Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Fermat Prime Search :
cpuGFN16_mt v4.01 -- crossing the b limit
| Author |
Message |
|---|
|
|
|
Why was cpuGFN16_mt moved to genefer22 yesterday?
The project is already at the b limit of genefer22's CPU transform. | |
|
Yves Gallot Volunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 820
ID: 164101
Credit: 305,989,513
RAC: 1,728
|
|
Close to the limit, not at the limit. Round-off error is 0.375, the limit is 0.5.
There is no difference between v3.21 and v4.01 on CPU.
Now the limit on GPU is 2,000,000,000. | |
|
|
|
|
Yves Gallot wrote: There is no difference between v3.21 and v4.01 on CPU.
Thanks; this bit of info was obscure (to me at least) until now.
| |
|
|
|
|
Related question:
Are the transforms which are called "sse2, sse4, avx, fma" in genefer22 v22.11.04 perhaps the same as the ones which are called "sse2i, sse4i, avxi, fmai" in genefer v3.4? | |
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 820
ID: 164101
Credit: 305,989,513
RAC: 1,728
|
Are the transforms which are called "sse2, sse4, avx, fma" in genefer22 v22.11.04 perhaps the same as the ones which are called "sse2i, sse4i, avxi, fmai" in genefer v3.4?
For all GFN except for GFN-22 ("sse2, sse4, avx, fma, 512" in genefer v3.4).
| |
|
 Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3233
ID: 50683
Credit: 151,443,349
RAC: 54,956
                         
|
|
So fast check is enabled fro GFN16 also?
If this is true then you can update front page...
____________
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: 14011
ID: 53948
Credit: 435,567,510
RAC: 868,496
                          
|
So fast check is enabled fro GFN16 also?
If this is true then you can update front page...
I don't believe GFN-16 is doing fast checking, especially since one of its apps isn't updated yet and hence can not do fast double checking. I'm trying to get a definitive answer on this from stream.
____________
My lucky number is 75898524288+1 | |
|
Crun-chiVolunteer tester
Send message
Joined: 25 Nov 09
Posts: 3233
ID: 50683
Credit: 151,443,349
RAC: 54,956
|
|
First task I found....
<core_client_version>7.14.2</core_client_version>
<![CDATA[
<stderr_txt>
genefer22g version 22.11.4 (win64, gcc-12.2.0, boinc-7.20.2)
Copyright (c) 2022, Yves Gallot
genefer22 is free source code, under the MIT license.
Command line: '-boinc -c -n 16 -b 175xxxxxxx -f gproof'
Running on device 'Quadro P620', vendor 'NVIDIA Corporation', version 'OpenCL 3.0 CUDA', driver '472.12', data size: 1.5 MB.
175xxxxxxx^{2^16} + 1 is checked, ckey = 0B350xxxxxxx, time = 00:00:45.
01:47:44 (30660): called boinc_finish(0)
</stderr_txt>
]]>
Also few other tasks of new GFN16 on another thread - seg fault on ryzen...
____________
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! | |
|
|
|
So fast check is enabled fro GFN16 also?
If this is true then you can update front page...
I don't believe GFN-16 is doing fast checking, especially since one of its apps isn't updated yet and hence can not do fast double checking. I'm trying to get a definitive answer on this from stream.
I’m not sure how exactly it is working, but the new app appears to be working well enough for most of us from the steady stream of double check tasks I am getting on GFN16.
____________
 | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 435,567,510
RAC: 868,496
|
So fast check is enabled fro GFN16 also?
If this is true then you can update front page...
I don't believe GFN-16 is doing fast checking, especially since one of its apps isn't updated yet and hence can not do fast double checking. I'm trying to get a definitive answer on this from stream.
I’m not sure how exactly it is working, but the new app appears to be working well enough for most of us from the steady stream of double check tasks I am getting on GFN16.
Well, it's not going to work on Mac CPUs, since the installed app is incapable of running these WUs. I'm pretty sure that any Mac CPUs will therefore only get assigned old non-fast-DC workunits. (It's not possible to mix old and new style tasks in the same WU.)
It's probably just an oversight and the missing v4 app will probably be installed tomorrow.
____________
My lucky number is 75898524288+1 | |
|
|
|
|
Yves Gallot wrote: xii5ku wrote: Are the transforms which are called "sse2, sse4, avx, fma" in genefer22 v22.11.04 perhaps the same as the ones which are called "sse2i, sse4i, avxi, fmai" in genefer v3.4?
For all GFN except for GFN-22 ("sse2, sse4, avx, fma, 512" in genefer v3.4).
Good. If I read the output correctly, genefer v3.4 reports a b limit at ≈ 200…240M (avxi) or ≈ 180…230M (sse2i) at n = 16. (More like the upper bound of those, given an error limit of 0.5.) So there is still a way to go until GFN-16 reaches cpuGFN's b limit. | |
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 1033
ID: 301928
Credit: 543,624,271
RAC: 5,083
                 
|
|
All projects except GFN-15 are on fast DC / apps 4.01
Converting GFN-15 will be tricky. GFN-15 is far above of CPU limit, it mean that server (which can use only CPU) have to use very slow special integer transform to create a certificate. For all other projects certificate can be generated in 0.5-1 sec. GFN-15 required up to 3 seconds in my tests. Considering high amount of tasks crunched every day even under normal rates, it'll create unacceptable load on server. Anyway, we have to deal with this problem somehow because old app will stop working at 400M quite soon. But later.
| |
|
Post to thread
Message boards :
Generalized Fermat Prime Search :
cpuGFN16_mt v4.01 -- crossing the b limit |
