Tiled Integrated Circuit Memory Access Synchronization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In parallel processing environments, managing memory access efficiently is challenging due to the high costs and performance limitations of Field Programmable Gate Arrays (FPGAs) compared to Application Specific Integrated Circuits (ASICs), and existing solutions do not effectively address the need for reconfigurability and performance.

Innovation Solution

A tiled integrated circuit architecture comprising multiple tiles, each with a processor and a switch for data forwarding, along with memory interface modules that ensure previous memory transactions are completed before executing new instructions, allowing for synchronized memory access and efficient data processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If FPGAs are used to achieve reconfigurability, then adaptability is improved, but performance and power efficiency deteriorate

Engineering Contradiction:
ImprovereconfigurabilityVSAvoidperformance
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The system is divided into multiple independent tiles, each capable of being individually configured and operated. This segmentation allows parallel processing across multiple tiles while maintaining reconfigurability at the tile level, thereby improving overall performance without sacrificing adaptability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Multiple tiles are merged into a unified parallel processing system with shared memory resources and inter-tile communication capabilities. This combining approach enables simultaneous execution of multiple processing tasks across tiles while maintaining efficient resource utilization, thus improving performance while preserving reconfigurability.

Inventive Principle:
Principle #5Merging (Combining)

2Adaptability or versatility

If FPGAs are used to achieve reconfigurability, then adaptability is improved, but power consumption increases

Engineering Contradiction:
ImprovereconfigurabilityVSAvoidpower consumption
Core Design Contradiction:
Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The system divides power-consuming operations across multiple independent tiles, allowing selective activation of only those tiles needed for current processing tasks. This segmentation reduces overall power consumption compared to a monolithic reconfigurable system by enabling granular power management at the tile level.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Instead of activating the entire reconfigurable system for all operations, only the necessary subset of tiles is activated for each specific processing task. This partial action approach reduces power consumption while maintaining full reconfigurability capability when needed.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If memory transactions are processed in parallel without synchronization, then productivity is improved, but reliability deteriorates

Engineering Contradiction:
Improveprocessing speedVSAvoidtransaction completion accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system establishes synchronization protocols and ordering rules for memory transactions before parallel execution begins. This preliminary action ensures that memory access patterns are pre-coordinated, allowing parallel processing to proceed while maintaining reliability through predetermined synchronization mechanisms.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements feedback mechanisms that monitor memory transaction completion status and coordinate between tiles. This feedback enables parallel processing to maintain reliability by detecting and resolving potential conflicts or incomplete transactions, ensuring accurate transaction completion while preserving processing speed.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS7805577B1Managing memory access in a parallel processing environment
Publication Date: 2010.09.28 MELLANOX TECHNOLOGIES LTD(IL)
  • US7805577B1 patent drawing
  • US7805577B1 patent drawing
  • US7805577B1 patent drawing

AI summary

An integrated circuit comprises a plurality of tiles. Each tile comprises a processor, and a switch including switching circuitry to forward data received over data paths from other tiles to the processor and to switches of other tiles, and to forward data received from the processor to switches of other tiles. The integrated circuit further comprises one or more memory interface modules including circuitry to access an external memory, each memory interface module coupled to a switch of at least one tile. At least some of the tiles are configured to send a message to a memory interface module to determine whether previous memory transactions associated with a tile have been completed.