Merge pull request #44 from Jinxto/cmake

Update cmake and cpplinter

.github/workflows/ci.yml

@@ -1,34 +1,34 @@
 name: CI
 
 on:
-    push:
-        branches: main
-        paths: ['**.cu','**.c','**.cpp', '**.h', '**CMakeLists.txt'] 
-    pull_request:
-        branches: main
-        paths: ['**.cu','**.c','**.cpp', '**.h', '**CMakeLists.txt']
+  push:
+    branches: main
+    paths: ["**.cu", "**.c", "**.cpp", "**.h", "**CMakeLists.txt"]
+  pull_request:
+    branches: main
+    paths: ["**.cu", "**.c", "**.cpp", "**.h", "**CMakeLists.txt"]
 
 jobs:
-    build-and-test:
-        runs-on: ubuntu-latest
+  build-and-test:
+    runs-on: ubuntu-latest
 
-        steps:
-        - name: Checkout code
-          uses: actions/checkout@v4
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
 
-        - name: Setup python
-          uses: actions/setup-python@v5
-          with:
-            python-version: '3.10' 
+      - name: Setup python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
 
-        - name: Install dependencies
-          run: |
-            pip install pandas
+      - name: Install dependencies
+        run: |
+          pip install pandas
 
-        - name: Build project
-          run: |
-            make build
-            
-        - name: Run test suite
-          run: |
-            make test
+      - name: Build project
+        run: |
+          make build
+
+      - name: Run test suite
+        run: |
+          make test_cpu

CMakeLists.txt

@@ -1,30 +1,36 @@
 cmake_minimum_required(VERSION 3.16)
 
-# Set the project name
-project(ichida-algo)
+project(ichida-algo LANGUAGES C CXX)
 
 set(CMAKE_C_FLAGS "-O3 -march=native -ffast-math -funroll-loops -fopenmp -Wall -Wextra")
-
 set(CMAKE_C_STANDARD 11)
 set(CMAKE_C_STANDARD_REQUIRED True)
-# set(CMAKE_VERBOSE_MAKEFILE ON)
+set(CMAKE_VERBOSE_MAKEFILE ON)
 
-set(SRC_DIR src)
 set(INC_DIR include)
-set(LIB_DIR lib)
-set(TEST_DIR test)
-set(BENCHMARK_DIR benchmark)
+set(SRC_DIR src)
+set(CUDA_SRC_DIR cudasrc)
 
-# Source files
-file(GLOB_RECURSE SOURCE_FILES ${SRC_DIR}/*.c)
+include_directories(${INC_DIR})
 
-include_directories(include)
+file(GLOB_RECURSE SOURCE_FILES ${SRC_DIR}/*.c)
 
 add_executable(speed_cpu ${SOURCE_FILES})
-# add_executable(benchmark ${SRC_DIR}/matrix.c ${BENCHMARK_DIR}/benchmark.c)
-
-target_link_libraries(speed_cpu m pthread)
-# target_link_libraries(benchmark m)
-
+target_link_libraries(speed_cpu m pthread gomp)
+
+find_package(CUDA)
+
+if(CUDA_FOUND)
+    enable_language(CUDA)
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xptxas -O3 --use_fast_math -Xcompiler -march=native -unroll-aggressive -arch=sm_80")
+    find_package(MPI REQUIRED)
+    include_directories(${MPI_INCLUDE_PATH})
+    file(GLOB_RECURSE CUDA_SOURCE_FILES ${CUDA_SRC_DIR}/*.cu)
+    add_executable(speed_gpu ${CUDA_SOURCE_FILES})
+    set_target_properties(speed_gpu PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
+    target_link_libraries(speed_gpu m ${MPI_LIBRARIES})
+else()
+    message(STATUS "CUDA not found, only CPU version will be built.")
+endif()
 
 

Makefile

@@ -1,37 +1,36 @@
-.PHONY: all test clean run build run_test
+.PHONY: all clean build run_cpu run_gpu test_cpu test_gpu bench stat
 
-all: rebuild
+# Default iterations
+iterations ?= 1000
+
+all: build
 
 clean:
 	rm -f test/results.csv
 	rm -f results.csv
 	rm -rf build
-	rm -f speed_cpu
+	rm -f speed_cpu speed_gpu
 
 build: clean
-	cmake -Bbuild
-	$(MAKE) -C ./build
-	mv ./build/speed_cpu ./
-
-rebuild:
-	$(MAKE) -C ./build
-	mv ./build/speed_cpu ./
-
-run: build
-	./speed_demo_cpu.sh ./weights_and_biases.txt ./tensors
-
-run_test: build
-	./speed_cpu ./weights_and_biases.txt ./tensors
-
-test: build
-	./speed_cpu ./weights_and_biases.txt ./tensors 1
-	mv ./results.csv ./test
-	python3 ./test/verify_csv.py
+	cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
+	$(MAKE) -C build
+	cp -u build/speed_cpu ./
+	if [ -f build/speed_gpu ]; then cp -u build/speed_gpu ./; fi
 
-bench: build
-	./build/benchmark
+run_cpu: build
+	./speed_cpu ./weights_and_biases.txt ./tensors $(iterations)
 
-stat: build
-	python3 ./benchmark/stat.py
+run_gpu: build
+	n_gpus=$(shell nvidia-smi --query-gpu=name --format=csv,noheader | wc -l); \
+	mpirun -np $$n_gpus ./speed_gpu ./weights_and_biases.txt ./tensors $(iterations)
 
+test_cpu: build
+	./speed_cpu ./weights_and_biases.txt ./tensors $(iterations)
+	mv ./results.csv ./test
+	python3 ./test/verify_csv.py
 
+test_gpu: build
+	n_gpus=$(shell nvidia-smi --query-gpu=name --format=csv,noheader | wc -l); \
+	mpirun -np $$n_gpus ./speed_gpu ./weights_and_biases.txt ./tensors $(iterations)
+	mv ./results.csv ./test
+	python3 ./test/verify_csv.py