Merge pull request LeelaChessZero#10 from LeelaChessZero/master

get latest

appveyor.yml

@@ -17,7 +17,8 @@ install:
 - cmd: IF %NAME%==opencl set OPENCL=true
 - cmd: IF %NAME%==blas set BLAS=true
 - cmd: IF %NAME%==blas set GTEST=true
-- cmd: IF %BLAS%==true nuget install OpenBLAS -Version 0.2.14.1 -OutputDirectory C:\cache
+- cmd: IF %BLAS%==true IF NOT EXIST C:\cache\OpenBLAS appveyor DownloadFile https://sjeng.org/ftp/OpenBLAS-0.3.3-win-oldthread.zip
+- cmd: IF %BLAS%==true IF NOT EXIST C:\cache\OpenBLAS 7z x OpenBLAS-0.3.3-win-oldthread.zip -oC:\cache\OpenBLAS
 - cmd: IF %OPENCL%==true nuget install opencl-nug -Version 0.777.12 -OutputDirectory C:\cache
 - cmd: IF %BLAS%==true IF NOT EXIST C:\cache\ispc-v1.9.2-windows appveyor DownloadFile https://sourceforge.net/projects/ispcmirror/files/v1.9.2/ispc-v1.9.2-windows.zip
 - cmd: IF %BLAS%==true IF NOT EXIST C:\cache\ispc-v1.9.2-windows 7z x ispc-v1.9.2-windows.zip -oC:\cache
@@ -52,12 +53,12 @@ cache:
   - C:\projects\lc0\subprojects\packagecache
 before_build:
 - cmd: git submodule update --init --recursive
-- cmd: meson build --backend vs2017 --buildtype release -Dgtest=%GTEST% -Dopencl=%OPENCL% -Dblas=%BLAS% -Dcudnn=%CUDA% -Dispc_native_only=false -Dpopcnt=false -Dcudnn_include="%CUDA_PATH%\include","%PKG_FOLDER%\cuda\include" -Dcudnn_libdirs="%CUDA_PATH%\lib\x64","%PKG_FOLDER%\cuda\lib\x64" -Dprotobuf_include="%PKG_FOLDER%\protobuf\include" -Dprotobuf_libdir="%PKG_FOLDER%\protobuf\lib" -Dopenblas_include="%PKG_FOLDER%\OpenBLAS.0.2.14.1\lib\native\include" -Dopenblas_libdirs="%PKG_FOLDER%\OpenBLAS.0.2.14.1\lib\native\lib\x64" -Dopencl_include="%PKG_FOLDER%\opencl-nug.0.777.12\build\native\include" -Dopencl_libdirs="%PKG_FOLDER%\opencl-nug.0.777.12\build\native\lib\x64" -Ddefault_library=static
+- cmd: meson build --backend vs2017 --buildtype release -Dgtest=%GTEST% -Dopencl=%OPENCL% -Dblas=%BLAS% -Dcudnn=%CUDA% -Dispc_native_only=false -Dpopcnt=false -Dcudnn_include="%CUDA_PATH%\include","%PKG_FOLDER%\cuda\include" -Dcudnn_libdirs="%CUDA_PATH%\lib\x64","%PKG_FOLDER%\cuda\lib\x64" -Dprotobuf_include="%PKG_FOLDER%\protobuf\include" -Dprotobuf_libdir="%PKG_FOLDER%\protobuf\lib" -Dopenblas_include="%PKG_FOLDER%\OpenBLAS\dist64\include" -Dopenblas_libdirs="%PKG_FOLDER%\OpenBLAS\dist64\lib" -Dopencl_include="%PKG_FOLDER%\opencl-nug.0.777.12\build\native\include" -Dopencl_libdirs="%PKG_FOLDER%\opencl-nug.0.777.12\build\native\lib\x64" -Ddefault_library=static
 build_script:
 - cmd: IF %APPVEYOR_REPO_TAG%==false msbuild "C:\projects\lc0\build\lc0.sln" /m /p:WholeProgramOptimization=true /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll"
 - cmd: IF %APPVEYOR_REPO_TAG%==true msbuild "C:\projects\lc0\build\lc0.sln" /m /p:WholeProgramOptimization=PGInstrument /logger:"C:\Program Files\AppVeyor\BuildAgent\Appveyor.MSBuildLogger.dll"
 - cmd: cd build
-- cmd: IF %APPVEYOR_REPO_TAG%==true IF %BLAS%==true copy C:\cache\OpenBLAS.0.2.14.1\lib\native\bin\x64\*.dll
+- cmd: IF %APPVEYOR_REPO_TAG%==true IF %BLAS%==true copy C:\cache\OpenBLAS\dist64\bin\libopenblas.dll
 - cmd: IF %APPVEYOR_REPO_TAG%==true IF %OPENCL%==true copy C:\cache\opencl-nug.0.777.12\build\native\bin\OpenCL.dll
 - cmd: IF %APPVEYOR_REPO_TAG%==true IF %CUDA%==true copy "%CUDA_PATH%"\bin\*.dll
 - cmd: IF %APPVEYOR_REPO_TAG%==true IF %CUDA%==true copy %PKG_FOLDER%\cuda\bin\cudnn64_7.dll
@@ -68,7 +69,7 @@ after_build:
 - cmd: IF %APPVEYOR_REPO_TAG%==true 7z a lc0-%APPVEYOR_REPO_TAG_NAME%-windows-%NAME%.zip %APPVEYOR_BUILD_FOLDER%\build\lc0.exe
 - cmd: IF %APPVEYOR_REPO_TAG%==true appveyor DownloadFile "https://ci.appveyor.com/api/projects/LeelaChessZero/lczero-client/artifacts/client.exe?branch=release&pr=false&job=Environment%%3A%%20NAME%%3D.exe%%2C%%20GOOS%%3Dwindows"
 - cmd: IF %APPVEYOR_REPO_TAG%==true 7z a lc0-%APPVEYOR_REPO_TAG_NAME%-windows-%NAME%.zip client.exe
-- cmd: IF %APPVEYOR_REPO_TAG%==true IF %BLAS%==true 7z a lc0-%APPVEYOR_REPO_TAG_NAME%-windows-%NAME%.zip C:\cache\OpenBLAS.0.2.14.1\lib\native\bin\x64\*.dll
+- cmd: IF %APPVEYOR_REPO_TAG%==true IF %BLAS%==true 7z a lc0-%APPVEYOR_REPO_TAG_NAME%-windows-%NAME%.zip C:\cache\OpenBLAS\dist64\bin\libopenblas.dll
 - cmd: IF %APPVEYOR_REPO_TAG%==true IF %OPENCL%==true 7z a lc0-%APPVEYOR_REPO_TAG_NAME%-windows-%NAME%.zip C:\cache\opencl-nug.0.777.12\build\native\bin\OpenCL.dll
 - cmd: IF %APPVEYOR_REPO_TAG%==true IF %CUDA%==true 7z a lc0-%APPVEYOR_REPO_TAG_NAME%-windows-%NAME%.zip "%CUDA_PATH%\bin\cudart64_100.dll" "%CUDA_PATH%\bin\cublas64_100.dll"
 - cmd: IF %APPVEYOR_REPO_TAG%==true IF %CUDA%==true 7z a lc0-%APPVEYOR_REPO_TAG_NAME%-windows-%NAME%.zip "%PKG_FOLDER%\cuda\bin\cudnn64_7.dll"
@@ -91,7 +92,7 @@ deploy:
       appveyor_repo_tag: true
 test_script:
 - cmd: cd build
-- cmd: IF %GTEST%==true copy C:\cache\OpenBLAS.0.2.14.1\lib\native\bin\x64\*.dll
+- cmd: IF %GTEST%==true copy C:\cache\OpenBLAS\dist64\bin\libopenblas.dll
 - cmd: IF %GTEST%==true xcopy /s /i C:\cache\syzygy syzygy
 - cmd: IF %GTEST%==true meson test --print-errorlogs
 - cmd: cd ..

