[ANN][GAP] Gapcoin - Prime Gap Search - New Math Algo - CPU / GPU - Zero Premine

[gjhiggins]

If nobody are developing it and nobody are paying for the explorer then it will be gone.

It's got some life in it yet. Wizz has generously funded the chainz block explorer until 2022-07-09. I'll continue making technical contributions where I consider it necessary to maintain the codebase. Freiexchange has been seeing some recent interest.

Cheers

Graham

[1714052000]

1714052000

[1714052000]

1714052000

[1714052000]

1714052000

[1714052000]

1714052000

[Dies diem docet]

I have looked a little more at the gpu miner code.

It seams to be a mess.

I looked at the following code and could not understand why the miner prints out "Error AMD accelerated parallel processing found". I saw that he did not care if it was a AMD or Nvidia but how could it drop down to the last else and get the error.


if (strcmp(platformId, "amd") == 0)
    platformName = "AMD Accelerated Parallel Processing";
  else if (strcmp(platformId, "nvidia") == 0)
    platformName = "NVIDIA CUDA";
  else {
    pthread_mutex_lock(&io_mutex);                            
    cout << get_time() << "ERROR: platform " << platformId << " not supported ";
    cout << " use amd or nvidia" << endl;
    pthread_mutex_unlock(&io_mutex);                      
    exit(EXIT_FAILURE);

Also he don't use the exit somewere else. He don't check for the cards before this code executes. So this piece of code is not useful for the program.

Then I started to read more. There are over 100 lines of code that are repeating itself. Checking the cards again and again. Also the fact that some old code in pthread also could course problems.  I'm not a c or c++ programmer so I maybe can't fix all of that. It's probably a cut and paste problem that left so much faulty code there.  

I was only trying to fix faulty math against graphic cards and compile it. If it don't make harm I leave it in there for now.

  
I could not find any later gpu miner than 4.1 to work on. So if I'm wrong about that then point me in the right direction. This one I'm looking at. https://github.com/gapcoin/GapMiner

I said that I probably will fail.   All that faulty code don't make it easier.

That binaries that are on github are diffrent in execution than that Graham are using. Maybe they have updated the phread so it works better. I'm cconfused.

[benxy031]

I have looked a little more at the gpu miner code.

It seams to be a mess.

I looked at the following code and could not understand why the miner prints out "Error AMD accelerated parallel processing found". I saw that he did not care if it was a AMD or Nvidia but how could it drop down to the last else and get the error.


if (strcmp(platformId, "amd") == 0)
    platformName = "AMD Accelerated Parallel Processing";
  else if (strcmp(platformId, "nvidia") == 0)
    platformName = "NVIDIA CUDA";
  else {
    pthread_mutex_lock(&io_mutex);                            
    cout << get_time() << "ERROR: platform " << platformId << " not supported ";
    cout << " use amd or nvidia" << endl;
    pthread_mutex_unlock(&io_mutex);                      
    exit(EXIT_FAILURE);

Also he don't use the exit somewere else. He don't check for the cards before this code executes. So this piece of code is not useful for the program.

Then I started to read more. There are over 100 lines of code that are repeating itself. Checking the cards again and again. Also the fact that some old code in pthread also could course problems.  I'm not a c or c++ programmer so I maybe can't fix all of that. It's probably a cut and paste problem that left so much faulty code there.  

I was only trying to fix faulty math against graphic cards and compile it. If it don't make harm I leave it in there for now.

  
I could not find any later gpu miner than 4.1 to work on. So if I'm wrong about that then point me in the right direction. This one I'm looking at. https://github.com/gapcoin/GapMiner

I said that I probably will fail.   All that faulty code don't make it easier.

Latest Gpu miner is rev4.1.
Where you get "Error AMD accelerated parallel processing found" ?
If you trying for NVIDIA you must use --platform

[Dies diem docet]

I have looked a little more at the gpu miner code.

It seams to be a mess.

