Merge pull request #81 from DemoGo/parallelmap

For parallel map implementation CUDA and openMP

Changes_list.txt

@@ -0,0 +1,11 @@
+lrparser.y : LINE 574 uncomment the line to activate the -o (optimize) option
+
+For fixing the segmentation fault in PageRank_DSL_V2 :-
+SymbolTableBuilder.cpp :   LINE 375 added a check to see if the forall statement is a for loop or a forall loop
+deviceVarsAnalyser.h   :   getTempVar() LINE 284 removed static cast and used strcpy since it previously generated random characters
+
+added fp_idnode to identifier
+analyser/deviceVars/getUsedVars.cpp: LINE 230 fixed fp for sssp
+src/symbolutil/SymbolTableBuilder.cpp: LINE 278 added fp node
+
+formal parameters generation:

README.md

@@ -5,11 +5,6 @@
 
 
 # StarPlat: A Versatile DSL for Graph Analytics
-
-```diff
-- Please use WorkingBranch
-```
-
 # GraphDSL MPI Backend
 
 ### How to compile Generated MPI Codes on Aqua Cluster
@@ -105,56 +100,10 @@ git clone -b WorkingBranch https://github.com/nibeditabh/GraphDSL.git
 cd GraphDSL/src
 make
 ```
-
-# How to Build (Mac OS | Tested on M1)
-
-By default, MacOS comes with the `clang` compiler, which can cause syntax errors when running the `make` command due to differences between `gcc` and `clang` semantics. To resolve this, follow these steps:
-
-## Installation:
-
-1. Install `gcc` using Homebrew:
-
-   ```sh
-   brew install gcc
-   ```
-
-2. Check the installation location of the `g++` binaries:
-
-   ```sh
-   which brew
-   ll /opt/homebrew/bin/g++-*
-   ```
-
-3. Create symbolic links to the installed version, (presumes that the above command returns gcc-13):
-
-   ```sh
-   sudo ln -sf /opt/homebrew/bin/gcc-13 /usr/local/bin/gcc
-   sudo ln -sf /opt/homebrew/bin/g++-13 /usr/local/bin/g++
-   sudo ln -sf /opt/homebrew/bin/c++-13 /usr/local/bin/c++
-   sudo ln -sf /opt/homebrew/bin/cpp-13 /usr/local/bin/cpp
-   ```
-
-   Note: Modifying anything in `/usr/bin/*` is discouraged and may require disabling SIP (System Integrity Protection). Therefore, we create symlinks to `/usr/local/bin/*`.
-
-4. After installation, check the versions:
-
-   ```sh
-   g++ --version # Should return clang version
-   g++-13 --version # Should return (Homebrew GCC)
-   ```
-
-5. Update the Makefile (`StarPlat/src/Makefile`) to use the symlinked `g++`:
-
-   ```make
-   CC = /usr/local/bin/g++
-   ```
-6. Add `%token return_func` in the file `lrparser.y` (`StarPlat/src/parser/lrparser.y`)
-
-
 # How to generate
 ```
 # Step 3. Generate files from DSL files.  
-./StarPlat [-s|-d] -f <dsl.sp> -b [cuda|omp|mpi|acc|sycl]
+./StarPlat [-s|-d] -f <dsl.sp> -b [cuda|omp|mpi|acc]
 
