Compact Arithmetic Elements Using Low-Precision HDR Computing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional computing systems inefficiently utilize transistors due to a focus on high precision arithmetic, limiting the ability to harness the full computing power of modern silicon-based processors, despite advancements in transistor count.

Innovation Solution

Implementing low precision high dynamic range (LPHDR) processing elements that perform arithmetic operations with a precision of about 0.1% error, allowing for a greater number of operations per unit time and area, and potentially using logarithmic or analog representations to optimize circuit design.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional high precision arithmetic processing elements are used, then measurement precision is improved, but productivity deteriorates

Engineering Contradiction:

Improvearithmetic precisionVSAvoidcomputing power delivery

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent changes the precision parameter from conventional high precision (32-bit or 64-bit floating point) to low precision (about 0.1% error or 10-bit mantissa). This parameter change allows significantly more processing elements to be packed into the same silicon area, thereby increasing productivity while maintaining adequate precision for many applications.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent segments the computing system into massively parallel low-precision processing elements rather than using fewer high-precision elements. By dividing the computational task across many simple processing elements, the system achieves higher overall productivity while each element maintains sufficient precision for its specific sub-tasks.

Inventive Principle:

Principle #1Segmentation

2Power

If transistor count is increased, then computing power is improved, but ease of operation deteriorates

Engineering Contradiction:

Improvecomputing powerVSAvoidsoftware-hardware interface efficiency

Core Design Contradiction:

PowerVSEase of operation

Solution Approach 1:

Instead of increasing precision to get more computing power, the patent inverts the approach by decreasing precision to enable massively parallel processing. This inversion allows billions of processing elements to operate simultaneously, delivering theoretical computing power while simplifying the software interface through uniform low-precision operations across all elements.

Inventive Principle:

Principle #13The other way round (Inversion)

3Productivity

If low precision processing elements are used, then productivity is improved, but measurement precision deteriorates

Engineering Contradiction:

Improvenumber of operations per unit timeVSAvoidarithmetic result accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent uses many copies of simple low-precision processing elements rather than fewer complex high-precision elements. Each processing element is a simple copy that performs basic arithmetic with about 0.1% precision, but the massive number of parallel copies compensates for individual imprecision through statistical aggregation and error cancellation in many applications.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS11354096B2Processing with compact arithmetic processing element

Publication Date: 2022.06.07 SINGULAR COMPUTING LLC

AI summary

Low precision computers can be efficient at finding possible answers to search problems. However, sometimes the task demands finding better answers than a single low precision search. A computer system augments low precision computing with a small amount of high precision computing, to improve search quality with little additional computing.