Pages:
Author

Topic: [BBR] Boolberry GPU Miner Discussion - AMD & Nvidia - OpenCL & CUDA - page 22. (Read 82987 times)

newbie
Activity: 26
Merit: 0
Hi,

I am having similar problems to Antonio with the miner. My environment is following:
-Ubuntu 14.04
-Biostar HB61 + 5 x 7950
- AMD drivers 14.6 beta
- AMD APP SDK installed version 2.9
- miner compiled with make -j=5

once I start mining in simple wallet using start_mining 5 I am getting following:

2014-Jul-26 23:02:23.210488 [RPC0]!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2014-Jul-26 23:02:23.217932 Miner thread was started
2014-Jul-26 23:02:23.218006 [miner 0][OCL 0/0] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:02:23.326086 Miner thread was started [1]
2014-Jul-26 23:02:23.326417 [miner 1][OCL 0/1] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:02:23.326614 [miner 1][OCL 0/1] error:  CL_INVALID_VALUE
2014-Jul-26 23:02:23.326769 [miner 1][OCL 0/1] platform index must me between 0 and 4294967295
2014-Jul-26 23:02:23.384927 [miner 0][OCL 0/0] error:  CL_INVALID_VALUE
2014-Jul-26 23:02:23.385091 [miner 0]ERROR /home/klemen/boolberry-opencl/src/currency_core/OCL_Device.cpp:222[_cl_program* OCL_Device::GetProgram(const char*)][OCL 0/0] build log: �
2014-Jul-26 23:02:23.385186 [miner 0][OCL 0/0] error:  CL_INVALID_PROGRAM
2014-Jul-26 23:02:23.385276 [miner 0][OCL 0/0] error:  CL_INVALID_PROGRAM
2014-Jul-26 23:02:23.385376 [miner 0][OCL 0/0] using platform: AMD Accelerated Parallel Processing
2014-Jul-26 23:02:23.385477 [miner 0][OCL 0/0] using device:Tahiti
2014-Jul-26 23:02:23.385563 [miner 0]Share target (diff 16777215) 0000010000010000
2014-Jul-26 23:02:23.430528 [miner 0][OCL 0/0] error:  CL_INVALID_KERNEL
2014-Jul-26 23:02:23.430582 [miner 0][OCL 0/0] error:  CL_INVALID_KERNEL
2014-Jul-26 23:02:23.430600 [miner 0][OCL 0/0] error:  CL_INVALID_KERNEL
2014-Jul-26 23:02:23.430615 [miner 0][OCL 0/0] error:  CL_INVALID_KERNEL
2014-Jul-26 23:02:23.430631 [miner 0][OCL 0/0] error:  CL_INVALID_KERNEL
2014-Jul-26 23:02:23.430648 [miner 0][OCL 0/0] error:  CL_INVALID_KERNEL
2014-Jul-26 23:02:23.430729 [miner 0]Miner thread stopped
2014-Jul-26 23:02:23.427891 Miner thread was started [2]
2014-Jul-26 23:02:23.431012 [miner 2][OCL 0/2] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:02:23.431219 [miner 2][OCL 0/2] device index must me between 0 and 0
2014-Jul-26 23:02:23.518074 Miner thread was started [3]
2014-Jul-26 23:02:23.518167 [miner 3][OCL 0/3] start with kernel wild_keccak.cl, type 0
2014-Jul-26 23:02:23.518214 [miner 3][OCL 0/3] device index must me between 0 and 0

any thoughts?

If necessary I can also downgrade to Ubuntu 13.10 with AMD 13.12 drivers. (althought it would take me several hours)

member
Activity: 81
Merit: 1002
It was only the wind.
Are these the proper instructions on how to mine on Windows?  Is there another method?

.....


No, those instructions are to build the windows programs, something most people (including me) can't do. MBK had the files ready for download, but the link no longer works. If he is unable to put them back up, I will get zoidberg to try to build them.

If MBK needs someone to host them, I guess I will.
sr. member
Activity: 519
Merit: 250
Got a 42.6% increase on my 860M, which is basically a 750Ti. That took a while.

NSFW
Proof: https://ottrbutt.com/tmp/bbropenclstock2.png vs. https://ottrbutt.com/tmp/bbropenclwolf2.png

Hi,can u help meu with a config of your bat file

Thx in advance
sr. member
Activity: 330
Merit: 252
Looks like the OPENCL_LIBRARY is undefined. It should be in line 149. Edit FindOpenCL.cmake. Put debug messages after the find_library

thanks... I hope that brings us a step further.
I will try this tomorrow when I'm back at home.
Have a good time.
sr. member
Activity: 330
Merit: 252
Got a 42.6% increase on my 860M, which is basically a 750Ti. That took a while.

