Centre For Development of Advanced Computing
National PARAM Super Computing Facility
Parallel Computing Optimizing Performance of Parallel Programs
(PCOPP-2002)
| PCOPP-2002 Schedule and Tentative Technical Programme |
| Time | Activity |
Day 1: JUNE 3rd (Monday) Technical Sessions |
|
08:15 Hrs - 08:45 Hrs Breakfast at C-DAC Roof-Top Pavilion |
|
| 08:30 - 08:45 | Registration at C-DAC Auditorium |
| 08:45 - 09:00 | Inauguration and the Overview of PCOPP-2002 workshop |
| 09:00 - 09:45 | Introduction to MPI:
MPI Basics; Features of MPI; Point-to-Point and Collective Communication library calls; Using MPI in simple programs Download presentation slides : pcopp02-Intro-MPI.pdf |
| 10:00 - 11:00 | Trends in Serial Processor Computing Features of the memory sub-system hierarchy Managing Memory overheads Optimization techniques to ease the memory access Download presentation slides : pcopp02-Memory-Hierarchy.pdf |
11:00 Hrs - 11:15 Hrs : Tea Break at C-DAC Roof-Top Pavilion |
|
| 11:15 - 12:00 | Explicit Parallelism: Shared Memory Programming - Introduction to OpenMP Download presentation slides : pcopp02-Intro-OpenMP.pdf |
| 12:15 - 13:00 | Basic Compiler Techniques:
What an Optimizing Compiler does to get maximum performance of your code?
What profiler can do for you?
Compiler role in loop optimization techniques Summary of performance of selective programs with/without compiler optimizations on PARAM 10000 Download presentation slides : pcopp02-Compiler-Tech.pdf |
13:00 Hrs - 14:00 Hrs : Lunch Break at C-DAC Roof-Top Pavilion
16:30 Hrs - 16:45 Hrs and 18:30 Hrs - 18:45 Hrs: Tea Break at C-DAC Roof-Top Pavilion/NPSF
|
|
| 14:00 - 18:30 | Hands-on Session (Day1): Performance of selective FORTRAN/C programs on Uni-processor of PARAM 10000 without compiler optimization features; With compiler optimization features; Using code restructuring techniques such as loop un-rolling, loop fission and loop fusion, loop distribution, loop interchange and other loop optimization techniques to ease the memory access pattern; Performance of parallel programs using tuned math libraries and their impact on performance |
19:30 Hrs - 20:30 Hrs : Dinner at C-DAC Roof-Top Pavilion |
|
Day 2: JUNE 4th (Tuesday) Technical Sessions |
|
08:30 Hrs - 09:00 Hrs : Breakfast at C-DAC Roof-Top Pavilion |
|
| 09:00 - 09:45 | Single processor optimization techniques Code restructuring techniques such as loop optimizations techniques (Loop collapsing; Loop alignment, Loop fission, Loop distribution, Loop unrolling, Loop interchange, Negatives of loop unrolling, and Loop fusion) Uni-Processor Benchmarks Performance of tuned mathematical libraries for matrix computations such as BLAS I, II, and III Download presentation slides : pcopp02-Code-Restructure.pdf |
| 10:00 - 11:00 | Models of Parallel Computers, An Overview of PARAM 10000 and Performance of Parallel Computers Download presentation slides : pcopp02-Parallel-Computing.pdf and pcopp02-PARAM10000.pdf |
11:00 Hrs - 11:15 Hrs : Tea Break at C-DAC Roof-Top Pavilion |
|
| 11:15 - 12:00 | Explicit Parallelism: Shared Memory Programming - Pthreads Download presentation slides : pcopp02-Pthreads.pdf |
| 12:15 - 13:00 | Basic Performance metrics (Workload & Speed Metrics) Performance of Parallel Computers - Computational Characteristics Parallelism and interaction overheads (Overhead Quantification and measurement methods) Performance of parallel programs (Performance metrics, Scalability &Speed-up Analysis) Download presentation slides :pcopp02-Scalability.pdf |
13:00 Hrs - 14:00 Hrs : Lunch Break at C-DAC Roof-Top Pavilion 16:30 Hrs - 16:45 Hrs : Tea Break at C-DAC Roof-Top Pavilion/NPSF |
|
| 14:00 - 18:30 | Hands-on Session (Day 2): Performance of serial programs for matrix computations using math libraries BLAS I, BLAS II, BLAS III; Performance of F90 programs using Sun-Performance libraries and Compliers; Parallel MPI Fortran 77/C/F90 programs on vector-vector, matrix-vector and matrix-matrix multiplication algorithms; Solution of matrix system of linear equations by Direct/Iterative Methods; Parallel Programs for Partial differential equations using proper MPI library calls and their impact on performance; Demonstration of benchmarks on one node of PARAM 10000 |
