Cub submodule

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "cub"]
+	path = cub
+	url = https://github.com/NVIDIA/cub.git

IM.h

@@ -10,7 +10,7 @@
 #include <thrust/execution_policy.h>
 
 #include "kernels.h"
-#include "./cub-1.8.0/cub/cub.cuh"
+#include "./cub/cub/cub.cuh"
 
 #define MAX_NUMBER_OF_RR_SETS 30000000
 

README.md

@@ -6,13 +6,23 @@ gIM is a GPU-accelerated RIS-based influence maximization (IM) algorithm.
 ### CUDA toolkit
 To install CUDA toolkit please use [this link](https://developer.nvidia.com/cuda-downloads).
 
+### Repository Preparation
+
+gIM uses cub library for some parallel algorithms like scan. Before compiling the code, initialize submodules using the following command:
+ ```
+ git submodule update --init
+ ```
+
 ### Compilation
 
 Compile the code using the following command:
  ```
  nvcc -std=c++11 -O3  -lcurand  *.cu  -o gIM
  ```
-
+ If you are using CUDA 7.5 and newer versions on Windows, you can remove the -std flag because these versions include the support for C++11 standard by default. Wildcard expansion does not work in cmd.exe, so you need to use the following command for compiling the code:
+ ```
+ nvcc -O3 -lcurand kernels.cu IM.cu testgraph.cu -o gIM
+ ``` 
 ### Execution
 
 Run the code using the following command:

kernels.cu

@@ -7,7 +7,7 @@
 
 #include <curand.h>
 #include <curand_kernel.h>
-#include "./cub-1.8.0/cub/cub.cuh"
+#include "./cub/cub/cub.cuh"
 
 __device__ uint32_t mix_hash(uint32_t a, uint32_t b, uint32_t c)
 {