Taiwan semiconductor manufacturing co., ltd. (20240331745). SYSTEMS AND METHODS FOR FLEXIBLE BANK ADDRESSING IN DIGITAL COMPUTING-IN-MEMORY (DCIM) simplified abstract

From WikiPatents
Revision as of 12:55, 8 October 2024 by Wikipatents (talk | contribs) (Creating a new page)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

SYSTEMS AND METHODS FOR FLEXIBLE BANK ADDRESSING IN DIGITAL COMPUTING-IN-MEMORY (DCIM)

Organization Name

taiwan semiconductor manufacturing co., ltd.

Inventor(s)

Haruki Mori of Hsinchu (TW)

Hidehiro Fujiwara of Hsinchu (TW)

Wei-Chang Zhao of Hsinchu (TW)

Kinshuk Khare of Hsinchu (TW)

SYSTEMS AND METHODS FOR FLEXIBLE BANK ADDRESSING IN DIGITAL COMPUTING-IN-MEMORY (DCIM) - A simplified explanation of the abstract

This abstract first appeared for US patent application 20240331745 titled 'SYSTEMS AND METHODS FOR FLEXIBLE BANK ADDRESSING IN DIGITAL COMPUTING-IN-MEMORY (DCIM)

Simplified Explanation: The patent application describes a method for flexible bank addressing in digital computing-in-memory (DCIM), where memory banks are grouped together to store input feature map data and perform multiply-accumulate operations.

  • Bank groups are created, each containing a number of memory banks.
  • Memory banks store specific portions of input feature map data.
  • During different clock cycles, portions of data are read from different bank groups to perform operations.
  • Multiply-accumulate operations are carried out using the data from different bank groups.

Key Features and Innovation:

  • Flexible bank addressing in digital computing-in-memory.
  • Bank groups for efficient storage and processing of input feature map data.
  • Multiplier-accumulate operations using data from different bank groups.

Potential Applications:

  • Image processing.
  • Machine learning algorithms.
  • Signal processing.

Problems Solved:

  • Efficient memory access and processing of large datasets.
  • Improved performance in computing-in-memory applications.

Benefits:

  • Faster processing of data.
  • Reduced memory access times.
  • Enhanced efficiency in memory utilization.

Commercial Applications:

  • AI hardware development.
  • Data centers.
  • Edge computing devices.

Prior Art: Prior research on computing-in-memory architectures and memory bank addressing methods.

Frequently Updated Research: Ongoing studies on optimizing memory access in computing-in-memory systems.

Questions about flexible bank addressing in digital computing-in-memory: 1. How does the method improve memory access efficiency in computing-in-memory systems? 2. What are the potential challenges in implementing this method in real-world applications?


Original Abstract Submitted

a method for flexible bank addressing in digital computing-in-memory (dcim). the method includes providing bank groups, each of the bank groups comprising a respective number of memory banks, each memory bank configured to store a corresponding portion of input feature map data. the method includes reading, during a first clock cycle, a first portion of the input feature map data from a first one of the bank groups and a second portion of the input feature map data from a second one of the bank groups. the method includes performing a first multiply-accumulate operation using the first portion and the second portion. the method includes reading, during a second clock cycle, a third portion of the input feature map data from the first bank group. the method includes performing a second multiply-accumulate operation using the second portion and the third portion.