blob: 3057f80fe72f66bcf40aff804e7a77946760de2a (
plain)
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
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
|
diff --git a/cuda/CMakeLists.txt b/cuda/CMakeLists.txt
index 8394b3aef..a91011666 100644
--- a/cuda/CMakeLists.txt
+++ b/cuda/CMakeLists.txt
@@ -1,5 +1,9 @@
cmake_minimum_required(VERSION 3.18 FATAL_ERROR)
add_library(ginkgo_cuda $<TARGET_OBJECTS:ginkgo_cuda_device> "")
+add_library(ginkgo_cuda_2 $<TARGET_OBJECTS:ginkgo_cuda_device> "")
+add_library(ginkgo_cuda_3 $<TARGET_OBJECTS:ginkgo_cuda_device> "")
+add_library(ginkgo_cuda_4 $<TARGET_OBJECTS:ginkgo_cuda_device> "")
+add_library(ginkgo_cuda_5 $<TARGET_OBJECTS:ginkgo_cuda_device> "")
include(${PROJECT_SOURCE_DIR}/cmake/template_instantiation.cmake)
add_instantiation_files(${PROJECT_SOURCE_DIR}/common/cuda_hip matrix/csr_kernels.instantiate.cpp CSR_INSTANTIATE)
add_instantiation_files(${PROJECT_SOURCE_DIR}/common/cuda_hip matrix/fbcsr_kernels.instantiate.cpp FBCSR_INSTANTIATE)
@@ -30,16 +34,32 @@ target_sources(
matrix/fft_kernels.cu
preconditioner/batch_jacobi_kernels.cu
solver/batch_bicgstab_kernels.cu
- ${BATCH_BICGSTAB_INSTANTIATE1}
- ${BATCH_BICGSTAB_INSTANTIATE2}
solver/batch_cg_kernels.cu
- ${BATCH_CG_INSTANTIATE1}
- ${BATCH_CG_INSTANTIATE2}
solver/lower_trs_kernels.cu
solver/upper_trs_kernels.cu
+)
+target_sources(
+ ginkgo_cuda_2
+ PRIVATE
${GKO_UNIFIED_COMMON_SOURCES}
+)
+target_sources(
+ ginkgo_cuda_3
+ PRIVATE
${GKO_CUDA_HIP_COMMON_SOURCES}
)
+target_sources(
+ ginkgo_cuda_4
+ PRIVATE
+ ${BATCH_BICGSTAB_INSTANTIATE1}
+ ${BATCH_BICGSTAB_INSTANTIATE2}
+)
+target_sources(
+ ginkgo_cuda_5
+ PRIVATE
+ ${BATCH_CG_INSTANTIATE1}
+ ${BATCH_CG_INSTANTIATE2}
+)
if(GINKGO_JACOBI_FULL_OPTIMIZATIONS)
set(GKO_CUDA_JACOBI_BLOCK_SIZES)
foreach(blocksize RANGE 1 32)
@@ -87,65 +107,74 @@ configure_file(
common/cuda_hip/preconditioner/jacobi_common.hpp
)
-if(CMAKE_CUDA_COMPILER_ID STREQUAL "NVIDIA")
- # remove false positive CUDA warnings when calling one<T>() and zero<T>()
- # and allows the usage of std::array for nvidia GPUs
- target_compile_options(
- ginkgo_cuda
- PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:--expt-relaxed-constexpr>
- )
- if(MSVC)
+function(confgure_cuda_target target_name)
+ if(CMAKE_CUDA_COMPILER_ID STREQUAL "NVIDIA")
+ # remove false positive CUDA warnings when calling one<T>() and zero<T>()
+ # and allows the usage of std::array for nvidia GPUs
target_compile_options(
- ginkgo_cuda
- PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:--extended-lambda>
- )
- else()
- target_compile_options(
- ginkgo_cuda
- PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:--expt-extended-lambda>
+ ${target_name}
+ PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:--expt-relaxed-constexpr>
)
+ if(MSVC)
+ target_compile_options(
