Adaptive Spatial Access Prefetcher for L2 Cache Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current prefetching technologies face challenges in achieving optimal timeliness, coverage, and accuracy while maintaining low resource requirements, particularly in predicting addresses for L2 cache prefetching, leading to inefficiencies in processor performance due to high latency in memory access.
Innovation Solution
The adaptive spatial access prefetcher (ASAP) optimizes prefetching by decoupling storage requirements from signatures, using a granularity selector to compress prefetch patterns from 64B to 128B or 256B, and employing prediction tables to generate prefetches at the earliest trigger points, thereby enhancing timeliness and coverage while maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional prefetching technologies are used to predict addresses for L2 cache, then memory access latency is reduced, but resource requirements (storage, complexity) increase while achieving suboptimal coverage and accuracy
Solution Approach 1:
The patent changes the fundamental parameters of prefetching by transitioning from signature-based prediction to delta-based prediction. This parameter change enables the system to achieve higher coverage and accuracy with reduced storage requirements, as deltas require minimal storage compared to full signatures, while still effectively predicting memory access patterns
Solution Approach 2:
The patent segments the memory access pattern prediction into multiple independent delta calculations (first delta, second delta, third delta) rather than relying on a single comprehensive signature. This segmentation allows the system to track different aspects of access patterns separately, improving overall prediction accuracy while using less resources than storing complete signatures
2Measurement precision
If prefetch patterns are stored at fine granularity (64B), then prediction accuracy is improved, but storage requirements increase
Solution Approach 1:
The patent changes the granularity parameter from fine-grained (64B) to coarse-grained (128B or 256B) prefetch patterns. This parameter change reduces storage requirements significantly while maintaining prediction accuracy through the use of multiple delta calculations that collectively capture the same level of detail as fine-grained patterns would provide
3Measurement precision
If prefetches are generated at later trigger points, then accuracy is improved, but timeliness deteriorates
Solution Approach 1:
The patent performs preliminary action by generating prefetches at the earliest possible trigger points (first load miss, second load miss, third load miss) rather than waiting for later confirmations of access patterns. The use of delta-based prediction allows accurate prefetch generation early in the access sequence, achieving both timeliness and accuracy simultaneously
Data Source
Figure 1A
Figure 1B
Figure 2A~2C
AI summary
An apparatus and method for adaptive spatial accelerated prefetching. For example, one embodiment of an apparatus comprises: execution circuitry to execute instructions and process data; a Level 2 (L2) cache to store at least a portion of the data; and a prefetcher to prefetch data from a memory subsystem to the L2 cache in anticipation of the data being needed by the execution unit to execute one or more of the instructions, the prefetcher comprising a buffer to store one or more prefetched memory pages or portions thereof, and signature data indicating detected patterns of access to the one or more prefetched memory pages; wherein the prefetcher is to prefetch one or more cache lines based on the signature data.