Z BUFFER BANDWIDTH REDUCTIONS VIA SPLIT TRANSACTIONS

Patent №

US 5,844,571

Granted

1998-12-01

Filed 1996

Owner

IBM CORPORATION

AI components

1

vision

Assignment

Recorded

Dataset

AIPD

2023_r1 edition

Application

08662426

In computer graphics systems, a view of scene of primitives is represented by pixel data associated with a set S of pixels. The present invention identifies whether a given primitive is visible at the set S of pixels by storing in a buffer, for each pixel P belonging to set S, a depth value Z.sub.old associated with the pixel P. The depth value Z.sub.old is partitioned into a plurality of portions including a most significant portion and at least one less significant portion. The buffer comprises a plurality of contiguous blocks each storing corresponding portions of the depth value Z.sub.old for a given set SP of consecutive pixels belonging to set S. A set SQ of consecutive pixels belonging to set S that cover the given primitive is determined. The most significant portion of the depth value Z.sub.old of the set SQ of consecutive pixels are fetched from the buffer. For each pixel Q belonging to set SQ, the most significant portion of the depth value Z.sub.old of the particular pixel is compared to the most significant portion of the computed depth value Z.sub.new of the particular pixel to determine if one of three conditions is satisfied: a) the given primitive is visible at the particular pixel, b) the given primitive is hidden at the particular pixel, or c) it is undetermined whether the given primitive is either visible or hidden at the particular pixel. If condition a) is satisfied, the blocks of the buffer associated with the particular pixel is updated to store the computed depth value Z.sub.new of the particular pixel. If condition b) is satisfied, the processing ends with respect to the particular pixel. If condition c) is satisfied, the fetching step and comparing step above is repeated for the next less significant portion of the depth values Z.sub.old and Z.sub.new associated with the particular pixel until the least significant portion of the depth values Z.sub.old and Z.sub.new has been processed.

AI classification

Vision0.86
AI hardware0.05
Knowledge representation0.01
Evolutionary computation0.00
Natural language0.00
Machine learning0.00
Speech0.00
Planning0.00

Ownership

IBM CORPORATION

assignment · 80520645

Assignors

NARAYANASWAMI, CHANDRASEKHAR

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

© 2026 NYSGPT2525 LLC