 #Example
 ./StarPlat -s -f ../graphcode/generated_cuda/sssp_dslV2 -b omp
@@ -188,107 +137,6 @@ pgc++ -std=c++14 ../graphcode/generated_OpenACC/sssp_dslV2.cu -I .. -o ssspAcc.o
 
 ```
 
-# Miscellaneous
-## Installing Boost Libraries to run StarPlat for MPI
-
-On the root directory of `StarPlat`, running the command:
-
-```sh
-grep -r '<boost/' .
-```
-
-shows the files that use boost. This section provides a walkthrough of building, installing, and linking the boost libraries and source to create a program. The workflow has been tested on an arm64 architecture with a 64-bit address type on a Mac machine. While this guide is compiled from various sources, for a comprehensive understanding, please refer to the official boost documentation.
-
-Our main goal is to build the following file with `mpic++` and run with `mpirun`:
-
-```c++
-#include <iostream>
-#include <boost/mpi/environment.hpp>
-#include <boost/mpi/communicator.hpp>
-
-namespace mpi = boost::mpi;
-
-int main() {
-    mpi::environment env;
-    mpi::communicator world;
-    std::cout << "I am process " << world.rank() << " of " << world.size() << "." << std::endl;
-    return 0;
-}
-```
-
-The default compiler that comes with MacOS is clang, and StarPlat is built on `gcc`. Note that a C++ library compiled with GCC is not compatible with Clang and vice versa. So, we will build the libraries and code exclusively on one platform.
-
-There are options to compile boost as a universal library agnostic to architecture (discussed [here](https://stackoverflow.com/questions/64553398/compile-boost-as-universal-library-intel-and-apple-silicon-architectures)), but for brevity, this walkthrough aims to run it on gcc/g++.
-
-We will be building the boost library from the source, providing more autonomy and control over packages and libraries, and installation directories from where we can include headers and link libraries.
-
-- **Step 1**: Download the tar file from boost/downloads.
-- **Step 2**: Extract the tar file to a location using the command `tar -xvf boost.tar.gz`.
-- **Step 3**: Create a directory anywhere with `mkdir -p /path/to/boost_version`.
-- **Step 4**: Navigate to the extracted boost directory.
-- **Step 5**: Run `./bootstrap.sh --prefix=/path/to/boost_version` to extract all the headers and binaries in the specified prefix location.
-- **Step 6**: The bootstrap.sh will create a file called `project-config.jam`.
-- **Step 7**: In the `project-config.jam` file, comment out the section related to the compiler configuration.
-
-```jam
-# Compiler configuration. This definition will be used unless
-# you already have defined some toolsets in your user-config.jam
-# file.
-# if ! clang in [ feature.values <toolset> ]
-# {
-#    using clang ;
-# }
-
-# project : default-build <toolset>clang ;
-```
-
-- **Step 8**: Configure the project-config.jam to use gcc.
-
-```jam
-using gcc : <version> : /path/to/gcc
-# Example
-using gcc : 13.2 : /usr/local/bin/gcc
-```
-
-- **Step 9**: Install MPI by following the steps [here](https://docs.open-mpi.org/en/v5.0.x/installing-open-mpi/quickstart.html).
-- **Step 10**: Add the following line to the `project-config.jam` file (notice the space between mpi and ;).
-
-```jam
-using mpi ;
-```
-
-- **Step 11**: Run the command `./b2 cxxflags=-std=c++17 install` to build boost with the standard library of choice and install it. If you encounter permission denied errors, use `sudo` before the command.
-
-- **Step 12**: If your editor's autocomplete looks for environment variables for autocompletion and code completion, run:
-
-```sh
-export DYLD_LIBRARY_PATH=/path/to/boost/lib/:$DYLD_LIBRARY_PATH
-```
-
-- **Step 13**: Copy the following content:
-
-```c++
-#include <iostream>
-#include <boost/mpi/environment.hpp>
-#include <boost/mpi/communicator.hpp>
-
-namespace mpi = boost::mpi;
-
-int main() {
-    mpi::environment env;
-    mpi::communicator world;
-    std::cout << "I am process " << world.rank() << " of " << world.size() << "." << std::endl;
-    return 0;
-}
-```
-
-```sh
-OMPI_CXX=g++ mpic++ -I/Users/durwasa/boost/include -L/Users/durwasa/boost/lib helloworld.cpp -o helloworld -lboost_mpi
-```
-
-The `OMPI_CXX` sets the environment variable to `gcc` instead of `clang` (since `mpic++` has been installed using Homebrew/MacPorts). `-I` includes the header files, `-L` links the library, and `-lboost_mpi` links the boost.mpi, which we configured using our .jam file.
-
-- **Step 14**: Run the command `OMPI_CXX=g++ mpirun helloworld`. If the default compiler in the system is `gcc`, omit the `OMPI_CXX` flag. `mpicxx` and other variants are technically compiler wrappers that translate the compiler commands and inject flags into it. Using the `OMPI_CXX` flag specifies which C++ compiler it should use while injecting the flags. Learn more about compiler wrappers [here](https://www.ibm.com/docs/en/smpi/10.2?topic=administering-compiling-applications).
 
 Graph DSL for basic graph algorithms 
 - Phase1 (static)  **SSSP, BC, PR, and TC**
@@ -297,6 +145,7 @@ Graph DSL for basic graph algorithms
 ```
 This project is funded by India's National Supercomputing Mission, whose timely funding and regular reviews are greatly appreciated.
 ```
+
 ## Licensing
 The StarPlat codebase is made available for academic and non-commercial use.
 Commercial use of the code needs licensing. The interested may contact rupesh@cse.iitm.ac.in.

TACode.h

@@ -0,0 +1,22 @@
+#include<stdio.h>
+#include<Symbol.h>
+
+using namespace std;
+
+class TACode 
+{
+  string opCode;
+  string target;
+  Symbol* destination;
+  Symbol* operand1;
+  Symbol* operand2;
+
+
+  TACode()
+  {
+    operand1=NULL;
+    operand2=NULL;
+    destination=NULL;
+  }
+
+};

comparison.py

@@ -0,0 +1,51 @@
+
+"""
+How to run  :  python name_of_python_script  <filename1> <filename2>
+Output  : Count of non-matching rows in both files
+"""
+
+
+import sys
+
+#first file name
+file = sys.argv[1] 
+#second file name
+file1 = sys.argv[2] 
+f=open(file,"r")
+f1=open(file1,"r")
+lines=f.readlines()
+lines1=f1.readlines()
+#print(lines);
+result=[]
+result1=[]
+for x in lines:
+    result.append(x.split("  ")[1])
+
+f.close()
+
+#print(result)
+
+for x in lines1:
+    result1.append(x.split("  ")[1])
+
+f1.close()
+#print(result1)
+
+new = []
+for sub in result:
+    new.append(sub.replace("\n", ""))
+#print(new)
+
+
+new1 = []
+for sub in result1:
+    new1.append(sub.replace("\n", ""))
+#print(new1)
+count=0;
+
+for i in range(len(result)):
+    if abs(round(float(result[i]),3) - round(float((result1[i])),3)) > 0.000001:
+        count = count+1
+
+print(count)
+