Method for obtaining maximum continuous resource block of GPU

By acquiring and processing GPU resource state sequences in parallel and combining them with a mapping table to quickly obtain the maximum number of contiguous resource blocks, the problem of uneven GPU resource scheduling is solved, thereby improving GPU computing performance and task processing efficiency.

CN116450332BActive Publication Date: 2026-05-29METAX INTEGRATED CIRCUITS (SHANGHAI) CO LTD

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
METAX INTEGRATED CIRCUITS (SHANGHAI) CO LTD
Filing Date
2022-01-10
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve balanced resource scheduling when GPUs execute tasks, leading to resource waste and decreased computing performance, especially inefficient in complex computing tasks.

Method used

By acquiring the GPU resource state sequence in parallel, performing bitwise AND-OR operations and OR-AND-NOT operations, and combining this with a preset mapping table, the maximum number of contiguous resource blocks can be quickly obtained, thus optimizing resource allocation.

Benefits of technology

It improves the efficiency and accuracy of resource scheduling, reduces resource waste, and enhances the computing performance and task processing efficiency of GPUs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116450332B_ABST
    Figure CN116450332B_ABST
Patent Text Reader

Abstract

The application relates to a GPU maximum continuous resource block acquisition method, which comprises the following steps: D1, reading a current resource state sequence S0={d1, d2,...d N},d n is the state identifier of the nth resource block of the resource to be searched, the value range of n is 1 to N, N is the total number of resource blocks of the resource to be searched; D2, acquiring state sequences S1, S2,...S N‑1 in parallel, wherein S i is a sequence obtained by moving S0 by i bits in a preset direction and setting the last i bits of the moved S0 in the preset direction as occupied identifiers, the value range of i is 0 to N-1; D3, acquiring S i in parallel, and performing bitwise AND operation or bitwise OR operation on S i ; D4, performing self-OR operation or self-AND operation on each SA i , and determining the current maximum continuous resource block number of the resource to be searched. The application can quickly and accurately acquire the current maximum continuous resource block number of the resource to be searched, and improves the resource scheduling efficiency.
Need to check novelty before this filing date? Find Prior Art