I looked at the following code and could not understand why the miner prints out "Error AMD accelerated parallel processing found". I saw that he did not care if it was a AMD or Nvidia but how could it drop down to the last else and get the error.


if (strcmp(platformId, "amd") == 0)
    platformName = "AMD Accelerated Parallel Processing";
  else if (strcmp(platformId, "nvidia") == 0)
    platformName = "NVIDIA CUDA";
  else {
    pthread_mutex_lock(&io_mutex);                            
    cout << get_time() << "ERROR: platform " << platformId << " not supported ";
    cout << " use amd or nvidia" << endl;
    pthread_mutex_unlock(&io_mutex);                      
    exit(EXIT_FAILURE);

Also he don't use the exit somewere else. He don't check for the cards before this code executes. So this piece of code is not useful for the program.

Then I started to read more. There are over 100 lines of code that are repeating itself. Checking the cards again and again. Also the fact that some old code in pthread also could course problems.  I'm not a c or c++ programmer so I maybe can't fix all of that. It's probably a cut and paste problem that left so much faulty code there.  

I was only trying to fix faulty math against graphic cards and compile it. If it don't make harm I leave it in there for now.

  
I could not find any later gpu miner than 4.1 to work on. So if I'm wrong about that then point me in the right direction. This one I'm looking at. https://github.com/gapcoin/GapMiner

I said that I probably will fail.   All that faulty code don't make it easier.

Latest Gpu miner is rev4.1.
Where you get "Error AMD accelerated parallel processing found" ?
If you trying for NVIDIA you must use --platform

The code is faulty. there are 100 lines of coide that should be reducet to around 20. It picks up that text in that mess but it seems not to have any impact at all. .

[benxy031]

Yes is a mess, its copy of xpmminer code.

[wizz13150]

Rysen 9 3900x , 12 cores 24 threads
I would expect around 3M PPS using the built-in miner, with this cpu. (will see)
Performance using CRT with high shifts should be good.

Okay, not bad. Running with a random 250w power supply, it does the job.
The motherboard is fine and allows to tune, but I propably won't.

Buit-in miner :
3.3M PPS & ~1M tests/sec  @67°c



CRT @shift 1024 :
gapminer-cpu.exe -o gap.suprnova.cc -p 2433 -u user.worker -x pass --stratum --shift 1024 --crt pdazzl-s1024.txt --threads 24 --fermat-threads 23 --sieve-primes 90000

Pool :


Solo :

[UsernameNumber7]

Nice Job Whizz, I don't ever test Shift 25............3 Million for 12 core is nice!  But Hash rate does not guarantee a winning block.  


16 Core AMD 3950X mines Shift 162 around 1 Milltion PPS.  Using 20 GB Ram.

I have lots of Prime Gap Records using Yitang Zhang mining algo using high RAM.


Prime Obsession owns the overall record!  Looks like Jonn9 might be late listing the new records from 2020 year

https://gapcoin.org/primegaps-length.php



Most all the records other than shift 25 are either mine or Prime Obsession's!  I don't think the CRT miner produces any good records!

[wizz13150]

 
It's Wizz.
It's almost guaranteed (at some point) when used in the right way.
But it's true that Gapcoin needs A LOT more computing power, and it's not an individual effort.

--shift 160 --crt crt/crt-22m-160s.txt --threads 12 --fermat-threads 11 --sieve-primes 5000  
This above gives me "16500 gaps/s 4M tests/s 0.09%", with my i9-9900kf with only -t 12, using CRT. (Shifts 128-256 are the minimum shifts to use CRT, for me)
 
How is it possible to have only 1M PPS ?! Without CRT ? Using standard miner ?
Sounds like a Joke, if Yes. Kind of inefficient/at the limit, I mean...you can improve your performances with this cpu. We could help IF YOU EVER JOIN DISCORD !
First Performance drop @shift 64 (I think there is another one around shift 128, not sure)
Anyway, PPS isn't a good value to look at, regarding the miner or the settings. (I can get more than 200M PPS, at some point, using CRT and the pool...?!)
To compare properly :
Tests/s is the universal/real performance value, as I saw, for every version. (even gpu)
Gaps/s is good to look at, with rev5.1, with or/and without using CRT. (Different ouput between rev5.0 and rev5.1, without CRT)

"I have lots of Prime Gap Records using Yitang Zhang mining algo using high RAM."
 
