mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Software (https://www.mersenneforum.org/forumdisplay.php?f=10)
-   -   Prime95 v30.3 (https://www.mersenneforum.org/showthread.php?t=25823)

jwnutter 2020-08-31 17:18

2 Attachment(s)
[QUOTE=Prime95;555555]It resumed from the last save file. Presumably prime95 will fail again creating the proof file. Your result will become a standard PRP without a proof.

BTW, in build 4 prime95 will try twice and give up -- no need to do the MaxProofgenWaits hack.[/QUOTE]

Good to know. Yes, it failed twice and has now moved on to the next assignment.

I've given P95 access to the vast majority of the system resources, so I'm not entirely sure why I received the error. At this point I'm going to assume the issue was Norton.

Yuno 2020-08-31 21:46

There used to be 2 settings related to extra error checking, now theres 1, is that intentional and they are merged together now or..?

Xyzzy 2020-08-31 22:36

[QUOTE=Yuno;555585]There used to be 2 settings related to extra error checking, now theres 1, is that intentional and they are merged together now or..?[/QUOTE][QUOTE=Prime95;554250]Sum-input-output checking was removed because no one uses it. Optimizations made for AVX and later FFTs made this error check impossible. If you have a Pentium 4 or maybe some of the early Core/Core2 CPUs, then you could get the sum-input error checking.

All the code is still there, only the menu choice is gone. That is, you can activate the error checking on ancient CPUs by editing prime.txt.[/QUOTE]:mike:

Chuck 2020-09-01 00:13

We set up build 4 following George's instructions and we haven't seen a cert job yet.

:sad:

??? I got 17 so far today.

Xyzzy 2020-09-01 00:55

We just got three, so maybe there wasn't enough before?

:mike:

moebius 2020-09-01 05:09

[QUOTE=Chuck;555600]I got 17 so far today.[/QUOTE]
I got my first Cert this Morning with Prime95 30.3 build 4.
[B][SIZE="2"][Tue Sep 1 04:45:57 2020 - ver 30.3]
PrimeNet success code with additional info:
Server assigned CERT work.
Got assignment xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx: CERT [URL="https://www.mersenne.org/report_exponent/?exp_lo=99875231&full=1"]M99875231[/URL]
Sending result to server: UID: *******/Ryzen7-3700X, M99875231 certification hash value *************************************. Wh4: ***********************,00000000, AID: *******************

PrimeNet success code with additional info:
Certification successfully completes double-check of M99875231 --
CPU credit is 1.4879 GHz-days.[/SIZE][/B]

pinhodecarlos 2020-09-01 17:27

Think I’m now consider a CERT volunteer since I’ve only done this type of work for the last 12 hours and the queue settings are now working. Fantastic!

Xyzzy 2020-09-01 17:50

If you want to do only CERT work, in [C]prime.txt[/C] put "[C]MaxExponents=0[/C]".

:mike:

DrobinsonPE 2020-09-02 02:06

1 Attachment(s)
Updated my Celeron J4105 computer with 4 GB of RAM to Prime 95 V30.3 build 4 and my Max Emergency Memory issues still happens. This time I took a screen shot. The only way to keep the default setting of 1 GB is to cancel out of the Advanced Resource Limits popup.

Prime95 2020-09-02 02:38

[QUOTE=DrobinsonPE;555730]Updated my Celeron J4105 computer with 4 GB of RAM to Prime 95 V30.3 build 4 and my Max Emergency Memory issues still happens. [/QUOTE]

My bad, I did not address this. Fixed in build 5 -- whenever that comes out.

jwnutter 2020-09-02 15:14

Not a big deal, but why did I receive Cert work for an exponent that had been previously factored many times over?

[url]http://www.mersenne.ca/exponent/2319257[/url]

James Heinrich 2020-09-02 15:29

[QUOTE=jwnutter;555771]Not a big deal, but why did I receive Cert work for an [url=https://www.mersenne.ca/M2319257]exponent[/url] that had been previously factored many times over?[/QUOTE]PRP test are performed on entire exponents, as well as the remaining cofactor for partially-factored exponents. I assume your cert on [M]M2319257[/M] falls into the latter category, certifying the PRP on the remaining cofactor.

James Heinrich 2020-09-02 15:49

I thought there was a way, but I can't remember how (nor find reference to it at the moment), to specify different worktypes and number of threads for different workers. Like if I want to run P-1 in worker1 and ECM in worker2, etc. Is that possible, if so how?

kriesel 2020-09-02 16:01

[QUOTE=James Heinrich;555775]I thought there was a way, but I can't remember how (nor find reference to it at the moment), to specify different worktypes and number of threads for different workers. Like if I want to run P-1 in worker1 and ECM in worker2, etc. Is that possible, if so how?[/QUOTE]
In prime95, Test, Worker Windows..., then at worker number, click the square button to expand the dropdown list, select a single worker, Then Type of Work to get, expand the dropdown list, select one. Repeat for as many as you want to change from the previous setting(s).

James Heinrich 2020-09-02 16:12

Ah, yes, I'd forgotten that, thank you kindly.

Uncwilly 2020-09-02 18:18

[QUOTE=jwnutter;555771]Not a big deal, but why did I receive Cert work for an exponent that had been previously factored many times over?

[url]http://www.mersenne.ca/exponent/2319257[/url][/QUOTE]
[QUOTE=James Heinrich;555772]PRP test are performed on entire exponents, as well as the remaining cofactor for partially-factored exponents. I assume your cert on [M]M2319257[/M] falls into the latter category, certifying the PRP on the remaining cofactor.[/QUOTE]
That is right. TJAOI recently dumped a bunch on new factors, so there is some new PRP-CF (cofactor) work on exponents with previously known other factors. Each time new factors are found a new PRP gets run on the remaining cofactor. ViliamF, myslef, and a few others doing PRP-CF have started using v30

jwnutter 2020-09-02 20:07

[QUOTE=Uncwilly;555793]That is right. TJAOI recently dumped a bunch on new factors, so there is some new PRP-CF (cofactor) work on exponents with previously known other factors. Each time new factors are found a new PRP gets run on the remaining cofactor. ViliamF, myslef, and a few others doing PRP-CF have started using v30[/QUOTE]

Understood, thanks for clarifying.

jwnutter 2020-09-03 15:37

I noticed that I've added two new CPUs to my account by upgrading from 29.8 b6 to 30.3 b3 and now 30.3 b4. I've merged CPUs before for a variate of reasons, but this generally comes with a reliability/confidence penalty that impacts my assignment category for a few months.

