Library
What Are the Arm Acceleration Libraries?
The Kunpeng BoostKit Library is optimized based on Arm instructions and developed based on the Kunpeng Accelerator Engine (KAE). It covers the system library, math library, compression, encryption & decryption, media, storage, and network libraries, and provides high-performance acceleration capabilities for application scenarios such as big data encryption and decryption, distributed storage and compression, and video transcoding.
What Is the BLAS Library?
KML_BLAS, a sub-library in the Kunpeng Math Library (KML), is a basic linear algebra operation library. Based on the Kunpeng architecture, KML_BLAS provides three levels of high-performance vector operations: vector-vector, vector-matrix, and matrix-matrix. It is the backbone of numerical calculation, and is widely used in fields such as manufacturing, machine learning, and big data. KML_BLAS performs in-depth mining on the computing efficiency of Basic Linear Algebra Subprograms (BLAS) by means of vectorization, data prefetch, compilation optimization, and data rearrangement. As a result, the performance of BLAS interface functions approaches the theoretical peak value.
How to Enable the Built-in RSA Acceleration Engine of Kunpeng?
KAE is a hardware-based acceleration solution based on the Kunpeng 920 processor. It includes KAE encryption and decryption as well as KAEzip, which are used to accelerate SSL/TLS applications and data compression, respectively. Therefore, it can significantly reduce processor consumption and improve the processor efficiency. In addition, KAE shields the internal implementation details from the application layer. You can quickly migrate services by using the standard OpenSSL and zlib interfaces.
KAE can be installed using an RPM package, dpkg package, or source code. Select one of the methods as required.
How to Install the KML?
The Kunpeng Math Library (KML) provides high-performance mathematical functions which are optimized based on the Kunpeng platform. All interfaces are implemented by C/C++ and assembly languages. Some interfaces are encapsulated using Java. The KML can be installed using an RPM package or DEB package. Select one of the methods as required.
How to Check Whether the KAE License Has Expired?
The KAE license has a validity period. You can check whether the license on the current server is within the validity period on the iBMC.
If the license is invalid, apply for a new one by referring to Huawei Server iBMC License User Guide (For Customer).
How to Migrate the libm Basic Math Library to the KML?
You can specify -lkm -lm during program linking, with no need to replace the header file math.h with km.h. When you invoke a math function, if it has been optimized by the KML, the optimized function is invoked; if no, the libm basic math library is used.
Does Hardware Random Number Generation Depend on the KAE License?
KAE does not support hardware random number generation.
Is FFmpeg Optimized on Kunpeng?
No.
How Does Kunpeng BoostKit Solve the Problem of Decrypting Data Transferred from Server A to Server B?
Kunpeng BoostKit provides the Kunpeng Acceleration Engine (KAE) for encryption and decryption. You can use KAE when symmetric or asymmetric encryption is required. This engine can offload OpenSSL software computing and improve encryption and decryption performance.
For details about how to install and use KAE, see Kunpeng Accelerator Engine User Guide.
What Are the Hardware Device Requirements of the KAE License?
The license required for installing KAE is applicable to Kunpeng processors, regardless of the detailed processor or server model.
What Are the Kunpeng Server K Series Models That Support KAE?
KAE is enabled by default on Kunpeng K series servers. You do not need to apply for a license. The following lists the specific models:
- Kunpeng 916 processor: 2280K
- Kunpeng 920 processor: 2180K, 2280K, 5280K, and 2480K
Which OSs Are Compatible with KAE?
KAE has two branches: KAE 1.0 and KAE 2.0, which support kernel versions 4.19 and 5.10, respectively. For details about the supported OSs, see Environment Requirements in the Kunpeng Accelerator Engine User Guide.
Can I Call KAEzip Using gzip?
gzip is optimized based on the gzip-1.10 release and cannot directly call KAEzip.
For details about how to use KAEzip, see Kunpeng Accelerator Engine User Guide.
Can the Kunpeng 920 7260 Processor Use AVX2KI to Accelerate LLM Inference?
The Kunpeng 920 7260 processor does not integrate SVE instruction sets and cannot accelerate the inference using the AVX2KI instruction set.