diff options
| -rw-r--r-- | CONTRIBUTING.md (renamed from CONTIBUTING.md) | 0 | ||||
| -rw-r--r-- | doc/compile.md | 24 | ||||
| -rw-r--r-- | doc/compile_Windows.md | 2 | ||||
| -rw-r--r-- | xmrstak/backend/amd/autoAdjust.hpp | 9 |
4 files changed, 20 insertions, 15 deletions
diff --git a/CONTIBUTING.md b/CONTRIBUTING.md index 1f0bd1b..1f0bd1b 100644 --- a/CONTIBUTING.md +++ b/CONTRIBUTING.md diff --git a/doc/compile.md b/doc/compile.md index 46d39cb..8ad410d 100644 --- a/doc/compile.md +++ b/doc/compile.md @@ -18,13 +18,13 @@ By default the miner will be build with all dependencies. Each optional dependen There are two easy ways to set variables for `cmake` to configure *xmr-stak* - use the ncurses GUI - - `ccmake .` + - `ccmake ..` - edit your options - end the GUI by pressing the key `c`(create) and than `g`(generate) - set Options on the command line - - enable a option: `cmake . -DNAME_OF_THE_OPTION=ON` - - disable a option `cmake . -DNAME_OF_THE_OPTION=OFF` - - set a value `cmake . -DNAME_OF_THE_OPTION=value` + - enable a option: `cmake .. -DNAME_OF_THE_OPTION=ON` + - disable a option `cmake .. -DNAME_OF_THE_OPTION=OFF` + - set a value `cmake .. -DNAME_OF_THE_OPTION=value` After the configuration you need to compile the miner, follow the guide for your platform: * [Compile in Windows](compile_Windows.md) @@ -33,17 +33,17 @@ After the configuration you need to compile the miner, follow the guide for your ## Generic Build Options - `CMAKE_INSTALL_PREFIX` install miner to the home folder - - `cmake . -DCMAKE_INSTALL_PREFIX=$HOME/xmr-stak-cpu` + - `cmake .. -DCMAKE_INSTALL_PREFIX=$HOME/xmr-stak-cpu` - you can find the binary and the `config.txt` file after `make install` in `$HOME/xmr-stak-cpu/bin` - `CMAKE_LINK_STATIC` link libgcc and libstdc++ libraries static (default OFF) - - disable with `cmake . -DCMAKE_LINK_STATIC=ON` + - disable with `cmake .. -DCMAKE_LINK_STATIC=ON` - `CMAKE_BUILD_TYPE` set the build type - valid options: `Release` or `Debug` - you should always keep `Release` for your productive miners - `MICROHTTPD_ENABLE` allow to disable/enable the dependency *microhttpd* - - there is no *http* interface available if option is disabled: `cmake . -DMICROHTTPD_ENABLE=OFF` + - there is no *http* interface available if option is disabled: `cmake .. -DMICROHTTPD_ENABLE=OFF` - `OpenSSL_ENABLE` allow to disable/enable the dependency *OpenSSL* - - it is not possible to connect to a *https* secured pool if option is disabled: `cmake . -DOpenSSL_ENABLE=OFF` + - it is not possible to connect to a *https* secured pool if option is disabled: `cmake .. -DOpenSSL_ENABLE=OFF` ## CPU Build Options @@ -61,7 +61,7 @@ After the configuration you need to compile the miner, follow the guide for your - `CUDA_ENABLE` allow to disable/enable the NVIDIA backend of the miner - `CUDA_ARCH` build for a certain compute architecture - this option needs a semicolon separated list - - `cmake . -DCUDA_ARCH=61` or `cmake . -DCUDA_ARCH=20;61` + - `cmake .. -DCUDA_ARCH=61` or `cmake .. -DCUDA_ARCH=20;61` - [list](https://developer.nvidia.com/cuda-gpus) with NVIDIA compute architectures - by default the miner is created for all currently available compute architectures - `CUDA_COMPILER` select the compiler for the device code @@ -70,12 +70,12 @@ After the configuration you need to compile the miner, follow the guide for your # compile host and device code with clang export CC=/usr/bin/clang export CXX=/usr/bin/clang++ - cmake . -DCUDA_COMPILER=clang + cmake .. -DCUDA_COMPILER=clang ``` - `XMR-STAK_LARGEGRID` use `32` or `64` bit integer for on device indices - default is enabled - - on old GPUs it can increase the hash rate if disabled: `cmake . -DXMR-STAK_LARGEGRID=OFF` + - on old GPUs it can increase the hash rate if disabled: `cmake .. -DXMR-STAK_LARGEGRID=OFF` - if disabled it is not allowed to use more than `1000` threads on the device - `XMR-STAK_THREADS` give the compiler information which value for `threads` is used at runtime - default is `0` (compile time optimization) - - if the miner is compiled and used at runtime with the some value it can increase the hash rate: `cmake . -DXMR-STAK_THREADS=32` + - if the miner is compiled and used at runtime with the some value it can increase the hash rate: `cmake .. -DXMR-STAK_THREADS=32` diff --git a/doc/compile_Windows.md b/doc/compile_Windows.md index dec1e75..ffe9e23 100644 --- a/doc/compile_Windows.md +++ b/doc/compile_Windows.md @@ -58,7 +58,7 @@ ## Compile -- download and unzip `xmr-stak +- download and unzip `xmr-stak` - open the command line terminal `cmd` - `cd` to your unzipped source code directory - execute the following commands (NOTE: path to VS2017 can be different) diff --git a/xmrstak/backend/amd/autoAdjust.hpp b/xmrstak/backend/amd/autoAdjust.hpp index 6157b25..ebe59df 100644 --- a/xmrstak/backend/amd/autoAdjust.hpp +++ b/xmrstak/backend/amd/autoAdjust.hpp @@ -79,6 +79,8 @@ private: configEditor configTpl{}; configTpl.set( std::string(tpl) ); + constexpr size_t byteToMiB = 1024u * 1024u; + std::string conf; int i = 0; for(auto& ctx : devVec) @@ -89,8 +91,11 @@ private: size_t perThread = (size_t(1u)<<21) + 224u; size_t max_intensity = availableMem / perThread; // 1000 is a magic selected limit \todo select max intensity depending of the gpu type - size_t intensity = std::min( size_t(1000u) , max_intensity ); - conf += std::string(" // gpu: ") + ctx.name + "\n"; + size_t possibleIntensity = std::min( size_t(1000u) , max_intensity ); + // map intensity to a multiple of the compute unit count, 8 is the number of threads per work group + size_t intensity = (possibleIntensity / (8 * ctx.computeUnits)) * ctx.computeUnits * 8; + conf += std::string(" // gpu: ") + ctx.name + " memory:" + std::to_string(availableMem / byteToMiB) + "\n"; + conf += std::string(" // compute units: ") + std::to_string(ctx.computeUnits) + "\n"; // set 8 threads per block (this is a good value for the most gpus) conf += std::string(" { \"index\" : ") + std::to_string(ctx.deviceIdx) + ",\n" + " \"intensity\" : " + std::to_string(intensity) + ", \"worksize\" : " + std::to_string(8) + ",\n" + |