Is there a way to merge CPUs without incurring this assignment penalty, which, as I understand, is linked to CPU reliability and confidence?

James Heinrich 2020-09-04 11:36

@George: I've been wondering if it would be possible to have a stage-2 resume-from-disk option for P-1.
I do mostly P-1 and whenever stage2 get interrupted (by [i]LowMemWhileRunning[/i] for example) the memory usage drops to zero quickly, but upon resuming it can take a very long time to rebuild whatever it is that takes up the RAM in stage2. In my case with 13GB allocated it takes 11 minutes to resume processing.
My thought was to have an optional (disabled-by-default) setting where I could specify a temp directory to which P-1 stage2 RAM snapshots could be saved and read back from disk rather than rebuilding. With an SSD especially it would seem potentially much faster. Is this possible, or is there a reason it wouldn't work?

Prime95 2020-09-04 15:39

[QUOTE=James Heinrich;555999]@George: I've been wondering if it would be possible to have a stage-2 resume-from-disk option for P-1.
I do mostly P-1 and whenever stage2 get interrupted (by [i]LowMemWhileRunning[/i] for example) the memory usage drops to zero quickly, but upon resuming it can take a very long time to rebuild whatever it is that takes up the RAM in stage2. In my case with 13GB allocated it takes 11 minutes to resume processing.
My thought was to have an optional (disabled-by-default) setting where I could specify a temp directory to which P-1 stage2 RAM snapshots could be saved and read back from disk rather than rebuilding. With an SSD especially it would seem potentially much faster. Is this possible, or is there a reason it wouldn't work?[/QUOTE]

I suppose it is possible. Have you tried using PauseWhileRunning? The OS should swap prime95's memory out to disk while your other program is running for the same effect.

James Heinrich 2020-09-04 16:02

[QUOTE=Prime95;556018]I suppose it is possible. Have you tried using PauseWhileRunning? The OS should swap prime95's memory out to disk while your other program is running for the same effect.[/QUOTE]PauseWhileRunning also dumps the RAM before pausing (I think this change might have actually been my own request from many years back, it may have not done that in the original implementation, but I'm not certain on this). LowMemWhileRunning is an extension of PauseWhileRunning; they both stop the current high-memory task (if any) and release the memory; then LowMemWhileRunning goes to look for something else to do instead of just pausing.

I prefer to run without a swap file, 64GB is plenty for pretty much everything (unless of course 40GB is taken up by P-1 :smile:).

Runtime Error 2020-09-04 22:22

"Offline" manual testing
 
Not sure if working as intended, but "UsePrimenet=0" alone does not seem to stop mprime from attempting to upload proof files and repeatedly throwing the 401 error.

[CODE]Unexpected error during pXXXXXX.proof upload: {"error_status":401,"error_message":"Unauthorized","error_description":"JSON result not yet submitted by your user ID or proof file MD5 mismatch"}
Unexpected error during pXXXXXX.proof upload: {"error_status":401,"error_message":"Unauthorized","error_description":"JSON result not yet submitted by your user ID or proof file MD5 mismatch"}
Unexpected error during pXXXXXX.proof upload: {"error_status":401,"error_message":"Unauthorized","error_description":"JSON result not yet submitted by your user ID or proof file MD5 mismatch"}
Unexpected error during pXXXXXX.proof upload: {"error_status":401,"error_message":"Unauthorized","error_description":"JSON result not yet submitted by your user ID or proof file MD5 mismatch"}[/CODE]

I suppose the following together should do it?
[CODE]
UsePrimenet=0
UploadRateLimit=0
DownloadDailyLimit=0
[/CODE]

Thank you!

James Heinrich 2020-09-05 19:05

A few random thoughts I'd like to see in Prime95:

1) Treat worktodo as a unified pool of work (if [c]!SequentialWorkToDo[/c]). This is perhaps more applicable to workers that have to fight for high-memory availability. Not infrequently I've seen workers stalled for lack of stage2 memory (due to [c]MaxHighMemWorkers[/c]) and sit idle while there is work available (that doesn't need RAM, LL/PRP/Cert/stage1) in worktodo under another worker's section. I would like to see if [c]SequentialWorkToDo=1[/c] then handle worktodo as it currently is with worker-specific sections and order, but otherwise just have a single-section worktodo and each worker can pick the next suitable available task to work on from the whole pool.

2) Allow small RAM use to ignore [c]MaxHighMemWorkers[/c], for example if I'm running P-1 at 20GB/ea in two workers, and have [c]MaxHighMemWorkers=2[/c], and a third worker wants 500MB for ECM I would like to be able to specify an amount of RAM that ECM (or maybe the last few RPs of a P-1, or anything else that might need a small amount of RAM) could use on top of the limits specified for [c]MaxHighMemWorkers[/c]. This could either be implemented as a specified config entry, or automatically as the difference between the overall [c]Memory[/c] setting and the sum of [c]Memory[/c] setting for individual workers (e.g. each worker has [c]Memory=20GB[/c], [c]MaxHighMemWorkers=2[/c], overall [c]Memory=45GB[/c], therefore 5GB is available for "other" uses.

chalsall 2020-09-05 19:20

[QUOTE=James Heinrich;556157]A few random thoughts I'd like to see in Prime95:[/QUOTE]

Excellent suggestions. I'd likely leverage on those myself in the future, if they were available.

George: An "easy win", I've asked for before: Could you please make "q|Q" exit out of the mprime TUI, in addition to "5"? (I know... annoying humans... :smile:)

Prime95 2020-09-05 19:58

[QUOTE=James Heinrich;556157]
2) Allow small RAM use to ignore [c]MaxHighMemWorkers[/c], for example if I'm running P-1 at 20GB/ea in two workers, and have [c]MaxHighMemWorkers=2[/c], and a third worker wants 500MB for ECM I would like to be able to specify an amount of RAM that ECM (or maybe the last few RPs of a P-1, or anything else that might need a small amount of RAM) could use on top of the limits specified for [c]MaxHighMemWorkers[/c]. This could either be implemented as a specified config entry, or automatically as the difference between the overall [c]Memory[/c] setting and the sum of [c]Memory[/c] setting for individual workers (e.g. each worker has [c]Memory=20GB[/c], [c]MaxHighMemWorkers=2[/c], overall [c]Memory=45GB[/c], therefore 5GB is available for "other" uses.[/QUOTE]

In 30.3 build 5, set HighMemThreshold=n in local.txt.

