Nvidia patents GPU memory safety checks using hardware metadata
Nvidia describes a method where a compiler adds code to track memory limits in a special table. The GPU's hardware then uses this table to instantly check if a memory request is safe, preventing errors without slowing down the main processor.
In the filing’s own words
While a compiler compiles source code to create an executable binary, code is added into the compiled source code that, when executed, identifies and stores in a metadata table base and bounds information associated with memory allocations. Additionally, additional code is added into the compiled source code that enables hardware to determine a safety of memory access requests during an implementation of the compiled source code by performing an out-of-bounds (OOB) check in hardware using the base and bounds information stored in the metadata table. This enables the identification and avoidance of unsafe memory operations during the implementation of the executable by a GPU.
What exactly they claim (claim 1)
1. A method comprising, at a device: maintaining a metadata table having entries corresponding to memory allocations made during an execution of compiled source code; for each of the memory allocations, modifying unused bits of a pointer to the memory allocation to include an index to its corresponding entry in the metadata table; determining, in hardware, a safety of memory access requests during the execution of the compiled source code, utilizing the metadata table, including for each of the memory access requests: determining from a pointer included with the memory access request the memory allocation pointed to by the pointer and the index to the memory allocation's corresponding entry in the metadata table, accessing the entry in the metadata table using the index determined from the pointer, and determining the safety of the memory access request using the entry.
Facts
- Company
- Nvidia (NVIDIA Corporation)
- Official title
- Implementing hardware-based memory safety for a graphic processing unit
- Patent number
- US12801724B2
- Application number
- 18902593
- Filed
- Sep 30, 2024
- Granted
- Nov 17, 2026
- Inventors
- 3
- Claims
- 27
Classified under
G06F 21/53 Physics & computingG06F 9/44589 Physics & computingG06F 12/1441 Physics & computingMore from Nvidia in Chips & hardware
- Nvidia stacks memory directly on chips to boost data speed Nov 10, 2026
- Nvidia patents method to skip zero data in neural network math Oct 13, 2026
- Nvidia patents hardware to speed up motion blur in ray tracing Oct 13, 2026
- Nvidia patents method to sync data across split processor chips Sep 17, 2026
- Nvidia patents chip math trick to speed up AI calculations Oct 8, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.
