Supercomputer Fugaku User Guide - Use and job execution - Version 1.53

Index:

• 1. Introduction
  • 1.1. The purpose of this document.
  • 1.2. Notation used in this document
  • 1.3. Abbreviations and aliases
  • 1.4. About trademarks
  • 1.5. Change log
• 2. Use direction
  • 2.1. Overview
  • 2.2. Manual
• 3. Usage rule
  • 3.1. Overview
  • 3.2. Use scale/Use environment
    • 3.2.1. Usable job scale
  • 3.3. Local account management
    • 3.3.1. Login
    • 3.3.2. Authentication by client certification
    • 3.3.3. System project manager responsiblity scope
    • 3.3.4. Update process
  • 3.4. Resource
    • 3.4.1. Compute node
    • 3.4.2. Confirmation of number of job submission
    • 3.4.3. How to use resource group
    • 3.4.4. Resource group use status
    • 3.4.5. Disk
    • 3.4.6. File creation and stripe setting
    • 3.4.7. File system
    • 3.4.8. Group
    • 3.4.9. Type of parallel job
    • 3.4.10. Login node
    • 3.4.11. I/O node
    • 3.4.12. Estimating the amount of memory available to user programs
    • 3.4.13. Definition of used computational resource
• 4. System Use
  • 4.1. Overview
  • 4.2. Client Certification Installation
    • 4.2.1. Installing the certificate to Firefox (Windows)
    • 4.2.2. Installing the certificate on Firefox (Mac)
    • 4.2.3. Installing the certificate to Chrome (Windows)
    • 4.2.4. Installing the certificate to Chrome (Mac)
  • 4.3. Accessing steps to the Fugaku website
  • 4.4. Login
    • 4.4.1. Private key/Public key creation
    • 4.4.2. Public key registration
    • 4.4.3. Accessing direction
    • 4.4.4. File transfer method
    • 4.4.5. Login shell
    • 4.4.6. E-mail distribution of Fugaku operation information
    • 4.4.7. Directory and File Permission Management
  • 4.5. Check node usage
  • 4.6. Module package
    • 4.6.1. Module command
    • 4.6.2. Example of command use
    • 4.6.3. How to change language environment
• 5. Job execution
  • 5.1. Overview
    • 5.1.1. About the jobs
    • 5.1.2. Types of job
    • 5.1.3. Create job script
    • 5.1.4. The command for creating template of job script
    • 5.1.5. Job allocation operation
  • 5.2. Normal job
    • 5.2.1. Submitting a job
    • 5.2.2. Refer to job execution result
    • 5.2.3. Statistical information
  • 5.3. Step job
    • 5.3.1. Job submitting
    • 5.3.2. Step job option
    • 5.3.3. Sub job ID
    • 5.3.4. Referencing job execution results
    • 5.3.5. Job submission example with dependency
  • 5.4. Bulk Job
    • 5.4.1. Job submission
    • 5.4.2. Bulk job script
    • 5.4.3. Sub job ID
    • 5.4.4. Referencing job execution results
  • 5.5. Interactive job
    • 5.5.1. Job submission
    • 5.5.2. Command format
    • 5.5.3. Command option block
    • 5.5.4. Script file block
  • 5.6. Workflow job
    • 5.6.1. Job submission
  • 5.7. Master-worker type job
    • 5.7.1. Job Submission
    • 5.7.2. Dynamic generation of worker processes
    • 5.7.3. Worker process creation request to Agent prosess
    • 5.7.4. Worker process generation by pjaexe command
    • 5.7.5. Notes on job creation
    • 5.7.6. Impact when system failure
  • 5.8. Environment setting
    • 5.8.1. Tools to reduce search load on dynamic libraries(sort_libp)
  • 5.9. Job execution command options
    • 5.9.1. Basic option
    • 5.9.2. Resource specification
    • 5.9.3. MPI option
    • 5.9.4. Job statistical information
    • 5.9.5. Environment variables in job scripts
    • 5.9.6. Environment variables in MPI processes
    • 5.9.7. Specifying an execution start time
  • 5.10. Large page space allocation
    • 5.10.1. Overview
    • 5.10.2. Environment variable for large page library setting
    • 5.10.3. Large page library linking
    • 5.10.4. Mitigating Memory Fragmentation
  • 5.11. CPU core and memory binding
  • 5.12. Displaying a job status
    • 5.12.1. Normal job
    • 5.12.2. Step job
    • 5.12.3. Bulk job
    • 5.12.4. Computer resources
    • 5.12.5. The job statistic information
    • 5.12.6. Job status display command options
  • 5.13. Deleting a job
    • 5.13.1. Normal job
    • 5.13.2. Step job
    • 5.13.3. Bulk job
    • 5.13.4. Job delete command options
  • 5.14. Notes on job operations
    • 5.14.1. Delete job (pjdel)
    • 5.14.2. Job hold (pjhold)
    • 5.14.3. Unpin a job (pjrls)
    • 5.14.4. Wait for job (pjwait)
    • 5.14.5. Send signal to job (pjsig)
    • 5.14.6. Job parameter change (pjalter)
  • 5.15. Use of custom resource
    • 5.15.1. Function overview
    • 5.15.2. How to check
    • 5.15.3. Assign method
  • 5.16. Checking when a job ends abnormally
    • 5.16.1. Job manager exit code
    • 5.16.2. Message output during job execution
    • 5.16.3. PJM 0079 ERROR REASON list
    • 5.16.4. CPU clock change (Power capping)
  • 5.17. Sample Scripts
    • 5.17.1. Normal job (sequential program)
    • 5.17.2. Normal job (MPI / 1D shape)
    • 5.17.3. Normal job (MPI / 2D shape)
    • 5.17.4. Normal job (MPI / 3D shape)
