Fixed-Point Alpha Blending to Cut GPU Power and Die Area

Solution Overview

Problem

Current graphics processing units (GPUs) face challenges in optimizing computations such as blending due to the overhead of dedicated floating-point hardware, which consumes power and occupies significant die area, and the increasing complexity of integrated circuits leads to heat generation and power management issues, limiting device usage and efficiency.

Innovation Solution

Implementing a combination of fixed-point and floating-point blending units, where fixed-point blending is used for simpler operations like alpha blending, reducing power consumption and gate utilization, and allowing for lower precision formats, while floating-point blending handles higher precision tasks, optimizing resource usage and reducing die area.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If dedicated floating-point blending hardware is used, then blending precision is improved, but power consumption increases and die area expands

Engineering Contradiction:

Improveblending precisionVSAvoidpower consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system dynamically selects between fixed-point and floating-point blending modes based on the color buffer format. For lower precision formats (565, 888), fixed-point blending is used; for higher precision formats (101010, 161616), floating-point blending is used. This dynamic adaptation resolves the contradiction by matching precision to actual requirements, reducing unnecessary power consumption while maintaining adequate blending precision for each format.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the precision parameter of the blending operation based on the color buffer format. By adjusting the blending precision to match the color buffer precision (e.g., using 5-bit precision for 565 format, 10-bit for 101010 format), the system avoids the overhead of high-precision floating-point operations when lower precision suffices, thereby reducing power consumption while maintaining adequate blending quality.

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If dedicated floating-point blending hardware is used, then blending precision is improved, but die area increases

Engineering Contradiction:

Improveblending precisionVSAvoiddie area

Core Design Contradiction:

Measurement precisionVSArea of stationary object

Solution Approach 1:

The system dynamically selects between fixed-point and floating-point blending modes based on the color buffer format. For lower precision formats (565, 888), fixed-point blending is used; for higher precision formats (101010, 161616), floating-point blending is used. This dynamic adaptation resolves the contradiction by matching precision to actual requirements, reducing unnecessary power consumption while maintaining adequate blending precision for each format.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the precision parameter of the blending operation based on the color buffer format. By adjusting the blending precision to match the color buffer precision (e.g., using 5-bit precision for 565 format, 10-bit for 101010 format), the system avoids the overhead of high-precision floating-point operations when lower precision suffices, thereby reducing power consumption while maintaining adequate blending quality.

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If additional components are integrated onto a single IC chip, then functionality increases, but heat generation and power consumption increase

Engineering Contradiction:

ImprovefunctionalityVSAvoidpower consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by stationary object

Solution Approach 1:

The system dynamically selects between fixed-point and floating-point blending modes based on the color buffer format. For lower precision formats (565, 888), fixed-point blending is used; for higher precision formats (101010, 161616), floating-point blending is used. This dynamic adaptation resolves the contradiction by matching precision to actual requirements, reducing unnecessary power consumption while maintaining adequate blending precision for each format.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the precision parameter of the blending operation based on the color buffer format. By adjusting the blending precision to match the color buffer precision (e.g., using 5-bit precision for 565 format, 10-bit for 101010 format), the system avoids the overhead of high-precision floating-point operations when lower precision suffices, thereby reducing power consumption while maintaining adequate blending quality.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11494968B2Dedicated fixed point blending for energy efficiency

Publication Date: 2022.11.08 INTEL CORP

AI summary

Briefly, in accordance with one or more embodiments, a processor receives an incoming data stream that includes alpha channel data, and a memory stores an application programming interface (API). The API is to route the alpha channel data to a fixed point blending unit to perform one or more blending operations using fixed point representation of the alpha channel data. The API is further to route the incoming data stream to a floating point blending unit to perform operations involving floating point representation of the incoming data.