Prime95 2020-09-05 20:00

[QUOTE=Runtime Error;556064]Not sure if working as intended, but "UsePrimenet=0" alone does not seem to stop mprime from attempting to upload proof files [/QUOTE]

I will change it.

ssybesma 2020-09-05 20:47

30.3 build 5?
 
[QUOTE=Prime95;556162]In 30.3 build 5, set HighMemThreshold=n in local.txt.[/QUOTE]


I checked the FTP site. You mentioned build 5 as it if existed, but I guess you're working on that still given there's some late fixes? Thank you.

James Heinrich 2020-09-05 20:52

[QUOTE=ssybesma;556170]I checked the FTP site. You mentioned build 5 as it if existed[/QUOTE]I interpreted it as George has changed it just now for build 5 (that he is currently working on), so when it (or a newer) build is available for download, then I can test the new feature.

ssybesma 2020-09-05 21:35

[QUOTE=James Heinrich;556172]I interpreted it as George has changed it just now for build 5 (that he is currently working on), so when it (or a newer) build is available for download, then I can test the new feature.[/QUOTE]


OK, gotcha. Thank you James. Currently using build 4. Looks like I got a CERT assigned to my most able machine. I recently brought online a Core i7 8700 3.2GHz (Dell Precision 3630 Tower) as my 18th machine.

James Heinrich 2020-09-06 12:38

[QUOTE=James Heinrich;556024]PauseWhileRunning also dumps the RAM before pausing. LowMemWhileRunning is an extension of PauseWhileRunning; they both stop the current high-memory task (if any) and release the memory; then LowMemWhileRunning goes to look for something else to do instead of just pausing.[/QUOTE]Perhaps this could be revised thus:[LIST][*]If a program is in [c]PauseWhileRunning[/c] then Prime95 pauses (the selected number of cores) but does not automatically free any stage2 RAM.[*]If the same program is also in [c]LowMemWhileRunning[/c] then RAM will be freed.[*]If [c]LowMemWhileRunningTempDir[/c] is set (and valid, and contains sufficient space) then the RAM contents will be written to a temp file prior to freeing memory so that it can be read back quickly on resume. Temp files should be written sequentially by worker, not simultaneously, to avoid disk thrashing (especially if temp disk is HDD).[*]When stage2 is allowed to resume, worker checks [c]LowMemWhileRunningTempDir[/c] for a matching RAMdump file and reads it into memory, if not then falls back to current system of rebuilding RAM contents. RAMdump file should probably contain a simple checksum (CRC32, MD5, etc) as the last few bytes of the file to ensure that the temp file was written and read completely and correctly[/LIST]

storm5510 2020-09-07 16:22

I came across something in my PRP results. Take a look at the page for [URL="https://www.mersenne.org/report_exponent/?exp_lo=10496897&full=1"]M10496897[/URL]. It has a bit of a long history. In there are three LL tests from long ago. [B]ATH[/B] and myself ran it on the same day, August 26. I ran it as a PRP, and [B]ATH[/B] ran it as a PRP-CF. Same residue for both. Is this not a duplication of effort? It seem to me like it would be...

James Heinrich 2020-09-07 16:34

[QUOTE=storm5510;556345][URL="https://www.mersenne.org/report_exponent/?exp_lo=10496897&full=1"]M10496897[/URL]... I ran it as a PRP, and [B]ATH[/B] ran it as a PRP-CF. Is this not a duplication of effort?[/QUOTE]Your run showed that the entire Mersenne number was composite, but we knew that already because a factor was found 12 days earlier (presumably prior to you starting the assignment). ATH shows that there is more than one factor; that the remaining cofactor is composite.

Viliam Furik 2020-09-07 19:08

I have checked the page of the exponent, and apart from noticing I was the one to find the factor :banana:, I have noticed that Ben Delo's certification looks stuck on 100%, with supposedly one day left to completion.

Prime95 2020-09-07 19:22

The server must be confused on that exponent. With two proofs for the same exponent, both with and without known factors, I was not thinking of that case! I'll investigate.

Prime95 2020-09-07 22:06

[QUOTE=storm5510;556345]I came across something in my PRP results. Take a look at the page for [URL="https://www.mersenne.org/report_exponent/?exp_lo=10496897&full=1"]M10496897[/URL]. It has a bit of a long history. In there are three LL tests from long ago. [B]ATH[/B] and myself ran it on the same day, August 26. I ran it as a PRP, and [B]ATH[/B] ran it as a PRP-CF. Same residue for both. Is this not a duplication of effort? It seem to me like it would be...[/QUOTE]

Storm your gpuowl proof was no good. You might want to contact Mihai with the details.

storm5510 2020-09-07 23:02

[QUOTE=Prime95;556374]Storm your gpuowl proof was no good. You might want to contact Mihai with the details.[/QUOTE]

As of late, I went back to using [I]Prime95 30.3 B4 [/I]for this work type. I have had a few miscellaneous problems with [I]gpuOwl[/I]. Those were with P-1 tests. I keep all my results, so I have the line from this test. I will pass this on.

[U]Edit[/U]: Mihai. I am not familiar with this name. What is the user name?

James Heinrich 2020-09-07 23:14

[QUOTE=storm5510;556376]Mihai. I am not familiar with this name. What is the user name?[/QUOTE]Mihai Preda, developer of gpuowl. Forum username [url=https://www.mersenneforum.org/member.php?u=14481]preda[/url].

storm5510 2020-09-07 23:27

[QUOTE=James Heinrich;556377]Mihai Preda, developer of gpuowl. Forum username [URL="https://www.mersenneforum.org/member.php?u=14481"]preda[/URL].[/QUOTE]

Thank you! :smile:

Prime95 2020-09-08 00:12

[QUOTE=storm5510;556376]As of late, I went back to using [I]Prime95 30.3 B4 [/I]for this work type. I have had a few miscellaneous problems with [I]gpuOwl[/I]. Those were with P-1 tests. I keep all my results, so I have the line from this test. I will pass this on.[/QUOTE]

They (the small expos you tested) were PRP tests with a proof uploaded.

