fixing cmake, adding back gpu targets in cmake script

example/44_elementwise_permute/CMakeLists.txt

-if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
-    add_example_executable(example_elementwise_permute_4D_fp16 elementwise_permute_4D_fp16.cpp)
-    add_example_executable(example_elementwise_permute_4D_fp16_2d elementwise_permute_4D_fp16_2d.cpp)
-    add_example_executable(example_elementwise_permute_5D elementwise_permute_5D.cpp)
-    add_example_executable(example_elementwise_permute_5D_3d elementwise_permute_5D_3d.cpp)
-endif()
+add_example_executable(example_elementwise_permute_4D_fp16 elementwise_permute_4D_fp16.cpp)
+add_example_executable(example_elementwise_permute_4D_fp16_2d elementwise_permute_4D_fp16_2d.cpp)
+add_example_executable(example_elementwise_permute_5D elementwise_permute_5D.cpp)
+add_example_executable(example_elementwise_permute_5D_3d elementwise_permute_5D_3d.cpp)
\ No newline at end of file

script/cmake-ck-dev.sh

@@ -12,7 +12,7 @@ cmake
 -save-temps=$PWD"                                                                                 \
 -D CMAKE_BUILD_TYPE=Release                                                                       \
 -D BUILD_DEV=ON                                                                                   \
--D GPU_TARGETS="gfx90a"                                                             \
+-D GPU_TARGETS="gfx908;gfx90a;gfx940"                                                             \
 -D CMAKE_VERBOSE_MAKEFILE:BOOL=ON                                                                 \
 -D USE_BITINT_EXTENSION_INT4=OFF                                                                  \
 ${MY_PROJECT_SOURCE}