Processor Instruction Grouping for Shared Data Buffer Reuse

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In processors with multiple execution units and data buffers, shared data is transferred to multiple buffers, reducing data reusability and increasing power consumption.

Innovation Solution

A processor design with a transfer control unit that detects shared data usage and transfers it to a single data buffer corresponding to an execution unit, optimizing data reusability and reducing power consumption by minimizing redundant data transfers.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If shared data is transferred to multiple data buffers corresponding to multiple execution units, then each execution unit can access the data independently, but data reusability decreases and power consumption increases

Engineering Contradiction:
Improvedata accessibilityVSAvoidpower consumption
Core Design Contradiction:
Ease of operationVSLoss of energy

Solution Approach 1:

Multiple data buffers corresponding to different execution units are merged into a single shared data buffer. This consolidation allows multiple execution units to access the same data through a common buffer, eliminating redundant data copies and reducing the energy consumption associated with multiple separate data transfers while maintaining data accessibility for all execution units.

Inventive Principle:
Principle #5Merging (Combining)

2Productivity

If shared data is transferred to multiple data buffers, then data can be distributed to multiple execution units, but data reusability is reduced due to redundant transfers

Engineering Contradiction:
Improvedata distribution efficiencyVSAvoidpower consumption
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The patent merges multiple data buffers into a single shared data buffer that serves multiple execution units. This approach maintains data distribution efficiency by allowing any execution unit to access the data through the unified buffer, while eliminating redundant data transfers that would occur if separate buffers were maintained for each execution unit, thereby reducing power consumption.

Inventive Principle:
Principle #5Merging (Combining)

3Reliability

If multiple data buffers are provided for multiple execution circuits, then data can be cached locally at each unit, but redundant data transfers occur and power consumption increases

Engineering Contradiction:
Improvedata cachingVSAvoidpower consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent combines multiple individual data buffers into a single shared data buffer that is accessed by multiple execution units. This maintains effective data caching by providing local access points for each execution unit while eliminating the redundancy of maintaining separate buffers, thereby reducing the energy consumption associated with multiple data transfer operations.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS20260010397A1Processor and operation control method of processor
Publication Date: 2026.01.08 FUJITSU LTD
  • US20260010397A1 patent drawing
  • US20260010397A1 patent drawing
  • US20260010397A1 patent drawing

AI summary

A processor includes execution circuits configured to execute computational instructions; a first instruction queue configured to hold the computational instructions; second instruction queues respectively provided corresponding to the execution circuits, the second instruction queues being configured to hold the computational instructions and issue the held computational instructions to the corresponding execution units; data buffers respectively provided corresponding to the execution circuits and configured to hold data used by the computational instructions; and a transfer control circuit configured to detect an address of a memory that holds data used by each of the computational instructions held in the first instruction queue, transfer, to a second instruction queue corresponding to one of the execution circuits, target computational instructions that use data at a same address, based on the detected address, and transfer the data at the same address to a data buffer corresponding to the one of the execution circuit.