Proof: https://ottrbutt.com/tmp/bbropenclstock2.png vs. https://ottrbutt.com/tmp/bbropenclwolf2.png

wow! get it on... you do a lot of cool stuff at the moment (tsiv CN).
If your BBR version would be easy to compile for nvidia linux guys I would be more than happy.
mbk
member
Activity: 106
Merit: 10
Looks like the OPENCL_LIBRARY is undefined. It should be in line 149. Edit FindOpenCL.cmake. Put debug messages after the find_library call.
Code:
  # Find the library
  find_library(OPENCL_LIBRARY OpenCL
    PATHS /lib${_OPENCL_LIB_DIR_SUFFIX}
    ${_OPENCL_LIB_OPTS}
    ${_OPENCL_ROOT_OPTS}
  )

  message("OpenCL Library: ${OPENCL_LIBRARY}")
  message("Path: ${CMAKE_FIND_ROOT_PATH}")
  message("Library suffix: ${_OPENCL_LIB_DIR_SUFFIX}")
  message("OPENCL_LIB_OPTS: ${_OPENCL_LIB_OPTS}")
  message("OPENCL_ROOT_OPTS: ${_OPENCL_ROOT_OPTS}")
Here is the correct output:
Code:
-- OpenCL: Searching for NVIDIA CUDA SDK
-- Looking for CL_VERSION_1_0
-- Looking for CL_VERSION_1_0 - found
-- Looking for CL_VERSION_1_1
-- Looking for CL_VERSION_1_1 - found
-- Looking for CL_VERSION_1_2
-- Looking for CL_VERSION_1_2 - not found
OpenCL Library: /opt/cuda/lib64/libOpenCL.so
Path: /opt/cuda
Library suffix: 64
OPENCL_LIB_OPTS:
OPENCL_ROOT_OPTS: ONLY_CMAKE_FIND_ROOT_PATH;NO_DEFAULT_PATH
-- Found OpenCL: /opt/cuda/include (found version "1.1")

In the logs you posted before I see
Code:
-- OpenCL: Searching in custom location
It means CUDA_PATH variable is not visible to cmake. Is it still the case? Delete the build directory or better clone the repository from scratch and run make again.

No. It was the case before as you advised to set OPENCL_ROOT (a few days ago).
But the defined path doesn't make a difference... the error is the same.

With set CUDA_PATH he is searching for the nvidia cuda SDK - and her found the CL Versions as you could read in my logs...
the problem is described in this post above, please take a closer look:
https://bitcointalksearch.org/topic/m.8016537


...and the error stays the same:

Code:
-- Looking for CL_VERSION_1_0
-- Looking for CL_VERSION_1_0 - found
-- Looking for CL_VERSION_1_1
-- Looking for CL_VERSION_1_1 - found
-- Looking for CL_VERSION_1_2
-- Looking for CL_VERSION_1_2 - not found
CMake Error at /usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:108 (message):
  Could NOT find OpenCL (missing: OPENCL_LIBRARY) (found version "1.1")
Call Stack (most recent call first):
  /usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:315 (_FPHSA_FAILURE_MESSAGE)
  FindOpenCL.cmake:159 (FIND_PACKAGE_HANDLE_STANDARD_ARGS)
  CMakeLists.txt:13 (find_package)
sr. member
Activity: 330
Merit: 252
In the logs you posted before I see
Code:
-- OpenCL: Searching in custom location
It means CUDA_PATH variable is not visible to cmake. Is it still the case? Delete the build directory or better clone the repository from scratch and run make again.

No. It was the case before as you advised to set OPENCL_ROOT (a few days ago).
But the defined path doesn't make a difference... the error is the same.

With set CUDA_PATH he is searching for the nvidia cuda SDK - and he found the CL Versions as you could read in my logs...
the problem is described in this post above, please take a closer look:
https://bitcointalksearch.org/topic/m.8016537


...and the error stays the same:

Code:
-- Looking for CL_VERSION_1_0
-- Looking for CL_VERSION_1_0 - found
-- Looking for CL_VERSION_1_1
-- Looking for CL_VERSION_1_1 - found
-- Looking for CL_VERSION_1_2
-- Looking for CL_VERSION_1_2 - not found
CMake Error at /usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:108 (message):
  Could NOT find OpenCL (missing: OPENCL_LIBRARY) (found version "1.1")
Call Stack (most recent call first):
  /usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:315 (_FPHSA_FAILURE_MESSAGE)
  FindOpenCL.cmake:159 (FIND_PACKAGE_HANDLE_STANDARD_ARGS)
  CMakeLists.txt:13 (find_package)
hero member
Activity: 938
Merit: 1001
Merged all updates from the main branch.
Source - https://github.com/mbkuperman/boolberry-opencl
Windows x64 binaries  - https://github.com/mbkuperman/boolberry-opencl/releases

Fixed previous binary release too (it was compressed with 7zip instead of zip).

