Persistent scratchpad memory for data exchange between programs

Abstract

The invention discloses a persistent scratchpad memory for data exchange between programs. Techniques are disclosed for sharing of data exchange among kernels (each a set of instructions) executing on a system having multiple processing units. In an embodiment, each processing unit includes an on-chip scratchpad memory that can be accessed by the kernels executing on the processing unit. All or a portion of the scratchpad memory can be allocated and configured, for example, such that the scratchpad is accessible to multiple kernels in parallel, to one or more kernels in serial, or a combination of both.

Description

technical field [0001] The present disclosure relates to sharing memory between programs, and more particularly, to persistent scratchpad memory for inter-program data exchange. Background technique [0002] Typically, an amount of on-chip scratch memory (e.g., 16KB, 32KB, or 64KB in size) is allocated from a scratch memory pool and assigned to a set of one or more execution contexts ("SECs") (e.g., Execution contexts are threads or processes) that execute kernels, which are sets of instructions, such as programs, in serial and / or parallel fashion. In one example, the set of one or more execution contexts is a Cooperative Thread Array (CTA). The scratchpad memory allocated to the SEC is private to the SEC, and data stored in the scratchpad memory is not persistent once the SEC completes kernel execution. In addition, the note memory does not automatically back up the memory. Thus, in conventional systems, data stored in scratch memory is resolved by having each SEC execut...

AI Technical Summary

Problems solved by technology

Exchanging data through global memory is at least an order of magnitude slower than the latency of accessing scratchpad memory

In addition, since the sticky note memory is private to each SEC, different SECs executing the same kernel must also exchange data through the global memory

Images