Product Overview

• You agree to comply with national laws, regulations, and public ethics when using the Kunpeng BoostKit OmniOperator JIT. You shall not use OmniOperator JIT to engage in any activities that violate the law, infringe on the rights and interests of others, interrupt social order, undermine social stability, or engage in any activities that endanger or attempt to endanger the computer system and network security.
• You agree that Huawei shall assume no liability for any indirect, incidental, special, or any form of punitive damages, or any loss of profits, revenue, data, or data use.
• You acknowledge and agree that you need to download and integrate the open source and third-party software on which the software package of OmniOperator JIT depends. Huawei does not assume any responsibility for the software vulnerabilities and security issues.

Conventional big data SQL engines involve a large number of repeated optimizations. To build competitiveness, the Kunpeng BoostKit OmniOperator uses one base to support different engines, reducing repeated optimization work, fully exploring common and heterogeneous computing power, and promoting the Kunpeng ecosystem.

OmniOperator JIT:

• Implements high-performance OmniOperators by using native code. It fully exploits the computing capabilities of hardware, especially the heterogeneous computing power. Compared with Java and Scala operators, OmniOperators greatly improve the performance of the compute engine.
• Uses data-driven optimized Just-In-Time (JIT) OmniJit to automatically identify the most performance-critical code sections in operators and optimizes them based on real-time data features to achieve higher performance. OmniJit works with OmniOperator to enable optimization policies such as parameter fixing, loop unrolling, and vectorized adaptive execution for some core methods, improving the performance of the compute engine.
• Uses OmniVec to organize data in an efficient way. OmniVec defines a column-oriented storage mode independent of languages and uses off-heap memory. OmniVec can read data with zero copy, and there is no serialization overhead, so that users can process data in the memory more efficiently.