| Author |
Message |
|---|
 Crun-chi Volunteer tester
Send message
Joined: 25 Nov 09
Posts: 3233
ID: 50683
Credit: 151,443,349
RAC: 73,965
                         
|
|
If somebody test, what is fastest app for process MEGA prime on PRPRNET?
LLR or PFGW ili something else?
____________
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! |
|
|
|
|
If somebody test, what is fastest app for process MEGA prime on PRPRNET?
LLR or PFGW ili something else?
I have planned to test it with llravx today, pfgw should be lower. |
|
|
|
|
|
Test done:
llr=1h45m
llravx=1h19m
i5-2500k@4Ghz |
|
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3233
ID: 50683
Credit: 151,443,349
RAC: 73,965
|
|
LLR 12000 seconds
PFGW 32 or 64 bit more then 20000 seconds
So LLR is winner :)
____________
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! |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
 
|
LLR 12000 seconds
PFGW 32 or 64 bit more then 20000 seconds
So LLR is winner :)
Is that llr or llravx? If llravx, then you need to wait for pfgw 3.6.2, which will have gwnum v27.2 and AVX support.
Without avx, pfgw tends to be faster for non power of 2 bases, which you can find on the GCW port. |
|
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3233
ID: 50683
Credit: 151,443,349
RAC: 73,965
|
LLR 12000 seconds
PFGW 32 or 64 bit more then 20000 seconds
So LLR is winner :)
Is that llr or llravx? If llravx, then you need to wait for pfgw 3.6.2, which will have gwnum v27.2 and AVX support.
Without avx, pfgw tends to be faster for non power of 2 bases, which you can find on the GCW port.
I have ordinary AMD without AVX support :)
I am soory but test show different at least on AMD CPU. I try every combination , and always llr was fastest choice...
Maybe some switch can boost PFGW but until then llr is fastest app.
On my custom Linux x64 MEGA WU is finished at 12060 seconds. And I put one WU with PFGW32 and with PFGW64, and they pass nearly 60 % with same time( 12000 seconds) , so they are much slower :(
____________
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! |
|
|
|
|
|
Thats exactly what rogue said:
For a primality test for a number of the form k*2^n+1 LLR is faster than pfgw.
For number of other forms (e.g. SR5 k*5^n+1) pfgw is faster than LLR ... |
|
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3233
ID: 50683
Credit: 151,443,349
RAC: 73,965
|
Thats exactly what rogue said:
For a primality test for a number of the form k*2^n+1 LLR is faster than pfgw.
For number of other forms (e.g. SR5 k*5^n+1) pfgw is faster than LLR ...
But since I in topic ask what is faster for MEGA I should get answer LLR :)
____________
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! |
|
|
KEPSend message
Joined: 10 Aug 05
Posts: 301
ID: 110
Credit: 12,352,853
RAC: 103
        
|
Is that llr or llravx? If llravx, then you need to wait for pfgw 3.6.2, which will have gwnum v27.2 and AVX support.
Without avx, pfgw tends to be faster for non power of 2 bases, which you can find on the GCW port.
Whats the prospect for getting pfgw 3.6.2 released? Also when using Prime95, using avx on my Sandy Bridge, there were quite a few problems running my S58 tests, so has the GWnum library been fixed or do pfgw search at a higher FFT length like it usually do if a test fail?
Take care
Kenneth! |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
Whats the prospect for getting pfgw 3.6.2 released? Also when using Prime95, using avx on my Sandy Bridge, there were quite a few problems running my S58 tests, so has the GWnum library been fixed or do pfgw search at a higher FFT length like it usually do if a test fail?
I have to ask George about the relative stability of gwnum v27.2.
pfgw has a lot of error checking in it. If it detects a roundoff or sumout error, it will automatically switch to a larger FFT and restart the test. |
|
|
KEPSend message
Joined: 10 Aug 05
Posts: 301
ID: 110
Credit: 12,352,853
RAC: 103
|
pfgw has a lot of error checking in it. If it detects a roundoff or sumout error, it will automatically switch to a larger FFT and restart the test.
Just as I thought then :) However if Prime95 had just had the same kind of error checking, where it automatically jumped to the next larger FFT length, in stead of backing off for 5 minutes, suspecting a hardware error and eventually abandone the test, that could have saved me some lost time :(
Anyway looking forward to your release :) |
|
|
