Accelerator Interface Task Queue Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current heterogeneous processor systems face inefficiencies due to fixed system memory allocation for accelerator function units (AFUs), requiring context switches that consume resources and leading to unbalanced task distribution among multiple AFUs, which hampers performance.
Innovation Solution
A processing system with a heterogeneous processor acceleration method that includes an accelerator interface to manage task queues in system memory, allowing AFUs to execute tasks without frequent mode switches and utilizing a round-robin method for balanced task distribution among multiple AFUs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If AFU is directly connected to bus and system memory, then memory access is simplified, but available memory space for AFU is fixed and cannot be dynamically allocated
Solution Approach 1:
The patent introduces an intermediary memory management mechanism between the AFU and system memory. The AFU accesses memory through a controlled interface that mediates between the fixed physical connection and the need for flexible virtual memory allocation, allowing dynamic space management while maintaining access simplicity.
2Reliability
If tasks are copied from user space to kernel space for AFU execution, then mode switching is enabled, but context switch operations consume excessive resources
Solution Approach 1:
The patent implements preliminary action by pre-establishing memory regions and access pathways for AFU task execution. By preparing the execution environment in advance and maintaining persistent memory mappings, the system eliminates the need for repeated context switches and resource-intensive mode transitions during task execution.
3Quantity of substance
If multiple AFUs are arranged in one chip to share among several applications, then hardware resource utilization increases, but task distribution becomes unbalanced without coordination
Solution Approach 1:
The patent introduces feedback mechanisms for task distribution management. The system monitors and tracks task allocation across multiple AFUs, using this feedback information to dynamically adjust task distribution strategies. This ensures balanced workload allocation while maintaining high hardware resource utilization efficiency.
4Reliability
If AFU executes tasks in kernel mode, then execution authority is enhanced, but task creation in user mode creates mode transition overhead
Solution Approach 1:
The patent applies preliminary action by preparing task definitions and execution parameters in user mode before the actual execution begins. The system pre-allocates memory regions, establishes execution contexts, and prepares all necessary task metadata, so that when execution starts, the AFU can immediately begin work without time-consuming mode transitions or initialization overhead.
Data Source
AI summary
A processing system includes a core, at least one accelerator function unit (AFU) and an accelerator interface. The core is utilized to develop at least one task. The AFU is utilized to execute the task. The accelerator interface is arranged between the core and the AFU to receive an accelerator interface instruction transmitted by the processing core and instruct the AFU to execute the task according to the accelerator interface instruction.