Vector Packed Multiply Rounding and Saturation Control

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current computer processor architectures face challenges in efficiently performing vector packed multiply operations for signed and unsigned words, particularly in handling rounding and saturation methods, which are crucial for precise data processing in various applications.

Innovation Solution

The introduction of specific vector packed multiply instructions, such as DVPMULUWR, DVPMULWR, and DVPMULWS, which allow for rounding and saturation of doubleword products, enabling efficient execution of vector packed multiply operations for unsigned and signed words, with control over rounding methods and saturation to specific word values.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional vector multiply instructions are used, then basic multiplication operations can be performed, but precision and control over rounding/saturation methods are insufficient

Engineering Contradiction:
Improvedata processing precisionVSAvoidrounding and saturation control
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent introduces new vector packed multiply instructions (DVPMULUWR, DVPMULWR, DVPMULWS) that incorporate rounding control and saturation capabilities as integral parameters. These instructions change the operational parameters by adding rounding mode selection and saturation control directly into the multiply operation, enabling precise control over the output format and range without requiring separate post-processing instructions.

Inventive Principle:
Principle #35Parameter changes

2Ease of operation

If multiple separate instructions are used for multiply, round, and saturate operations, then each operation can be controlled independently, but execution time and instruction complexity increase

Engineering Contradiction:
Improveinstruction simplicityVSAvoidexecution time
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The patent merges multiple operations (vector packed multiplication, rounding, and saturation) into single integrated instructions. For example, DVPMULUWR combines unsigned word multiplication with rounding to store high results, while DVPMULWS combines signed word multiplication with saturation to lower word. This merging eliminates the need for separate instructions for each operation, reducing instruction complexity and execution time while maintaining full control over rounding and saturation behavior.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4109246A1Apparatus and method for vector packed multiply of signed and unsigned words
Publication Date: 2022.12.28 INTEL CORP
  • EP4109246A1 patent drawingFigure 1A
  • EP4109246A1 patent drawingFigure 1B
  • EP4109246A1 patent drawingFigure 2A~2C

AI summary

An apparatus and method for performing a vector packed multiplication of signed and unsigned words. For example, one embodiment of a processor includes a decoder to decode a vector packed multiply instruction having operands to identify a first and a second plurality of packed words, first and second source registers to store the first and second plurality of packed words, and execution circuitry to execute the decoded instruction. The execution circuitry includes multiplier circuitry to multiply each packed word in the first source register with a corresponding packed word in the second source register to generate a plurality of doubleword products and rounding circuitry to round each of the doubleword products according to a rounding method to generate a plurality of rounded doubleword products. Each upper word of the rounded doubleword results is then stored into a corresponding word data element positions of a destination register.