Graphics Transposer Read Scheduler Using Minimum Cost Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing graphics systems face inefficiencies in scheduling read operations for transposers, leading to additional passes and clock cycles due to the conventional greedy algorithm, which results in inefficient data transfer.

Innovation Solution

A read scheduler utilizing minimum cost analysis is employed to determine the optimal sequence of block read operations for the transposer, minimizing the total number of passes required to process input vectors by adhering to constraints on data attributes and transfer sizes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If a greedy algorithm is used to schedule read operations for the transposer, then the scheduling process is simple and fast, but the data transfer efficiency deteriorates resulting in additional passes and clock cycles

Engineering Contradiction:
Improvedata transfer efficiencyVSAvoidscheduling algorithm complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent pre-calculates and stores optimal read schedules in lookup tables before runtime. The minimum cost analysis determines the optimal sequence of block read operations in advance, considering all constraints on data attributes and transfer sizes. During actual transposition, the system simply retrieves pre-computed schedules from the lookup tables, avoiding complex real-time calculations while achieving optimal data transfer efficiency and minimizing the number of passes required

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements a hybrid approach that adapts to different input data characteristics. The system dynamically selects between using pre-computed lookup tables for common cases and performing minimum cost analysis for less common scenarios. This dynamic adaptation allows the system to maintain high efficiency for typical workloads while handling edge cases optimally, balancing between pre-computed efficiency and real-time adaptability

Inventive Principle:
Principle #15Dynamics

2Reliability

If additional passes are used to process input vectors, then the transposer can handle constrained data attributes, but the processing time increases

Engineering Contradiction:
Improveconstraint satisfactionVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The minimum cost analysis pre-determines the optimal read schedule that satisfies all transposer constraints while minimizing the number of passes required. By calculating the optimal sequence of block read operations in advance and storing it in lookup tables, the system ensures constraint satisfaction is achieved with the minimum necessary passes, preventing time loss from suboptimal scheduling

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system uses feedback from the minimum cost analysis to continuously improve scheduling decisions. The analysis evaluates the current state of input data and constraint satisfaction, then adjusts the read schedule accordingly to minimize additional passes. This feedback mechanism ensures that the system adapts to maintain optimal performance while satisfying all transposer constraints

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS7797258B1Graphics system transposer read scheduler
Publication Date: 2010.09.14 NVIDIA CORP
  • US7797258B1 patent drawing
  • US7797258B1 patent drawing
  • US7797258B1 patent drawing

AI summary

A graphics system includes a transposer. A read scheduler utilizes a minimum cost analysis to schedule a read transfer order for the transposer to minimize the total number of passes required to process a set of input vectors.