Local and global data share

Abstract

A graphics processing unit is disclosed, the graphics processing unit having a processor having one or more SDVID processing units, and a local data share corresponding to one of the one or more SIMD processing units, the local data share comprising one or more low latency accessible memory regions for each group of threads assigned to one or more execution wavefronts, and a global data share comprising one or more low latency memory regions for each group of threads.

Description

technical field [0001] The present invention generally relates to optimizing data sharing between threads of execution in a graphics processing unit. Background technique [0002] A graphics processing unit (graphics processing unit; GPU) is an application-specific integrated circuit optimized for graphics processing operations. Graphics processing units (GPUs) are often included in computing devices (such as personal computers, rendering farms or servers, handheld devices, digital televisions, etc.) that execute applications with demanding graphics processing requirements, such as video game applications. . [0003] To improve processing efficiency, a Graphics Processing Unit (GPU) usually executes parallel threads using single instruction multiple data (single instruction multiple data; "SIMD", or "vector") instructions to achieve data-level parallelism. This enables a SIMD processor to execute the same instruction on multiple pixel data, for example by running independe...

AI Technical Summary

Problems solved by technology

However, data generated within any one SIMD lane is often inaccessible to other SIMD lanes without executing computationally complex and costly data storage and retrieval instructions.

Images