Scalable software tools to support the optimization of applications for HPC systems

Screenshot Performance-Analyse-Softwaretool

Performance optimization and analysis tools are increasingly important for scientists and code developers given the increasing the complexity of parallel computers and applications. These tools are the foundation for the performance optimization and analysis of application codes. The availability and use of a broad range of scalable, user-friendly, simple to handle software tools can save an enormous amount of time in the development of applications and the diagnosis of their behavior, in view of achieving a more efficient utilization of the computing resources. ZIH has a long experience in event-based monitoring and analysis of program execution traces with Nvidia CUDA, OpenCL or AMD ROCm.

Current Key Topics and Projects

Vampir: Visualization and analysis of parallel applications
Score-P: Scalable performance measurement infrastructure for parallel codes
Perf: System and application tracing on Linux
Virtual Institute - High Productivity Supercomputing (VI-HPS): Integrated programming tools for high-performance computing
Prestige: Development, application and validation of innovative multi-disziplinary simulation and design methods for an improved understading of airplane turbines

Completed Projects

CRESTA: Collaborative Research into Exascale Systemware, Tools and Applications
DataHeap: Embeding external performance data in program traces
ECOUSS: Efficient open compiler environment for semantically annotated parallel simulationen
eeClust: Energy efficient cluster computing
ELP: Efficient Runtime Support for Future Programing Standards
EP-Cache
H4H: Optimise HPC applications on heterogeneous architectures
HI-CFD: Hocheffiziente Implementierung von CFD-Codes für HPC-Many-Core-Architekturen
HOlistic Performance System Analysis (HOPSA): Diagnostic infrastructur for combined application and system tuning
LMAC: Performance dynamics of massively parallel codes
Marmot / UniMCI: Runtime correctness checking of MPI applications
MUST: Scalable runtime correctness checking for MPI application
NextGenIO: Next Generation I/O for Exascale
Open Trace Format (OTF): Scalable file format for performance event data
ParMA: Parallel programming for multi-core architectures
ProPE: Process-Oriented Performance Engineering Service Infrastructure
for Scientific Software at German HPC Centers
Scout: A Source-to-Source Transformator for SIMD-Optimizations
SILC: Scalable infrastructure for automated performance analysis of parallel codes
VampirTrace: Performance data monitor for HPC applications