Bad proofs was for 10496897.

storm5510 2020-09-08 00:57

[QUOTE=Prime95;556385]They (the small expos you tested) were PRP tests with a proof uploaded.

Bad proofs was for 10496897.[/QUOTE]

[B]preda [/B]asked if I still had the proof file. I do. I have placed it where he can download it.

P-A 2020-09-08 14:57

I have upgraded to build 4 just after release. I am doing PRP first time. Since upgrading fro build 3 I have not received any CERT work even though I have set it to 90%, download limit 5000MB/ day and bandwidth 500Mbps. Before used to get CERT work. I have not tried do any modification, just wanted to see how it worked without advanced workarounds. Is this expected?

drew 2020-09-08 15:34

1 Attachment(s)
Hi, I just wanted to let you know that I downloaded this latest version (30.3 build 4), and now when I run the .exe I get the attached dialog (not sure if the attachment will show up in this post, so I'll included text here:

"Search for app in the Store?

You need to install an app for this task.
Would you like to search for one in the store?"

If I hit Yes or No nothing happens, but the program seems to work ok regardless.

Prime95 2020-09-08 20:11

[QUOTE=P-A;556421]Since upgrading fro build 3 I have not received any CERT work even though I have set it to 90%, download limit 5000MB/ day and bandwidth 500Mbps. Before used to get CERT work. I have not tried do any modification, just wanted to see how it worked without advanced workarounds. Is this expected?[/QUOTE]

Not too surprising. Build 4 requests CERT work less often. You can turn on "Debug=1" in the [PrimeNet] section of prime,txt to see every time Prime95 requests CERT work.


[QUOTE=drew;556426]Hi, I just wanted to let you know that I downloaded this latest version (30.3 build 4), and now when I run the .exe I get the attached dialog (not sure if the attachment will show up in this post, so I'll included text here:

"Search for app in the Store?

You need to install an app for this task.
Would you like to search for one in the store?"

If I hit Yes or No nothing happens, but the program seems to work ok regardless.[/QUOTE]

How weird. Some Windows malfunction -- anyone else see this?

pinhodecarlos 2020-09-08 20:24

Fine under Windows 7, isn’t that Windows 10?

DrobinsonPE 2020-09-08 20:33

[QUOTE=Prime95;556457]How weird. Some Windows malfunction -- anyone else see this?[/QUOTE]

I have two Windows 10 computers running 30.3 build 4 and have not seen that screen.

PhilF 2020-09-08 21:05

[QUOTE=drew;556426]Hi, I just wanted to let you know that I downloaded this latest version (30.3 build 4), and now when I run the .exe I get the attached dialog (not sure if the attachment will show up in this post, so I'll included text here:

"Search for app in the Store?

You need to install an app for this task.
Would you like to search for one in the store?"

If I hit Yes or No nothing happens, but the program seems to work ok regardless.[/QUOTE]

Could it be malware or a virus?

James Heinrich 2020-09-08 21:28

[QUOTE=drew;556426]"Search for app in the Store?
You need to install an app for this task.
Would you like to search for one in the store?"[/QUOTE]It seems to be a rare, but not-unknown, problem that affects a small number of Windows 10 installations. [url=https://answers.microsoft.com/en-us/windows/forum/all/search-for-app-in-the-store-always-appearing-when/c6a313e1-a63b-40e1-898d-5512ebfa54f9]This guy[/url] for example reports (from 2019) pretty much the exact same problem (launch program, get Yes/No dialog that seems to do nothing except go away whether you click yes or no). One [i]possible[/i] thing to look at would be to make sure that [c]Start > Settings > Apps > Allow apps from the Store only[/c] is not set.

Mark Rose 2020-09-08 21:42

So I got some literally free AWS hardware and over the past week I've been running mprime v30b3 on an m5.12xlarge. The 8 GB disk I gave it filled up at some point during the past week with .residues files. Would it be possible to have mprime/Prime95 configure itself appropriately based on the available disk space at the time a PRP test is started? It would have been fine with a Proof Power of 7 for the two 110M exponents it's working on.

After increasing the disk size to 50 GB, mprime wrote out a 3.3 GB and a 14 GB file. It should have never thought it would be possible to write out a 14 GB file in the first place.

Seems processing was completely stuck for 5 or 6 days. The system has 186 GB of memory free, so it could have kept processing...

Just bugs...

Prime95 2020-09-08 22:35

[QUOTE=Mark Rose;556471]So I got some literally free AWS hardware and over the past week I've been running mprime v30b3 on an m5.12xlarge. The 8 GB disk I gave it filled up at some point during the past week with .residues files. Would it be possible to have mprime/Prime95 configure itself appropriately based on the available disk space at the time a PRP test is started? It would have been fine with a Proof Power of 7 for the two 110M exponents it's working on.

After increasing the disk size to 50 GB, mprime wrote out a 3.3 GB and a 14 GB file. It should have never thought it would be possible to write out a 14 GB file in the first place.

Seems processing was completely stuck for 5 or 6 days. The system has 186 GB of memory free, so it could have kept processing....[/QUOTE]

There are settings for these things (disk space to use and emergency memory). Assuming you left settings at the default (6GB disk / worker), I'd be interested in why prime95 was trying to write a 14GB file. Can you send screen outputs, log files, etc. that might shed some light on what happened.

Prime95 2020-09-08 22:50

Build 5
 
Build 5 is now available. Other than a couple of little fixes, this version uses a new proof file format that is, on average, 0.000128% faster with default settings on a 100Mbit exponent!

OK, that's not very impressive. The proof files are about 5% easier to process on the server plus they allow us to calculate the res64 value from the proof file -- a small extra level of checking.

