Vector Instruction Fusion for Processing Unit Execution Delay

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In high-performance processing units, the execution delay between vector configuration and vector operation instructions leads to low efficiency due to the need for the vector operation instruction to wait for the completion of the vector configuration instruction, resulting in a significant delay and reduced processing efficiency.

Innovation Solution

A processing unit and method that fuse adjacent vector configuration and vector operation instructions into a single fusion instruction, allowing simultaneous execution by decoding the fusion instruction to obtain necessary execution information, modifying the vector control register, and bypassing the modified value to the vector operation unit, enabling concurrent execution without waiting for the configuration to complete.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the vector operation instruction waits for the vector configuration instruction to complete before execution, then the vector control register can be properly configured, but the execution delay increases and processing efficiency decreases

Engineering Contradiction:
Improveproper configuration of vector control registerVSAvoidprocessing efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent merges the vector configuration instruction and vector operation instruction into a single fused instruction. The instruction decoding unit decodes the fused instruction to extract both configuration parameters (element width, element number, register grouping, rounding type) and operation parameters simultaneously. This allows the vector control register to be configured and the vector operation to be executed in the same clock cycle, eliminating the sequential waiting delay while ensuring proper configuration through the unified decoding process.

Inventive Principle:
Principle #5Merging (Combining)

2Ease of operation

If the vector configuration instruction and vector operation instruction are executed sequentially, then the instruction execution order is simple to manage, but the execution delay between instructions increases

Engineering Contradiction:
Improveinstruction execution managementVSAvoidexecution delay
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The patent combines configuration and operation instructions into one fused instruction that is decoded once to produce both configuration data and operation data. This maintains simplicity in instruction management at the software level while achieving parallel execution at the hardware level, thus reducing time loss without significantly complicating the execution management mechanism.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The instruction decoding unit performs preliminary decoding of the fused instruction to extract configuration parameters before the vector operation executes. The vector control register is configured in advance within the same clock cycle, so when the vector operation unit retrieves the configuration, it is already ready, eliminating the waiting delay.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If the vector operation instruction waits for configuration completion, then the configuration can take effect properly, but the overall instruction execution speed decreases

Engineering Contradiction:
Improveconfiguration effect validationVSAvoidinstruction execution speed
Core Design Contradiction:
ReliabilityVSSpeed

Solution Approach 1:

The fused instruction ensures that configuration and operation are tightly coupled in a single execution cycle. The decoding unit extracts configuration parameters and validates them simultaneously with operation parameters, ensuring the configuration takes effect properly before the operation uses it, all within the same clock cycle thereby maintaining high execution speed.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS12175250B2Computing device and method for fusing and executing vector instructions
Publication Date: 2024.12.24 C SKY MICROSYST CO LTD
  • US12175250B2 patent drawing
  • US12175250B2 patent drawing
  • US12175250B2 patent drawing

AI summary

The embodiments of the present application provide a processing unit. The processing unit comprises: an instruction fetching unit configured for fusing instruction of vector configuration instruction and vector operation instruction that are adjacent in order to obtain fusion instruction; an instruction decoding unit configured to decode the fusion instruction to obtain first execution information and second execution information; a vector configuration unit configured to execute the vector configuration instruction according to the first execution information, modify vector control register, and bypass the value of the modified vector control register to the vector operation unit; the vector operation unit configured to execute the vector operation instruction according to the second execution information and the value of the modified vector control register.