C-DAC,Pune : High-Perf. Comp. Frontier Technologies Exploration Group and CMSD, University of Hyderabad, Technology Workshop hyPACK (October 15-18), 2013

hyPACK-2013 Mode-1 : Software tools on Multi-Core Processor Systems

The High Performance Computing (HPC) tool simplifies the multi-core development, which increases the application runtime performance and enhances developer productivity. It is important for developer to identify performance hotspots in serial and parallel applications including distributed MPI applications on Multi-core processor systems. The tools such as highly optimized compilers and libraries, code-level and Memory debuggers, Thread analyzer, Application Performance analyzer, and system-profiling tools help developer to take full advantage of multi-core architecture systems.

The commercial HPC tools, which are provided by Intel Software tools, AMD, IBM POWER-6/POWER-7 AIX, HP, Sun Studio 12, PGI Compilers, open source tools PAPI, HPCToolkit, Jumpshot and other HPC tools provide an optimized environment for the development of single, multi-threaded, and distributed MPI applications, leading to more robust scalable software. These tools help the programmer to understand the scalable behavior of sequential / parallel programs as well as performance of applications on Multi-Core Processors.

Example programs using different APIs, Compilation and execution of programs, programs based on numerical and non-numerical computations are executed on multi-core processors and behaviour of these codes are analyzed using different tools in hypack-2013. Some example programs on dense matrix computations have been analyzed using PAPI (Performance Application Programming Interface) tool. PAPI is being developed at the University of Tennessee (Innovative Computing Laboratory).

List of Programs using software tools :

Write a Pthreads program for addition of two vectors & and analyze the Performance using thread Visualization tool.
Write an Multi-threaded (Pthreads) program to compute the value of PI pie function f(x) = 4/(1+x²) between the limits 0 and 1 by numerical integration. Analyze the Performance using thread Visualization tool.
Write an Multi-threaded (OpenMP) program to compute the value of PI pie function f(x) = 4/(1+x²) between the limits 0 and 1 by numerical integration. Analyze the Performance using thread Visualization tool.
Write an MPI program to compute the value of PI pie function f(x) = 4/(1+x²) between the limits 0 and 1 by numerical integration using MPI point-to-point communication library calls. Analyze the Performance using thread Visualization tool.
Write an MPI-Pthreads program to calculate Infinity norm of a matrix using block striped partitioning with row wise data distribution using p processes and t threads. Analyze the Performance using thread Visualization tool.
MPI program to compute Matrix - Matrix Multiplication using block checkerboard partitioning & MPI Cartesian topology. Analyze the Performance using MPI Visualization tool.