LLR Version 3.8.20 released

[Jean Penné]

Originally Posted by pinhodecarlos

Have an issue with this latest version.
First let me share my llr.ini content:

WorkDone=0
Pid=0
Work=0
PgenInputFile=input.txt
PgenOutputFile=output.txt
PgenLine=20
OutputIterations=100000
ResultsFileIterations=99999999
DiskWriteTime=15
TwoBackupFiles=1
Priority=1
HeaderLine=0
Error_Count=2

My issue is when I stop the client and resume it afterwards the client will start from first line of the input.txt and not resume where it was previously. Can anyone doublecheck if this occurs? I've replaced client back to previous version (3.8.17) and it works, picks up where it was left.

TIA

It seems I reproduced the issue and it is machine dependant...
Please let me know on what machine you are working, and send me more details about your input data.
Maybe a problem with IniGetString and/or IniWriteString functions...

Regards,
Jean

Sorry, my test was fool! I did not reproduce the issue...
Jean

[ATH]

Maybe you are using LLR in a folder without write permissions? So it cannot update the ini file.

You should probably not use a sub folder to C:\Program Files or C:\Program Files(x86) on Windows 8 or 10.

[pinhodecarlos]

I'm on Windows 7 Enterprise SP1 64 bits, Processor is i7-3630QM.
Example attached, you just need to add the client into the folder. Try to run version 3.8.20 on it.

[Jean Penné]

Originally Posted by pinhodecarlos

I'm on Windows 7 Enterprise SP1 64 bits, Processor is i7-3630QM.
Example attached, you just need to add the client into the folder. Try to run version 3.8.20 on it.

Thank you for your attachment,
I tested your data :

I reproduced the issue while using your llr.ini file which is abnormal (it has not the "OldInputFile=input.txt" line).

If I remove your llr.ini and restart from scratch, all becomes normal, so, I think that ATH is right!

Regards,
Jean

[pinhodecarlos]

Originally Posted by Jean Penné

Thank you for your attachment,
I tested your data :

I reproduced the issue while using your llr.ini file which is abnormal (it has not the "OldInputFile=input.txt" line).

If I remove your llr.ini and restart from scratch, all becomes normal, so, I think that ATH is right!

Regards,
Jean

Thank you.

1) I've been running on this folder outside the program files since 2013, so it is not a permission issue. Haven't run any windows update since that time.
2) can you better clarify your statement, should I add the line "OldInputFile=input.txt" into llr.ini file?

[pinhodecarlos]

I've just done one simple thing, start all over again through the GUI version, set every options and then moved to the command line. Because each number takes 9000 secs I can only test the stop and resume latter in the evening.

[Jean Penné]

Originally Posted by pinhodecarlos

Thank you.

1) I've been running on this folder outside the program files since 2013, so it is not a permission issue. Haven't run any windows update since that time.
2) can you better clarify your statement, should I add the line "OldInputFile=input.txt" into llr.ini file?

Yes, (or add -oOldInputFile=input.txt in the command line) because the new version of LLR needs to know you are using the same input file as before, in order to take into account the PgenLine parameter.

Regards,
Jean

[pinhodecarlos]

Originally Posted by Jean Penné

Yes, (or add -oOldInputFile=input.txt in the command line) because the new version of LLR needs to know you are using the same input file as before, in order to take into account the PgenLine parameter.

Regards,
Jean

Jean,

I've done that but it didn't sort out things so I did what I said on the other message, started GUI, set all settings, and moved to command line. With this way the OldInputFile=input.txt now shows up on the ini file.
I'll give you an update later on when I will be trying the stop and resume to see if it picks us from last checkpoint and not start all over again from the first line of the input.txt.

On every LLR update I was just replacing the command line version, not taking into consideration what was inside the ini file.

Carlos

[Jean Penné]

Warning to all users!

If you are using an input file that was processed with an older version of LLR, and do not want to process it from the beginning, you must add this line in the .ini file :

OldInputFile=<your input file name>, or add this option in the command line :

-oOldInputFile=<your input file name>

Sorry for this drawback, due to missing a default value in a call to IniGetString() function...

I corrected this locally, it works, and it will be fixed in the next release.

Best Regards,
Jean

[DeleteNull]

I am testing 3.8.20 with the new AMD Ryzen 1700X and 8 threads in parrallel.
llr -m reports me that the CPU is capable for AVX/KMA3, but the calculations are made with:
Using Jean Penne's llr (64 bit)
LLR Program - Version 3.8.20, using Gwnum Library Version 28.13
Using all-complex AMD K10 FFT length 512K, Pass1=512, Pass2=1K, 8 threads, a = 3

Is there a way to force LLR to use the these features, or do we have to wait for a new version of GWNUM and/or LLR?

[rebirther]

As read at primegrid the new Ryzen CPUs doesnt use any AVX/FMA. George should check the gwnum lib.