Dynamic IGPU DGPU Z-Value Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computing systems with integrated (IGPU) and discrete graphics processing units (DGPU), the IGPU often becomes a bottleneck in the graphics pipeline due to its lower processing power, hindering overall system performance since it must finish pre-computing Z-values before the DGPU can begin rendering, especially when the DGPU is more powerful.
Innovation Solution
A method for temporal load balancing across GPUs, where the graphics driver dynamically adjusts the IGPU percentage to pre-compute Z-values for a varying number of pixels, allowing the DGPU to render efficiently using pre-computed Z-values, and redistributes work based on processing times to optimize performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If the IGPU pre-computes Z-values for all pixels before the DGPU renders, then the DGPU can use pre-computed Z-values to avoid unnecessary shading operations, but the IGPU becomes a bottleneck in the graphics pipeline due to its lower processing power
Solution Approach 1:
The patent implements dynamic workload distribution where the IGPU processes a percentage of pixels for Z-value pre-computation rather than all pixels. The graphics driver dynamically adjusts the IGPU percentage based on performance metrics, allowing the system to adapt between fully utilizing the IGPU for pre-computation (when it helps) and reducing its workload (when it becomes a bottleneck), thus resolving the contradiction between energy efficiency and productivity
Solution Approach 2:
The system changes the parameter of IGPU workload from a fixed 100% pixel processing to a variable percentage. By monitoring frame times and performance metrics, the graphics driver adjusts the IGPU percentage parameter to optimize the balance between pre-computation benefits and pipeline throughput, preventing the IGPU from becoming a bottleneck while still achieving energy efficiency when conditions permit
2Reliability
If the IGPU runs one frame ahead of the DGPU to pre-compute Z-values, then the DGPU can benefit from pre-computed depth information, but the IGPU must finish processing before the DGPU can begin, creating a sequential dependency
Solution Approach 1:
The patent applies partial action by having the IGPU process only a percentage of pixels rather than all pixels for Z-value pre-computation. This partial processing reduces the time the IGPU occupies the pipeline, allowing the DGPU to start rendering sooner while still obtaining sufficient depth information for many pixels, thus reducing the sequential dependency and overall processing time while maintaining acceptable rendering accuracy
3Power
If the IGPU is substantially less powerful than the DGPU, then the DGPU has higher rendering capability, but the IGPU takes longer to pre-compute Z-values than the DGPU takes to finish rendering the previous frame
Solution Approach 1:
The system dynamically adjusts the IGPU's processing percentage based on the observed performance gap between IGPU pre-computation time and DGPU rendering time. When the IGPU takes too long, the percentage is reduced to prevent pipeline stalls. This dynamic adjustment allows the system to leverage the IGPU's capabilities when appropriate while avoiding the bottleneck effect when the power disparity causes timing issues
Data Source
AI summary
One embodiment of the present invention sets forth a method for dynamically load balancing rendering operations across an IGPU and a DGPU. For each frame, the graphics driver configures the IGPU to pre-compute Z-values for a portion of the display surface and to write feedback data to the system memory indicating the time that the IGPU used to process the frame. The graphics driver then configures the DGPU to use the pre-computed Z-values while rendering to the complete display surface and to write feedback data to the system memory indicating the time that the DGPU used to process the frame. The graphics driver uses the feedback data from the IGPU and DGPU in conjunction with the percentage of the display surface that the IGPU Z-rendered for the frame to scale the portion of the display surface that the IGPU Z-renders for one or more subsequent frames. In this fashion, overall processing within the graphics pipeline is optimized across the IGPU and DGPU.