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

[joblo]

I'm using precompiled windows binaries with a Ryzen 7 1700, trying to mine Verium.

My bat file starts with cpuminer-aes-avx2 -a scrypt:1048576 -o stratum+tcp://eu2.poolinat0r.com:7103

However, it only uses SSE 2 as far as I can see.



What am I doing wrong?

Where can I download precompiled binaries optimized for ryzen?

Thanks.

because the algo supports only sse2

Oh, I see. Is this the correct algo, then?

I'm asking because my old Xeon E5645 with v3.4.6. gives me almost double H/s per core than my Ryzen 1700, so I guess something is wrong.

Yes, thats the correct algo from what i remember.
And the speed might depend on instruction implementation in the processor, hyperthreading, core speed and cache size.
I don't know how much the specific algo benefit from each of those, but it might be logical

On some algos i get higher hash rate on core i7-6700@3.4GHz probably due to HT and larger cache, while some are faster on core i5-7600@4.5Ghz due to higher clock

The Windows binaries are pretty bad on AMD, not so great on Intel either. You'll need Linux and compile your own to get the best
performance on any platform.

As far as scrypt algo goes, I haven't paid much attention to it due to the existance of scrypt ASICs. However the verium
version uses a different N factor that can't be mined with the scrypt ASIC.
 
The implementation in cpuminer-opt was taken from Pooler and I haven't (intentionally) changed it. You could try his version
to see if it gives better performance. I think cpuminer-multi also uses the same version. Their Windows binaries might be better.

[1714030589]

1714030589

[1714030589]

1714030589

[1714030589]

1714030589

[nizzuu]

Hi, there's no link to legacy binaries now (link points to regular binaries). Is it useless now? I mean non-AES-NI cpus...

[lncm]

I'm using precompiled windows binaries with a Ryzen 7 1700, trying to mine Verium.

My bat file starts with cpuminer-aes-avx2 -a scrypt:1048576 -o stratum+tcp://eu2.poolinat0r.com:7103

However, it only uses SSE 2 as far as I can see.



What am I doing wrong?

Where can I download precompiled binaries optimized for ryzen?

Thanks.

because the algo supports only sse2

Oh, I see. Is this the correct algo, then?

I'm asking because my old Xeon E5645 with v3.4.6. gives me almost double H/s per core than my Ryzen 1700, so I guess something is wrong.

Yes, thats the correct algo from what i remember.
And the speed might depend on instruction implementation in the processor, hyperthreading, core speed and cache size.
I don't know how much the specific algo benefit from each of those, but it might be logical

On some algos i get higher hash rate on core i7-6700@3.4GHz probably due to HT and larger cache, while some are faster on core i5-7600@4.5Ghz due to higher clock

The Windows binaries are pretty bad on AMD, not so great on Intel either. You'll need Linux and compile your own to get the best
performance on any platform.

As far as scrypt algo goes, I haven't paid much attention to it due to the existance of scrypt ASICs. However the verium
version uses a different N factor that can't be mined with the scrypt ASIC.
 
The implementation in cpuminer-opt was taken from Pooler and I haven't (intentionally) changed it. You could try his version
to see if it gives better performance. I think cpuminer-multi also uses the same version. Their Windows binaries might be better.

Thanks. I could do that, install Linux in a secondary partition.

Any recommended distro and compile instructions?


PS: And for older CPU's like my 6-core Xeon E5645, will I see a difference in hashrate using Linux?

[joblo]

Hi, there's no link to legacy binaries now (link points to regular binaries). Is it useless now? I mean non-AES-NI cpus...

Aha!, Now I know how I screwed up the 3.7.2 links upon its release. Fixed the legacy links now.

Your post was timely, I'm rethinking the usefullnes of the legacy release.

Technically it only affects algo that use the Groestl function on CPUs that don't have AES_NI. These include
algos like X11 which are either dominated by ASICs or GPUs. Spefically the legacy version continues to use
the Groestl macro implementation which may be slightly faster than the SPH version in the current release.

I have been considering dropping the legacy release because:
   - lack of AES_NI is the real killer for Groestl
   - the macro version probably isn't really much faster than SPH
   - it only affects algos where CPU mining is obsolete
   - anyone who actually uses it already has it
   - it's getting dusty (no new algos) and I have no  intentions of ever updating it
   - some users are probably using it who shouldn't

[joblo]

The Windows binaries are pretty bad on AMD, not so great on Intel either. You'll need Linux and compile your own to get the best
performance on any platform.

