Multi-VPU Memory Mapping via GART Virtual Addressing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current graphics and video processing systems face challenges in producing high-quality, complex video at affordable costs for mass production, while also requiring excessive memory and computational complexity, and are not seamlessly compatible with existing computer systems.
Innovation Solution
A video processing system that utilizes multiple graphics or video processing units (VPUs) sharing rendering tasks in parallel, with each VPU accessing the local memories of others through a virtual addressing scheme, allowing for cooperative processing and minimizing the need for unique addressing, and combining data from multiple VPUs to produce image data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple VPUs share rendering tasks in parallel with virtual addressing scheme, then productivity and video quality are improved, but device complexity increases
Solution Approach 1:
The patent implements a universal memory mapping approach where each VPU is configured with identical address ranges in its GART that map to the same physical memory locations of other VPUs. This universal addressing scheme allows any VPU to access any other VPU's local memory using the same address space, eliminating the need for complex unique addressing schemes while enabling parallel cooperative rendering.
Solution Approach 2:
The patent creates virtual copies of memory address spaces by mapping each VPU's local memory into the GART of every other VPU with identical address ranges. This copying approach allows VPUs to access each other's memory as if it were their own local memory, simplifying the addressing mechanism while enabling efficient inter-VPU data access for parallel processing.
2Quantity of substance
If multiple VPUs access each other's local memories through GART, then memory requirements are reduced, but manufacturing precision requirements increase
Solution Approach 1:
The patent changes the addressing parameter from unique per-VPU address spaces to identical shared address ranges. By configuring each VPU's GART to map other VPUs' local memories using the same address ranges that each VPU uses for its own memory, the system reduces total memory requirements while maintaining precise address mapping through standardized parameter configuration.
3Ease of operation
If VPUs use identical address ranges for accessing own and other VPUs' memories, then ease of operation improves, but reliability challenges arise
Solution Approach 1:
The patent introduces the GART (Graphics Address Remapping Table) as an intermediary layer between virtual addresses and physical memory locations. The GART translates identical address ranges from multiple VPUs into distinct physical memory locations, allowing simple identical addressing while maintaining reliable correct memory access through the mediating translation mechanism.
Data Source
AI summary
A system and method for memory mapping in a multiple video processor (multi VPU) system is described. In various embodiments, rendering tasks are shared among multiple VPUs in parallel to provide improved performance and capability with minimal increased cost. In various embodiments, multiple VPUs in a system access each other's local memories to facilitate cooperative video processing. In one embodiment, each VPU in the system has the local memories of each other VPU mapped to its own graphics aperture relocation table (GART) table to facilitate access via a virtual addressing scheme. Each VPU uses the same virtual addresses for this mapping to other VPU local memories. This allows the driver to send exactly the same write commands to each VPU, including the numeric value of the destination address for operations such as writing rendered data. Thus, unique addresses need not be generated for each VPU.