Prime95 v30.4/30.5/30.6

[kriesel]

Originally Posted by Prime95

Try renaming gwnum.txt. Restart prime95, will the exponent now run? If so, email/PM the gwnum.txt file to me. Thanks.

On Xeon Phi 7210 that ran M660M P-1 until a prime95-initiated benchmark interrupted progress, then won't resume:

Stop all prime95 workers
rename gwnum.txt wasgwnum.txt
reintstate M660M P-1 as first line of worker 1 worktodo by notepad edit
continue all workers

Issue with fft for M660M P-1 reoccured. Maybe it only reads gwnum.txt on startup. Try again, more thoroughly.

Stop prime95 workers
verify there's no gwnum.txt by name sort in the working directory in explorer
Exit prime95 completely
Use task manager to verify prime95.exe is not running at all
Relaunch prime95 v30.5b1 with a double click

Prime95 resumes M660M P-1, ETA ~1.5 weeks

Check for a gwnum.txt email attachment, George, & thanks!
(at 122KB, much too big for a PM. FYI, Xeon Phi 7250 system that could not even start it, has a 133KB gwnum.txt. Laptop that could continue it has a 9KB gwnum.txt. Stealth size limit?)

[Prime95]

Originally Posted by kriesel

Prime95 resumes M660M P-1, ETA ~1.5 weeks

Check for a gwnum.txt email attachment,

Fixed in 30.6 build 4

[Prime95]

30.6 build 4 is ready

Several bug fixes: Torture tests, AVX-512 P+1 on FFT lengths that are a multiple of 7, best 36M FFT selection

Download links:

Before downloading, make sure prime.spl is uploaded to the server when upgrading from early versions of 30.6. There was a bug in spool file format for versions 30.6b1 and 30.6b2. Version 30.6b4 will have no trouble creating and reading spool files that are compatible with version 30.5 and earlier plus versions 30.6b3 and later.


Windows 64-bit: https://mersenne.org/ftp_root/gimps/p95v306b4.win64.zip
Linux 64-bit: https://mersenne.org/ftp_root/gimps/...linux64.tar.gz
Source code: https://mersenne.org/ftp_root/gimps/....source.tar.gz

[Falkentyne]

Originally Posted by Prime95

30.6 build 4 is ready

Several bug fixes: Torture tests, AVX-512 P+1 on FFT lengths that are a multiple of 7, best 36M FFT selection

Download links:

Before downloading, make sure prime.spl is uploaded to the server when upgrading from early versions of 30.6. There was a bug in spool file format for versions 30.6b1 and 30.6b2. Version 30.6b4 will have no trouble creating and reading spool files that are compatible with version 30.5 and earlier plus versions 30.6b3 and later.


Windows 64-bit: https://mersenne.org/ftp_root/gimps/p95v306b4.win64.zip
Linux 64-bit: https://mersenne.org/ftp_root/gimps/...linux64.tar.gz
Source code: https://mersenne.org/ftp_root/gimps/....source.tar.gz

Super fast test on 11900k. Large FFT AVX disabled no longer instacrashes all threads after one iteration. Thank you (can't do more conclusive testing because I'm gaming).

[ATH]

Weird error in mprime 30.5b1 (or primenet server error).

Factors for these 2 exponents got reported twice each: 3021587 and 3310193

and for some reason they got added twice to the PRP-CF worktodo line by the server:
PRP=0C618B35F23DD4EB9B3E5D7094A1____,1,2,3021587,-1,99,0,3,1,"93415939664009,19611913284769456132423,19611913284769456132423"
PRP=165545938287D7EBB30C87F04923____,1,2,3310193,-1,99,0,3,1,"2755537060921,555881755075443360919,555881755075443360919"

which of course resulted in:
19611913284769456132423 does not divide M3021587
555881755075443360919 does not divide M3310193

and the 2 assignments were skipped.

[Prime95]

Originally Posted by ATH

Weird error in mprime 30.5b1 (or primenet server error).

Factors for these 2 exponents got reported twice each: 3021587 and 3310193.

I've seen this before -- no clue as to why it happens.
The database has been repaired.

[LaurV]

Got 306.b4, it works fine for P+1 in 13M. Work resumed.

Good job Gorge!

[rainchill]

I still think the main title bar should show the version and build so it is in screenshots here and anywhere they show up around the web.

[LaurV]

Originally Posted by LaurV

Got 306.b4, it works fine for P+1 in 13M. Work resumed.

Good job Gorge!

Update: b4 still has the issue with hanging when exiting (it was doing P+1, it may be related or not). It needs to be killed from taskman on both Win7 and Win10. When you "stop" it (from the menu, or right-clicking the icon), two or three workers (from 6) still won't stop, but also do nothing ("stopped" or "exited" P95 takes about 4% CPU). "Stopped" P95 is then not responsive (you can not "stop" it, because it is obviously stopped, and you can not "continue", because some tasks are running ).

Hopefully as long as you don't exit it, and don't stop it, it works fine.

[kruoli]

On my Windows 7 laptop, I now had it multiple times that Prime95 "forgot" a lot of assignments. I set MaxExponents=1024 and UnreserveDays=1024 in prime.txt. Using WellBehavedWork=1 makes no difference. This happened with every 30.6 build so far. I put lots of (but far less than 1,000) assignments of P+1 in the worktodo file (4 workers).

[nordi]

Originally Posted by kruoli

I now had it multiple times that Prime95 "forgot" a lot of assignments.

Maybe you should try "UnreserveExponents=999", which does the opposite of what the name implies: It protects 999 exponents per worker from bein unreserved. I set it on my system because mprime kept unreserving assignments.