As far as scrypt algo goes, I haven't paid much attention to it due to the existance of scrypt ASICs. However the verium
version uses a different N factor that can't be mined with the scrypt ASIC.
 
The implementation in cpuminer-opt was taken from Pooler and I haven't (intentionally) changed it. You could try his version
to see if it gives better performance. I think cpuminer-multi also uses the same version. Their Windows binaries might be better.

Thanks. I could do that, install Linux in a secondary partition.

Any recommended distro and compile instructions?

PS: And for older CPU's like my 6-core Xeon E5645, will I see a difference in hashrate using Linux?

Compile instructions are in RELEASE_NOTES. Ubuntu is favoured by many.
I can't really coment on performance as I said I haven't paid much attention to this
algo. But a native compile with a newer compiler can't hurt.

Please post your results as I am curious also.

[bl0atw4re]

Hello joblo, How to use install-sh file on linux to install cpuminer to system ? im using arch linux.Thanks in advance.

[ronnylov]

Hi!

Older AMD CPU than Piledriver is not supported including Phenom II.
But what about Bulldozer?

I have a server with four Opteron 6276 on Linux.
Can I use this miner go mine Hodl (or ROI coin) with my Opterons?
Bulldozer and Piledriver is not very different architecture.

Edit: I am using Arch Linux

[lncm]

The Windows binaries are pretty bad on AMD, not so great on Intel either. You'll need Linux and compile your own to get the best
performance on any platform.

As far as scrypt algo goes, I haven't paid much attention to it due to the existance of scrypt ASICs. However the verium
version uses a different N factor that can't be mined with the scrypt ASIC.
 
The implementation in cpuminer-opt was taken from Pooler and I haven't (intentionally) changed it. You could try his version
to see if it gives better performance. I think cpuminer-multi also uses the same version. Their Windows binaries might be better.

Thanks. I could do that, install Linux in a secondary partition.

Any recommended distro and compile instructions?

PS: And for older CPU's like my 6-core Xeon E5645, will I see a difference in hashrate using Linux?

Compile instructions are in RELEASE_NOTES. Ubuntu is favoured by many.
I can't really coment on performance as I said I haven't paid much attention to this
algo. But a native compile with a newer compiler can't hurt.

Please post your results as I am curious also.

I just did. It's consistently 14% faster in Linux (Xeon E5645), with compiled binary. That was with Mint 18.2. But I'm going to install Ubuntu 16.4.3 because I need blockchain drivers to mine ethereum, and I couldn't figure out how to install amdgpu pro drivers on Mint 18.2.

[My9bot]

pls add POLYTIMOS algo
https://bitcointalk.org/index.php?topic=2409413.0

[joblo]

Hello joblo, How to use install-sh file on linux to install cpuminer to system ? im using arch linux.Thanks in advance.

I've never used install-sh so I can't help you. You could probably google it for help if you really want to use it, but you
don't ned to install to run.

[joblo]

Hi!

Older AMD CPU than Piledriver is not supported including Phenom II.
But what about Bulldozer?

I have a server with four Opteron 6276 on Linux.
Can I use this miner go mine Hodl (or ROI coin) with my Opterons?
Bulldozer and Piledriver is not very different architecture.

Edit: I am using Arch Linux

I don't know about specific CPUs, you need to try for yourself.
Hodl is a special case. If you're CPU doesn't have AES_NI you should try
the legacy version of cpuminer-opt. The non-AES code for hodl has been removed
from the master release.
 

[joblo]

pls add POLYTIMOS algo
https://bitcointalk.org/index.php?topic=2409413.0

I'll put it on my todo list for next release but it may take a while since I'm in the middle of some other heavy
development.

[CryptoGoddess]

Gosh, I'm so lost with all this. I think I need a dummy's guide, I have no technical skills whatsoever.

[Bannedseller]

Gosh, I'm so lost with all this. I think I need a dummy's guide, I have no technical skills whatsoever.

Me too. a quick manual will surely help

[joblo]

Gosh, I'm so lost with all this. I think I need a dummy's guide, I have no technical skills whatsoever.

Me too. a quick manual will surely help

A manual for what? cpuminer-opt works like any other fork of minerd, cpuminer or ccminer.
The only difference is you need to know your CPU archietcture and its capabilities. Use google for that.

[comchay]

how to config multi cpu thead. I have 8 cpu core but i want miner 4 core.

[Elder III]

how to config multi cpu thead. I have 8 cpu core but i want miner 4 core.

All you have to do is add "-t 4" at the end of your .bat file and it will use 4 threads. If you leave it blank it will use all your CPU has. You can specify anything from 1 to 8 threads this way.