Mark Rose 2020-09-09 00:25

5 Attachment(s)
[QUOTE=Prime95;556474]There are settings for these things (disk space to use and emergency memory). Assuming you left settings at the default (6GB disk / worker), I'd be interested in why prime95 was trying to write a 14GB file. Can you send screen outputs, log files, etc. that might shed some light on what happened.[/QUOTE]

local.txt was configured entirely by using the menu system or automatically
prime.txt was primarily configured by me
ls -l.txt shows the system as it is now
screen snapshot.txt shows the error messages I was seeing. I don't have screen logs from when the errors started, but the errors went away once I added more space

It's possible the system wasn't actually idle during past week. These 110M exponents take a long time.

Mark Rose 2020-09-09 00:29

4 Attachment(s)
[QUOTE=Prime95;556474]There are settings for these things (disk space to use and emergency memory). Assuming you left settings at the default (6GB disk / worker), I'd be interested in why prime95 was trying to write a 14GB file. Can you send screen outputs, log files, etc. that might shed some light on what happened.[/QUOTE]

Attached here are the logs.

gwnum.txt was truncated at 16 KB; I guess running out of space for writes will do that

One strange thing I did notice was that the expected completion at PrimeNet had these exponents as finishing yesterday and today. Once I added disk space, I did stop and restart mprime, which then communicated the current expected time for the exponents to finish. That's why I thought processing may have completely stopped.

Prime95 2020-09-09 01:50

[QUOTE=Mark Rose;556486]local.txt was configured entirely by using the menu system or automatically
prime.txt was primarily configured by me
ls -l.txt shows the system as it is now.[/QUOTE]

From the file sizes, it is clear that worker #1 is trying to generate a power=10 proof (14GB) and worker #2 is generating the expected power=8 (3GB) proof.

The question is, was the temp disk space mprime ever set to more than 6GB -- specifically at the time worker #1 started it's PRP assignment? If not, and I suspect it never was, then mprime has a bug that under some conditions it miscalculates the proper proof power to use.

Prime95 2020-09-09 02:03

Another weird thing. At start of the PRP test, the entire interim residues file is allocated. If an error occurs, mprime drops the proof power down. Why did this fail-safe not work?

LaurV 2020-09-09 02:36

[U][B]stupid question:[/B][/U] Can it be that the resources on those aws clusters vary in some cases by occupancy or by method of interrogation? (i.e. when you ask how space is free on disk or memory, you get larger values, but in reality, or later as the cluster gets busier, the values are smaller / restricted, etc. - you know what I mean, that's why is called "elastic" computing, haha, well, sorry for the stupidity, I am the layman here, I never played with ec2).

Mark Rose 2020-09-09 02:49

[QUOTE=Prime95;556498]From the file sizes, it is clear that worker #1 is trying to generate a power=10 proof (14GB) and worker #2 is generating the expected power=8 (3GB) proof.

The question is, was the temp disk space mprime ever set to more than 6GB -- specifically at the time worker #1 started it's PRP assignment? If not, and I suspect it never was, then mprime has a bug that under some conditions it miscalculates the proper proof power to use.[/QUOTE]

[QUOTE=Prime95;556499]Another weird thing. At start of the PRP test, the entire interim residues file is allocated. If an error occurs, mprime drops the proof power down. Why did this fail-safe not work?[/QUOTE]

This is a brand new installation. I had never configured it to be anything other than 6 GB, and I believe I had done that via the menus when I first set it up.

I had initially copied my config file which tells mprime to do DC work, with WorkPreference=151 iirc. mprime fetched 6 LLDC assignments for 6 workers. I then ran some benchmarks and determined the optimal number of workers to run on this hardware was 2. I stopped mprime, edited worktodo.txt for two workers and modified local.txt from 6 workers 4 cores each to 2 workers 12 cores each.

It hadn't started any PRP work until three or so days later, well after I made changes to config files. I didn't configure it to do PRP, but somehow the WorkPreference got changed to 0.

It does appear worker #1 is significantly ahead of worker #2 and started doing its PRP assignment 16 hours earlier based on the NF-PM1 results I see at mersenne.org.

Also, mersenne.org is still showing 6 workers for the CPU instead of the 2 that are configured. Don't know if that's a bug with mprime or mersenne.org.

Mark Rose 2020-09-09 02:52

[QUOTE=LaurV;556502][U][B]stupid question:[/B][/U] Can it be that the resources on those aws clusters vary in some cases by occupancy or by method of interrogation? (i.e. when you ask how space is free on disk or memory, you get larger values, but in reality, or later as the cluster gets busier, the values are smaller / restricted, etc. - you know what I mean, that's why is called "elastic" computing, haha, well, sorry for the stupidity, I am the layman here, I never played with ec2).[/QUOTE]

Not in this case. This instance is using the ext4 filesystem, with an initial size of 8 GB. I manually grew it. Log files and whatnot can consume a little space, but nothing that would make it think there was more disk available.

You could get weird stuff like that using btrfs in raid1.

AWS also has an NFS implementation they call EFS, where I suppose that could happen.

Prime95 2020-09-09 03:59

[QUOTE=Mark Rose;556503]I didn't configure it to do PRP, but somehow the WorkPreference got changed to 0.[/QUOTE]

We're not going to be able to reproduce this problem. It looks like memory corruption. WorkPreference and Allowable temp disk space are both stored in global variables. Given AWS' proven reliablilty we conclude it is a rather rare program bug.

I'll try testing running out of disk space on a Windows box to see if I can discover this was not handled better.

Prime95 2020-09-09 04:52

[QUOTE=Prime95;556514]I'll try testing running out of disk space on a Windows box to see if I can discover this was not handled better.[/QUOTE]

Aha - a bug.

From your results.txt:

[CODE][Mon Aug 31 09:07:44 2020]
Error pre-allocating proof interim residues file
Errno: 2, No such file or directory
Will use proof power 10 instead of 8.[/CODE]

The code that was supposed to reduce the proof power when an error occurs preallocating the disk space was in fact increasing the proof power.

Now the only unexplained phenomenon is the work preference changing.

Mark Rose 2020-09-09 05:53

[QUOTE=Prime95;556517]Aha - a bug.[/quote]

I'm glad I decided to test a low disk space situation. :smile:

[Quote]Now the only unexplained phenomenon is the work preference changing.[/QUOTE]

I've never had that one happen to me before. Considering mprime started doing PRP assignments immediately after the LLDC work finished, it probably already had a modified configuration beforehand when it fetched assignments. I had been running benchmarks, and control-c'ed out of them, so that may be the place to look?

