gemmlib/CMakeLists.txt at main · chenfucn/gemmlib · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
# Set the minimum required CMake version
cmake_minimum_required(VERSION 3.25)

set(CMAKE_CUDA_ARCHITECTURES 80)

# Set the project name
project(ms_blkq4gemm_proj LANGUAGES CUDA CXX)

set(CUTLASS_NVCC_ARCHS "80;90a")

# Set the C++ standard
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)

include(FetchContent)

# fetch the cutlass library from github
FetchContent_Declare(
  cutlass
  GIT_REPOSITORY https://github.com/NVIDIA/cutlass.git
  GIT_TAG        a75b4ac483166189a45290783cb0a18af5ff0ea5 # v3.3.0
)

FetchContent_GetProperties(cutlass)
if(NOT cutlass_POPULATED)
  FetchContent_Populate(cutlass)
endif()

FetchContent_Declare(
  gsl
  URL https://github.com/microsoft/GSL/archive/refs/tags/v4.0.0.zip
  URL_HASH SHA1=cf368104cd22a87b4dd0c80228919bb2df3e2a14
)

FetchContent_Declare(
  googletest
  URL https://github.com/google/googletest/archive/03597a01ee50ed33e9dfd640b249b4be3799d395.zip
)
# For Windows: Prevent overriding the parent project's compiler/linker settings
set(gtest_force_shared_crt ON CACHE BOOL "" FORCE)

FetchContent_MakeAvailable(gsl googletest)

list(APPEND INC_PATH ${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES} ${cutlass_SOURCE_DIR}/include ${cutlass_SOURCE_DIR}/tools/util/include ${gsl_SOURCE_DIR}/include ${PROJECT_SOURCE_DIR}/include ${PROJECT_SOURCE_DIR}/src/mickey)

message(STATUS "include paths: ${INC_PATH}")

add_library(ms_blkq4gemm ${PROJECT_SOURCE_DIR}/src/blkq4_f16_gemm_sm80.cu)
target_include_directories(ms_blkq4gemm PUBLIC ${INC_PATH})

# Since we need to link this library to a shared library, to be loaded by python, we
# have to use this to avoid the error: "relocation R_X86_64_32 against `some local symbol' can not be used when making a shared object; recompile with -fPIC"
set_property(TARGET ms_blkq4gemm PROPERTY POSITION_INDEPENDENT_CODE ON)


# Tests

enable_testing()

add_executable(ms_blkq4gemm_test
  ${PROJECT_SOURCE_DIR}/test/blkq4_fp16_quant_sm80_test.cu
  ${PROJECT_SOURCE_DIR}/test/blkq4_fp16_gemm_sm80_test.cu)
target_include_directories(ms_blkq4gemm_test PUBLIC ${INC_PATH})

target_link_libraries(ms_blkq4gemm_test PRIVATE ms_blkq4gemm GTest::gtest_main)

include(GoogleTest)
gtest_discover_tests(ms_blkq4gemm_test)


#find_package(Python REQUIRED COMPONENTS Development)
#find_package(Torch REQUIRED)

#add_library(ms_blkq4linear_ext SHARED
#  ${PROJECT_SOURCE_DIR}/src/blkq4_f16_gemm_sm80.cu
#  ${PROJECT_SOURCE_DIR}/python/ms_blkq4linear_ext/ms_blkq4linear.cpp
#)
#target_compile_features(pytorch_cmake_example PRIVATE cxx_std_11)
#target_include_directories(ms_blkq4linear_ext PUBLIC ${INC_PATH})
#target_link_libraries(ms_blkq4linear_ext PRIVATE ${TORCH_LIBRARIES} Python::Python)

# Use if the default GCC version gives issues
#target_compile_options(pytorch_cmake_example PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:-ccbin g++-9>)

# Use a variant of this if you're on an earlier cmake than 3.18
# target_compile_options(pytorch_cmake_example PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:-gencode arch=compute_61,code=sm_61>)