IBM Research CDT Static Analysis Features CDT Developer

IBM Research CDT Static Analysis Features CDT Developer Summit - Ottawa Beth Tibbitts September 20, 2006 tibbitts@us. ibm. com This work has been supported in part by the Defense Advanced Research Projects Agency (DARPA) under contract No. NBCH 30390004. © 2006 IBM Corporation

IBM Research The Problem § Static Analysis of C programs is useful § Existing Abstract Syntax Tree (AST) in Eclipse CDT provides basic navigation and information, but needs more 2 © 2006 IBM Corporation

IBM Research CDT AST Extensions § Enhance existing CASTNode and Visitor – bottom-up traversal § Add other additional graphs: – Call Graph – Control Flow Graph – Data Dependence Graph § Traversal of these new graphs available in: – Topological Order – Reverse Topological Order 3 © 2006 IBM Corporation

IBM Research Bottom-up AST traversal org. eclipse. cdt. core. dom. ast. ASTVisitor org. eclipse. cdt. core. dom. ast. c. CASTVisitor Existing: public int visit(IASTxxx. . ){ return PROCESS_CONTINUE; } New: public int leave(IASTxxx. . ){ return PROCESS_CONTINUE; } 4 © 2006 IBM Corporation

IBM Research #include "mpi. h" #include "stdio. h" Call Graph void foo(int x); void gee(int x); void kei(int x); void foo(int x){ x ++; gee(x); } main foo a gee kei void gee(int x){ x *= 3; kei(x); } void kei(int x){ x = x % 10; foo(x); } void a(int x){ x --; } int main 3(int argc, char* argv[]){ int x = 0; foo(x); a(x); } Recursive calls detected… A cycle is detected on foo, gee and kei 5 © 2006 IBM Corporation

IBM Research Control Flow Graph & Data Flow Dependence Graph – sample program if #include <stdio. h> #include <string. h> if (my_rank !=0){ #include "mpi. h" /* create message */ // Sample MPI program sprintf(message, "Greetings from process %d!", my_rank); int main(int argc, char* argv[]){ dest = 0; printf("Hello MPI World the original. n"); /* use strlen+1 so that '' get transmitted */ int my_rank; /* rank of process */ MPI_Send(message, strlen(message)+1, MPI_CHAR, int p; /* number of processes */ dest, tag, MPI_COMM_WORLD); int source; /* rank of sender */ int dest; /* rank of receiver */ } int tag=0; /* tag for messages */ else{ char message[100], *tmp; /* storage for message */ MPI_Status status ; */ /* return status for receive MPI_Barrier(MPI_COMM_WORLD); for printf("From process 0: Num processes: %dn", p); for (source = 1; source < p; source++) { MPI_Recv(message, 100, MPI_CHAR, source, tag, int * array; MPI_COMM_WORLD, &status); printf("%sn", message); /* start up MPI */ } array = (int *)malloc(sizeof(int) * 10); MPI_Barrier(MPI_COMM_WORLD); MPI_Init(&argc, &argv); } /* shut down MPI */ /* find out process rank */ MPI_Finalize(); MPI_Comm_rank(MPI_COMM_WORLD, &my_rank); /* find out number of processes */ MPI_Comm_size(MPI_COMM_WORLD, &p); free(array); return 0; } MPI_Barrier(MPI_COMM_WORLD); 6 © 2006 IBM Corporation

IBM Research Control Flow Graph entry A printf MPI_Comm_rank() Int my_rank Int p Int source Int dest Int tag = 0 B free(array) MPI_Comm_size if return 0 MPI_Barrier() exit my_rank != 0 sprintf dest = 0 source = 1 Char message[100], *tmp MPI_Send() source < p MPI_Status status MPI_Barrier() for MPI_Recv Int *array = malloc() MPI_Init() A 7 (join block) MPI_Finalize() printf source ++ B © 2006 IBM Corporation

IBM Research Work in Progress Data Dependence Graph (DDG) entry A printf free(array) MPI_Comm_size Int p return 0 MPI_Barrier() exit Int source Int tag = 0 B MPI_Comm_rank() Int my_rank Int dest (graph not complete) my_rank != 0 sprintf dest = 0 Control flow Data flow source = 1 Char message[100], *tmp MPI_Send() source < p MPI_Status status MPI_Barrier() MPI_Recv Int *array = malloc() MPI_Init() A 8 (join block) MPI_Finalize() printf source ++ B © 2006 IBM Corporation

IBM Research Summary § MPI Barrier Analysis uses these structures § Is this valuable as an addition to CDT? § Other Future plans – Parallel Tools Platform (PTP) Analysis: • Static and Dynamic Analysis of MPI, Open. MP, and LAPI programs for detection of common errors • Code Refactorings for Performance Optimization, e. g. refactoring for improved computation / communication overlap 9 © 2006 IBM Corporation