I am certain I left mprime set to do LLDC because I figured others would be testing the new PRP hotness and thought someone should test the old.

stippix 2020-09-09 08:03

Hi,

FYI and just for fun,
32-bit linux version works fine for my very, very old 32-bit machines (Athlon XPs ;) using ubuntu-mate 16.04 and Linux Mint Mate 19.2,
but 32-bit Windows version 30.3beta4 and beta5 does not start on Windows XP and Windows 7,

greetings.

James Heinrich 2020-09-09 10:25

[QUOTE=stippix;556529]32-bit Windows version 30.3beta4 and beta5 does not start on Windows XP and Windows 7[/QUOTE]I can confirm that [c]p95v303b5.win32[/c] crashes on startup on Win7-64:[code]Problem signature:
Problem Event Name: APPCRASH
Application Name: prime95.exe
Application Version: 30.3.1.0
Application Timestamp: 5f56f582
Fault Module Name: libhwloc-15.dll
Fault Module Version: 0.0.0.0
Fault Module Timestamp: 5ab8b310
Exception Code: c0000005
Exception Offset: 00007837
OS Version: 6.1.7601.2.1.0.256.48
Locale ID: 1033
Additional Information 1: 0a9e
Additional Information 2: 0a9e372d3b4ad19135b953a78882e789
Additional Information 3: 0a9e
Additional Information 4: 0a9e372d3b4ad19135b953a78882e789[/code]

James Heinrich 2020-09-09 16:42

1 Attachment(s)
Something odd with the [c]Test > Status...[/c] dialog. I realize that Prime95 will attempt to do Cert work first and therefore out-of-order compared to how it's shown here and in worktodo.txt but I don't understand why the ETA for all the Cert work is exactly the same. Bug?

jwnutter 2020-09-09 17:59

1 Attachment(s)
A question and a comment related to v30.3 b4.

Question: Can someone direct me to the HT options in v30.3 b4? For some reason I'm unable to find this menu in b4, but I recall seeing it during the initial setup.

Comment: I recall someone had an issue with a missing check in the "get occasional proof certification work" field. I too had the same issue but caught the problem before I closed the worker window. My specific issue resulted from an errant mouse click at the bottom of the worker window. After a little investigation I found that the user can modify this checkbox by clicking up to (from left to right) approximately the red vertical line shown in the image below.

James Heinrich 2020-09-09 18:09

[QUOTE=jwnutter;556576]Can someone direct me to the HT options in v30.3 b4?[/QUOTE][c]Test > Worker Windows...[/c] sets the number of workers, [c]CPU cores to use (multithreading)[/c] in that dialog set the number of threads per worker.

The settings for actual hyperthreading is under [c]Options > Resource Limits... > Advanced... > Use hyperthreading for ___[/c] checkboxes at the bottom.
Note however that hyperthreading is explicitly not recommended (other than for TF, which itself is not recommended for CPUs):[quote=readme.txt]Except for trial factoring, which is best left for GPUs to do, hyperthreading often offers no performance benefit while using more electricity. You can try test if hyperthreading speeds up your worker windows by selecting these options.[/quote]

jwnutter 2020-09-09 18:15

[QUOTE=James Heinrich;556577]
The settings for actual hyperthreading is under [c]Options > Resource Limits... > Advanced... > Use hyperthreading for ___[/c] checkboxes at the bottom.
Note however that hyperthreading is explicitly not recommended (other than for TF, which itself is not recommended for CPUs):[/QUOTE]

Thanks, James. I had overlooked the advanced button. And, understood on the HT/TF comment.

Prime95 2020-09-09 18:18

[QUOTE=James Heinrich;556537]I can confirm that [c]p95v303b5.win32[/c] crashes on startup on Win7-64:[/QUOTE]

Hmm, old hwloc DLL. Try this one [url]https://www.dropbox.com/s/dmw9nbb0vebteuq/libhwloc-15.dll?dl=0[/url] or get it from the hwloc website: [url]https://www.open-mpi.org/software/hwloc/v2.2/[/url]

[QUOTE=James Heinrich;556566]Something odd with the [c]Test > Status...[/c] dialog. I realize that Prime95 will attempt to do Cert work first and therefore out-of-order compared to how it's shown here and in worktodo.txt but I don't understand why the ETA for all the Cert work is exactly the same. Bug?[/QUOTE]

I'll see if I can't to better for build 6.

kriesel 2020-09-09 18:19

[QUOTE=James Heinrich;556577]The settings for actual hyperthreading is under [c]Options > Resource Limits... > Advanced... > Use hyperthreading for ___[/c] checkboxes at the bottom.
Note however that hyperthreading is explicitly not recommended (other than for TF, which itself is not recommended for CPUs):[/QUOTE]Hyperthreading in primality testing typically has a speed penalty. Rarely it will show an advantage. It seems to depend on fft length, processor model, and number of workers. Benchmark and know what's best for your own setup. The guidance seems to be a good general rule, and there are exceptions. Some examples: [URL]https://www.mersenneforum.org/showpost.php?p=504218&postcount=4[/URL]
[URL]https://www.mersenneforum.org/showpost.php?p=504219&postcount=5[/URL]

Prime95 2020-09-09 18:26

[QUOTE=jwnutter;556576] I found that the user can modify this checkbox by clicking up to (from left to right) approximately the red vertical line shown in the image below.[/QUOTE]

A new Windows annoyance! As a developer one must leave extra space after the text so that the text is not truncated when different system fonts are used (maybe not as much extra space as I allocated). MS also has an annoying habit of changing the system font with new Windows releases that can lead to dialog box text getting truncated.

stippix 2020-09-10 06:20

[QUOTE=Prime95;556579]Hmm, old hwloc DLL. Try this one [url]https://www.dropbox.com/s/dmw9nbb0vebteuq/libhwloc-15.dll?dl=0[/url] or get it from the hwloc website: [url]https://www.open-mpi.org/software/hwloc/v2.2/[/url]



I'll see if I can't to better for build 6.[/QUOTE]

Hi,
new libhwloc from their homepage solved the problem for beta 4 and 5,
greetings.

jwnutter 2020-09-10 17:52

Understood, thank you George.

