ECM users - version 30.9-30.13 (see post#316) - Page 10

chalsall · 2022-12-11, 02:57

Quote:

Originally Posted by LaurV

P95 30.9 build 3, i7-6950X, win7, 48G RAM allocated (from 64), 1 worker 10 cores. ECM doesn't like the exponent 2859071. Seriously!

All exponents around it, work well, it takes minutes to stage 1 and minutes to stage 2 (see my daily result reports). For this particular exponent stage 1 works well, but then stage 2 is stuck somehow.

Wow! Serious edge case!

Only someone in this community would notice, and then report, something like this.

And... That's why we get stuff done...

James Heinrich · 2022-12-11, 03:23

Quote:

Originally Posted by LaurV

P95 30.9 build 3, i7-6950X, win7, 48G RAM allocated (from 64), 1 worker 10 cores.
ECM doesn't like the exponent 2859071

You're apparently the first to try with v30.9, in which case it might actually an issue with the new ECM implementation.
[Moved the posts from the thread about 30.8, making the next sentence appear irrelevant :] And yet, you posted the bug report in the v30.8 thread (probably should post here instead).[/end of edit]

For reference, previous ECM curves (all B1=50k,B2=5M) have been run successfully on all these versions:

Code:

Windows32, v25.8, build 5
Windows32, v25.9, build 4
Windows32, v26.6, build 3
Linux64,   v28.5, build 2
Windows64, v29.4, build 7
Windows64, v29.4, build 8
Windows32, v29.8, build 3
Windows64, v29.8, build 5
Linux64,   v29.8, build 6
Windows64, v30.3, build 6
Linux32,   v30.3, build 6
Linux64,   v30.3, build 6
Linux64,   v30.6, build 4

chalsall · 2022-12-11, 03:24

Quote:

Originally Posted by LaurV

P95 30.9 build 3, i7-6950X, win7, 48G RAM allocated (from 64), 1 worker 10 cores.

ECM doesn't like the exponent 2859071.

Just because I'm not going to be sleeping for a little while... Stuff to do...

@LaurV... Could you please provide the exact worktodo.txt line that failed?

Some might like to put in "on the bench", for in-depth testing...

Prime95 · 2022-12-11, 05:20

Quote:

Originally Posted by LaurV

P95 30.9 build 3, i7-6950X, win7, 48G RAM allocated (from 64), 1 worker 10 cores.

ECM doesn't like the exponent 2859071.

Seriously!

Bug found and fixed. When a 30.10 version comes out, grab it.
In the meantime, delete the line from worktodo.txt.

LaurV · 2022-12-11, 13:32

Thanks a billion George! You are the man!

@James, sorry for posting in the wrong place and thanks for moving it. Reality is that I was looking for a 30.9 thread which I remembered I saw in the past, but I could not find it, because I was missing the point that is ECM related. The other threads are called "Prime95 [version] blah blah", that is what I was looking for in the list. My mistake I didn't use the search.

bayanne · 2022-12-12, 11:40

Any chance of a Mac version, ever hopeful

tuckerkao · 2022-12-14, 18:13

I'm wondering when Prime95 30.9 will start to run smoothly with Ryzen i9 7950X on Windows 11. The 30.8.17 version kept getting the FFT errors during the PRP testing or even roundoffs during P-1. System virtual memory and page size already increased to 17 GBs ~ 25 GBs.

Prime95 · 2022-12-21, 00:23

Quote:

Originally Posted by bayanne

Any chance of a Mac version, ever hopeful

Try https://mersenne.org/ftp_root/gimps/....MacOSX.tar.gz
Totally untested.

Prime95 · 2022-12-21, 00:39

Quote:

Originally Posted by tuckerkao

I'm wondering when Prime95 30.9 will start to run smoothly with Ryzen i9 7950X on Windows 11. The 30.8.17 version kept getting the FFT errors during the PRP testing or even roundoffs during P-1. System virtual memory and page size already increased to 17 GBs ~ 25 GBs.

Go back to the thread where people are trying to help you with your hardware problems

congsz · 2022-12-21, 04:20

Quote:

Originally Posted by Prime95

Try https://mersenne.org/ftp_root/gimps/....MacOSX.tar.gz
Totally untested.

30.10? Will there be any new features?

Prime95 · 2022-12-21, 05:44

No.

Version numbers are a little weird. They can go up for one of two reasons.
1) Major new features in prime95.
2) Changes in the gwnum library -- even relatively small changes. LLR and OpenPFGW are built on the gwnum library and print out the gwnum version they are linked to. If bugs are reported, the gwnum version lets me pin down the gwnum code that needs debugging.

So what's new in gwnum 30.10. A significant bug in zero-padded FFTs was fixed (bug introduced in 30.5 or so). Default implementations for Ryzen AVX-512 FFTs were added.

2022-12-14, 18:13	#106
tuckerkao "Tucker Kao" Jan 2020 Head Base M168202123 2⁴·5·11 Posts	I'm wondering when Prime95 30.9 will start to run smoothly with Ryzen i9 7950X on Windows 11. The 30.8.17 version kept getting the FFT errors during the PRP testing or even roundoffs during P-1. System virtual memory and page size already increased to 17 GBs ~ 25 GBs. Last fiddled with by tuckerkao on 2022-12-14 at 18:26

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
That's a Lot of Users!!!	jinydu	Lounge	9	2006-11-10 00:14
Beta version 24.6 - Athlon users wanted	Prime95	Software	139	2005-03-30 12:13
For Old Users	Citrix	Prime Sierpinski Project	15	2004-08-22 16:43
Opportunity! Retaining new users post-M40	GP2	Lounge	55	2003-11-21 21:08
AMD USERS	ET_	Lounge	3	2003-10-11 16:52

2022-12-11, 13:32	#104
LaurV Romulan Interpreter "name field" Jun 2011 Thailand 41·251 Posts	Thanks a billion George! You are the man! @James, sorry for posting in the wrong place and thanks for moving it. Reality is that I was looking for a 30.9 thread which I remembered I saw in the past, but I could not find it, because I was missing the point that is ECM related. The other threads are called "Prime95 [version] blah blah", that is what I was looking for in the list. My mistake I didn't use the search.

2022-12-12, 11:40	#105
bayanne "Tony Gott" Aug 2002 Yell, Shetland, UK 373 Posts	Any chance of a Mac version, ever hopeful

2022-12-21, 05:44	#110
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 10000001010111₂ Posts	No. Version numbers are a little weird. They can go up for one of two reasons. 1) Major new features in prime95. 2) Changes in the gwnum library -- even relatively small changes. LLR and OpenPFGW are built on the gwnum library and print out the gwnum version they are linked to. If bugs are reported, the gwnum version lets me pin down the gwnum code that needs debugging. So what's new in gwnum 30.10. A significant bug in zero-padded FFTs was fixed (bug introduced in 30.5 or so). Default implementations for Ryzen AVX-512 FFTs were added.