Table of Contents

1. Introduction
1.1. A Guide to Using this Document
2. Hardware Considerations
2.1. Hardware Terms and Definitions
2.2. CPU, Memory, and I/O Balance
3. Understanding Mechanical APDL Computing Demands
3.1. Computational Requirements
3.1.1. Shared Memory Parallel vs. Distributed Memory Parallel
3.1.2. Parallel Processing
3.1.3. GPU Accelerator Capability
3.2. Memory Requirements
3.2.1. Specifying Memory Allocation
3.2.2. Memory Considerations for Parallel Processing
3.3. I/O Requirements
3.3.1. I/O Hardware
3.3.2. I/O Considerations for Distributed Memory Parallel Processing
4. Memory Usage and Performance
4.1. Linear Equation Solver Memory Requirements
4.1.1. Direct (Sparse) Solver Memory Usage
4.1.2. Iterative (PCG) Solver Memory Usage
4.1.3. Modal (Eigensolvers) Solver Memory Usage
5. Parallel Processing Performance
5.1. What Is Scalability?
5.2. Measuring Scalability
5.3. Hardware Issues for Scalability
5.3.1. Multicore Processors
5.3.2. Interconnects
5.3.3. I/O Configurations
5.3.4. GPUs
5.4. Software Issues for Scalability
5.4.1. Program Architecture
5.4.2. Distributed ANSYS
5.4.3. GPU Accelerator Capability
6. Measuring Performance
6.1. Understanding Overall Performance
6.2. Understanding Solver Performance
6.3. Sparse Solver Performance Output
6.4. Distributed Sparse Solver Performance Output
6.5. PCG Solver Performance Output
6.6. PCG Lanczos Solver Performance Output
6.7. Supernode Solver Performance Output
6.8. Identifying CPU, I/O, and Memory Performance
6.9. Solver Performance Guidelines
6.9.1. Sparse Direct Solver
6.9.2. PCG Iterative Solver
A. Glossary

Release 18.2 - © ANSYS, Inc. All rights reserved.