| 18:30 Hrs - 18:45 Hrs : Tea Break at C-DAC Roof-Top Pavilion | |
| 19:30 - 20:00 | Popular Lecture: Grid Computing |
20:15 Hrs - 22:00 Hrs: Banquet Dinner |
|
| 4:30 PM-6:00 PM | Hands-on Session:
in Lab-1, & Lab-2 Performance Issues: PAPI - tool on Multi Cores; compiler Optimization Techniques; Example programs using Multi-Threaded I/O, MPI-OpenMP, MPI-Pthreads |
Day 3: JUNE 5th (Wednesday) Technical Sessions |
|
08:30 Hrs - 09:00 Hrs Registration and Breakfast at C-DAC Roof-Top Pavilion |
|
| 09:00 - 09:45 | Explicit Parallelism: Data Parallel Programming - F90/F95/HPF Performance issues for F90/F95 & Case Studies Download presentation slides :pcopp02-F90.pdf |
| 09:45 - 10:30 | Explicit Parallelism: Message Passing Interface - Advance features of MPI Download presentation slides :pcopp02-Adv-MPI.pdf |
| 10:30 - 11:00 | Parallel Paradigms and Programmability Programming Models - Implicit Parallelism Download presentation slides :pcopp02-Paradigms-Programmability.pdf |
11:00 Hrs - 11:15 Hrs : Tea Break at C-DAC Roof-Top Pavilion |
|
| 11:15 - 12:00 | Explicit Parallelism: Shared Memory Programming - Advanced features of OpenMP Mixed mode programming MPI - OpenMP Download presentation slides :pcopp02-Adv-OpenMP.pdf |
| 12:15 - 13:00 | An Overview of Day-3 and Day-4 Hands-on Session |
13:00 Hrs - 14:00 Hrs : Lunch Break at C-DAC Roof-Top Pavilion 16:30 Hrs - 16:45 Hrs and 18:30 Hrs - 18:45 Hrs: Tea Break at C-DAC Roof-Top Pavilion/NPSF |
|
| 14:00 - 18:30 | Hands-on Session (Day 3): Performance of selective FORTRAN/C programs on PARAM 10000 using Pthreads, OpenMP and MPI on vector-vector, matrix-vector, matrix-matrix multiplication and Sparse matrix vector multiplication algorithms; Solution of matrix system of linear equations by Direct/Iterative Methods; Performance of parallel programs using combination of Pthreads and MPI; OpenMP and MPI; Examples on comparative study of Performance of parallel programs on matrix computations using different programming models -Pthreads, OpenMP, and MPI |
18:30 Hrs - 18:45 Hrs : Tea Break at C-DAC Roof-Top Pavilion |
|
| 18:45 - 19:30 | Popular Lecture : High Performance Computing in BioInformatics |
| 19:30 Hrs - 20:30 Hrs : Dinner at C-DAC Roof-Top Pavilion | |
| DAY 4 - JUNE 06 (Thursday) | |
| 08:30 Hrs - 09:00 Hrs Breakfast at C-DAC Roof-Top Pavilion | |
| 09:00 - 09:30 | Principles of Algorithms design - Decomposition techniques; Static and Dynamic load balancing features; Overheads in algorithm design and their impact on performance Download presentation slides :pcopp02-Parallel-Algorithm-Design.pdf |
| 09:30 - 10:15 | High Performance Visualization tools for Parallel Programs & Case Studies Download presentation slides :pcopp02-Perf-Tools.pdf |
| 10:15 - 11:00 | Performance of Application and System Benchmarks PARAM 10000 Download presentation slides :pcopp02-System-Benchmarks.pdf |
11:00 Hrs - 11:15 Hrs : Tea Break at C-DAC Roof-Top Pavilion |
|
| 11:15 - 12:15 | An Overview of Performance of Parallel Programs on Parallel Computers Download presentation slides :pcopp02-Parallel-Programs.pdf |
| 12:15 - 13:00 | Executive Director Remarks, Open Session, Feedback and Conclusions |
13:00 Hrs - 14:00 Hrs : Lunch Break at C-DAC Roof-Top Pavilion 16:30 Hrs - 16:45 Hrs and 18:30 Hrs - 18:45 Hrs: Tea Break at C-DAC Roof-Top Pavilion/NPSF |
|
| 14:00 - 19:00 | Hands-on Session (Day 4): Performance of parallel programs using combination of Pthreads and MPI; OpenMP and MPI; Performance of MPI parallel programs for matrix computations using different decomposition techniques from algorithm point of view; Demonstration of MPI parallel programs using parallel visualization tools; Demonstration of MPI parallel programs to measure communication overheads on clusters using MPI; Demonstration of Performance for selective application and system benchmarks (NAS, ScaLAPACK, LINPACK) on PARAM 10000 using different networks (Fast Ethernet), C-DAC High Performance Computing Software (HPCC) and optimized Message Passing Library (MPI) |
| 19:30 Hrs - 20:30 Hrs : Dinner at C-DAC Roof-Top Pavilion | |
Closure of PCOPP- 2002 |
|