Multi-Socket Memory Decompression for Physically Scattered Buffers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current signal processing systems experience latency issues due to the inefficiency in decompressing large physically scattered data across a multi-socket symmetric multiprocessing architecture, as existing methods do not account for the physical layout of data and processors, leading to longer decompression times.
Innovation Solution
The method offloads scattered fragments of data to local hardware accelerators based on their physical proximity, using a decompression agent to identify the compression method and assign data to the closest accelerator, thereby reducing latency and clock cycles required for decompression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is decompressed using traditional CPU-based methods, then system compatibility is maintained, but decompression time and latency increase significantly
Solution Approach 1:
The patent segments the decompression task by dividing scattered data fragments across multiple sockets and assigning them to local accelerators. Each socket processes its local fragments independently through dedicated hardware accelerators, then results are combined to form the complete decompressed data, resolving the contradiction between maintaining system-wide compatibility and achieving fast decompression.
Solution Approach 2:
The patent introduces hardware accelerators as intermediary components between the CPU and memory. These accelerators handle the decompression operations locally at each socket, acting as mediators that offload time-consuming decompression tasks from the main CPU while maintaining system compatibility through standardized interfaces.
2Device complexity
If scattered data fragments are processed sequentially by a single processor, then system complexity is minimized, but productivity and decompression throughput decrease
Solution Approach 1:
The patent divides the processing workload into segments distributed across multiple sockets, each with its own local accelerator. This segmentation enables parallel processing of scattered data fragments, significantly improving decompression throughput while keeping each individual socket's complexity manageable through modular design.
Solution Approach 2:
The patent merges multiple local accelerator operations across different sockets into a coordinated parallel processing system. By combining the capabilities of multiple accelerators working simultaneously on different data fragments, the system achieves high throughput decompression without requiring complex centralized control.
3Quantity of substance
If data is stored in physically scattered locations across memory, then memory utilization is optimized, but access time and decompression latency increase
Solution Approach 1:
The patent applies local quality by assigning data fragments to local accelerators based on their physical location in memory. Each socket's accelerator processes data stored in its local memory space, ensuring that data access and processing occur in the same physical location, thereby optimizing both memory utilization and access time.
Solution Approach 2:
The patent implements preliminary action by pre-assigning scattered data fragments to specific local accelerators based on their memory location. This pre-establishment of data-accelerator mappings ensures that when decompression is needed, the data is already positioned for optimal local processing, reducing access latency while maintaining efficient memory utilization.
Data Source
AI summary
Aspects of the invention include identifying a first subsystem and a second subsystem of a plurality of subsystems respectively storing a first compressed data and a second compressed data, wherein the first compressed data and the second compressed data are fragments of a requested data. A compression method used to compress the first compressed data and second compressed data is identified. A first accelerator of first subsystem and a second accelerator of the second subsystem is identified. The first compressed data from a first local memory of the first subsystem is offloaded to the first accelerator, and the second compressed data from a second local memory of the second subsystem is offloaded to the second accelerator, wherein offloading comprises provided a decompression method for the first compressed data and the second compressed data.