*If you have an 8 core CPU with hyperthreading (or whatever AMD calls it these days) then you actually have 16 threads (AMD Ryzen 7 series for example).

[joblo]

A little teaser of what I'm working on. It's only the beginning.

Code:

cpuminer-opt/3.7.3	d=4	keccak	4		27 MH/s

Code:

CPU: Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
CPU features: SSE2 AES AVX AVX2
SW built on Nov 20 2017 with GCC 4.8.4
SW features: SSE2 AES AVX AVX2
Algo features: SSE2 AVX2
Start mining with SSE2 AVX2

[2017-11-20 11:58:16] Starting Stratum on stratum+tcp://keccak.mine.zpool.ca:5133
[2017-11-20 11:58:16] 8 miner threads started, using 'keccak' algorithm.
[2017-11-20 11:58:16] Stratum difficulty set to 4
 1 nonces submitted
[2017-11-20 11:58:31] Accepted 1/1 (100%), 24.21 MH, 23.45 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:34] Accepted 2/2 (100%), 52.09 MH, 23.61 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:39] Accepted 3/3 (100%), 235.98 MH, 24.21 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:39] Accepted 4/4 (100%), 200.89 MH, 24.21 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:41] Accepted 5/5 (100%), 199.70 MH, 24.19 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:42] Accepted 6/6 (100%), 202.41 MH, 24.19 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:48] Accepted 7/7 (100%), 45.96 MH, 24.20 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:48] Accepted 8/8 (100%), 43.01 MH, 24.21 MH/s, 63C

Code:

CPU: Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
CPU features: SSE2 AES AVX AVX2
SW built on Nov  4 2017 with GCC 4.8.4
SW features: SSE2 AES AVX AVX2
Algo features: SSE2
Start mining with SSE2

[2017-11-20 11:56:35] Starting Stratum on stratum+tcp://keccak.mine.zpool.ca:5133
[2017-11-20 11:56:35] 8 miner threads started, using 'keccak' algorithm.
[2017-11-20 11:56:35] Stratum difficulty set to 4
[2017-11-20 11:56:54] Accepted 1/1 (100%), 26.41 MH, 7735.68 kH/s, 62C
[2017-11-20 11:57:02] Accepted 2/2 (100%), 44.40 MH, 7736.24 kH/s, 62C
[2017-11-20 11:57:29] Accepted 3/3 (100%), 199.67 MH, 7745.76 kH/s, 62C
[2017-11-20 11:57:36] Accepted 4/4 (100%), 194.64 MH, 7737.53 kH/s, 62C
[2017-11-20 11:58:00] Accepted 5/5 (100%), 220.11 MH, 7741.79 kH/s, 63C
[2017-11-20 11:58:02] Accepted 6/6 (100%), 222.54 MH, 7745.23 kH/s, 63C

[4ward]

A little teaser of what I'm working on. It's only the beginning.

Code:

cpuminer-opt/3.7.3	d=4	keccak	4		27 MH/s

Code:

CPU: Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
CPU features: SSE2 AES AVX AVX2
SW built on Nov 20 2017 with GCC 4.8.4
SW features: SSE2 AES AVX AVX2
Algo features: SSE2 AVX2
Start mining with SSE2 AVX2

[2017-11-20 11:58:16] Starting Stratum on stratum+tcp://keccak.mine.zpool.ca:5133
[2017-11-20 11:58:16] 8 miner threads started, using 'keccak' algorithm.
[2017-11-20 11:58:16] Stratum difficulty set to 4
 1 nonces submitted
[2017-11-20 11:58:31] Accepted 1/1 (100%), 24.21 MH, 23.45 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:34] Accepted 2/2 (100%), 52.09 MH, 23.61 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:39] Accepted 3/3 (100%), 235.98 MH, 24.21 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:39] Accepted 4/4 (100%), 200.89 MH, 24.21 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:41] Accepted 5/5 (100%), 199.70 MH, 24.19 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:42] Accepted 6/6 (100%), 202.41 MH, 24.19 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:48] Accepted 7/7 (100%), 45.96 MH, 24.20 MH/s, 63C
 1 nonces submitted
[2017-11-20 11:58:48] Accepted 8/8 (100%), 43.01 MH, 24.21 MH/s, 63C

Code:

CPU: Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz
CPU features: SSE2 AES AVX AVX2
SW built on Nov  4 2017 with GCC 4.8.4
SW features: SSE2 AES AVX AVX2
Algo features: SSE2
Start mining with SSE2

