GPU to 72 status...

[Uncwilly]

Originally Posted by Uncwilly

It looks like some Colab results are not being noticed and posted to PrimeNet.

This again for my TF, P-1 looks ok.

[storm5510]

One of my WorkFetch processes has stopped working for no apparent reason. The Low setting is 98,000,000. The High is 125,000,000. As a test, I adjusted the Pledge to 77. No change. I know I have a connection because Firefox is working normally.

[chalsall]

Originally Posted by storm5510

One of my WorkFetch processes has stopped working for no apparent reason. The Low setting is 98,000,000. The High is 125,000,000. As a test, I adjusted the Pledge to 77. No change. I know I have a connection because Firefox is working normally.

No change on the server-side.

Could you please give me more data to debug from? MISFIT or Manual, etc.

Developers hate "it doesn't work".

[storm5510]

Originally Posted by chalsall

No change on the server-side.

Could you please give me more data to debug from? MISFIT or Manual, etc.

Developers hate "it doesn't work".

I knew better...

Below is GPU72config:

StagingFile:C:\mfaktc\worktodo.txt
ReloadWhenBelow:1
FetchCount:10
Low:98000000
High:125000000
Pledge:75

#OPTION
#WhatMakesSense = 0
#LowestTFLevel = 1
#HighestTFLevel = 2
#LowestExponent = 3
#OldestExponent = 4

Option:2

ReplaceNAwithDate:False

URL:https://www.gpu72.com/account/getassignments/lltf/
UserID:storm5510
UserPWD:xxxxxx

Below is a batch file I use to automate the process:

@echo off
:top
cls
call mfaktc-2047.exe
if %errorlevel% gtr 0 goto errHandler
timeout 4

if exist c:\mfaktc\GPU72FETCH_logs\*.html (
del c:\mfaktc\GPU72FETCH_logs\*.html
)

call c:\gpu72\gpu72workfetch c:\gpu72\gpu72config.txt
if %errorlevel% gtr 0 goto errHandler
echo.
goto top


:errHandler
echo.
echo %date% %time% - Either mfaktc or GPU72 fetch has produced a non-zero exit code.
echo.

I know that mfaktc will not generate an error code for simply running out of work despite what it places on the screen. I do not know about gpu72workfetch.exe.

[storm5510]

This is an additional to my above posting. I was unable to edit it.

I did more digging on my own and I found something. I looked at the compiled code of gpu72workfetch.exe with a hex editor. Way down, I found a reference to MS Visual Studio 2010 -and- MS .Net Framework 4.0. That particular machine did not have .Net Framework 4 installed because I had to do a HD wipe and reload two weeks ago. gpu72workfetch.exe does not display anything about a missing dependency when started. It should. Yes, it is mentioned in the "readme" files, but I never looked at them. My automatic fetch is working now.

[DrobinsonPE]

I have been gone for a week but today I got Colab to do something for just long enough to complete a TF about 12 hours ago, 2020-11-29 15:41:20, Unfortunately the completed assignment is still sitting in the notebook results. Colab has not been working very nice lately so is the server holding a grudge against it?

[kladner]

Originally Posted by DrobinsonPE

I have been gone for a week but today I got Colab to do something for just long enough to complete a TF about 12 hours ago, 2020-11-29 15:41:20, Unfortunately the completed assignment is still sitting in the notebook results. Colab has not been working very nice lately so is the server holding a grudge against it?

On multiple runs today on different accounts, 35 minutes seemed to be the magic duration. These were CPU only. The few GPU runs did worse, though I did get a few completions.

[chalsall]

Hey everyone.

Just a heads up... I made a mistake with a manual SQL statement which corrupted 2.6 million historical assignment records.

If anyone sees anything strange with their overall summary reports, this is the reason. I'll restore from backups today.

No impact on current assignments nor workflow...

[ZFR]

Originally Posted by chalsall

If anyone sees anything strange with their overall summary reports, this is the reason. I'll restore from backups today.

No impact on current assignments nor workflow...

Thanks. I was going to ask why mine went wrong suddenly.

[KEP]

It seems something has changed for my manual colab. Currently I'm pulling about 520 GHzd/day on a T4. This means that I can do about 4 tests at n~103M from 76-77 bit and about 7.5 tests from 75-76 bit. What is the desired bit level we want to achieve currently?

I'm asking, because in 3 days, I need to reserve new work and I would like to do a manual GPU72 reservation of what makes most sense.

If 77 bit is desired target, I will reserve 50 and if 76 bit is the desired target as of now, I'll reserve 100 tests. In both scenarious I will return work once a day or at most every second day

Take care and lets hope my colab adventure last

... Now time to watch some SpaceX and hopefully a 12.5KM hop for starship

[Uncwilly]

Talk about SpaceX test moved to:
https://www.mersenneforum.org/showthread.php?p=565794