This may not be the correct place for this comment, but I have a P95 feature request. P95 appears to run a quick benchmark prior to starting PRP work - but I'm not sure if this is due to a user setting somewhere or if this occurs for everyone. Could P95 also start work following a user initiated benchmark? Maybe a new check box in the benchmark menu that allows the user to toggle this option on/off?

henryzz 2020-09-10 19:03

[QUOTE=jwnutter;556652]Understood, thank you George.

This may not be the correct place for this comment, but I have a P95 feature request. P95 appears to run a quick benchmark prior to starting PRP work - but I'm not sure if this is due to a user setting somewhere or if this occurs for everyone. Could P95 also start work following a user initiated benchmark? Maybe a new check box in the benchmark menu that allows the user to toggle this option on/off?[/QUOTE]

I think the way to do this would be to start work and then initiate a benchmark. The work would be resumed after the benchmark.

jwnutter 2020-09-10 19:05

[QUOTE=henryzz;556656]I think the way to do this would be to start work and then initiate a benchmark. The work would be resumed after the benchmark.[/QUOTE]

Ahhhh, very nice. I'll try that next time. Thanks!

kriesel 2020-09-10 21:07

Windows Vista 64 bit Home Basic Core2 Duo 6GB ram
(Prime95 V29.8b6 x64 ran on this system again after I woke it up from a long summer nap.)

Prime95 V30.3b5 win64 first impression:

[CODE]Prime95.exe Entry Point Not Found
The procedure entry point K32GetModuleFileNameExA could not be
located in the dynamic link library KERNEL32.dll[/CODE][CODE]Problem signature:
Problem Event Name: APPCRASH
Application Name: prime95.exe
Application Version: 30.3.1.0
Application Timestamp: 5f56f570
Fault Module Name: KERNEL32.dll!K32GetModuleFileNameExA
Fault Module Version: 6.0.6002.18005
Fault Module Timestamp: 49e0421d
Exception Code: c0000139
Exception Offset: 00000000000b8fb8
OS Version: 6.0.6002.2.2.0.768.2
Locale ID: 1033
Additional Information 1: fa3e
Additional Information 2: ac0507478d1c5bd693cfc4fe3987e900
Additional Information 3: fa3e
Additional Information 4: ac0507478d1c5bd693cfc4fe3987e900

Read our privacy statement:
http://go.microsoft.com/fwlink/?linkid=50163&clcid=0x0409[/CODE] libhwloc-15.dll appears to be the current version.
09/10/2020 03:56 PM 1,753,095 libhwloc-15.dll (created 3/30/2020 5:31 pm)
Issue duplicated with b3, b2, b1.
Retreated to 29.8b6 for now.

Prime95 2020-09-10 21:49

[QUOTE=kriesel;556666]Windows Vista Home Basic Core2 Duo 6GB ram
(Prime95 V29.8b6 x64 ran on this system again after I woke it up from a long summer nap.)

Prime95 V30.3b5 win64 first impression:

[CODE]Prime95.exe Entry Point Not Found
The procedure entry point K32GetModuleFileNameExA could not be
located in the dynamic link library KERNEL32.dll[/CODE].[/QUOTE]

This is a result of upgrading to Visual Studio 2019 (from MSVC 2005)

MS docs at [url]https://docs.microsoft.com/en-us/windows/win32/api/psapi/nf-psapi-getmodulefilenameexa[/url] suggest a work-around. I will try that for build 6.

Prime95 2020-09-11 03:27

Build 6 is now available.

kriesel 2020-09-11 06:19

[QUOTE=kriesel;556666]Windows Vista 64 bit Home Basic Core2 Duo 6GB ram
...
Issue duplicated with b3, b2, b1.[/QUOTE]
For v30.3b5, same applies to Windows Vista Business 32-bit on Pentium M 2GB ram.
And XP 32 bit on Pentium M 2GB ram says it is not a valid Win32 application.
Will retry with b6 when time allows.

James Heinrich 2020-09-11 10:23

The [c]Test > Status...[/c] dialog can only display a certain number (depending on worker count) of assignments per worker, however I speculate that the ETA is [i]calculated[/i] for all the assignments in worktodo. P-1 ETA is slightly expensive to calculate. If I have a large number (hundreds) of small P-1 assignments it takes a long time to display the status dialog, even though it's only displaying the ETA status for 8x2 of them.

storm5510 2020-09-11 13:33

[QUOTE=Prime95;556694]Build 6 is now available.[/QUOTE]

I always look at the dates of the libraries to see if any are new. I have not been extracting any of them with each new release. Perhaps you could mention if any need to be replaced. :smile:

JuanTutors 2020-09-12 04:12

I'm starting a new 100M PRP test in about 30 days. This version of P95 seems to be coming along strongly. Should I switch to this version of P95 when the next exponent starts?

Prime95 2020-09-12 04:47

[QUOTE=JuanTutors;556783]I'm starting a new 100M PRP test in about 30 days. This version of P95 seems to be coming along strongly. Should I switch to this version of P95 when the next exponent starts?[/QUOTE]

Switch now. It should be backward compatible. (but make a backup just to be safe).

James Heinrich 2020-09-12 11:59

1 Attachment(s)
[QUOTE=Prime95;556579][QUOTE=James Heinrich;556566][c]Test > Status...[/c] dialog... ETA for all the Cert work is exactly the same.[/QUOTE]I'll see if I can't to better for build 6.[/QUOTE]Much better in 30.3b6, thank you.

Runtime Error 2020-09-13 14:03

Much to my dismay, I accidentally re-ran a PRP on [M]107972569[/M] that I had already completed with proof now verified. Prime95 will not upload the second proof file. Instead, it throws error 409, proof already uploaded. My guess is this is working as intended to save space and prevent redundant computations sever side, but I thought it might be worth posting since I doubt anyone else has made the same dumb mistake that I did. Thanks!

storm5510 2020-09-13 14:23

This is probably going to be off-topic, but I don't know where else to put it. I am having a difficulty with [I]mprime[/I] and I suspect it may have something to do with my recent forced update to Windows 10 v2004. This is for the latest [I]mprime[/I] build. I am trying to run it with Ubuntu on Windows. In the past, I have been able to get it started like so:

[QUOTE]sudo ./mprime -d[/QUOTE]Using "sudo" has something to do with permissions which I don't know how to set. If I don't use it then I get a permission denied message. I get the following result:

[QUOTE]sudo ./mprime: command not found.[/QUOTE]I know I am in the right folder because I can give it a "ls" command and I can see the binary is there. I know my password is correct because I tried one which was incorrect and it refused.

