[LOCKED] cpuminer-opt v3.12.3, open source optimized multi-algo CPU miner

[joblo]

No CPU mining here but I like your style,

0cc24f5a79ad5200fc546bd534523d2cb25da73420830d59eed7c9cf4732e99b

Cheers!

WOW, my first donation. Many thanks. I'll have to make sure I don't post any sarcastic replies
to any of your mining questions, CPU or otherwise.

[1685551903]

1685551903

[1685551903]

1685551903

[1685551903]

1685551903

[1685551903]

1685551903

[1685551903]

1685551903

[th3.r00t]

cpuminer-opt v3.1.10 is available for download.

https://drive.google.com/file/d/0B0lVSGQYLJIZckxmTS1aRWQ5OEU/view?usp=sharing

This release adds support for the HOdl algo for CPUs with and without AES_NI. See the first post
in this thread for details.

Is it me ot the new algo is missing from help?

Code:

~/cpuminer-opt-3.1.10/cpuminer --help

         **********  cpuminer-opt 3.1.10  ***********
     A CPU miner with multi algo support and optimized for CPUs
     with AES_NI extension.
     BTC donation address: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT
     Forked from TPruvot's cpuminer-multi with credits
     to Lucas Jones, elmad, palmd, djm34, pooler, ig0tik3d,
     Wolf0 and Jeff Garzik.

Checking CPU capatibility...
        Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz
CPU Supports AES_NI: YES.
SW  Supports AES_NI: YES.
Start mining with AES_NI optimisations...

Usage: cpuminer-opt [OPTIONS]
Options:
  -a, --algo=ALGO       specify the algorithm to use
                          scrypt       scrypt(1024, 1, 1) (default)
                          scrypt:N     scrypt(N, 1, 1)
                          sha256d      SHA-256d
                          pentablake   Pentablake
                          quark        Quark
                          qubit        Qubit
                          shavite3     Shavite3
                          x11gost      sib (SibCoin)
                          skein        Skein+Sha (Skeincoin)
                          skein2       Double Skein (Woodcoin)
                          s3           S3
                          vanilla      blake256r8vnl
                          x11          X11
                          x13          X13
                          x14          X14
                          x15          X15
                          x17
                          yescrypt
                          zr5          ZR5
  -o, --url=URL         URL of mining server

[ryen123]

cpuminer-opt v3.1.10 is available for download.

https://drive.google.com/file/d/0B0lVSGQYLJIZckxmTS1aRWQ5OEU/view?usp=sharing

This release adds support for the HOdl algo for CPUs with and without AES_NI. See the first post
in this thread for details.

Is it me ot the new algo is missing from help?

It works for me, hodl algo is there and works. Maybe you compiled without libboost. See the OP about requirement.

Code:

         **********  cpuminer-opt 3.1.10  ***********
     A CPU miner with multi algo support and optimized for CPUs
     with AES_NI extension.
     BTC donation address: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT
     Forked from TPruvot's cpuminer-multi with credits
     to Lucas Jones, elmad, palmd, djm34, pooler, ig0tik3d,
     Wolf0 and Jeff Garzik.

Checking CPU capatibility...
               Intel(R) Core(TM) i5-2500K CPU @ 3.30GHz
CPU Supports AES_NI: YES.
SW  Supports AES_NI: YES.
Start mining with AES_NI optimisations...

Usage: cpuminer-opt [OPTIONS]
Options:
  -a, --algo=ALGO       specify the algorithm to use
                          scrypt       scrypt(1024, 1, 1) (default)
                          scrypt:N     scrypt(N, 1, 1)
                          sha256d      SHA-256d
                          axiom        Shabal-256 MemoHash
                          blake        Blake-256 (SFR)
                          blakecoin    blake256r8
                          blake2s      Blake-2 S
                          bmw          BMW 256
                          c11          flax
                          cryptolight  Cryptonight-light
                          cryptonight  Monero
                          decred
                          drop         Dropcoin
                          fresh        Fresh
                          groestl      groestl
                          hodl         hodlcoin
                          heavy        Heavy
                          keccak       Keccak
                          luffa        Luffa
                          lyra2re      lyra2
                          lyra2rev2    lyrav2
                          myr-gr       Myriad-Groestl
                          neoscrypt    NeoScrypt(128, 2, 1)
                          nist5        Nist5
                          pluck        Pluck:128 (Supcoin)
                          pentablake   Pentablake
                          quark        Quark
                          qubit        Qubit
                          shavite3     Shavite3
                          x11gost      sib (SibCoin)
                          skein        Skein+Sha (Skeincoin)
                          skein2       Double Skein (Woodcoin)
                          s3           S3
                          vanilla      blake256r8vnl
                          x11          X11
                          x13          X13
                          x14          X14
                          x15          X15
                          x17
                          yescrypt
                          zr5          ZR5

[th3.r00t]

I have the dependencies installed:
 libboost-dev
 libboost-system-dev
 libboost-thread-dev

Can you she your ./configure ?

