| Author |
Message |
|---|
GDBSend message
Joined: 15 Nov 11
Posts: 304
ID: 119185
Credit: 4,289,124,211
RAC: 1,725,094
                      
|
|
I didn't see any announcement about a LLR TRP app update.
____________
 |
|
|
Reggie Volunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 10 May 14
Posts: 241
ID: 311759
Credit: 270,763,471
RAC: 165,626
                  
|
|
TRP is the first project we're trying with LLR2's fast double check. It is out, but old workunits still use the old app. The new app is being used for n>=10.81M. We're monitoring it closely to make sure nothing breaks because of it. We will likely make an announcement once everything is confirmed to be working correctly. Offhand, I know the pending credit and badge pages are inaccurate for TRP right now. |
|
|
 Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14043
ID: 53948
Credit: 481,202,847
RAC: 504,835
                  
|
|
Also, if anyone is using Macs on TRP, you should switch to a different subproject, at least temporarily. There's no Mac version of LLR2 yet, and you can't use the old LLR with the new tasks. They're incompatible. If you have a Mac, once the old tasks run out, you'll stop getting TRP tasks.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Looking forward to the announcement and seeing LLR2 used elsewhere. Will there be a way to know that you are the double checker vs. the primary checker?
____________
 |
|
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3250
ID: 50683
Credit: 152,646,050
RAC: 10,054
                  
|
Looking forward to the announcement and seeing LLR2 used elsewhere. Will there be a way to know that you are the double checker vs. the primary checker?
There is no DC any more.
____________
92*10^1585996-1 NEAR-REPDIGIT PRIME :) :) :)
4 * 650^498101-1 CRUS PRIME
2022202116^131072+1 GENERALIZED FERMAT
Proud member of team Aggie The Pew. Go Aggie! |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14043
ID: 53948
Credit: 481,202,847
RAC: 504,835
|
Looking forward to the announcement and seeing LLR2 used elsewhere. Will there be a way to know that you are the double checker vs. the primary checker?
The double check tasks are MUCH shorter.
If I'm reading Pavel's table correctly, the TRP double check tasks should be 128 times shorter than the main tasks. Even if I'm wrong about the exact ratio, they're just a small fraction of the size of the main task. It will be plainly obvious when you have a double check task. Or I should say "when you HAD a double check task", because you have to be looking at just the right time to catch it running.
My Haswell just finished an LLR2 TRP task in 4:50. The double check task for that should take a bit more than 2 minutes.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14043
ID: 53948
Credit: 481,202,847
RAC: 504,835
|
There is no DC any more.
No, there's still double checking. It's just that the double check tasks are so fast as to be inconsequential.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Will that also mean that double check tasks will receive less credit than the initial checks?
____________
|
|
|
|
|
It will be plainly obvious when you have a double check task. Or I should say "when you HAD a double check task", because you have to be looking at just the right time to catch it running.
I am sitting here watching 8/9 of my TRP instances so that I don't miss the first one. Very excited to see that roll through!
Will that also mean that double check tasks will receive less credit than the initial checks?
On the Private GFN Server where this has been implemented for some time, the tasks will receive the proper ratio of credit. The certification tasks are 1/64 of a full task for GFN15-Mega and they receive precisely 1/64 of the credit :) |
|
|
|
|
|
Hi,
i got two double check tasks today:-).
https://www.primegrid.com/workunit.php?wuid=670048130
Stderr output
<core_client_version>7.16.6</core_client_version>
<![CDATA[
<stderr_txt>
BOINC PrimeGrid wrapper 2.01 (Aug 11 2020 20:39:32)
running ../../projects/www.primegrid.com/sllr2_1.0.0_linux64_200814 -v
LLR2 Program - Version 1.0.0, using Gwnum Library Version 29.8
running ../../projects/www.primegrid.com/sllr2_1.0.0_linux64_200814 -oGerbicz=1 -oProofName=proofc -pVerifyCert -q215443*2^10810515-1 -d -t1 -oDiskWriteTime=10
Gerbicz check is requested, switching to PRP.
Starting probable prime test of 215443*2^10810515-1
Using FMA3 FFT length 1M, Pass1=1K, Pass2=1K, clm=1, a = 3, L2 = 207*207
Testing complete.
22:15:19 (1439): called boinc_finish(0)
</stderr_txt>
]]>
https://www.primegrid.com/workunit.php?wuid=670048134
Stderr output
<core_client_version>7.16.6</core_client_version>
<![CDATA[
<stderr_txt>
BOINC PrimeGrid wrapper 2.01 (Aug 11 2020 20:39:32)
running ../../projects/www.primegrid.com/sllr2_1.0.0_linux64_200814 -v
LLR2 Program - Version 1.0.0, using Gwnum Library Version 29.8
running ../../projects/www.primegrid.com/sllr2_1.0.0_linux64_200814 -oGerbicz=1 -oProofName=proofc -pVerifyCert -q9221*2^10810626-1 -d -t1 -oDiskWriteTime=10
Gerbicz check is requested, switching to PRP.
Starting probable prime test of 9221*2^10810626-1
Using FMA3 FFT length 896K, Pass1=896, Pass2=1K, clm=1, a = 3, L2 = 242*242
Testing complete.
22:12:29 (36284): called boinc_finish(0)
</stderr_txt>
]]>
Are they running longer, if the Number is prime, or are they always short?
|
|
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3250
ID: 50683
Credit: 152,646,050
RAC: 10,054
|
There is no DC any more.
No, there's still double checking. It's just that the double check tasks are so fast as to be inconsequential.
I was read ( obvious wrong) that only one server has certificate so he can do that.
Like now I see many got DC task, short one. But main idea stay still: if you find prime it is your :)
____________
92*10^1585996-1 NEAR-REPDIGIT PRIME :) :) :)
4 * 650^498101-1 CRUS PRIME
2022202116^131072+1 GENERALIZED FERMAT
Proud member of team Aggie The Pew. Go Aggie! |
|
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3250
ID: 50683
Credit: 152,646,050
RAC: 10,054
|
Will that also mean that double check tasks will receive less credit than the initial checks?
Yes from 38-60 like I see on examples
____________
92*10^1585996-1 NEAR-REPDIGIT PRIME :) :) :)
4 * 650^498101-1 CRUS PRIME
2022202116^131072+1 GENERALIZED FERMAT
Proud member of team Aggie The Pew. Go Aggie! |
|
|
|
|
|
I just finished my first TRP task since the change was rolled out, and realized that it looks like the first pass are being split off from the second pass, so you don't actually get to see the double check happen on your tasks.
This whole new system is nice, it means that we will now be able to keep small caches of work around on computers since we will no longer need to race to be the first to complete a task. It also means that I will be more likely to try and find a megaprime in PPS MEGA since it will be a function of time invested and not also a function of luck.
____________
|
|
|
|
|
|
Very nice!
Not running that subproject myself now, but looked at some of your tasks.
One can recognize the DC tasks from having a much lower credit, and in the Stderr output there is a switch like -pVerifyCert (on the "first" tasks, it is -pSavePoints instead).
With the old LLR wrapper, you could not see the actual number being tested. But with this LLR2 wrapper, you can (for example, "Starting probable prime test of 121889*2^10811312-1", and there is a -q121889*2^10811312-1 switch). I assume Stream has ensured that a composite result looks identical to a prime result, so that a find is not revealed before it has been properly submitted to Top 5000.
/JeppeSN |
|
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 1051
ID: 301928
Credit: 563,881,725
RAC: 768
           