+ ${target_name}
+ PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:--extended-lambda>
+ )
+ else()
+ target_compile_options(
+ ${target_name}
+ PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:--expt-extended-lambda>
+ )
+ endif()
endif()
-endif()
-ginkgo_compile_features(ginkgo_cuda)
-target_compile_definitions(
- ginkgo_cuda
- PRIVATE GKO_COMPILING_CUDA GKO_DEVICE_NAMESPACE=cuda
-)
-if(GINKGO_CUDA_CUSTOM_THRUST_NAMESPACE)
+ ginkgo_compile_features(${target_name})
target_compile_definitions(
- ginkgo_cuda
- PRIVATE THRUST_CUB_WRAPPED_NAMESPACE=gko
+ ${target_name}
+ PRIVATE GKO_COMPILING_CUDA GKO_DEVICE_NAMESPACE=cuda
)
-endif()
+ if(GINKGO_CUDA_CUSTOM_THRUST_NAMESPACE)
+ target_compile_definitions(
+ ${target_name}
+ PRIVATE THRUST_CUB_WRAPPED_NAMESPACE=gko
+ )
+ endif()
-# include path for generated headers like jacobi_common.hpp
-target_include_directories(ginkgo_cuda PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
-target_link_libraries(
- ginkgo_cuda
- PRIVATE
- CUDA::cudart
- CUDA::cublas
- CUDA::cusparse
- CUDA::curand
- CUDA::cufft
- nvtx::nvtx
-)
-# NVTX3 is header-only and requires dlopen/dlclose in static builds
-target_link_libraries(ginkgo_cuda PUBLIC ginkgo_device ${CMAKE_DL_LIBS})
+ # include path for generated headers like jacobi_common.hpp
+ target_include_directories(${target_name} PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
+ target_link_libraries(
+ ${target_name}
+ PRIVATE
+ CUDA::cudart
+ CUDA::cublas
+ CUDA::cusparse
+ CUDA::curand
+ CUDA::cufft
+ nvtx::nvtx
+ )
+ # NVTX3 is header-only and requires dlopen/dlclose in static builds
+ target_link_libraries(${target_name} PUBLIC ginkgo_device ${CMAKE_DL_LIBS})
-ginkgo_default_includes(ginkgo_cuda)
-ginkgo_install_library(ginkgo_cuda)
+ ginkgo_default_includes(${target_name})
+ ginkgo_install_library(${target_name})
-if(GINKGO_CHECK_CIRCULAR_DEPS)
- set(check_header_def "GKO_COMPILING_CUDA;GKO_DEVICE_NAMESPACE=cuda")
- if(GINKGO_CUDA_CUSTOM_THRUST_NAMESPACE)
- set(check_header_def
- "${check_header_def};THRUST_CUB_WRAPPED_NAMESPACE=gko"
- )
+ if(GINKGO_CHECK_CIRCULAR_DEPS)
+ set(check_header_def "GKO_COMPILING_CUDA;GKO_DEVICE_NAMESPACE=cuda")
+ if(GINKGO_CUDA_CUSTOM_THRUST_NAMESPACE)
+ set(check_header_def
+ "${check_header_def};THRUST_CUB_WRAPPED_NAMESPACE=gko"
+ )
+ endif()
+ ginkgo_check_headers(${target_name} "${check_header_def}")
endif()
- ginkgo_check_headers(ginkgo_cuda "${check_header_def}")
-endif()
+endfunction()
+
+confgure_cuda_target(ginkgo_cuda)
+confgure_cuda_target(ginkgo_cuda_2)
+confgure_cuda_target(ginkgo_cuda_3)
+confgure_cuda_target(ginkgo_cuda_4)
+confgure_cuda_target(ginkgo_cuda_5)
+target_link_libraries(ginkgo_cuda PUBLIC ginkgo_cuda_2 ginkgo_cuda_3 ginkgo_cuda_4 ginkgo_cuda_5)
if(GINKGO_BUILD_TESTS)
add_subdirectory(test)
|