- Infos im HLRS Wiki sind nicht rechtsverbindlich und ohne Gewähr -
- Information contained in the HLRS Wiki is not legally binding and HLRS is not responsible for any damages that might result from its use -
Programming How-To's, Tips & Tricks: Difference between revisions
From HLRS Platforms
Jump to navigationJump to search
No edit summary |
|||
| (4 intermediate revisions by the same user not shown) | |||
| Line 15: | Line 15: | ||
=== Debugging === | === Debugging === | ||
* [https://kb.hlrs.de/platforms/index.php/GDB#Obtaining_a_backtrace_of_a_crashing_MPI_program Using GDB to obtain a backtrace of a crashing MPI program] | * [https://kb.hlrs.de/platforms/index.php/GDB#Obtaining_a_backtrace_of_a_crashing_MPI_program Using GDB to obtain a backtrace of a crashing MPI program] | ||
=== Big Data, AI Applications, and Frameworks === | |||
* Containers: [[How to use AI containers on GPU-accelerated compute partitions?]] | |||
* Conda: [[How to use Conda environments on the clusters]] | |||
* Web-Based Services: [[How to use Web Based Services on HLRS Compute Platforms]] | |||
* Ray: [https://code.hlrs.de/SiVeGCS/ray_template How to launch a Ray Cluster on Hawk?] | |||
=== Miscalleneous === | === Miscalleneous === | ||
Latest revision as of 08:41, 11 April 2024
Project Management and Collaboration
General
- POSIX compliant shell scripting
- Best practices for I/O, Parallel I/O, aka MPI-IO
- Best practices for Porting & Migration of Code
- Best practices for Debugging
- Best practices for Performance Analysis
Performance Optimization
- Useful compiler options on x86 processors
- Pinning threads and memory on multicore/processor system under Linux
Debugging
Big Data, AI Applications, and Frameworks
- Containers: How to use AI containers on GPU-accelerated compute partitions?
- Conda: How to use Conda environments on the clusters
- Web-Based Services: How to use Web Based Services on HLRS Compute Platforms
- Ray: How to launch a Ray Cluster on Hawk?
Miscalleneous
- SX and configure
- PACX-MPI on clusters
- How to measure Flop/s etc. of massively parallel jobs with perf stat
- How to detect out of memory events