Not sure what that means(and what RAM has to do with gapminers...I asked myself and tested it but I don't get it), but do you know that we see it all, right?
(in live, even some of the stale/orphan/bad blocks)


Here is what we can see for the shift 160-162   :
7648,0,'C','?','P','Gapcoin',2021,26.4307,126,'4651426099...
7884,0,'C','?','P','Gapcoin',2021,27.2315,126,'5444089368...
7918,0,'C','?','P','Gapcoin',2021,27.3661,126,'4539078678...
7928,0,'C','?','P','Gapcoin',2021,27.3733,126,'6065679372...
7940,0,'C','?','P','Gapcoin',2021,27.4324,126,'5031042489...
8066,0,'C','?','P','Gapcoin',2021,27.8441,126,'6432676602...
8376,0,'C','?','P','Gapcoin',2021,28.9516,126,'4426672039...
9738,0,'C','?','P','Gapcoin',2021,33.6448,126,'5016592138... gg for this one.
0 for 2020

"Prime Obsession owns the overall record!"
 
Hum maybe, but I doubt that PrimeObsession is Andrey Balyakin. If yes okay, good to know.
I know Prime Obsession's Gapcoin main address.

https://gapcoin.org/primegaps-length.php
 
is deprecated and shouldn't be used anymore. But the PrimeGap List :
https://raw.githubusercontent.com/primegap-list-project/prime-gap-list/master/allgaps.sql

"Most all the records other than shift 25 are either mine or Prime Obsession's!"
 
Nop Sorry. At least, not anymore. M. Loizides r*ped us  
https://primegaps.cloudygo.com/graphs?max=32000

"I don't think the CRT miner produces any good records!"
 
Shares, Blocks, Records. It does.
FYI
Shift 768 without CRT :


Shift 928 using CRT :


Seems pretty obvious.


Getting back to the records:
We'll keep getting r*ped. (Yes I'm abusing a bit, it's true.)

As M. Higgins said :
"They're oriented on finding new records, Gapcoin merely uses prime gap search as a Proof-of-Work. The size of the record known first occurrence prime gap list has expanded dramatically since 2014."

And it's true. We are just raising the minimum threshold (minimum numerical limits) for records. This is not the most motivating aspect of Gapcoin, for me.
It's like cleaning up after the others have had fun at the party.  

Gapcoin NeEd PoWeR !! WaY, wAaYy MoRe pOwEr !!
1000M PPS or so... not 30 or 48 ... (nor 100 or 300, btw) to find/improve records, now.
(300M PPS would correspond to a difficulty of ~24.5, "only".)
Blocks with merit of 27 or 28 should be much more frequent. I'm hungry, I want more.


(PS:
Settings for gpu miners should appear soon. -w and -n seems to be useful.
A little 'surprise'(I hope lol) could also(I hope too lol) happen soon. Shh)

[wizz13150]

Whiz, if you tried raping me I would snap your neck in a heart beat!

But your just a pussy talking shit.........I will wait till Jonn9 posts your records next year!

Stop google translate. Dumbass. No one r*ped no one, that was a joke.
I tried to be constructive, to help you. To help Gapcoin.
We are all making an effort, wasting time with you. You must take a step towards us.
I leave it here for others.



The long lines are the blocks that I produce (for the pool) now @shift 1024.
As you can see, the rate (compared to the whole network) is good for 2 cpus using CRT.
I would be happy to have a "challenger", it would be fun. But we need to cooperate.
So we'll see records in a while. As you say. Let's wait together.

I renew my offer to reward the improved records in higher shifts. (Except Username7, unless he apologizes.)

Edit:

Looks like Jonn9 might be late

I just noticed that.  
What a sense of observation.

Two fresh ones for me.
24334,C??,27.442008,Gapcoin,27-04-2021,386,128071..
23290,C??,26.280766,Gapcoin,28-04-2021,385,743900..
And a fresh one for you, nice.
8418,C??,29.083531,Gapcoin,28-04-2021,126,5048105.. Block 1419552

[PrimeObsession]

Hey Wizz,

Thank you for your efforts and for posting your CRT@Ryzen stats! Please notice that tests/sec metric reported by the original CRT miner is plain wrong, so please add "n_test = 0;" after "cur_tests = (cur_tests + 3 * n_test) / 4;" in ChineseSieve.cpp and recompile.

