PiM Error Logging and Selective Reset for Debug Recovery

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current PiM architectures lack mechanisms for error detection and logging, leading to data loss when errors occur, and require full DRAM array resets, which are inefficient and disruptive.

Innovation Solution

Implementing a status register, counter, and command history register in PiM architectures to detect errors, log commands, and enable selective power gating of PiM logic, allowing precise error classification and recovery without affecting the entire DRAM array.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If full DRAM array reset is performed when error occurs, then system reliability is improved, but data loss increases and productivity decreases

Engineering Contradiction:
Improvesystem reliabilityVSAvoiddata loss
Core Design Contradiction:
ReliabilityVSLoss of substance

Solution Approach 1:

The patent segments the PiM device into distinct functional blocks: memory array, PiM logic, status register, command history register, and debug interface. This segmentation allows selective resetting of only the PiM logic block when an error occurs, rather than resetting the entire DRAM array. The status register and command history register capture error information at the block level, enabling targeted recovery that preserves data in the memory array while restoring only the faulty logic component.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements preliminary error detection and logging mechanisms before full system failure occurs. Status registers and command history registers continuously monitor and record PiM operations, detecting errors early in the execution sequence. This preliminary detection allows the system to log the specific command that caused the error and reset only the PiM logic block, preventing cascading failures that would require a complete array reset and cause total data loss.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If full DRAM array reset is performed when error occurs, then system reliability is improved, but latency increases

Engineering Contradiction:
Improvesystem reliabilityVSAvoidlatency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent segments the reset operation to affect only the PiM logic block rather than the entire DRAM array. This segmented approach reduces reset latency significantly because only the computational logic needs to be reinitialized, not the entire memory structure. The status register and command history register provide fast error identification at the block level, enabling rapid targeted recovery that minimizes time loss while maintaining system reliability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a feedback mechanism where the status register and command history register continuously report PiM operation status to the host system. When an error is detected, this feedback loop provides immediate information about the error type and location, enabling the host to initiate a targeted PiM logic reset rather than a conservative full array reset. This feedback-driven approach reduces latency by eliminating unnecessary reset operations on healthy memory portions.

Inventive Principle:
Principle #23Feedback

3Reliability

If error detection mechanisms are added to PiM architecture, then reliability is improved, but device complexity increases

Engineering Contradiction:
Improveerror detection capabilityVSAvoidarchitecture complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges error detection and logging functionality directly into the existing PiM architecture blocks. The status register and command history register are integrated with the PiM logic block, sharing control signals and data pathways. This merging approach enables comprehensive error detection without adding separate complex monitoring systems, as the existing register structures are repurposed to serve both normal operation and error detection functions.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The status register and command history register serve multiple functions: they track normal PiM operation status, log error conditions, and provide debugging information to the host system. This multi-functionality reduces the need for dedicated error detection hardware, as existing universal register structures handle both operational monitoring and error detection tasks, thereby limiting the increase in device complexity while improving reliability.

Inventive Principle:
Principle #6Universality (Multi-functionality)

4Difficulty of detecting and measuring

If command logging is implemented in PiM architecture, then debug capability is improved, but device complexity increases

Engineering Contradiction:
Improvedebug capabilityVSAvoidarchitecture complexity
Core Design Contradiction:
Difficulty of detecting and measuringVSDevice complexity

Solution Approach 1:

The patent merges command logging functionality with the existing status register infrastructure. The command history register uses the same control and data pathways as normal PiM operations, combining debugging capabilities with operational monitoring. This merging allows comprehensive command logging without duplicating existing register structures, thereby improving debug capability while limiting the increase in device complexity.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The command history register creates a simplified copy of the command sequence executed by the PiM logic block. Rather than implementing complex tracing mechanisms, the system copies essential command information into the status register structure, which can then be read by the host system for debugging. This copying approach provides adequate debug capability using existing register hardware, avoiding the need for more complex logging infrastructure.

Inventive Principle:
Principle #26Copying

Data Source

PatentUS20260044408A1Error detection and debug techniques for processing-in-memory architectures
Publication Date: 2026.02.12 GOOGLE LLC
  • US20260044408A1 patent drawing
  • US20260044408A1 patent drawing
  • US20260044408A1 patent drawing

AI summary

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for error detection of a processing-in-memory (PiM) subsystem that includes one or more integrated compute elements and one or more memory devices. One method can include executing, by the one or more integrated compute elements, instructions to perform a set of PiM operations on data stored in the one or more memory devices; upon execution of each PiM operation, storing, in a set of registers included in the PiM subsystem, PiM operation execution data, the PiM operation execution data including (1) a count indicating a number of PiM operations that have been executed and (2) data identifying previously-executed PiM operations; determining, during execution of the set of PiM operations, that a first error has occurred; in response to determining that the first error has occurred, stopping execution of further PiM operations.