SX-4 Series General Description
Contents
Proprietary Notice
Documentation Log
Introduction
Chapter 1, Developmental Background
- 1.1 Developmental Philosophy
- 1.2 Features of the SX-4 Series
- 1.2.1 Scalable Parallel Vector Systems
- 1.2.2 SUPER-UX, an Enhanced UNIX-Based Operating System
- 1.2.3 Software Development Environment Vector and Parallel Requirements
Chapter 2, Overview of the SX-4 Series
- 2.1 SX-4 Series Applications Systems
- 2.1.1 High Performance Applications Systems
- 2.2 SUPER-UX Overview
- 2.3 Hardware Overview
- 2.3.1 System Configuration
- 2.3.2 Architecture
- 2.4 LSI Technology
- 2.4.1 CMOS LSI
- 2.4.2 Packaging and Cooling Technologies
- 2.4.3 Highly Reliable System
Chapter 3, SUPER-UX Orginazation of SUPER-UX
- 3.1 Organization of SUPER-UX
- 3.2 IOX Software
- 3.3 Initial System Loader (ISL)
- 3.4 SUPER-UX Kernel
- 3.4.1 Basic Kernel Services
- 3.4.1.1 Memory Management
- 3.4.1.2 Process Management
- 3.4.1.3 Execution Management
- 3.4.1.4 Parallel Processing Services
- 3.4.2 File Management
- 3.4.2.1 Organization of the SUPER-UX File System
- 3.4.2.2 Supercomputing File System (SFS)
- 3.4.2.3 Hybrid Supercomputing File System (SFS/H)
- 3.4.2.4 Easy-to-use File System Environment
- 3.4.2.5 Intelligent I/O Accelerator Subsystem (IAS)
- 3.4.2.6 SFS and IAS Utilities
- 3.4.2.7 Network File System (NFS)
- 3.4.2.8 Tape Library
- 3.4.2.9 SX-BackStore
- 3.4.3 Batch Processing
- 3.4.4 Network Functions
- 3.4.4.1 Overview
- 3.4.4.2 Local Area Network (LAN) Connection
- 3.4.4.3 Wide-Area Network (WAN) Connection
- 3.4.4.4 Streams
- 3.4.4.5 Internet Protocol
- 3.4.4.6 OSI Protocol
- 3.4.4.7 Network Utilities
- 3.4.5 Operation Management Services
- 3.4.6 High Reliability
- 3.5 SUPER-UX Utilities
- 3.5.1 Overview
- 3.5.2 Shell
- 3.5.3 Basic Commands
- 3.5.4 BSD Commands
- 3.5.5 NQS
- 3.5.6 X-Window
- 3.5.7 Distributed Computing Environment
- 3.6 Multi-node Systems
Chapter 4, Software Development Support
- 4.1 FORTRAN90/SX
- 4.1.1 Language Specifications
- 4.1.2 Automatic Vectorization
- 4.1.3 Parallelization
- 4.1.4 Optimization
- 4.1.5 High-speed Input/Output Features
- 4.1.6 XMU Array Function
- 4.2 FORTRAN77/SX
- 4.2.1 Language Specifications
- 4.2.2 Automatic Vectorization, Parallelization, and Optimization Features
- 4.3 C/SX
- 4.3.1 Overview
- 4.3.2 Automatic Vectorization Features
- 4.3.3 Parallelization Features
- 4.3.4 Optimization Features
- 4.4 C++ Language
- 4.5 HPF/SX
- 4.6 MPI/SX
- 4.7 Performance Enhancement Support Tools
- 4.7.1 ANALYZER-P/SX
- 4.7.2 ANALYZER90/SX
- 4.7.3 PARALLELIZER/SX
- 4.7.4 C-ANALYZER/SX
- 4.8 Development Support Tools
- 4.9 Debugging Support Tools
- 4.10 Cross Development Environment
- 4.11 Conversion Support Tool
Chapter 5, Application Software
- 5.1 Mathematical Libraries
- 5.1.1 Advanced Scientific Library ASL/SX
- 5.1.2 Advanced Scientific Library C Language Interface ASLCINT/SX
- 5.1.3 Advanced Scientific Library External Memory Extension ASLEME/SX
- 5.1.4 Mathematical Library MATHLIB/SX
- 5.2 Graphics Software
- 5.2.1 Graphical Kernel System (GKS)
- 5.2.2 Three-Dimensional Graphics System (PHIGS PLUS)
- 5.2.3 Visual Simulation System (SXview)
- 5.3 Molecular Science Software
- 5.3.1 Ab initio Molecular Orbital Calculation System (AMOSS/SX)
- 5.4 Flow Simulation Software
- 5.4.1 ThreeDimensional Real-Time Flow Simulation System (
-FLOW/SX)
- 5.5 Third-party Software
Chapter 6, Hardware
- 6.1 Central Processing Unit
- 6.1.1 Configuration of Central Processing Unit
- 6.1.2 Vector Unit
- 6.1.3 Scalar Unit
- 6.2 Main Memory Unit
- 6.3 Extended Memory Unit
- 6.4 Internode Crossbar Switch
- 6.5 Input/Output Processor
- 6.5.1 Input/Output Processor Configuration
- 6.5.2 Maximum Number of Channels
- 6.6 Input/Output Multiplexer
- 6.7 Automatic Operation Controller
- 6.8 Peripheral Devices
- 6.8.1 High Speed RAID Disk Subsystem
- 6.8.2 SCSI Disk Units
- 6.8.3 Tape Systems
- 6.8.4 Network Facilities
Appendix A, SX-4 Series Specifications
Index
Figures
- 1-1 Configuration of a Single-node Model
- 1-2 Configuration of a Multi-node Model
- 1-3 Detailed Configuration of the SX-4 Processor
- 2-1 Features of SUPER-UX
- 2-2 Configuratoin of a Single-nde System (Maximum Configuration)
- 2-3 Configuration of an H type (Maximum Configuration)
- 2-4 Configuratoin of an M type (Maximu Configuration)