[ryen123]

./configure CFLAGS="-Ofast -march=corei7-avx" CXXFLAGS="-Ofast -march=corei7-avx" --with-crypto --with-curl

[th3.r00t]

./configure CFLAGS="-Ofast -march=corei7-avx" CXXFLAGS="-Ofast -march=corei7-avx" --with-crypto --with-curl

Thanks.
Nope, that does'nt work for me.
Still no HOdl in list.

Maybe I should include somehow required boost libs?

[ryen123]

Just to point out, my pc is running ubuntu 14.04 and I install the entire libboost lib (libboost1.55-all-dev). Not individually.

[th3.r00t]

Same here - Ubuntu 14.04 Server.
Installed libboost1.55-all-dev and still no luck.

[ryen123]

Ok. Just to be clear these are my own steps.

1) Download cpuminer-opt-3.1.10.tar.gz and extract
2) ./autogen.sh
3) ./configure CFLAGS="-Ofast -march=native" CXXFLAGS="-Ofast -march=native" --with-crypto --with-curl
4) make -j4
5) strip cpuminer

Works for me.

[th3.r00t]

Well...
After reboot it compiled OK.

Code:

  -a, --algo=ALGO       specify the algorithm to use
                          scrypt       scrypt(1024, 1, 1) (default)
                          scrypt:N     scrypt(N, 1, 1)
                          sha256d      SHA-256d
                          axiom        Shabal-256 MemoHash
                          blake        Blake-256 (SFR)
                          blakecoin    blake256r8
                          blake2s      Blake-2 S
                          bmw          BMW 256
                          c11          flax
                          cryptolight  Cryptonight-light
                          cryptonight  Monero
                          decred
                          drop         Dropcoin
                          fresh        Fresh
                          groestl      groestl
                          hodl         hodlcoin
                          heavy        Heavy
                          keccak       Keccak
                          luffa        Luffa
                          lyra2re      lyra2
                          lyra2rev2    lyrav2
                          myr-gr       Myriad-Groestl
                          neoscrypt    NeoScrypt(128, 2, 1)
                          nist5        Nist5
                          pluck        Pluck:128 (Supcoin)
                          pentablake   Pentablake
                          quark        Quark
                          qubit        Qubit
                          shavite3     Shavite3
                          x11gost      sib (SibCoin)
                          skein        Skein+Sha (Skeincoin)
                          skein2       Double Skein (Woodcoin)
                          s3           S3
                          vanilla      blake256r8vnl
                          x11          X11
                          x13          X13
                          x14          X14
                          x15          X15
                          x17
                          yescrypt
                          zr5          ZR5

[Epsylon3]

the build.sh script is made for that, it does a "make clean" before

tested, and i see a good hashrate on the miner side, but seems to be only 150H/s on pools side (both suprnova and blockquary)

[th3.r00t]

the build.sh script is made for that, it does a "make clean" before

tested, and i see a good hashrate on the miner side, but seems to be only 150H/s on pools side (both suprnova and blockquary)

Yeah... 
But recently we was discussing build.sh

I haven't looked at build.sh since the fork, I should clean it up.

It got me errors on AMD SSE2 compile and slower binary on Intel i7.

Epsylon3, thanks for cpuminer-multi, which is the main base for this great cpuminer!

[joblo]

the build.sh script is made for that, it does a "make clean" before

tested, and i see a good hashrate on the miner side, but seems to be only 150H/s on pools side (both suprnova and blockquary)

Yes that is a problem. I don't understand, the hashrate calculation is very simple and the hashes_done calculation
is all self contained in the algo source file. scanhash_hodl is counting the collisions and the miner thread measures
the time and it's a simple division. I don't see where it could go wrong.

It happens with both AES_NI and NON_AES_NI implementations of hodl. It doesn't
happen with any other algo.

If the hashes_done calculation is wrong then it is in code was copied as-is from the original hodlminer.
If the time is wrong it is in code that is the same for all algos.

The only other thing I can think of is phantom accepts or silent rejects. Either the miner is dsplaying shares that
were never submitted or not reporting shares that were rejected.

It happens with both AES_NI and NON_AES_NI implementations of hodl. It doesn't happen with any other algo.

The pool reported hashrate appears to be inline with both precedessors, although it's hard to be sure with the flunctuations.
I've seen it as high as 400 H/s with both Wolf and opt/wolf on an i7-4790K and i7-6700K repsectively.

I will do some instrumenting to gather data to see if I can figure out what's going on.


FYI, I haven't maintained build.sh, so YMMV

[joblo]

Epsylon3, thanks for cpuminer-multi, which is the main base for this great cpuminer!

Hear! Hear!

He has made my job much easier.

[joblo]

Here's some data i colllected for hashrates

Code:

[2016-04-07 14:22:25] 1 miner threads started, using 'hodl' algorithm.
[2016-04-07 14:22:28] Stratum difficulty set to 1
[2016-04-07 14:22:38] scanh start hashes done= 0
[2016-04-07 14:22:44] scanh ret 1 hashes done= 328
[2016-04-07 14:22:44] diff.tv_sec= 6 usec= 16424
[2016-04-07 14:22:44] thr rate= 54.517434
[2016-04-07 14:22:44] CPU #0: 54.52 H/s
[2016-04-07 14:22:44] accepted: 1/1 (100%), 54.52 H/s yes!

