Kunpeng HPC Progressive Knowledge

Master core concepts and navigate modern computational challenges.

Deep dive into the functional framework and technical design.

Execute the end-to-end process for setup, benchmarking, porting, and optimization.

Prepare the essential foundation for porting HPC industry applications to the Kunpeng computing platform.

Deploy the GNU development toolchain, featuring the GCC compiler, assembler, and linker.

Configure the high-throughput computing framework to parallelize compute-intensive tasks.

Implement the parallel file system designed for large-scale clusters and cluster computing.

Integrate the community-driven FOSS toolset for Linux-based HPC environments.

Deploy the high-performance message-passing library to enable efficient parallel processing.

Utilize the open-source, scalable cluster management and job scheduling system for Linux clusters of any scale.

Measure MPI performance benchmarks across various message sizes for point-to-point and collective communication operations.

Evaluate parallel I/O performance at POSIX and MPI-IO levels by measuring read/write throughput under diverse conditions.

Assess practical memory bandwidth using industry-standard metrics, including Copy, Scale, Add, and Triad operations.

Port key weather and climate modeling applications, including WRF, CAMx, CESM, NEMO, and ROMS.

Port industrial engineering and simulation software such as OpenFOAM, SU2, PALABOS, deal.II, and AMG.

Port essential scientific computing tools and libraries, including UDUNITS, GROMACS, LAMMPS, QUANTUM ESPRESSO, and NAMD.

Port core genomic and bioinformatics applications such as BLAST, BWA, CNVnator, Cufflinks, and GATK.

Access porting guides for versatile tools such as Hadoop, SGE, Conda, NPB, and double-conversion.

Leverage the Kunpeng processor's NUMA architecture for optimized system performance.

Fine-tune system efficiency by configuring memory, high-speed network interfaces, and server BIOS settings.

Enhance stability and throughput by configuring OS kernels, adjusting page sizes, and enabling huge pages.

Boost data throughput by refining dirty data flush policies, drive prefetching, I/O scheduling, and XFS mounting.

Maximize computational efficiency through specialized compilers, math libraries, and MPI/OpenMP core pinning.

Refine software performance using parallel I/O adjustments and top-down model analysis.

Utilize dedicated on-chip accelerators for high-speed encryption/decryption and compression/decompression.