24.0.T50

Kunpeng DevKit

• Changed the validity period of the server certificate to 10 years.
• Added support for CTyunOS 22.06.

DevKit-Affinity-Analyzer

• Capability in the command line tool: precision analysis of the Fortran language and Fortran code instrumentation on x86 and Arm.

• Capability of scanning BC files generated by Clang 16 in vectorization check and static memory consistency check.

• Seven domain optimization technologies used in SME for C/C++.
  • Cut-off radius branch elimination
  • Iterative calculation and lookup splitting
  • Inter-particle forces iteration optimization
  • Unrolling of intermolecular forces iteration
  • Full unrolling of fixed-length loops
  • Adjacency table aggregation
  • Vectorized calculation of adjacency particle forces

DevKit-Devtools

Data compression and distributed SDKs.

DevKit-Debugger

Conditional breakpoint expressions.

DevKit-Porting-Advisor

• Capability of scanning source code of remote nodes in System Migration.
• Single-point and combined migration scenarios.
• Migration to the Kingbase database.
• Latest AVX2KI and SSE2Neon libraries for source code porting.
• Middleware migration from Tomcat to BES and automatic migration of Elasticsearch and Docker.
• Container scenario in System Migration.
• GCC for openEuler 2.3.8 and BiSheng Compiler 4.1.0 as new target compiler version options for source code porting.

DevKit-Sys-Diagnosis

• Capability of collecting memory allocation information of processes and threads in application mode of memory diagnosis.
• Automatic UTgen generation.

DevKit-Sys-Perf

• AutoTuner CLI.
  • devkit kat template command for simpler tuning configuration.
  • devkit kat train command for application tuning in user-defined scenarios.
  • devkit kat use command for result display, retesting, and one-click setting.
  • Microarchitecture parameter tuning.
  • Ten applications: MySQL, RocksDB, openGauss, Vastbase, Flink, Hive, Spark, Kafka, Redis, and PostgreSQL.
• HBM hardware included in roofline analysis.
• Python 3.7 to Python 3.11 support by Python-Perf.
• Thread-level memory access included in NUMA refined analysis.
• MPI information collection of HPC applications in the CLI.
  • MPI data collection.
  • Data table analysis.

DevKit-Java-Perf

• Capability of analyzing the performance of Java applications in pods of a namespace.
• Search box and inverted flame graph for Java hotspot flame graph analysis.
• Analysis of Java processes in Containerd.
• Capability of analyzing Parallel Scavenge (PS) GC logs.

DTS2024121703496

[CC] [Tuning with DevKit] After a Donau tenant runs the dsub command to submit a DevKit container job, the job is suspended and affects the information collection for the container job.

Minor

Use the UTF-8 encoding format instead. For other encoding formats, you can transcode the files and modify the LANG environment variable to set the encoding format to UTF-8.

The problem has not been reproduced in an environment other than Tuanbowa. If the conditions described in the problem description are met, the job is inevitably suspended. Users have to press Ctrl+C or user another method to stop the suspension. The problem has been delayed to the next version.

Further actions to be taken.

DTS2024122618065

[Kunpeng] [DevKit] When a user submits a large rank debugging process using the DonauKit in VS Code, the UI is suspended and no result is returned.

Minor

1. Release the sockets that have been used.
2. The problem with the 500-rank scenario has been resolved and the rectification has been fully tested.

1. This problem does not recur when there are no more than 500 ranks.
2. The result displayed on the UI cannot be ensured when there are more than 500 ranks.

The solution has been verified for 600 ranks, but failed the verification for more than 600 ranks due to an unknown cause. The current environment is not available.

DTS2025010211916

[System Diagnosis] [Memory usage] [Backend] The collection task has continued for three and a half hours when 120-second memory usage statistics of an HPC application are being collected using BPF Compiler Collection (BCC). The application path is /opt/test/usr/ompi/bin/mpirun, and the application parameter is --allow-run-as-root -np 2 -H xx.xx.xx.xx:28 -H xx.xx.xx.xx:28 -wdir /opt/test/lammps/RUN/airebo/ /opt/test/lammps/src/lmp_mpi -in in.tension -v model_name data.

Minor

1. Reduce the collected time of memory usage.

2. If the collected time is relatively long, use libbpf for the collection.

When the memory usage data of a large application is collected, a larger data volume requires a long collection time.

The collection duration decreases significantly but the CPU usage increases when BCC is used to collect data of a large application in multiple processes.

DTS2024123106544

[Source code porting] Due to a defect in the preprocessing module, no message is displayed by default when the source code porting function is used in the first-layer macro definition, while a message is displayed when this function is used in the second-layer macro definition.

Minor

If the functions of macro definitions on the AArch64 and x86 platforms are different, messages may not be displayed.

When the memory usage data of a large application is collected, a larger data volume requires a long collection time.

Add the macro in the incorrect preprocessing branch to the macro list in the normal conditional branch. However, for some source packages, constant expressions are incorrectly parsed due to an unknown cause.

DTS2024102218659

[VS Code] [Compiler and Debugger Plugin - HPC debugging] The variable values are abnormal. All variable values are displayed as 0 when the debugging demo is yyy_test_variable90.a.

Minor

Replace the compiler in use with the BiSheng Compiler (Flang) for compilation and debugging.

According to the analysis of 3,667 Fortran source files, 236 files define 694 derived types, among which the sizes of nine derived types exceed 3,400 bytes.

Proportion of derived types whose sizes exceed 3,400 bytes: 9/694 < 1.5%; Proportion of files containing derived types whose sizes exceed 3,400 bytes: 9/3667 < 0.3%.

According to the preceding data, there is a low probability that this problem occurs and the impact on the UI is limited.

In this scenario, all variables are set to 0, according to which users would be able to realize that there is something wrong with the debugging variable information.

This problem has been converted into a service requirement.

DTS2024112909019

[Database] [Static check] [MySQL] During a static check with the DevKit, the swap partition does not take effect, and an error message is displayed indicating that the memory space is insufficient.

Minor

Adjust the lower limit of the free memory space to 10%.

When scanning large BC files, the tool predicts that the memory space is insufficient and does not start the scan.

Further actions to be taken.

DTS2024112908199

[Database] [Static check] [MySQL] During a static check with the DevKit, it takes a long time to check a large file (more than 60 MB) and no result is generated even after 10 days.

Minor

Add the loop exit mechanism (timeout duration) to the internal logic of symbolic execution. Set the timeout duration to 12 hours.

It takes a long time to scan large files.

Further actions to be taken.