[2017-11-20 11:56:35] Starting Stratum on stratum+tcp://keccak.mine.zpool.ca:5133
[2017-11-20 11:56:35] 8 miner threads started, using 'keccak' algorithm.
[2017-11-20 11:56:35] Stratum difficulty set to 4
[2017-11-20 11:56:54] Accepted 1/1 (100%), 26.41 MH, 7735.68 kH/s, 62C
[2017-11-20 11:57:02] Accepted 2/2 (100%), 44.40 MH, 7736.24 kH/s, 62C
[2017-11-20 11:57:29] Accepted 3/3 (100%), 199.67 MH, 7745.76 kH/s, 62C
[2017-11-20 11:57:36] Accepted 4/4 (100%), 194.64 MH, 7737.53 kH/s, 62C
[2017-11-20 11:58:00] Accepted 5/5 (100%), 220.11 MH, 7741.79 kH/s, 63C
[2017-11-20 11:58:02] Accepted 6/6 (100%), 222.54 MH, 7745.23 kH/s, 63C

thats a VERY nice improvement!
Can't wait...

By the way, please consider releasing cross-compiled binaries for windows, they give much better performance that mingw on windows.
I have described the process earlier in the post, but sadly I have could not make neoscrypt work in that build (just crashes every time), probably something in the compile flags or algo settings.
If something is missing in what I posted and you want full instructions for the cross build, just say )

[filoozom]

A pull request has been received to change the dockerfile.

There are 2 changes, one to address an issue with line continuation, but also a change that claims to
reduce the image size.

Since I know very little about docker I defer to the community for review of the change. If there are any
concerns with the change, please raise them. If there are no objections the changes will be included in
the next release.

https://github.com/JayDDee/cpuminer-opt/pull/41

The Docker image could be smaller by using Alpine Linux as the base. Alpine uses libressl by default but also have openssl-1.0.2 in its repo so I'm not sure about any impact on performance.

Yes, it could, but didn't quite manage to compile it on alpine yet. If you have more info, please share!

I just tried compiling but got the error

Code:

In file included from algo/echo/aes_ni/hash.c:21:0:
algo/echo/aes_ni/vperm.h:53:7: note: called from here
  x  = _mm_shuffle_epi8(*((__m128i*)table + 0), x);\
       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
algo/echo/aes_ni/hash.c:385:4: note: in expansion of macro 'TRANSFORM'
    TRANSFORM(_state[i][j], _k_opt, t1, t2);
    ^~~~~~~~~
In file included from algo/echo/aes_ni/vperm.h:20:0,
                 from algo/echo/aes_ni/hash.c:21:
/usr/lib/gcc/x86_64-alpine-linux-musl/6.4.0/include/tmmintrin.h:136:1: error: inlining failed in call to always_inline '_mm_shuffle_epi8': target specific option mismatch
 _mm_shuffle_epi8 (__m128i __X, __m128i __Y)
 ^~~~~~~~~~~~~~~~
In file included from algo/echo/aes_ni/hash.c:21:0:
algo/echo/aes_ni/vperm.h:52:5: note: called from here
  t1 = _mm_shuffle_epi8(*((__m128i*)table + 1), t1);\

algo/echo/aes_ni/hash.c:385:4: note: in expansion of macro 'TRANSFORM'
    TRANSFORM(_state[i][j], _k_opt, t1, t2);
    ^~~~~~~~~
source='algo/hodl/aes.c' object='algo/hodl/cpuminer-aes.o' libtool=no \
DEPDIR=.deps depmode=none /bin/sh ./depcomp \
gcc -DHAVE_CONFIG_H -I.  -Iyes/include -Iyes/include -fno-strict-aliasing  -I. -Iyes/include -Iyes/include -Wno-pointer-sign -Wno-pointer-to-int-cast   -g -O2  -Iyes/include -Iyes/include -c -o algo/hodl/cpuminer-aes.o `test -f 'algo/hodl/aes.c' || echo './'`algo/hodl/aes.c
make[2]: *** [Makefile:2491: algo/echo/aes_ni/cpuminer-hash.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[2]: Leaving directory '/cpuminer-opt'
make[1]: *** [Makefile:3506: all-recursive] Error 1
make[1]: Leaving directory '/cpuminer-opt'
make: *** [Makefile:665: all] Error 2

Got it to compile, you can try filoozom/cpuminer-opt:alpine.
It's quite a bit smaller but also a little bit slower. Also needs a fix in `algo/hodl/sha512_avx.c`.

More info: https://github.com/JayDDee/cpuminer-opt/pull/41#issuecomment-345828950
Help would be appreciated!