LOCAL STALL/HAZARD DETECT IN SUPERSCALAR, PIPELINED MICROPROCESSOR TO AVOID RE-READ OF REGISTER FILE

Patent №

US 6,587,940

Granted

2003-07-01

Filed 2000

Owner

HEWLETT-PACKARD COMPANY

Lab

AI components

1

hardware

Assignment

Recorded

Dataset

AIPD

2023_r1 edition

Application

09483774

A method and apparatus that utilizes a simplified, localized version (“a local data-dependency stall”) of a global data-dependency stall to avoid re-reading of a register file to improve the performance of a pipelined microprocessor. A non-asserted local data-dependency stall indicates that source operand for an instruction is correct. Accordingly, when a global data-dependency stall arrives, the instruction is stalled in a stage without re-reading the register file. Without the simplified, localized version of the global data-dependency stall, the source operand data is not known to be correct and is indeed assumed to be incorrect. Therefore, when the global data-dependency stall arrives, a complete re-computation of the source operand data must be performed, including a re-read of the register file. Likewise, an asserted local data-dependency stall indicates that source operand for an instruction is not correct. Accordingly, the instruction is stalled when the local data-dependency stall is asserted and until the data for the instruction is available. The available data is forwarded directly to the stalled instruction.

AI hardwareG06F 9/3867G06F 9/3828G06F 9/3838

AI classification

AI hardware0.78
Evolutionary computation0.02
Planning0.01
Natural language0.00
Knowledge representation0.00
Vision0.00
Machine learning0.00
Speech0.00

Ownership

HEWLETT-PACKARD COMPANY

assignment · 108100635

Assignors

SOLTIS JR., DONALD CHARLES, BHATIA, ROHIT

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

© 2026 NYSGPT2525 LLC