Genefer data sizes

Join PrimeGrid

Read our rules and policies.
Download, install and run the BOINC software used by PrimeGrid.
When prompted, enter the URL:
http://www.primegrid.com/
Select which subprojects to run via the PrimeGrid Preferences.

If you are not using graphics enabled BOINC client, you may use manual registration form.

Author

Message

mackerel

Volunteer tester
Avatar

Send message
Joined: 2 Oct 08
Posts: 2645
ID: 29980
Credit: 568,565,361
RAC: 147

We have n = 21, N = 2^21, data size = N * 8 = 16 MB.
We also have the cos/sin table (read only): 10 MB (I can try to reduce this).

From: https://www.primegrid.com/forum_thread.php?id=8168&nowrap=true#119783

Above was from when I asked about the data sizes in use by genefer over 4 years ago. I'm trying to understand how it may be optimal to run on systems without directly testing them. Does the above still apply or has things changed since then?

Specifically regarding the cos/sin table, is that data shared between multiple tasks? Or does each task have its own copy?

ID: 159847 | Reply Quote

Yves Gallot

Volunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 820
ID: 164101
Credit: 305,989,513
RAC: 1,728

Does the above still apply or has things changed since then?
Specifically regarding the cos/sin table, is that data shared between multiple tasks? Or does each task have its own copy?

The current values are

GFN-16: 1.19 MB GFN-17: 2.38 MB GFN-18: 5.07 MB GFN-19: 10.1 MB GFN-20: 20.3 MB GFN-21: 40.5 MB GFN-22: 40.5 MB

Data are not shared between tasks.

ID: 159853 | Reply Quote

Yves Gallot

Volunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 820
ID: 164101
Credit: 305,989,513
RAC: 1,728

Note that because of its four memory channels and memory bandwidth, the 16.5 MB L3 cache of the i9-7920X is not the limit and the throughput may be better with 12 GFN-17 tasks rather than 6.

ID: 159854 | Reply Quote

mackerel

Volunteer tester
Avatar

Send message
Joined: 2 Oct 08
Posts: 2645
ID: 29980
Credit: 568,565,361
RAC: 147

Data are not shared between tasks.

Thanks for the data.

I was referring to the previously mentioned read only sin/cos table. Is that no longer required at all? I take it a lot has changed over the years.

Note that because of its four memory channels and memory bandwidth, the 16.5 MB L3 cache of the i9-7920X is not the limit and the throughput may be better with 12 GFN-17 tasks rather than 6.

For the recent challenge I did choose to run GFN-17 with 1 task per core on that CPU. That was determined by test and observation. Particularly for Skylake-X CPUs, the relatively large L2 cache may contribute more than with other CPUs.

Now the focus is on the upcoming Tour de Primes and many are looking at running GFN-16. I'm seeing reports that running 2 cores per task may be optimal but I have not seen this on my own limited systems. 1 task per core still seems optimal.

ID: 159855 | Reply Quote

Crun-chi

Volunteer tester
Avatar

Send message
Joined: 25 Nov 09
Posts: 3233
ID: 50683
Credit: 151,443,349
RAC: 54,956

OS Win 10
CPU AMD Ryzen 7 5700X ( 4 GHz)

8*1 task ( HT disabled) 745 seconds = 927 WU per day
4*2 task ( HT disabled) 440 seconds = 785 WU per day

So combination 8*1 is clear 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!

ID: 159864 | Reply Quote

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other