|
i got two double check tasks today:-).
Are they running longer, if the Number is prime, or are they always short?
They always have same length, on TRP it's 1/128 of original task. It look like standard prime test but really it's not, it's just a calculation of some magic value to confirm that proofs of correctness, uploaded during first test, are valid.
|
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1963
ID: 352
Credit: 6,418,141,603
RAC: 2,733,162
                               
|
|
Great, run via soft_link.
Eventually, I would like to see error rate comparation old LLR vs LLR2.
____________
My stats |
|
|
|
|
|
Gerbicz check should catch most errors. Certificate fails should be extremely rare. |
|
|
|
|
|
Does the double check task also use FMA/AVX, and run MT?
____________
Reno, NV
 |
|
|
|
|
Does the double check task also use FMA/AVX, and run MT?
Yes. It runs with the same settings and performs similiar operations, just with a different goal. |
|
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3250
ID: 50683
Credit: 152,646,050
RAC: 10,054
|
Does the double check task also use FMA/AVX, and run MT?
Any much shorter: I made one DC task it is finished for 217 seconds on 2 cores
____________
92*10^1585996-1 NEAR-REPDIGIT PRIME :) :) :)
4 * 650^498101-1 CRUS PRIME
2022202116^131072+1 GENERALIZED FERMAT
Proud member of team Aggie The Pew. Go Aggie! |
|
|
|
|
|
I’m really eager to get a double check task through, but so far I have only received initial checks, which makes sense given the new time requirements. Unfortunately it will probably be unlikely that I will see one happen live with my own eyes anytime soon. I’m still very excited for this. Huge advancement.
____________
|
|
|
|
|
|
My question is....how will credit for LLR2 be calculated for contests? ;) |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14043
ID: 53948
Credit: 481,202,847
RAC: 504,835
|
My question is....how will credit for LLR2 be calculated for contests? ;)
No change.
____________
My lucky number is 75898524288+1 |
|
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3250
ID: 50683
Credit: 152,646,050
RAC: 10,054
|
I’m really eager to get a double check task through, but so far I have only received initial checks, which makes sense given the new time requirements. Unfortunately it will probably be unlikely that I will see one happen live with my own eyes anytime soon. I’m still very excited for this. Huge advancement.
One suggestion: now when you are always first you dont need to put 6 cores on one task:there is no hurry anymore.Test your configuration and see what will you the best output, not the fastest one :)
____________
92*10^1585996-1 NEAR-REPDIGIT PRIME :) :) :)
4 * 650^498101-1 CRUS PRIME
2022202116^131072+1 GENERALIZED FERMAT
Proud member of team Aggie The Pew. Go Aggie! |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14043
ID: 53948
Credit: 481,202,847
RAC: 504,835
|
|
The "first" statistics on the private user page have been modified to exclude the fast-DC tasks.
____________
My lucky number is 75898524288+1 |
|
|
|
|
The "first" statistics on the private user page have been modified to exclude the fast-DC tasks.
Thanks!
The certificate tasks also have a "c" after the 9 digit task name
____________
My lucky number is 6219*2^3374198+1
 |
|
|
|
|
|
Boinc use "srw_win32_201105.exe" when I crunch TRP. but my PC win 10 64 and CPU AMD 2700X. Something wrong how do I get the 64bit app for TRP ? any help appreciated.
____________
My Lucky Number is 9037*2^1301022+1
 |
|
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 1051
ID: 301928
Credit: 563,881,725
RAC: 768
|
Boinc use "srw_win32_201105.exe" when I crunch TRP. but my PC win 10 64 and CPU AMD 2700X. Something wrong how do I get the 64bit app for TRP ? any help appreciated.
This is a not an app, this is a "wrapper" - a small program which controls big main program. Main program does all real work. Its name is "llr2" and this program is 64-bit.
|
|
|
