SYSTEM AND METHOD OF OPERAND VALUE BASED PROCESSOR OPTIMIZATION BY DETECTING A CONDITION OF PRE-DETERMINED NUMBER OF BITS AND SLECTIVELY DISABLING PRE-DETERMINED BIT-FIELDS BY CLOCK GATING

Patent №

US 6,745,336

Granted

2004-06-01

Filed 2000

Owner

PRINCETON UNIVERSITY

Lab

AI components

1

hardware

Assignment

Recorded

Dataset

AIPD

2023_r1 edition

Application

09574793

Circuitry reduces power consumption by a microprocessor with operand-value-based clock gating. A bit detect unit detects the condition of a pre-determined number of bits of an operand. If the pre-determined number of bits are not necessary for executing the operand, a condition detect signal is generated. Gating logic receives the condition detect signal and initiates a gated clock signal. Latching circuitry or pre-charge circuitry receives the gated clock signal and disables the pre-determined number of bits, preventing the execution of unnecessary bits by the microprocessor and reducing the power consumed during execution. Operation packing improves microprocessor performance by packing narrow-width operations for parallel execution by the microprocessor. A bit detect unit detects the condition of a pre-determined number of bits of an operand and initiates a condition detect signal. Issue logic detects common operations within execution instructions and receives the condition detect signal, initiating an operation packing signal when common operations are ready to issue and the operands involved contain a pre-determined number of bits unnecessary for execution. Multiplexers receive the operation packing signal and move data from the lowermost bits of the operands to the upper sub-words of the source operand bus, creating a parallel sub-word operation. After execution, multiplexers move data from upper sub-words of the result onto the lowermost bit boundaries of the individual result operands.

AI hardwareG06F 9/30025G06F 1/3203G06F 1/3237G06F 1/3243G06F 1/3287G06F 7/48G06F 9/30014G06F 9/30036+1 more

AI classification

AI hardware1.00
Knowledge representation0.07
Planning0.00
Natural language0.00
Evolutionary computation0.00
Speech0.00
Vision0.00
Machine learning0.00

Ownership

PRINCETON UNIVERSITY

assignment · 110150754

Assignors

MARTONOSKI, MARGARET, BROOKS, DAVID

On an employer assignment, the assignors are typically the inventors.

© 2026 NYSGPT2525 LLC