- 2-5 Parallel Processing
- 2-6 Reducing I/O Processing Time Through the Use of an Extended Memory Unit
- 2-7 Vectorization Ratio and the Effect of Vectorization
- 2-8 Various Vector Processing Fucntions
- 2-9 Effect of Parallel Execution of Vector and Scalar Instructions
- 2-10 Benefits of a Superscalar Architecture
- 2-11 Summary of Architecture-driven Performance Improvements
- 2-12 Basic Data Formats
- 2-13 Principal Instruction Formats
- 2-14 Effective Address for a Vector Function
- 2-15 Effective Address for Scalar Instruction
- 2-16 Generating an Absolute Address
- 2-17 LSI
- 2-18 External View of an LSI Chip
- 2-19 Bi-CMOS Static RAM
- 2-20 CPU Card
- 2-21 MMU Card
- 3-1 SUPER-UX Software System
- 3-2 Organization of the SUPER-UX File System
- 3-3 File Hierarchy
- 3-4 Configuration of i List and i Node
- 3-5 Conceptual Diagram of a Virtual Volume
- 3-6 Virual Volume Cache Configuration
- 3-7 Parallel Input/Output Service
- 3-8 Sample Network Configuration Using FDDI
- 3-9 Example ATM Network
- 3-10 Example HIPPI Network
- 3-11 Example Wide Area Network
- 3-12 Protocols Supported by SUPER-UX
- 3-13 SUPER-UX API Support
- 3-14 OSI Architecture
- 3-15 SUPER-UX Distributed Processing
- 3-16 SUPER-UX RAS Features
- 3-17 Source and Destination Redirection
- 3-18 Sample Batch Request
- 3-19 Example NQS Configuration
- 3-20 Example NQS Status Display
- 3-21 NQS Load Distribution Feature
- 3-22 X-Windos System Configuration
- 3-23 DCE Architecture
- 4-1 Vector Types
- 4-2 Vector Gathering and Scattering
- 4-3 Flow of Asynchronous Input/Output
- 4-4 Processing Flow of XMU Array Function
- 4-5 Data Distribution in HPF
- 4-6 HPF Program Example
- 4-7 Data Distribution in Terms of Nodes
- 4-8 Data Distribution in Terms of CPUs
- 4-9 MPI Overview
- 4-10 Differences in Performance According to Vectorization Ratio
- 4-11 Relationship Between Vectorization Ratio and Performance Improvement Ratio
- 4-12 Steps for Performance Improvement by Vectorization
- 4-13 Example of a Program Structure List Report by ANALYZER-P/SX (Static Analysis)
- 4-14 Example of Macrotasking Cross Reference List Report by ANALYZER-P/SX (Static Analysis)
- 4-15 Example of a Summary List Report by ANALYZER-P/SX (Dynamic Analysis)
- 4-16 Example of a Macrotasking Summary List Report by ANALYZER-P/SX (Dynamic Analysis)
- 4-17 Example of a Program Unit List Report by ANALYZER-P/SX (Dynamic Analysis)
- 4-18 Example of a Calling Patch List Report by ANALYZER-P/SX (Dynamic Analysis)
- 4-19 Example of a Edited Program List Report by ANALYZER-P/SX (Dynamic Analysis)
- 4-20 Example of a Program Unit Summary List Report by ANALYZER-P/SX
- 4-21 Example of a Program Unit Detail List Report by ANALYZER-P/SX
- 4-22 Relationship Between Compilers and Performance Analysis Tools
- 4-23 PARALLELIZER/SX Display Example
- 4-24 Summary List Report Example (Execution Time Analysis)
- 4-25 Function Summary List Report Example (Execution Time Analysis)
- 4-26 Portion of Sample Report of an Edited Program List (Execution Frequency Analysis)
- 4-27 Flow of CONVERTER/SX
- 6-1 Vector Operation Pipelines
- 6-2 Operation Pipelines and Automatic Chaining
- 6-3 Array Data and its Placement in Main Storage
- 6-4 Scalar Unit Configuration
- 6-5 Effects of Scalar Operatoin Pipelining
- 6-6 Character of Iteration Loops and Branch Instructions
- 6-7 Single-node Main Memory Unit Configuration
- 6-8 Configuration of Extended Memory Unit
- 6-9 Input/Output Processor Configuration
- 6-10 Input/Output Processor Configuration by Model Group
- 6-11 Automatic Operation Controller Configuration
Table
- 2-1 Specifications for Compact Models
- 2-2 Specifications for Single-node Models (Representative Models)
- 2-3 Specifications for Mult-node Models (Representative Models)
- 2-4 Principal Vector Instructions
- 4-1 Outline of FORTRAN90/SX Automatic Vectorization
- 4-2 Outline of Automatic Parallelization
- 4-3 Basic Conditions for the Automatic Vectorization Features of the C/SX Compiler
- 4-4 Features for Vectorizing Loops that Do Not Statisfy Basic Conditions
- 4-5 Features that Improve the Efficiency of Generated Vector Instructions\
- 4-6 Outline of C/SX Parallelization
- 4-7 Outline of the Optimization Features of the C/SX Compiler
- 6-1 Specifications of Main Memory Unit
- 6-2 Specifications of Extended Memory Unit