Add the dynamic parallelism config to the triangle counting lab

labs/triangle_counting/README.md

@@ -157,4 +157,9 @@ docker build . -t raiproject/pumps2018:triangle-counting-amd64-cuda100
 docker push raiproject/pumps2018:triangle-counting-amd64-cuda100
 ```
 
+## Enabling Dynamic Parallelism
+Please override the configurations files with the files in dynamic\_parallelism\_enabled. In the new CMakeLists.txt, make sure the following line reflect the compute capability of the device RAI uses. For example, V100 is corresponding to 7.0 and that means you don't need to make changes.
 
+```
+list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_70,code=sm_70;)
+```

labs/triangle_counting/dynamic_parallel_enabled/CMakeLists.txt

@@ -0,0 +1,81 @@
+include(cmake/HunterGate.cmake)
+
+# use config in cmake/Hunter/config.cmake
+HunterGate(
+  URL "https://github.com/ruslo/hunter/archive/v0.23.112.tar.gz"
+  SHA1 "4b894e1d5d203f0cc9a77431dbb1b486ab6f4430"
+  LOCAL
+)
+
+cmake_minimum_required(VERSION 3.8 FATAL_ERROR)
+
+
+
+project(tc LANGUAGES CXX CUDA)
+
+
+#include(FindCUDA/select_compute_arch)
+find_package(CUDAToolkit REQUIRED)
+#find_package(CUDA REQUIRED)
+
+hunter_add_package(fmt)
+hunter_add_package(spdlog)
+find_package(spdlog CONFIG REQUIRED)
+hunter_add_package(cub)
+find_package(cub CONFIG REQUIRED)
+hunter_add_package(Catch)
+find_package(pangolin CONFIG REQUIRED)
+
+set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
+set(CMAKE_COLOR_MAKEFILE ON)
+set(VERBOSE_BUILD ON)
+set(DEFAULT_BUILD_TYPE "Release")
+
+
+
+
+set_property(GLOBAL PROPERTY USE_FOLDERS ON)
+
+include(CTest)
+
+add_executable(tc
+  eval.cu
+  helper.hpp
+  template.cu
+  template.hu
+  common/catch.hpp
+  common/clara.hpp
+  common/utils.hpp
+)
+
+
+if(DEFINED GRAPH_PREFIX_PATH)
+  message(STATUS "compiling with GRAPH_PREFIX_PATH=${GRAPH_PREFIX_PATH}")
+  target_compile_definitions(tc PRIVATE GRAPH_PREFIX_PATH=${GRAPH_PREFIX_PATH})
+else()
+  message(FATAL_ERROR "please define GRAPH_PREFIX_PATH")
+endif()
+
+#CUDA_SELECT_NVCC_ARCH_FLAGS(CUDA_ARCH_FLAGS Auto)
+#list(APPEND CUDA_NVCC_FLAGS ${CUDA_ARCH_FLAGS})
+list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_70,code=sm_70;)
+
+target_compile_features(tc PUBLIC cxx_std_11)
+
+target_compile_options(tc PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:${CUDA_NVCC_FLAGS}>)
+set_target_properties( tc PROPERTIES
+                       CUDA_SEPARABLE_COMPILATION ON
+                       #CUDA_RESOLVE_DEVICE_SYMBOLS ON
+                      )
+
+target_link_libraries(tc pangolin::pangolin32 ${CUDA_cudadevrt_LIBRARY} ${CUDA_LIBRARIES})
+
+# target_link_options(tc PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:${CUDA_NVCC_FLAGS}>)
+
+
+include_directories(tc
+  ${PROJECT_SOURCE_DIR}/src
+  ${CUDA_INCLUDE_DIRS}
+)
+
+enable_testing()

labs/triangle_counting/dynamic_parallel_enabled/rai_build.yml

@@ -0,0 +1,20 @@
+# lab: Triangle Counting
+rai:
+  version: 0.2
+  image: tonywukun/pumps2018:triangle-counting-amd64-cuda100-cmake3.9test2
+resources:
+  cpu:
+    architecture: amd64
+  gpu:
+    count: 1
+  limits:
+    network: true
+commands:
+  build:
+    - cp -r /src . # ensure source code is in upload folder
+    - /root/spack/opt/spack/linux-ubuntu18.04-skylake/gcc-7.3.0/cmake-3.26.3-yuk5ln5gyf345uqiqb44jzuzaczldhz2/bin/cmake /src -DCMAKE_PREFIX_PATH=${PANGOLIN_INSTALL_DIR} -DGRAPH_PREFIX_PATH=\"/graphs\"
+    #- /root/spack/opt/spack/linux-ubuntu18.04-skylake/gcc-7.3.0/cmake-3.9.6-nsyti5bha6bunwre2xxkbw5j3aplpx7o/bin/cmake /src -DCMAKE_PREFIX_PATH=${PANGOLIN_INSTALL_DIR} -DGRAPH_PREFIX_PATH=\"/graphs\"
+    - make VERBOSE=1
+    - ./tc -c LINEAR
+    - ./tc -c OTHER
+