Please also be advised that both CRT-miner and GPU-miner have been provided by j0nn9 as a proof of concept and were not intended for production runs! CRT-sieve may need to be rewritten, otherwise running the original CRT-miner makes little practical sense. In the meanwhile, it may turn out to be more efficient to concentrate on improving CRT offsets for higher shifts. CRT offset generator is also a proof of concept to some extent, but it may produce useful results nevertheless.

... It's definitely possible, with the appropriate skills.
Be aware that only one person claims to have run an nvidia miner. But I don't believe this person.

[2014-11-12 17:34:40] pps: 0 / 0.0000 10g/h 0.0000 / 0.0000  15g/h 0.0000 / 0.0000
We found our gap...

I have reasons to believe without doubt that Supercomputing's claims are perfectly legit. It is extremely unfortunate that other commitments prevented him from contributing to the GapCoin project openly.

[wizz13150]

Hey,

thanks for your intervention.
Indeed, tests/sec increment instead of displaying correctly, but we can get used to it with some patience.
The CRT miner is far from its full potential, but J0nn9 couldn't improve it further, from what I read.
But anyway, we can now largely confirm that it's "efficient", with the right parameters.

Thanks for the tip, I'll adjust that.

If "Supercomputing's claims are perfectly legit", then I hate him. What a waste..  
Anyway.

I know this is a bad comparison, but seeing what other researchers manage to produce by focusing on record improvements.
I see (potentially) bigger things for Gapcoin.
By being on the right point in space-time, and if a few stars can align, who knows.


I'm hopeful for Gapcoin in 2021.
Who knows what can happen.

[BitcoinFX]

AWOL, but glad to see good progress and that everyone is getting along so swimmingly!  


@gjhiggins - Can we all now safely assume (as per https://gapcoin-project.github.io/), that;

1. This repository: https://github.com/gapcoin-project/gapcoin-core.git is now the main Gapcoin Core v0.16.3 source / release?

2. This repository: https://github.com/gjhiggins/gapcoin-core is very much a 'work in progress'.

3. This repository: https://github.com/gapcoin-project/gapcoin is for v0.9.2-3 and that the experimental v0.9.4 source should generally be avoided.

4. The v0.16.3 source / binary releases would generally now be advisable over running v0.9.2-3 for anyone starting afresh with Gapcoin / Core / Solo  / Pool Mining ?


Cheers!

P.S. I'd also like to know the odds on @UsernameNumber7 being an alt. or crony? of @Vlad2Vald ? (which would explain a lot ...)  

[ir.hn]

The problem with releasing a coin without an optimized miner is the person who develops and optimizes a GPU or FPGA miner will not release it publicly lol.  They will just eat everyone else's lunch.  This is what is currently happening with RandomX, but good luck nonetheless.

[wizz13150]

Okay.

How to say that.... ping @Everyone


HE DID IT !
The current GPUminer now works with NVIDIA cards. All you have to do is to replace the current fermat.cl file by the 'new' one.
This new file still seems to work with AMD cards.

We can properly thank benxy031 for his solution to make the current miner work, with the NVIDIA cards.


How To :
You just have to use this fermat.cl instead of the default one. That's all !!
This is not an Update. This is a Fix.
(in the gpuminer's "gpu" directory, no need to (re)compile)
(Temporary link, before integrating it properly into the project and releases.)


Obviously, that doesn't change the problem of a new miner or an evolution of the current miner.
But after so long without a working NVIDIA miner, it's still Awesome !


Enjoy, Plug in whatever you have  , and don't forget benxy031 !


I'll run a GTX660 5 CUs (1M PPS), GTX970 with 13 CUs (1.6M PPS), 750ti with 5 CUs (700K PPS).
I've tested a Quadro 410 512mb, it gives me 200K PPS for 1 CUs    That's the low limit.

Good Luck & Have Fun
2021 smells good.

PS: Just saw this :
6186,C??,31.594989,Gapcoin,28-04-2021,86,10733691.. gg !

