Data processing system, method of operating data processing system, and computer-readable medium

Abstract

Data processing systems, methods of operating data processing systems, and computer-readable media. A data processing system comprising one or more processors (4, 5, 6, 7) operable to initiate an atomic memory request of a thread of execution and a processor configured to store the atomic memory operation when the atomic memory operation is to be performed Multiple data caches (8, 9, 10, 11) for required data. When an atomic operation is to be performed against a data cache, the results of the atomic operation to access the same memory location are accumulated in a temporary cache in the data cache until the "real" cache line from memory arrives in the cache in line. Then, when the cache line arrives in the cache, the cumulative result of the atomic operation stored in the temporary cache line is combined with the cache line from memory. A single atomic value can also be reconstructed once the cache line arrives in the cache.

Description

technical field [0001] The present invention relates to data processing systems, and in particular, to the operation of graphics processing systems that include one or more programmable processing stages ("shaders"). Background technique [0002] As is known in the art, graphics processing is typically performed in a pipeline, where one or more pipeline stages operate on data to generate a final rendered output, eg, a frame that is displayed. Many graphics processing pipelines now include one or more programmable processing stages, commonly referred to as "shaders." For example, a graphics processing pipeline may include one or more, and often all, of geometry shaders, vertex shaders, and fragment (pixel) shaders. These shaders are programmable processing stages that execute shader programs against input data values ​​to produce the desired output data set (e.g., in the case of vertex shaders, transforming and lighting vertex data appropriately) for the graphics pipeline T...

AI Technical Summary

Problems solved by technology

[0027] It can be seen that in this operation, if a large number of listener requests are received while performing atomic operations, the throughput of atomic operations can be significantly delayed

Images