diff options
| author | Martin Junczys-Dowmunt <Marcin.JunczysDowmunt@microsoft.com> | 2020-01-06 22:14:00 +0300 |
|---|---|---|
| committer | Martin Junczys-Dowmunt <Marcin.JunczysDowmunt@microsoft.com> | 2020-01-06 22:14:00 +0300 |
| commit | 164d26cc36204316324a2fc76f7538a044dd192a (patch) | |
| tree | 630189bdeca50fe62c1b76b4e9c6b3e7ba24fca5 /CMakeLists.txt | |
| parent | 88d998058917648510ebe66bf764a418ed1d3a8f (diff) | |
Merged PR 10999: Splitting up add_all.h into *.h, *.cu and *.inc
Splitting up header file into header and *.cu, comes with the price of having to include specializations for combinations of types as for element.inc and add.inc. No code changes otherwise.
Add CMake options to disable specific compute capabilities.
When run with `make -j16` this compiles in about 6 minutes instead of 7 minutes. Selecting only SM70 during compilation brings down the time to 3 minutes.
Diffstat (limited to 'CMakeLists.txt')
| -rw-r--r-- | CMakeLists.txt | 23 |
1 files changed, 17 insertions, 6 deletions
diff --git a/CMakeLists.txt b/CMakeLists.txt index ab5460e6..c442931f 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -13,6 +13,10 @@ set(BUILD_ARCH native CACHE STRING "Compile for this CPU architecture.") # Custom CMake options option(COMPILE_CPU "Compile CPU version" ON) option(COMPILE_CUDA "Compile GPU version" ON) +option(COMPILE_CUDA_SM35 "Compile GPU version with SM35 support" ON) +option(COMPILE_CUDA_SM50 "Compile GPU version with SM50 support" ON) +option(COMPILE_CUDA_SM60 "Compile GPU version with SM60 support" ON) +option(COMPILE_CUDA_SM70 "Compile GPU version with SM70 support" ON) option(COMPILE_EXAMPLES "Compile examples" OFF) option(COMPILE_SERVER "Compile marian-server" OFF) option(COMPILE_TESTS "Compile tests" OFF) @@ -181,8 +185,6 @@ set(EXT_LIBS ${EXT_LIBS} ${CMAKE_DL_LIBS}) if(COMPILE_CUDA) -LIST(APPEND COMPUTE -arch=sm_35; -gencode=arch=compute_35,code=sm_35; -gencode=arch=compute_50,code=sm_50; -gencode=arch=compute_52,code=sm_52; -gencode=arch=compute_60,code=sm_60; -gencode=arch=compute_61,code=sm_61;) - if(USE_STATIC_LIBS) # link statically to stdlib libraries set(CMAKE_EXE_LINKER_FLAGS "-static-libgcc -static-libstdc++") @@ -202,10 +204,19 @@ if(CUDA_FOUND) if((CUDA_VERSION VERSION_EQUAL "10.0" OR CUDA_VERSION VERSION_GREATER "10.0") AND (CMAKE_VERSION VERSION_LESS "3.12.2")) message(WARNING "On some Unix systems CUDA 10.0+ requires CMake 3.12.2+; you use CMake ${CMAKE_VERSION}") endif() - - if(CUDA_VERSION VERSION_GREATER "8.0") - LIST(APPEND COMPUTE -gencode=arch=compute_70,code=sm_70; -gencode=arch=compute_70,code=compute_70) - endif() + + if(COMPILE_CUDA_SM35) + LIST(APPEND COMPUTE -arch=sm_35; -gencode=arch=compute_35,code=sm_35;) # Tesla K40 and above + endif(COMPILE_CUDA_SM35) + if(COMPILE_CUDA_SM50) + LIST(APPEND COMPUTE -gencode=arch=compute_50,code=sm_50; -gencode=arch=compute_52,code=sm_52;) # Maxwell GPUs + endif(COMPILE_CUDA_SM50) + if(COMPILE_CUDA_SM60) + LIST(APPEND COMPUTE -gencode=arch=compute_60,code=sm_60; -gencode=arch=compute_61,code=sm_61;) # Pascal GPUs + endif(COMPILE_CUDA_SM60) + if(COMPILE_CUDA_SM70) + LIST(APPEND COMPUTE -gencode=arch=compute_70,code=sm_70; -gencode=arch=compute_70,code=compute_70) # Volta GPUs + endif(COMPILE_CUDA_SM70) if(USE_STATIC_LIBS) find_library(CUDA_culibos_LIBRARY NAMES culibos PATHS ${CUDA_TOOLKIT_ROOT_DIR}/lib64) |