Everything adds up. This is exactly what it's getting back from the scan. The problem must be happening later
with the submission.

Edit: I think I have found a long standing bug.

When a share is submitted the miner reports the sum of the hashrates of the last scan of each thread. That's is
usually a good approximation but doesn't take into account failed scans. It also doesn't take into account
super-scans where a solution is found by 2 threads at the same time.

A share is submitted only when the scan returns success. Otherwise the scan was wasted time. This wasted time
doesn't get measured by the miner but the pool sure notices the delay since the last submit. That appears to
account for discrepency between the miner rate and the pool rate.

Fixing that will take some thought. The displayed share submission rate should be the sum of all the scans since
the last submission. This will cause volatility in the displayed hashrates but they will be accurate and should match
the pool.

The per-thread hashrate could also be improved. The display could indicate when a scan has suceeded or failed.

That problem seems understood but now the issue is why so many scans fail. I'm going to try to fix that before
redesigning the hashrate display.

[joblo]

I think I need to redeesign the hashrate display in order to troubleshoot the scan failures, first to quantify
and compare with other miners.

I've added a quick efficiency factor by simply using the inverse of the number of fails between successes.
100% means every thread is successful on every scan. The efficiency is also already indicated by the number
of thread hashrate lines between shares.

Here's what it looks like.

Code:

[2016-04-07 17:12:20] CPU #7: 32.76 H/s
[2016-04-07 17:12:20] scan efficiency 12.50%
[2016-04-07 17:12:20] accepted: 38/38 (100%), 415.59 H/s yes!
[2016-04-07 17:12:21] CPU #0: 64.15 H/s
[2016-04-07 17:12:21] scan efficiency 100.00%
[2016-04-07 17:12:21] accepted: 39/39 (100%), 427.02 H/s yes!
[2016-04-07 17:12:22] CPU #3: 65.58 H/s
[2016-04-07 17:12:22] CPU #4: 53.58 H/s
[2016-04-07 17:12:22] CPU #2: 49.79 H/s
[2016-04-07 17:12:22] CPU #6: 53.61 H/s
[2016-04-07 17:12:22] CPU #5: 54.44 H/s
[2016-04-07 17:12:22] CPU #1: 50.27 H/s
[2016-04-07 17:12:25] CPU #2: 67.54 H/s
[2016-04-07 17:12:25] scan efficiency 14.29%
[2016-04-07 17:12:25] accepted: 40/40 (100%), 441.93 H/s yes!
[2016-04-07 17:12:27] Stratum difficulty set to 4.96669
[2016-04-07 17:12:27] CPU #6: 58.60 H/s
[2016-04-07 17:12:27] CPU #5: 58.98 H/s
[2016-04-07 17:12:27] CPU #4: 53.22 H/s
[2016-04-07 17:12:27] CPU #1: 45.70 H/s

[joblo]

Something is bugging me about hashrates. First I will point out that the current method the miner
reports hashrate generally agrees with the pools.

This bugs me for the simple reason that one thread finds the solution but the reported hashrate
is the sum of all threads. The hashrate is only correct when a single thread finds a solution each scan.
If all 8 threads find a solution my pool rate would be 800% what the miner reports. This explains somewhat
why the pool sometimes reports hashrates much higher than the miner.

Now to the pools. They don't know how many threads are running, they just get a solution and the number of
hashes to find it (an assumption). They look at the size of the solution and the time it took to find it and
calculate a hashrate. How in hell could the miner ever agree? I must be missing something.

I've been running for about 120 accepts calculating an average efficiency and it's running at about
7%.

That's 7% of 800% = 56%.

Miner reports 450 H/s, multiply by 56% = 252 H/s.

That should be my actual hashrate at the pool.

I think I need to let that sink in for a while. I'm getting dizzy.

[joblo]

My theory was wrong, at least with respect to hodl. With hodl the threads work cooperatively
by spliting the scratchpad into sections with each thread searching their own section. First thread
to find it scores a collision.

I had a bug that was getting new work in every thread.

That seems to have solved the most important part of the problem, the hashrate reported at the pool.
Unfortunately it's not the same high hashrate still being reported by the miner.

I will likely build another release with this fix and continue to investigate the local hashrate display issue.

[joblo]

        Release 3.1.11 of cpuminer-opt is available for download.

https://drive.google.com/file/d/0B0lVSGQYLJIZZ3VCOERfMHZiQmM/view?usp=sharing

New in 3.1.11

   - hodl algo hashrate issue partially fixed.

     Errata: The hash rate reported at the pool should now
             be the same as hodlminer-wolf although the miner
             stilll reports an erroneously high rate. The
             performance chart uses the pool reported rate.

   - argon2 algo added, 6% faster than encel version

[citronick]

Hi All,

I recently inherited an AMD FD8350FRHKBOX FX-8350 FX-Series 8-Core Black Edition Processor.

My mining experience are GPU and ASICs, and a forum member recommended I visit this thread to get help and ideas.

Any ideas how can I use this CPU to mine the most profitable CPU only alt-coins?

Cheers!