Phase V: ecm wrap

[rebirther]

I install the latest stable Client Version 6.4.5 and the ECM-WU's works fine. It was the pre-released installed Client from the Software-Packag, that they produced this Error.
THX for Help.

VG Volker

Iam running 6.2.15 (stable!) without any errors, 6.2.12 is also a dev version so dont expect too much from it.

[yoyo]

Again
- a 51 digit factor was found by marodeur6
- a 47 digit factor by Pitheque and macverk

yoyo

[Dennis Kautz]

Again
- a 51 digit factor was found by marodeur6
- a 47 digit factor by Pitheque and macverk

yoyo

Is there an overall list of found factors? If not, maybe we should create it in the wiki...

[Sysadm@Nbg]

this host crashed the last 6 workunits

stderr out
<core_client_version>6.4.5</core_client_version>
<![CDATA[
<message>
process exited with code 195 (0xc3, -61)
</message>
<stderr_txt>
wrapper: starting
wrapper: running ecm (-v -nn -timestamp -chkpnt checkpnt -inp in 110000000)
ecm: mul_fft.c:2534: mpn_mul_fft_full_a: Assertion `cc == 0' failed.
app exit status: 0x6
called boinc_finish
</stderr_txt>
]]>

any idea why ??
(meanwhile i stoped crunching via BAM, because I cant reached this host tody)

[yoyo]

No idea currently. I'll contact the ecm developers.
yoyo

[Ungelovende]

Today 5700 sec WUs arrived for ecm. They seems memoryloving, and uses up to 800 MB RAM EACH! On my AMDQuad with 4 GB I managed to get it working after closing all other programs, and edit the memorysettings in "General preferences" to 100%. Some of my teammates have problems getting enough RAM to get it work - boinc pauses the WUs. A 8 core or HT i7 needs more than 6 GB RAM for those WUs.

We are trying to heat our homes here in Norway. We are freezing with our cpus idle! Any chance of fixing this, or mix the long WUs with shorter WUs?

[X1900AIW]

We are trying to heat our homes here in Norway. We are freezing with our cpus idle!

In Germany we had - first time in this winter - to use our GPUs to help their CPU-brothers warming up our workrooms.

[yoyo]

Hi and Greetings to Norway,
I'm not realy sure why this is a problem with the big wus. The memory requirements for each wu are set to 800 MB (1GB in future). Boinc shoud not start such wu if less memory is free and boinc should also not download such wu.
I have some more of these monster wus. But now I will create some smaller batches between them. So work generation should generate round about
- 4 x 900 short wus
- 1 x 1800 monster wus
If you want to mix the wus, you can also run some ogr, muon or evo in parallel.
yoyo

[yoyo]

this host crashed the last 6 workunits

stderr out
<core_client_version>6.4.5</core_client_version>
<![CDATA[
<message>
process exited with code 195 (0xc3, -61)
</message>
<stderr_txt>
wrapper: starting
wrapper: running ecm (-v -nn -timestamp -chkpnt checkpnt -inp in 110000000)
ecm: mul_fft.c:2534: mpn_mul_fft_full_a: Assertion `cc == 0' failed.
app exit status: 0x6
called boinc_finish
</stderr_txt>
]]>

any idea why ??

Seems to be a real ecm bug.
If any body else has such Assertion error, please contact me.
yoyo

[Matthias Lehmkuhl]

I got this error too
resultid=2204558

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<message>
process exited with code 195 (0xc3, -61)
</message>
<stderr_txt>
wrapper: starting
wrapper: running ecm (-v -nn -timestamp -chkpnt checkpnt -inp in 110000000)
wrapper: delete file checkpoint_file
wrapper: running ecm (-v -nn -timestamp -chkpnt checkpnt -inp in 110000000)
wrapper: starting
wrapper: running ecm (-resume checkpnt -v -nn -timestamp -chkpnt checkpnt -inp in 110000000)
app exit status: 0x9
called boinc_finish

</stderr_txt>

My problem maybe was that I had two results at the same time that need round 1 GB memory at the end of the 20% step, one on my Vista 32 (Host 2GB RAM), and the other in my virtual Linux EM64T Kubuntu 8.04 Client (1GB RAM). The virtual machine runs in swapping modus.
But on the other hand, I suspended the virtual machine so that the Vista 32 could finish the 20% step and the crash in the Linux Box appears 10 mins after resuming the virtual machine.

Does the ECM need near 1 GB real RAM, or is Virtual RAM enough?

for now I changed the project preferences settings, and upgraded the virtual memory for the Linux box to 1152 MB (+128 MB)

[Matthias Lehmkuhl]

one point more
On my Linux Kubuntu 8.10 64 KDE4.1 client all 6 result crashed with this error

<core_client_version>6.2.15</core_client_version>
<![CDATA[
<message>
process exited with code 22 (0x16, -234)
</message>
<stderr_txt>
execv: No such file or directory

</stderr_txt>
results for hostid=22404
ldd don't show any missing libraries.

ogr runs without errors on this machine

[yoyo]

I got this error too
resultid=2204558

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<message>
process exited with code 195 (0xc3, -61)
</message>
<stderr_txt>
wrapper: starting
wrapper: running ecm (-v -nn -timestamp -chkpnt checkpnt -inp in 110000000)
wrapper: delete file checkpoint_file
wrapper: running ecm (-v -nn -timestamp -chkpnt checkpnt -inp in 110000000)
wrapper: starting
wrapper: running ecm (-resume checkpnt -v -nn -timestamp -chkpnt checkpnt -inp in 110000000)
app exit status: 0x9
called boinc_finish

</stderr_txt>

My problem maybe was that I had two results at the same time that need round 1 GB memory at the end of the 20% step, one on my Vista 32 (Host 2GB RAM), and the other in my virtual Linux EM64T Kubuntu 8.04 Client (1GB RAM). The virtual machine runs in swapping modus.
But on the other hand, I suspended the virtual machine so that the Vista 32 could finish the 20% step and the crash in the Linux Box appears 10 mins after resuming the virtual machine.

Does the ECM need near 1 GB real RAM, or is Virtual RAM enough?

for now I changed the project preferences settings, and upgraded the virtual memory for the Linux box to 1152 MB (+128 MB)

I saw this error on some hosts, but do not really know why it crashes. In each 20% step one ecm runs. At the end of this ecm run round about 850MB RAM is needed. I set this as requirement for the wu. It must not be real RAM, Virtual RAM is enough, but slow downs the performance.
yoyo