[Dies diem docet]

I just found a nice gap with a good merit with cpu mining..

{
   "hash": "b0ceadb1e6b9db2da0e7546f11d647391636a419882702ec300a8f63181cd920",
   "confirmations": 16,
   "size": 199,
   "height": 1420016,
   "version": 2,
   "merkleroot": "db615857c809222c68c8e73edfd290d9b45b21f5b852d886e3e9528a75d086ee",
   "tx": [
      "db615857c809222c68c8e73edfd290d9b45b21f5b852d886e3e9528a75d086ee"
   ],
   "time": 1619647683,
   "nonce": 953,
   "difficulty": 22.39325715,
   "shift": 27,
   "adder": "7713753",
   "gapstart": "1073369110017342124068557053141144742394912274462591624383511656087314960589041 4195673",
   "gapend": "1073369110017342124068557053141144742394912274462591624383511656087314960589041 4201859",
   "gaplen": 6186,
   "merit": 31.59498997,
   "chainwork": "6966836342016086",
   "previousblockhash": "983a28118b9e809037cea45f2c154b83bd940e9bb79fe273ab32d84f22ed223c",
   "nextblockhash": "933fc934589df828b80c281f3c518f8bf08699dd51f60df16cbb63dd1ef8b831"
}

[wizz13150]

Indeed, I just saw this and shared it on the Discord !!  

Your record (shift 27) is :
6186,C??,31.594989,Gapcoin,28-04-2021,86,10733691..
The current record (shift 64) is :
6186,0,'C','?','P','Loizides',2021,27.9727,97,'20062693329181*211#/30-3044'

Congrats !
Any settings recommendations? or just default?

If you post an address (or PM me), I would be happy to send you something !

Don't forget to try the NVIDIA miner

[gjhiggins]

Indeed, I just saw this and shared it on the Discord !!  

Now recorded on the prime gap list, details on the community web site: https://gapcoin-project.github.io/primegaplist/2021-04-29-0-updates/

Cheers

Graham

[minerja]

Okay.

How to say that.... ping @Everyone


HE DID IT !
The current GPUminer now works with NVIDIA cards. All you have to do is to replace the current fermat.cl file by the 'new' one.
This new file still seems to work with AMD cards.

We can properly thank benxy031 for his solution to make the current miner work, with the NVIDIA cards.


How To :
You just have to use this fermat.cl instead of the default one. That's all !!
This is not an Update. This is a Fix.
(in the gpuminer's "gpu" directory, no need to (re)compile)
(Temporary link, before integrating it properly into the project and releases.)


Obviously, that doesn't change the problem of a new miner or an evolution of the current miner.
But after so long without a working NVIDIA miner, it's still Awesome !


Enjoy, Plug in whatever you have  , and don't forget benxy031 !


I'll run a GTX660 ?(not plugged yet?), GTX970 with 13 CUs (1.6M PPS), 750ti with 5 CUs (700K PPS).
I've tested a Quadro 410 512mb, it gives me 200K PPS for 1 CUs    That's the low limit.

Good Luck & Have Fun
2021 smells good.

PS: Just saw this :
6186,C??,31.594989,Gapcoin,28-04-2021,86,10733691.. gg !

Can you please link to the miner you are using, cos the one i use, i get these errors

[2021-04-29 11:44:42] Found platform[0] name = NVIDIA CUDA
[2021-04-29 11:44:42] Found 1 device(s)
[2021-04-29 11:44:46] Compiling ...
[2021-04-29 11:44:46] Source: 642216 bytes
[2021-04-29 11:46:05] <kernel>:4329:1: warning: type specifier missing, defaults to 'int'
sqrProductScan320(uint32_t *out, uint32_t *op)
^~~~~~~~~~~~~~~~~
<kernel>:4670:1: warning: control reaches end of non-void function
}
^
<kernel>:4671:1: warning: type specifier missing, defaults to 'int'
sqrProductScan352(uint32_t *out, uint32_t *op)
^~~~~~~~~~~~~~~~~
<kernel>:5068:1: warning: control reaches end of non-void function
}
^
<kernel>:5815:1: error: expected identifier or '('
<!DOCTYPE html>
^
<kernel>:5981:27: error: unknown type name 'up'
                Sign&nbsp;up
                          ^
<kernel>:5982:15: error: expected identifier or '('
              </a>
              ^
<kernel>:5988:27: error: unknown type name 'up'
                Sign&nbsp;up
                          ^
<kernel>:5989:15: error: expected identifier or '('
              </a>
              ^
<kernel>:6015:275: error: expected identifier or '('
                    <a href="/features" class="py-2 lh-condensed-ultra d-block Link--primary no-underline h5 Bump-lin
hover" data-ga-click="(Logged out) Header, go to Features">Features <span class="Bump-link-symbol float-right text-no
l color-text-tertiary pr-3">&rarr;</span></a>