changelog.txt

@@ -1,3 +1,72 @@
+v0.20.0-rc1 (2018-12-22)
+~~~~~~~~~~~
+
+* Squeeze-and-Excitation Networks are now supported! (lc0.org/se)
+* Older text network files are no longer supported.
+* Various performance fixes (most major being having fast approximate math
+  functions).
+* For systems with multiple GPUs, in addition to "multiplexing" backend
+  we now also have "demux" backend and "roundrobin" backend.
+* Compiler settings tweaks (use VS2017 for windows builds, always have LTO
+  enabled, windows releases have PGO enabled).
+* Benchmark mode has more options now (e.g. movetime) and saner defaults.
+* Added an option to prevent engine to resign too early (used in training).
+* Fixed a bug when number of visits could be too high in collision nodes.
+  The fix is pretty hacky, there will be better fix later.
+* 32-bit version compiles again.
+
+v0.19.1 (2018-12-10)
+~~~~~~~
+
+(no changes relative to v0.19.1-rc2)
+
+v0.19.1-rc2 (2018-12-07)
+~~~~~~~~~~~
+
+* Temperature and FPU related params. (#568)
+* Rework Cpuct related params. (#567)
+
+v0.19.1-rc1 (2018-12-06)
+~~~~~~~~~~~
+
+* Updated cpuct formula from alphazero paper. (#563)
+* remove UpdateFromUciOptions() from EnsureReady() (#558)
+* revert IsSearchActive() and better fix for one of #500 crashes (#555)
+
+v0.19.0 (2018-11-19)
+~~~~~~~
+
+* remove Wait() from EngineController::Stop() (#522)
+
+v0.19.0-rc5 (2018-11-17)
+~~~~~~~~~~~
+
+* OpenCL: replace thread_local with a resource pool. (#516)
+* optional wtime and btime (#515)
+* Make convolve1 work with workgroup size of 128 (#514)
+* adjust average depth calculation for multivisits (#510)
+
+v0.19.0-rc4 (2018-11-12)
+~~~~~~~~~~~
+
+* Microseconds have 6 digits, not 3! (#505)
+* use bestmove_is_sent_ for Search::IsSearchActive() (#502)
+
+v0.19.0-rc3 (2018-11-07)
+~~~~~~~~~~~
+
+* Fix OpenCL tuner always loading the first saved tuning (#491)
+* Do not show warning when ComputeBlocking() takes too much time. (#494)
+* Output microseconds in log rather than milliseconds. (#495)
+* Add benchmark features (#483)
+* Fix EncodePositionForNN test failure (#490)
+
+v0.19.0-rc2 (2018-11-03)
+~~~~~~~~~~~
+
+* Version v0.19.0-rc1 reported it's version as v0.19.0-dev
+  Therefore v0.19.0-rc2 is released with this issue fixed.
+
 v0.19.0-rc1 (2018-11-03)
 ~~~~~~~~~~~
 

meson.build

@@ -15,7 +15,7 @@
 # along with Leela Chess.  If not, see <http://www.gnu.org/licenses/>.
 
 project('lc0', 'cpp',
-        default_options : ['cpp_std=c++14', 'b_ndebug=if-release', 'b_lto=true'],
+        default_options : ['cpp_std=c++14', 'b_ndebug=if-release'],
         meson_version: '>=0.45')
 
 cc = meson.get_compiler('cpp')
@@ -26,7 +26,6 @@ endif
 if cc.get_id() == 'clang' or cc.get_id() == 'gcc'
   add_project_arguments('-Wextra', language : 'cpp')
   add_project_arguments('-pedantic', language : 'cpp')
-  add_project_arguments('-ffast-math', language : 'cpp')
 
   if get_option('buildtype') == 'release'
     add_project_arguments('-march=native', language : 'cpp')
@@ -51,7 +50,10 @@ else
 endif
 protoc = find_program('protoc', required : false)