Thank you! I updated the first post.
mbk
member
Activity: 106
Merit: 10
Yes, you can use as many GPUs as you like with solo miner (single thread in readme is about pool miner). Try larger values for thread_delay if config file (ex. 2000). It's in milliseconds.


"thank you for your "answer", but in the readme file, i read :
"--mining-threads=[n], n - number of GPUs to use"

mbk
member
Activity: 106
Merit: 10
In the logs you posted before I see
Code:
-- OpenCL: Searching in custom location
It means CUDA_PATH variable is not visible to cmake. Is it still the case? Delete the build directory or better clone the repository from scratch and run make again.

When building on Linux you should:
1. Run 'make clean' if you got back here after some problems.
1. Set CUDA_PATH or AMDAPPSDKROOT, run 'printenv' to check.
2. Run 'make' and watch for OpenCL: Searching for NVIDIA CUDA SDK or OpenCL: Searching for AMD APP SDK. Back to 1st step if it's not in the log.
3. Watch for -- Looking for CL_VERSION_1_x - found. Check the environment variable points to the actual SDK directory if it's not in the log and get back to 1st step.
4. Check the other dependencies were found (Boost).
5. Compilation should start.

first steps clear. CUDA_PATH set. printenv checked.
3. is the problem. -- Looking for CL_VERSION_1_1 - found - is ok - but afterwards the compiler shows my beloved message:
CMake Error at /usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:108 (message):
  Could NOT find OpenCL (missing: OPENCL_LIBRARY) (found version "1.1")


with the boolberry miner version from the 22. July (you fixed backward compatibility after for openCL 1.1/1.0) he shows the right path afterwards:
-- Found OpenCL: /usr/local/cuda/include (found version "1.1")
...and failed later (https://bitcointalksearch.org/topic/m.7972653)
but with the actual version I did not get so far.

compiling cudaminer and ccminer forks were no problem so far.
mbk
member
Activity: 106
Merit: 10
Merged all updates from the main branch.
Source - https://github.com/mbkuperman/boolberry-opencl
Windows x64 binaries  - https://github.com/mbkuperman/boolberry-opencl/releases

Fixed previous binary release too (it was compressed with 7zip instead of zip).
hero member
Activity: 1426
Merit: 506
Are these the proper instructions on how to mine on Windows?  Is there another method?

.....


No, those instructions are to build the windows programs, something most people (including me) can't do. MBK had the files ready for download, but the link no longer works. If he is unable to put them back up, I will get zoidberg to try to build them.
Much appreciated!
hero member
Activity: 938
Merit: 1001
I added a brief summary to the first post of this thread. Let me know what else to add. Probably needs more Linux build info.
hero member
Activity: 938
Merit: 1001
Are these the proper instructions on how to mine on Windows?  Is there another method?

.....


No, those instructions are to build the windows programs, something most people (including me) can't do. MBK had the files ready for download, but the link no longer works. If he is unable to put them back up, I will get zoidberg to try to build them.
sr. member
Activity: 378
Merit: 250
Are these the proper instructions on how to mine on Windows?  Is there another method?
"Windows

Dependencies: MSVC 2012, CMake 2.8.6 or later, and Boost 1.55.

You may download them from:
http://www.microsoft.com/en-US/download/details.aspx?id=34673
http://www.cmake.org/files/v3.0/cmake-3.0.0-win32-x86.exe
http://sourceforge.net/projects/boost/files/boost-binaries/1.55.0-build2/boost_1_55_0-msvc-11.0-64.exe/download

OpenCL SDK (you can install CUDA too):
http://developer.amd.com/tools-and-sdks/opencl-zone/opencl-tools-sdks/amd-accelerated-parallel-processing-app-sdk/

To build, edit paths to Boost in cmake-vs2012.bat and run it.
Open the generated build/boolberry.sln in VS2012, change the solution configuration to Release and build the targets you need in the prog group.
You'll find executables in build/src/Release directory.
Copy *.cl kernels from src/cl to executable's directory.

Good luck!"
full member
Activity: 182
Merit: 100
sr. member
Activity: 330
Merit: 252
mbk...
I edited the findopenCL.cmake file to spit out some "messages" between the codelines.
Its obvious that everything works fine till the end. He found the cuda_paths & openCL, versions. yes, we already know that because of the logs.

but at the end goes something wrong:

Code:
  # Find the library
  find_library(OPENCL_LIBRARY OpenCL
    PATHS /lib${_OPENCL_LIB_DIR_SUFFIX}
    ${_OPENCL_LIB_OPTS}
    ${_OPENCL_ROOT_OPTS}
  )