I really don't know where to go with this... :confused:

kriesel 2020-09-13 14:47

Win some, lose some
 
[QUOTE=kriesel;556700]For v30.3b5, same applies to Windows Vista Business 32-bit on Pentium M 2GB ram.
And XP 32 bit on Pentium M 2GB ram says it is not a valid Win32 application.
Will retry with b6 when time allows.[/QUOTE]
V30.3build 6 resolved:
[URL="https://www.mersenneforum.org/showpost.php?p=556666&postcount=268"]..EXA issue[/URL], on both Vista Home Basic 64 bit and Vista Business 32 bit.

V30.3build 6 did not resolve:
Prime95 failure to launch on Windows XP 32bit. (v29.8b6 runs fine there.)

kruoli 2020-09-13 15:07

[QUOTE=storm5510;556859]I know I am in the right folder because I can give it a "ls" command and I can see the binary is there.[/QUOTE]

Would you mind posting the result of [c]ls -lah[/c] in that folder? Maybe it's a issue of rights. Normally, mprime should not need root privileges (i.e. "sudo").

axn 2020-09-13 15:08

[QUOTE=storm5510;556859]Using "sudo" has something to do with permissions which I don't know how to set. If I don't use it then I get a permission denied message. [/QUOTE]

I'm guessing [C]mprime[/C] doesn't have "execute" permission. I'm also guessing mprime's owner is not your login account.

If you do [C]ls -l[/C] and paste it here, we can confirm.

Use [C]sudo chown[/C] to change the ownership of the files in that folder. And [C]chmod[/C] to set the execute bit for [C]mprime[/C]

Prime95 2020-09-13 15:43

[QUOTE=kriesel;556861]
V30.3build 6 did not resolve:
Prime95 failure to launch on Windows XP 32bit. (v29.8b6 runs fine there.)[/QUOTE]

Same "KERNEL32.dll!K32GetModuleFileNameExA" error message?

kriesel 2020-09-13 17:37

3 Attachment(s)
[QUOTE=Prime95;556875]Same "KERNEL32.dll!K32GetModuleFileNameExA" error message?[/QUOTE]No, completely separate. [QUOTE=kriesel;556700]For v30.3b5
...
And XP 32 bit on Pentium M 2GB ram says it is not a valid Win32 application.
Will retry with b6 when time allows.[/QUOTE]Here's what it looks like, and v29.8b6 on the same system.
(Old slow system, low priority, but there it is. I have older...)

Prime95 2020-09-13 21:39

[QUOTE=Runtime Error;556856]Much to my dismay, I accidentally re-ran a PRP on [M]107972569[/M] that I had already completed with proof now verified. Prime95 will not upload the second proof file. Instead, it throws error 409, proof already uploaded. My guess is this is working as intended to save space and prevent redundant computations sever side, but I thought it might be worth posting since I doubt anyone else has made the same dumb mistake that I did. Thanks![/QUOTE]

Try again with the proof upload. I think it will now work.

Prime95 2020-09-13 23:02

[QUOTE=kriesel;556892]No, completely separate. Here's what it looks like, and v29.8b6 on the same system.
(Old slow system, low priority, but there it is. I have older...)[/QUOTE]

A quick scan of VS2019 documentation shows no easy or obvious solution. I don't intend to work on this any more. As you can imagine 15 year old 32-bit OSes are not a high priority, but I do appreciate the test. We at least found and fixed a bug for the remaining one or two Vista users.

Jwb52z 2020-09-14 04:13

I hope this question is ok because I didn't have time to read all of the current 26 pages. Is there a speed increase at all from version 29.8 build 5? Thank you in advance!

Prime95 2020-09-14 04:36

[QUOTE=Jwb52z;556923]I hope this question is ok because I didn't have time to read all of the current 26 pages. Is there a speed increase at all from version 29.8 build 5? Thank you in advance![/QUOTE]

Only if you count no longer having to do double-checking.

To more directly answer your question, no speed increase.

stippix 2020-09-14 06:38

[QUOTE=Prime95;556909]A quick scan of VS2019 documentation shows no easy or obvious solution. I don't intend to work on this any more. As you can imagine 15 year old 32-bit OSes are not a high priority, but I do appreciate the test. We at least found and fixed a bug for the remaining one or two Vista users.[/QUOTE]

Hello,
Windows XP needs older compiler version VS-2017 ...
[url]https://docs.microsoft.com/en-us/cpp/build/configuring-programs-for-windows-xp?view=vs-2019[/url]
Greetings

moebius 2020-09-14 08:32

[QUOTE=Prime95;556925]Only if you count no longer having to do double-checking.To more directly answer your question, no speed increase.[/QUOTE]
Somehow I have the feeling, that most of the normal (smaller) primenet users (like me) do not even know about the changes regarding PRP / CERTS, because they are required to read the forum for this purpose. I'm just wondering who will later do all of these PRP-DC's of the 92-110M exponents who currently come in. The CPU'S / GPU'S / RAM unfortunately won't be faster by a factor of 10 every year.

James Heinrich 2020-09-14 10:58

[QUOTE=Prime95;556925]Only if you count no longer having to do double-checking. To more directly answer your question, no speed increase.[/QUOTE]For [u]identical worktypes[/u] there is no speed increase. But by eliminating double-checks the overall GIMPS throughput is, while not quite double, it's maybe 70%(?) increased. But only when we get the bulk of users moved to v30 from v29 and older.

And for users like [i]Jwb52z[/i] (and myself) who like to do P-1, there is some slightly optimized bounds selection improvments in v30.

Uncwilly 2020-09-14 12:45

[QUOTE=moebius;556930][I][B]Somehow I have the feeling, that most of the normal (smaller) primenet users (like me) do not even know about the changes regarding PRP / CERTS, because they are required to read the forum for this purpose[/QUOTE]
George put a note on Mersenne.org. That is where the average user is likely to see it.


All times are UTC. The time now is 01:47.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.