    • 5.17.5. Normal job (thread parallel (automatic parallel))
    • 5.17.6. Normal job (thread parallel (OpenMP))
    • 5.17.7. Normal job (Flat MPI / 1D shape)
    • 5.17.8. Normal job (hybrid parallel)
  • 5.18. Obtaining TofuD TNR statistics
  • 5.19. Job list display command affected by the failure
    • 5.19.1. job_events
    • 5.19.2. show_evict_node
  • 5.20. Note
    • 5.20.1. How to use pjrsh command
    • 5.20.2. Job Execution in Startup Project (trial)
• 6. MPI job execution
  • 6.1. Overview
  • 6.2. Execution command format
  • 6.3. How to execute back ground job
    • 6.3.1. Back ground execution
    • 6.3.2. Job creation example
    • 6.3.3. Notes on back ground execution
  • 6.4. Standard output / Standard error output / Standard input
    • 6.4.1. How to specify standard output / standard error output / standard input
    • 6.4.2. Output list by option specification
    • 6.4.3. About standard output / standard error output when executing large-scale jobs
  • 6.5. Specification when submitting an MPI job
    • 6.5.1. Static process shape specification
    • 6.5.2. Specify the maximum number of static processes
  • 6.6. Rank assignment
    • 6.6.1. Rank-map-bynode
    • 6.6.2. rank-map-bynode (1 dimention)
    • 6.6.3. rank-map-bynode (2 dimentions)
    • 6.6.4. rank-map-bynode (3 dimentions)
    • 6.6.5. rank-map-bynode (the multiple process specification in node (1 dimention))
    • 6.6.6. rank-map-bychip
    • 6.6.7. rank-map-hostfile
    • 6.6.8. rank-map-hostfile (1 dimention)
    • 6.6.9. rank-map-hostfile (2 dimentions)
    • 6.6.10. rank-map-hostfile (3 dimentions)
    • 6.6.11. rank-map-hostfile (the multiple rank specification in node)
    • 6.6.12. rank-map-hostfile/rank-map-bychip (the multiple rank specification in node)
    • 6.6.13. Notes on when using host map file
  • 6.7. Job shape (Torus)
  • 6.8. Job submitting command option (MPI)
  • 6.9. Dynamic process creation (MPI_COMM_SPAWN)
    • 6.9.1. How to execute a job
    • 6.9.2. The execution from Java program
    • 6.9.3. Notes
  • 6.10. Rank inquiry functions
    • 6.10.1. Inquiry function of logical dimention number and logical form specified to MPI_COMM_WORLD
    • 6.10.2. Inquiry function of logical coordinates of MPI program by rank number
    • 6.10.3. Inquiry function of rank number of MPI program by logical coordinates
    • 6.10.4. Function use example
  • 6.11. MPI statistic information
    • 6.11.1. Whole mode
    • 6.11.2. Section mode
  • 6.12. Collective Communication Algorithms(knomial, rabenseifner) in MPI
    • 6.12.1. How to Specify Algorithms
• 7. Power control function
  • 7.1. Overview
  • 7.2. Power mode (Power knob operation at job submisson)
    • 7.2.1. Function overview
    • 7.2.2. Use example
  • 7.3. Power API
    • 7.3.1. Function overview
    • 7.3.2. How to create Power API program
    • 7.3.3. Use direction Power API from within the program
    • 7.3.4. Electric power measurement point
    • 7.3.5. Power control point
    • 7.3.6. Note
• 8. Layered storage
  • 8.1. Overview
    • 8.1.1. IO time reduction and area selection
    • 8.1.2. Simultaneous access to common files from all processes
  • 8.2. About File Operations to FEFS/LLIO
    • 8.2.1. When writing from one process to one file
    • 8.2.2. When multiple processes write to a file
    • 8.2.3. Rename File
    • 8.2.4. Unlink File
  • 8.3. Cache Area of Second-Layer Storage
    • 8.3.1. The Cache Area of Second-Layer Storage size
    • 8.3.2. Writing timing to second-layer storage
    • 8.3.3. Stripe
    • 8.3.4. Asynchronous close / synchronous close
    • 8.3.5. Common file distribution function (llio_transfer)
    • 8.3.6. Option when Job submitting (pjsub --llio)
  • 8.4. Node Temporary Area
    • 8.4.1. Node temporary area size
    • 8.4.2. How to use
    • 8.4.3. Job submitting option (pjsub --llio)
    • 8.4.4. Option related to cashe in compute node
    • 8.4.5. Usage example of node temporary area
  • 8.5. Shared temporary area
    • 8.5.1. Stripe setting for shared temporary area
    • 8.5.2. Shared temporary area size
    • 8.5.3. How to use
    • 8.5.4. Job submitting option (pjsub –llio)
  • 8.6. Important Notices
    • 8.6.1. Notes on High Parallel Jobs (1000 or more parallel)
    • 8.6.2. MPI-IO
  • 8.7. I/O profiling
    • 8.7.1. LLIO performance information
    • 8.7.2. Darshan
  • 8.8. I/O optimization
    • 8.8.1. Analysis of bottlenecks using LLIO performance information
    • 8.8.2. Analysis for Darshan
  • 8.9. Selecting a usage file system (volume)
    • 8.9.1. Environment Variables and pjsub Options
    • 8.9.2. Job submission method
• 9. Migration from other systems
  • 9.1. Overview
  • 9.2. Migration from K
  • 9.3. Migration from other Fujitsu systems
    • 9.3.1. Option difference
    • 9.3.2. Optimization indicator (OCL) difference
  • 9.4. Migration from other systems
  • 9.5. Endian
    • 9.5.1. Overview
    • 9.5.2. Target compiler