CCminer M7 (XCN) by djm34, fixed + optimized for cuda 8 and new cards by PALLAS

[sp_]

If you do manage to make a version 50% faster, that's definitely worth paying for.

The m7 contains kernals that can be optimized 50%, but they use little time so the real effect will only be a few percent. If I fix the windows build it will be without any optimalizations. I will just add a pull request to pallas with a fix.

Keccak512 in the m7m:

https://github.com/pallas1/ccminer-m7-branch/blob/40445a3aecfbbf1c0122b221867c6400b2dcb771/x13/m7_keccak512.cu

Keccak512 should be rewritten to uint2 like I have done in my fork: (+40%)

https://github.com/sp-hash/ccminer/blob/windows/quark/cuda_quark_keccak512.cu

NVIDIA doesn't have a bitselect instruction, but the compiler likes this structure and generate faster code. The first iteration of the loop should be unrolled because many instructions will be removed. (working on constant data)

[1714093038]

1714093038

[1714093038]

1714093038

[1714093038]

1714093038

[pallas]

If you do manage to make a version 50% faster, that's definitely worth paying for.

The m7 contains kernals that can be optimized 50%, but they use little time so the real effect will only be a few percent. If I fix the windows build it will be without any optimalizations. I will just add a pull request to pallas with a fix.

Thanks Sp_!
Please release the windows binary as well.

[pallas]

Pushed new commit: "Improvements and cleanup for ripemd and tiger, better default intensity, fixed high cpu usage"

If you have a card with less than 2GB, or if it hangs in any way, try a lower intensity.
About cpu usage: it was using 100% before, now it's very low, power consumption reduced significantly.
Hashrate decreased a little bit as well, but overall it's much better.
If you prefer the old 100% cpu behavior, just comment line 242 of x13/m7.cu:

// cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);

[Amph]

i can install vs2015 and try to compile, but i don't remeber now all the things needed, aside from cuda 8 and vs2015 what else is needed? all the dependencies i mean

[dwarf0]

This is what I got while building process using vs2013 and cuda6.5:

Error 1 error MSB3721: The command ""C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\bin\nvcc.exe" -gencode=arch=compute_52,code=\"sm_52,compute_52\" -gencode=arch=compute_50,code=\"sm_50,compute_50\" -gencode=arch=compute_35,code=\"sm_35,compute_35\" --use-local-env --cl-version 2013 -ccbin "C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\bin"  -I"..\mpir-2.6.0\build.vc10\Win32\Release" -I. -Icompat -Icompat\jansson -Icompat\getopt -I"..\pthreads\Pre-built.2\include" -I"..\curl-7.29.0\include" -I"..\OpenSSL-Win32\include" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\include" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\include"    --keep --keep-dir .\Release\PrepFile -maxrregcount=80 --ptxas-options=-v --machine 32 --compile -cudart static     -DWIN32 -DNDEBUG -D_CONSOLE -D_CRT_SECURE_NO_WARNINGS -DCURL_STATICLIB -DSCRYPT_KECCAK512 -DSCRYPT_CHACHA -DSCRYPT_CHOOSE_COMPILETIME -D_MBCS -Xcompiler "/EHsc /W3 /nologo /O2 /Zi  /MT  " -o Release\cuda_blake256.cu.obj "C:\eth\ccminer-m7-branch-master\Algo256\cuda_blake256.cu"" exited with code 1

any guess?

[Nikolaj]

I can do it tonight. correct and make the windows build. for free.

Thank you very much

[sp_]

This is what I got while building process using vs2013 and cuda6.5:

Error 1 error MSB3721: The command ""C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\bin\nvcc.exe" -gencode=arch=compute_52,code=\"sm_52,compute_52\" -gencode=arch=compute_50,code=\"sm_50,compute_50\" -gencode=arch=compute_35,code=\"sm_35,compute_35\" --use-local-env --cl-version 2013 -ccbin "C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\bin"  -I"..\mpir-2.6.0\build.vc10\Win32\Release" -I. -Icompat -Icompat\jansson -Icompat\getopt -I"..\pthreads\Pre-built.2\include" -I"..\curl-7.29.0\include" -I"..\OpenSSL-Win32\include" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\include" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\include"    --keep --keep-dir .\Release\PrepFile -maxrregcount=80 --ptxas-options=-v --machine 32 --compile -cudart static     -DWIN32 -DNDEBUG -D_CONSOLE -D_CRT_SECURE_NO_WARNINGS -DCURL_STATICLIB -DSCRYPT_KECCAK512 -DSCRYPT_CHACHA -DSCRYPT_CHOOSE_COMPILETIME -D_MBCS -Xcompiler "/EHsc /W3 /nologo /O2 /Zi  /MT  " -o Release\cuda_blake256.cu.obj "C:\eth\ccminer-m7-branch-master\Algo256\cuda_blake256.cu"" exited with code 1

any guess?

Use visual studio 2010?

[rednoW]

i can install vs2015 and try to compile, but i don't remeber now all the things needed, aside from cuda 8 and vs2015 what else is needed? all the dependencies i mean

to compile with cuda 7.5 or older you need vs2013. vs2015 isn't supported. Don't know for  cuda 8.