  # Restore the original search paths
  set(CMAKE_FIND_ROOT_PATH ${_CMAKE_FIND_ROOT_PATH})

XXX

  include(FindPackageHandleStandardArgs)
  FIND_PACKAGE_HANDLE_STANDARD_ARGS(OpenCL
    REQUIRED_VARS OPENCL_INCLUDE_DIR OPENCL_LIBRARY
    VERSION_VAR OPENCL_VERSION_STRING
  )
  if(OPENCL_FOUND)
    set(OPENCL_INCLUDE_DIRS ${OPENCL_INCLUDE_DIR})
    set(OPENCL_LIBRARIES ${OPENCL_LIBRARY})
  endif()
endif()

the cmake funktion after "XXX" include FIND_PACKAGE_HANDLE_STANDARD_ARGS returns the error.
but why FIND_PACKAGE_HANDLE_STANDARD_ARGS could not deal with the paths found before?

is there a possibility I could print out the value of the variables while compiling (message XYZ)?
then I could check what path's he exactly found before.

hero member
Activity: 518
Merit: 500
full member
Activity: 182
Merit: 100
Hi,
i'm trying to use windows binaries, but i got this :
Quote
2014-Jul-25 09:00:14.707796 [P2P4]!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!
2014-Jul-25 09:00:14.727796 Miner thread was started
2014-Jul-25 09:00:14.737796 [miner 0][OCL 0/0] start with kernel wild_keccak.cl,
 type 0
2014-Jul-25 09:00:14.827797 Miner thread was started [1]
2014-Jul-25 09:00:14.837797 [miner 1][OCL 0/1] start with kernel wild_keccak.cl,
 type 0
2014-Jul-25 09:00:14.837797 [miner 1][OCL 0/1] error:  CL_INVALID_VALUE
2014-Jul-25 09:00:14.837797 [miner 1][OCL 0/1] platform index must me between 0
and 4294967295
2014-Jul-25 09:00:14.927797 Miner thread was started [2]
2014-Jul-25 09:00:14.937797 [miner 2][OCL 0/2] start with kernel wild_keccak.cl,
 type 0
2014-Jul-25 09:00:14.937797 [miner 2][OCL 0/2] error:  CL_INVALID_VALUE
2014-Jul-25 09:00:14.937797 [miner 2][OCL 0/2] platform index must me between 0
and 4294967295
2014-Jul-25 09:00:15.027797 Miner thread was started [3]
2014-Jul-25 09:00:15.037797 [miner 3][OCL 0/3] start with kernel wild_keccak.cl,
 type 0
2014-Jul-25 09:00:15.037797 [miner 3][OCL 0/3] error:  CL_INVALID_VALUE
2014-Jul-25 09:00:15.037797 [miner 3][OCL 0/3] platform index must me between 0
and 4294967295
2014-Jul-25 09:00:15.127797 [P2P4]Mining has started with 5 threads, good luck!
2014-Jul-25 09:00:15.127797 [P2P4][81.31.38.5:10101 OUT] COMMAND_HANDSHAKE INVOK
ED OK
2014-Jul-25 09:00:15.137797 [P2P7]Connecting to 54.90.184.240:10101(white=1, las
t_seen: d0.h0.m9.s45)...
2014-Jul-25 09:00:15.127797 Miner thread was started [4]
2014-Jul-25 09:00:15.187797 [miner 4][OCL 0/4] start with kernel wild_keccak.cl,
 type 0
2014-Jul-25 09:00:15.207797 [miner 4][OCL 0/4] error:  CL_INVALID_VALUE
2014-Jul-25 09:00:15.217797 [miner 4][OCL 0/4] platform index must me between 0
and 4294967295

So, only one card out of 5 can mine.

i run this command line : boolbd --start-mining=1JNsCgTA1pK7ksYpZPsDDV7HihrAN5JKMTNA17fwk6aQD2CknCd34A5UKaZBxft8B9Z6mKFxKuWkheccMdbda2iARS88Y6r --mining-threads=5

In the .json file, i got this :
Quote
{
  "current_extra_message_index": 0,
  "device_index": 0,
  "difficulty": 16777215,
  "donation_decision": true,
  "donation_decision_made": false,
  "donation_descision": true,
  "donation_percent_opencl": "2.0",
  "kernel": "wild_keccak.cl",
  "kernel_type": 0,
  "platform_index": 0,
  "thread_delay": 100,
  "work_size": 262144
}

What did i do wrong?
sr. member
Activity: 342
Merit: 250
ok then you can add to the test matrix 6990 1160 kh/s
6950 540 kh/s

if there is any hardware matrix i would add the results

(note  the results are with the 2 % fee added. dunno if devel makes the hardware hash slower with fees are disabled)

Specs:
OS = Ubuntu 12.04 64bit
GPU = 7850 1GB Pitcairn
driver =  14.6
Hashrate: 490 kh/s
compiled without issues on my end

is there a way to tweak memclock or engine clock on linux?
Pages:
Jump to: