International business machines corporation (20240135158). PROCESSING TENSORS simplified abstract
Contents
- 1 PROCESSING TENSORS
PROCESSING TENSORS
Organization Name
international business machines corporation
Inventor(s)
Julian Heyne of Stuttgart (DE)
Razvan Peter Figuli of Remchingen (DE)
Cedric Lichtenau of Stuttgart (DE)
Holger Horbach of Aidlingen (DE)
PROCESSING TENSORS - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240135158 titled 'PROCESSING TENSORS
Simplified Explanation
The present disclosure describes a method for accessing a multidimensional tensor of elements in a computer memory, consisting of two-dimensional arrays called pages, each containing a predefined number of one-dimensional arrays known as sticks. The method involves loading each page of the tensor linearly, loading non-empty sticks from memory using a base address, determining a base address for the next page based on the number of loaded sticks and an address offset for potential empty sticks, and reinitializing the loading process when a chunk size is reached.
- Explanation of the patent:
- Accessing n-dimensional tensor elements in memory using a linear loading method. - Utilizing two-dimensional arrays (pages) with one-dimensional arrays (sticks) to organize data. - Loading non-empty sticks sequentially and efficiently determining addresses for subsequent pages. - Reinitializing the loading process at chunk boundaries for optimized memory access.
Potential Applications
This technology could be applied in: - Data processing systems - Image and video processing - Machine learning algorithms
Problems Solved
- Efficient access to multidimensional data structures - Optimized memory usage and access patterns - Streamlined processing of large datasets
Benefits
- Improved performance in handling complex data structures - Reduced memory overhead and access latency - Scalability for large-scale data processing tasks
Potential Commercial Applications
Optimized Memory Access Method for Multidimensional Tensors
Possible Prior Art
Prior art related to efficient memory access methods for multidimensional data structures.
Unanswered Questions
How does this method compare to existing memory access techniques in terms of performance and efficiency?
This article does not provide a direct comparison with other memory access techniques, leaving the reader to infer the advantages of the proposed method.
What specific types of computer systems or applications would benefit most from implementing this linear loading method for accessing tensors?
The article does not delve into specific use cases or industries where this technology could have the most significant impact.
Original Abstract Submitted
the present disclosure relates to a method of accessing a n-dimensional tensor of elements in a memory by a computer system. the multidimensional tensor comprises two-dimensional arrays, herein referred to as pages, each page being configured to comprise a predefined number of one-dimensional arrays of elements, herein referred to as sticks. the method includes linearly loading page per page of the tensor, and doing the following for each page: loading the non-empty sticks of the page from the memory using a base address of the page and determining a base address for the subsequent page using the number of loaded sticks and using an address offset indicative of potential empty sticks of the page. in case the number of loaded pages reaches a chunk size, the chunk page counter may be reinitialized and the linear loading may be continued with a subsequent page.