[Amph]

i can install vs2015 and try to compile, but i don't remeber now all the things needed, aside from cuda 8 and vs2015 what else is needed? all the dependencies i mean

to compile with cuda 7.5 or older you need vs2013. vs2015 isn't supported. Don't know for  cuda 8.

i remember i need something else besides those two things, but maybe i'm wrong, last time i compiled something was in 2014...

some google search lead me to, vs2015 being supported only up to version 1 and with cuda 8 only, version 2 and 3 are not supported

[Elun]

i can install vs2015 and try to compile, but i don't remeber now all the things needed, aside from cuda 8 and vs2015 what else is needed? all the dependencies i mean

to compile with cuda 7.5 or older you need vs2013. vs2015 isn't supported. Don't know for  cuda 8.

i remember i need something else besides those two things, but maybe i'm wrong, last time i compiled something was in 2014...

some google search lead me to, vs2015 being supported only up to version 1 and with cuda 8 only, version 2 and 3 are not supported

mpir, jansson, pthreads, curl, openssl. For this fork. in sp's fork everything in repo already.

[sp_]

Required dependencies for building on Windows:
pthreads: http://sourceware.org/pthreads-win32/
OpenSSL-Win32: http://slproweb.com/download/Win32OpenSSL-1_0_1e.exe
curl-7.29.0: http://curl.haxx.se/download/curl-7.29.0.zip
or in precompiled form for Visual Studio 2010 SP1 cudaminer_vc2010_prerequisites.7z (49.3 MB) https://mega.co.nz/#!cQQziBrb!HB1fEEcVM7HQuE3HI1ZcV00GCOA6kV4rwmAWkBaMI84

So just add extract these files, and add includepath and libpath into the c++ project file.

You also need to copy the dll's into the release folder to be able to start the miner.

[Nikolaj]

Required dependencies for building on Windows:
pthreads: http://sourceware.org/pthreads-win32/
OpenSSL-Win32: http://slproweb.com/download/Win32OpenSSL-1_0_1e.exe
curl-7.29.0: http://curl.haxx.se/download/curl-7.29.0.zip
or in precompiled form for Visual Studio 2010 SP1 cudaminer_vc2010_prerequisites.7z (49.3 MB) https://mega.co.nz/#!cQQziBrb!HB1fEEcVM7HQuE3HI1ZcV00GCOA6kV4rwmAWkBaMI84

So just add extract these files, and add includepath and libpath into the c++ project file.

You also need to copy the dll's into the release folder to be able to start the miner.

Thank you

[Dr_Victor]

I can do it tonight. correct and make the windows build. for free.

Wish you no bugs!

[sp_]

I have a build ready...

https://github.com/sp-hash/ccminer/releases/tag/1.2

Compiled with cuda 6.5. Unmodified sourcecode from Pallas github exept lower the intensity to 256*256*32 to fit 2GB cards.

It does around 5MHASH on the 750ti

[pallas]

I have a build ready...

https://github.com/sp-hash/ccminer/releases/tag/1.2

Compiled with cuda 6.5. Unmodified sourcecode from Pallas github exept lower the intensity to 256*256*32 to fit 2GB cards.

It does around 5MHASH on the 750ti

Thanks!
Now that everyone can use the miner, please donate to myself and sp_.
That would encourage me to continue optimising.

[maxim000]

I have a build ready...

https://github.com/sp-hash/ccminer/releases/tag/1.2

Compiled with cuda 6.5. Unmodified sourcecode from Pallas github exept lower the intensity to 256*256*32 to fit 2GB cards.

It does around 5MHASH on the 750ti

unfortunatly 1070 not working correctly - only "booo" for suprnova and Stratum authentication failed for 1GH.com

p.s.
also same for 970

p.p.s  and for 750Ti too

[reb0rn21]

I have a build ready...

https://github.com/sp-hash/ccminer/releases/tag/1.2

Compiled with cuda 6.5. Unmodified sourcecode from Pallas github exept lower the intensity to 256*256*32 to fit 2GB cards.

It does around 5MHASH on the 750ti

unfortunatly 1070 not working correctly - only "booo" for suprnova and stratum error for 1GH.com

+1
 WARNING !!! You seem to be sending a large amount of INVALID shares

1070 same on suprnova

[klondike_bar_recovery]

I have a build ready...

https://github.com/sp-hash/ccminer/releases/tag/1.2

Compiled with cuda 6.5. Unmodified sourcecode from Pallas github exept lower the intensity to 256*256*32 to fit 2GB cards.

It does around 5MHASH on the 750ti

unfortunatly 1070 not working correctly - only "booo" for suprnova and stratum error for 1GH.com

+1
 WARNING !!! You seem to be sending a large amount of INVALID shares

1070 same on suprnova

+1. just an obscene amount of "BOO", something like 90% invalids

using "-f 0.03 helps to slow down the process, but has little or no effect on the invalid rate

[pallas]

I have a build ready...

https://github.com/sp-hash/ccminer/releases/tag/1.2

Compiled with cuda 6.5. Unmodified sourcecode from Pallas github exept lower the intensity to 256*256*32 to fit 2GB cards.

It does around 5MHASH on the 750ti

unfortunatly 1070 not working correctly - only "booo" for suprnova and stratum error for 1GH.com

+1
 WARNING !!! You seem to be sending a large amount of INVALID shares

1070 same on suprnova

+1. just an obscene amount of "BOO", something like 90% invalids

now for whatever reaon i just get "stratum authentication failed". dont know what i did to cause that change?

You got banned for too many invalid shares.
Sp_ please help with this issue.

[sp_]

I noticed that cpu verifcation was disabled in the sourcecode. I will take a look