Tessellation Shader Attribute Handling via Local Memory Redistribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The imbalance of workload in graphics processing pipelines during tessellation leads to inefficiency and increased power consumption due to the need to copy attributes between local and system memory, which is particularly problematic in low-power applications like mobile devices.
Innovation Solution
A method that redistributes workload between phases of graphics processing within a single streaming multiprocessor system without copying attributes to and from external memory, utilizing shared memory to maintain data residency and reduce power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If attributes are copied between local memory and cache or system memory to rebalance workload, then workload distribution is improved, but power consumption increases
Solution Approach 1:
The patent merges the attribute storage and processing functions within the same streaming multiprocessor by maintaining attribute data in local shared memory buffers. This eliminates the need to copy attributes between separate memory systems (local memory and cache/system memory), thereby resolving the contradiction by achieving workload rebalancing without the power penalty of cross-memory transfers.
Solution Approach 2:
The patent introduces intermediate buffers within the local shared memory of the streaming multiprocessor as a mediator between attribute storage and processing stages. These buffers allow workload rebalancing to occur locally without requiring data to traverse to and from external cache or system memory, thus improving productivity while minimizing power consumption.
2Productivity
If attributes are transferred to cache or system memory for rebalancing, then workload balance is improved, but access time increases
Solution Approach 1:
The patent combines attribute storage and processing operations within the same local memory hierarchy, eliminating the time-consuming data transfer to and from cache or system memory. This allows workload rebalancing to occur rapidly using local shared memory buffers.
Solution Approach 2:
The patent performs preliminary actions by pre-allocating and pre-positioning attribute data in local shared memory buffers before processing stages begin. This eliminates the need for time-consuming memory access during actual processing, as all necessary data is already resident in fast local memory.
3Ease of operation
If off-chip memory components are energized for attribute access, then data retrieval is enabled, but power consumption increases
Solution Approach 1:
The patent extracts the attribute storage function from external cache or system memory and places it directly within the local shared memory of the streaming multiprocessor. This extraction eliminates the need to energize off-chip memory components for attribute access, thereby enabling data retrieval while minimizing power consumption.
Solution Approach 2:
The streaming multiprocessor serves itself by maintaining and accessing attribute data locally within its own shared memory resources. This self-service approach eliminates dependency on external memory systems, allowing data retrieval without the power overhead of energizing off-chip memory components.
Data Source
AI summary
Attributes of graphics objects are processed in a plurality of graphics processing pipelines. A streaming multiprocessor (SM) retrieves a first set of parameters associated with a set of graphics objects from a first set of buffers. The SM performs a first set of operations on the first set of parameters according to a first phase of processing to produce a second set of parameters stored in a second set of buffers. The SM performs a second set of operations on the second set of parameters according to a second phase of processing to produce a third set of parameters stored in a third set of buffers. One advantage of the disclosed techniques is that work is redistributed from a first phase to a second phase of graphics processing without having to copy the attributes to and retrieve the attributes from the cache or system memory, resulting in reduced power consumption.