                                  ^
<kernel>:6017:251: error: expected identifier or '('
                        <li class="edge-item-fix"><a href="/mobile" class="py-2 lh-condensed-ultra d-block Link--seco
ry no-underline f5 Bump-link--hover">Mobile <span class="Bump-link-symbol float-right text-normal color-text-tertiary
-3">&rarr;</span></a></li>

[wizz13150]

I don't like this. The problem with adding to much mining power without proper testing is that false primes can be accepted.Then network can run into problems. Instable.

Hey,

not sure what you mean by 'false primes'.

The only change in the fermat.cl is :
-  uint4 q[2] = {0, 0}; this is for amd
+ uint4 q[2] = {0, 0, 0, 0}; this nvidia

This is the exact same miner as before.
And how much is "too much" ? It seems to be kind of subjective.
As I tried to say, 1000M PPS would be 'only' a difficulty of ~24.5 .. Gapcoin was designed to reach 35 or 40.

I just doing this for puting my name on a finding and not for the coins. I know there is better ways to find the records.

And the Gapcoin community thanks you !   (at least me, ahah)

Edit:

I was just using it to make a working miner for the rest of you to use. No need now for that.

10 733691 100173 421240 685570 531411 447423 949122 744625 916243 835116 560873 149605 890414 201859 (86 digits) is prime
10 733691 100173 421240 685570 531411 447423 949122 744625 916243 835116 560873 149605 890414 195673 (86 digits) is prime

6186 = 2 × 3 × 1031
Number of divisors: 8

Sum of divisors: 12384

Euler's totient: 2060

Awesome ! A gamechanger still can happen, who knows.

Can you please link to the miner you are using, cos the one i use, i get cpp:420 opencl errors

Hum, curious. Your issue is when compiling. We could already compile without errors, usually.
Tested on linux, windows and for several cards.
We are talking about the rev4 or rev4.1 of the gpuminer.
https://gapcoin-project.github.io/downloads/
Not the oldest releases. It should works just by replacing the file.
Here is my windows miner (x64), rezipped. (temporary)
What start line and card are you using ?

[Dies diem docet]

I don't like this. The problem with adding to much mining power without proper testing is that false primes can be accepted.Then network can run into problems. Instable.

Hey,

not sure what you mean by 'false primes'.

The only change in the fermat.cl is :
-  uint4 q[2] = {0, 0}; this is for amd
+ uint4 q[2] = {0, 0, 0, 0}; this nvidia

This is the exact same miner as before.
And how much is "too much" ? It seems to be kind of subjective.
As I tried to say, 1000M PPS would be 'only' a difficulty of ~24.5 .. Gapcoin was designed to reach 35 or 40.

I just doing this for puting my name on a finding and not for the coins. I know there is better ways to find the records.

And the Gapcoin community thanks you !   (at least me, ahah)

Can you please link to the miner you are using, cos the one i use, i get cpp:420 opencl errors

Hum, curious. Tested on linux, windows and for several cards.
We are talking about the rev4 or rev4.1 of the gpuminer.
https://gapcoin-project.github.io/downloads/
Not the oldest releases. It should works just by replacing the file.
Here is my miner, rezipped. (temporary)
What start line are you using ?

It was not an attack on you. My thinking is that it could spam the network if it's untested.

It don't work on my old laptop for some reasons. I have to check it out later. I can't update the blockchain on that one at the moment. Warning about some miners have problems.