDATA[
<stderr_txt>
<search_application> milkyway_nbody_0.80_windows_x86_64__mt.exe 0.80 Windows x86 double OpenMP, Crlibm </search_application>
Using OpenMP 2 max threads on a system with 2 processors
Could not load Ktm32.dll (1815): (null)<search_likelihood>-85.378413103860581</search_likelihood>
17:55:31 (1508): called boinc_finish

</stderr_txt>
]]>

Named 64bit app?

In the MW@h project folder is the milkyway_nbody_0.80_windows_intelx86__mt.exe app.

In BOINC it was shown that 2 CPUs are used for 1 WU.

(no app_info.xml file, all stock)

Everything O.K. here?

---

Everything O.K. here?

The names were wrong in the 0.80s but it's fine.

So I understand correct, at Duo CPUs (2 threads) the whole CPU is used for one n-body WU?
BOINC said 2 CPUs. (After looking to client_state.xml there are entries: <avg_ncpus>2.000000</avg_ncpus> and <max_ncpus>2.000000</max_ncpus>)
Or all n-body CPU WUs need 2 threads, independent which CPU?

Is really the whole CPU @ 100 % used (Duo CPU), or it's only a calculation thing for BOINC how to calculate the WU cache?

It was too fast for me to look in Task Manager.

If BOINC say 2 CPUs/1 WU for n-body, then the GPU stop the calculation, because stock it's 0.05 CPU/1 GPU for the MW@h GPU WUs.

So maybe it's possible to cancel the thread usage for CPU WUs (upper mentioned entries)? So the GPU wouldn't stop the calculation.

---

Got my first 0.84 WUs and they ran perfectly.

I also got some 0.80 WUs which seem to have gone "missing" on the server.
I remember reporting them one or two days ago. They don't show up on my BM anymore either. When I look online it says "in progress".

can someone please help me? maybe reassingn them?

http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31703697
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31477467
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31436242
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31702928
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31702298
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31690965
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31532889
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31700222
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31700221
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31700216
http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=31700189

Thanks in advance!

Additional.. little bit off topic.. sorry.

Since I calculate MW@h WUs, I find it a pity that on my E7600 + GTX260 I have max 12 WUs/whole machine.
It's 6 WUs/CPU(-thread), right?

It's possible to make it little bit better?
CPU and GPU WU limit in progress separately?

In past, sometimes I have 12 CPU WUs and no GPU WUs in BOINC, because of the combined WU limit in progress.
There I needed to un-/check always the CPU in the project settings, for to have CPU and GPU busy - very cumbersome..

It would be nice to have a CPU and GPU WU limit in progress separately here at MW@h.

E.g. currently S@h have 50 WUs/CPU(-thread) and 400 WUs/GPU (limit of tasks in progress).
So I guess here at MW@h it could be also possible (if the server software is the same). ;-)
I don't mean the same limit, maybe 12 WUs/GPU and 4 WUs/CPU(-thread).
But if currently the n-body lasts only 10 seconds.. then maybe currently also 12 WUs/CPU(-thread).

---

I'm running an i7 32 bit processor and was having trouble with V 0.84 WU's. I'm now getting V0.80's and they all run well! They are very, very short!!! 8 Sec, at best. Is this what's to be expected. I remember running some for several Hrs. for 2500 credits. Are these test units or is that all there is? Pick

Hi Matt,
I'm still using milkyway_nbody 86 and I get the same problem. it's not redownloaded :(

Relatively new to MW and have been crunching away until version 0.60 was upgraded to 0.80. Now none of my NBODY tasks work. All are end with computation error. My O/S is Linux 2.6.38-13-generic (UBUNTU 11.04) and my cpu type is AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 4400+ [Family 15 Model 107 Stepping 2] I've provided the results from the last one that failed.

Stderr output

<core_client_version>6.10.59</core_client_version>
<![CDATA[
<message>
process exited with code 193 (0xc1, -63)
</message>
<stderr_txt>
<search_application> milkyway_nbody 0.86 Linux x86 double OpenMP, Crlibm </search_application>
Using OpenMP 2 max threads on a system with 2 processors
SIGSEGV: segmentation violation
Stack trace (22 frames):
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt(boinc_catch_signal+0x123)[0x80bd27b]
[0x9c7400]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x8096fcc]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x80970f1]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x8084348]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x80822a1]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x8082c61]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x8081f1b]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x80843f0]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x807c761]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x807cb22]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x807ddcb]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt(lua_load+0x51)[0x807ae41]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt(luaL_loadstring+0x38)[0x808c468]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt(dostringWithArgs+0x2f)[0x806f5af]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt(nbOpenLuaStateWithScript+0x14b)[0x80692db]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt(nbSetup+0x24)[0x8069bd4]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x805e68d]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt(nbMain+0x34)[0x805ea64]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt(main+0x2c1)[0x805c8a1]
/lib/i386-linux-gnu/libc.so.6(__libc_start_main+0xe7)[0x126e37]
../../projects/milkyway.cs.rpi.edu_milkyway/milkyway_nbody_0.86_i686-pc-linux-gnu__mt[0x805cb45]

Exiting...

</stderr_txt>
]]>

Please keep in mind that I do NOThave anywhere close to the technical knowledge the others seem have on this forum. With that said, I need somebody's help out there so I can get back to normal.

Thanks,
ED.R

So I understand correct, at Duo CPUs (2 threads) the whole CPU is used for one n-body WU?
BOINC said 2 CPUs. (After looking to client_state.xml there are entries: <avg_ncpus>2.000000</avg_ncpus> and <max_ncpus>2.000000</max_ncpus>)
Or all n-body CPU WUs need 2 threads, independent which CPU?

mt means multi-threaded, it will use all cores you have.

Is really the whole CPU @ 100 % used (Duo CPU), or it's only a calculation thing for BOINC how to calculate the WU cache?

It should use both cores @100%, or lets say what's left of it, depending on what else the machine is doing, i.e. on your DualCore 1 MW_nbody WU should behave like 2 S@H WUs.

If BOINC say 2 CPUs/1 WU for n-body, then the GPU stop the calculation, because stock it's 0.05 CPU/1 GPU for the MW@h GPU WUs.

Should actually not happen, CPU is supposed to be overcommited by BOINC in that case. I'm also running 2 S@H WUs on the CPU and 2 MW WUs on GPU which makes 2.1 CPU total.

Since I calculate MW@h WUs, I find it a pity that on my E7600 + GTX260 I have max 12 WUs/whole machine.
It's 6 WUs/CPU(-thread), right?

It's possible to make it little bit better?
CPU and GPU WU limit in progress separately?

IIRC it should be 12 WUs per device (GPU or CPU core)... I think, but I'm not sure about the CPU, I didn't run Milkyway on CPU since... very long ago.
And besides the fact, that new WUs are created based on the returned ones and crunching old stuff is not so useful as fresh WUs, S@H is not the only project with hardware issues, so we are not getting more WUs, but since a while they stay at least long enough in the task list, so we can see if something was wrong.

---

thanks Matt!
Goshy, I have another error and I don't know if you could help as before.
I've a missing library (libgomp), 'cause my linux environment is on Fedora10 based :(
you'd statically link the executable, but I guess it's not what you want to do, so
I'll ask linuxulator people!

thanks again
Luca

libgomp ships with it? It shouldn't be a problem

Apparently it didn't in the last one. See if it redownloads for you with it.

With 0.88 everything here is fine!
Thanks!