 # For tensorflow skip system protobuf, chances are it will not work.
-if not protobuf_dep.found() or not protoc.found() or get_option('tensorflow')
+if get_option('protobuf-3-6-0')
+  deps += subproject('protobuf-3.6.0').get_variable('protobuf_dep')
+  protoc = subproject('protobuf-3.6.0').get_variable('protoc')
+elif not protobuf_dep.found() or not protoc.found() or get_option('tensorflow')
   deps += subproject('protobuf').get_variable('protobuf_dep')
   protoc = subproject('protobuf').get_variable('protoc')
 else
@@ -98,9 +100,11 @@ files += [
   'src/neural/factory.cc',
   'src/neural/loader.cc',
   'src/neural/network_check.cc',
+  'src/neural/network_demux.cc',
   'src/neural/network_legacy.cc',
   'src/neural/network_mux.cc',
   'src/neural/network_random.cc',
+  'src/neural/network_rr.cc',
   'src/neural/network_st_batch.cc',
   'src/neural/writer.cc',
   'src/selfplay/game.cc',
@@ -155,6 +159,7 @@ if get_option('build_backends')
       tensorflow_include,
       tensorflow_include[0] + '/bazel-genfiles',
       tensorflow_include[0] + '/tensorflow/contrib/makefile/downloads',
+      tensorflow_include[0] + '/tensorflow/contrib/makefile/downloads/absl',
       tensorflow_include[0] + '/tensorflow/contrib/makefile/downloads/eigen',
       tensorflow_include[0] + '/tensorflow/contrib/makefile/downloads/gemmlowp',
       tensorflow_include[0] + '/tensorflow/contrib/makefile/downloads/nsync/public',
@@ -179,6 +184,9 @@ if get_option('build_backends')
 
   mkl_libdirs = get_option('mkl_libdirs')
   mkl_lib = cc.find_library('mkl_rt', dirs: mkl_libdirs, required: false)
+  if not mkl_lib.found()
+    mkl_lib = cc.find_library('mklml', dirs: mkl_libdirs, required: false)
+  endif
 
   openblas_libdirs = get_option('openblas_libdirs')
   openblas_lib = cc.find_library('openblas.dll', dirs: openblas_libdirs, required: false)
@@ -394,9 +402,9 @@ endif # if get_option('build_backends')
 if not has_backends and get_option('build_backends')
   error('''
 
-        No usable computation backends (cudnn/tensorflow/etc) are found.
-        If you want to build it with random only backend, pass
-        -D build_backends=false to a meson build.''')
+        No usable computation backends (cudnn/opencl/blas/etc) enabled.
+        If you want to build with the random backend only, add
+        -Dbuild_backends=false to the build command line.''')
 endif
 
 

meson_options.txt

@@ -117,3 +117,8 @@ option('gtest',
        type: 'boolean',
        value: true,
        description: 'Build gtest tests')
+
+option('protobuf-3-6-0',
+       type: 'boolean',
+       value: false,
+       description: 'Use the protobuf 3.6.0 subproject')

src/benchmark/benchmark.cc

@@ -56,17 +56,6 @@ void Benchmark::Run() {
   options.Add<IntOption>(kNNCacheSizeId, 0, 999999999) = 200000;
   options.Add<IntOption>(kThreadsOptionId, 1, 128) = kDefaultThreads;
 
-  auto defaults = options.GetMutableDefaultsOptions();
-
-  defaults->Set<int>(SearchParams::kMiniBatchSizeId.GetId(), 256);
-  defaults->Set<float>(SearchParams::kFpuReductionId.GetId(), 1.2f);
-  defaults->Set<float>(SearchParams::kCpuctId.GetId(), 3.4f);
-  defaults->Set<float>(SearchParams::kPolicySoftmaxTempId.GetId(), 2.2f);
-  defaults->Set<int>(SearchParams::kMaxCollisionVisitsId.GetId(), 9999);
-  defaults->Set<int>(SearchParams::kMaxCollisionEventsId.GetId(), 32);
-  defaults->Set<int>(SearchParams::kCacheHistoryLengthId.GetId(), 0);
-  defaults->Set<bool>(SearchParams::kOutOfOrderEvalId.GetId(), true);
-
   if (!options.ProcessAllFlags()) return;
 
   try {