# Hybrid Branch Predictor Corpus Analysis

- Generated: 2026-05-07T09:00:09+00:00
- Papers: 365
- Embedding provider: openai
- Embedding model: `text-embedding-3-large`
- Chunks: 5847
- Estimated embedding tokens: 3012920
- Estimated OpenAI cost: 0.39168 USD

## Scope Buckets

- Application-Adjacent Branch Behavior: 5
- Core Predictor Architecture: 263
- Power / Embedded / Energy: 11
- Security / Side-Channel: 49
- Simulator / FPGA / Soft Processor: 8
- Software / Static / WCET: 21
- Testing / Verification / Faults: 8

## application_adjacent:0 - Sorting and Microservice Control-Flow Behavior

- Scope: Application-Adjacent Branch Behavior
- Size: 4
- Year span: 2008 to 2025
- Top terms: sorting, algorithms, cyprus, will, algorithm, microservices, instruction, modern

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2008 | An experimental study of sorting and branch prediction | 14 | n/a | 0.9593 | 0.2114 |
| 2 | 2022 | Improving mergesort performance by optimizing branch prediction | 2 | n/a | 0.4396 | 0.0000 |
| 3 | 2022 | Improving quicksort performance by optimizing branch prediction | 2 | n/a | 0.4279 | 0.0514 |
| 4 | 2025 | Leveraging control flow similarity to reduce branch predictor cold effects in microservices | 0 | n/a | 0.0000 | 1.0000 |

Top-1 note: An experimental study of sorting and branch prediction is ranked highly because it combines cluster impact 0.959, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.211, so it is treated as a canonical family member within this cluster.

Top-2 note: Improving mergesort performance by optimizing branch prediction is ranked highly because it combines cluster impact 0.440, downstream semantic support 0.227, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

Top-3 note: Improving quicksort performance by optimizing branch prediction is ranked highly because it combines cluster impact 0.428, downstream semantic support 0.222, and citation metadata status `ok`. Its novelty score is 0.051, so it is treated as a canonical family member within this cluster.

## application_adjacent:1 - Sobel Edge-Detection Branch Behavior

- Scope: Application-Adjacent Branch Behavior
- Size: 1
- Year span: 2002 to 2002
- Top terms: sobel, edge, conditional, multi-dimensional, execution, such, retiming, sobel edge

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2002 | Predicting conditional branch outcomes on a sobel edge detecting filter | 1 | n/a | 0.8077 | 1.0000 |

Top-1 note: Predicting conditional branch outcomes on a sobel edge detecting filter is ranked highly because it combines cluster impact 0.808, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

## core_architecture:0 - Modern Industrial / Security-Aware Predictor Structures

- Scope: Core Predictor Architecture
- Size: 22
- Year span: 2009 to 2026
- Top terms: instruction, instructions, history, accuracy, wkh, data, university, target

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2019 | Brb mitigating branch predictor side channels | 37 | n/a | 0.9460 | 0.1123 |
| 2 | 2019 | Branch prediction is not a solved problem measurements opportunities and future directions | 25 | n/a | 0.9113 | 0.1415 |
| 3 | 2019 | Towards the adoption of local branch predictors in modern out of order superscalar processors | 6 | n/a | 0.6707 | 0.1930 |
| 4 | 2009 | Experiment flows and microbenchmarks for reverse engineering of branch predictor structures | 50 | n/a | 0.6655 | 0.6396 |
| 5 | 2020 | Securing branch predictors with two level encryption | 17 | n/a | 0.6608 | 0.2658 |

Top-1 note: Brb mitigating branch predictor side channels is ranked highly because it combines cluster impact 0.946, downstream semantic support 0.988, and citation metadata status `ok`. Its novelty score is 0.112, so it is treated as a canonical family member within this cluster.

Top-2 note: Branch prediction is not a solved problem measurements opportunities and future directions is ranked highly because it combines cluster impact 0.911, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.142, so it is treated as a canonical family member within this cluster.

Top-3 note: Towards the adoption of local branch predictors in modern out of order superscalar processors is ranked highly because it combines cluster impact 0.671, downstream semantic support 0.753, and citation metadata status `ok`. Its novelty score is 0.193, so it is treated as a canonical family member within this cluster.

## core_architecture:1 - Early Correlation / Two-Level Foundations

- Scope: Core Predictor Architecture
- Size: 51
- Year span: 1991 to 2008
- Top terms: history, instruction, schemes, instructions, dynamic, accuracy, static, address

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 1992 | Improving the accuracy of dynamic branch prediction using branch correlation | 365 | 18 | 0.9078 | 0.1643 |
| 2 | 1993 | A comparison of dynamic branch predictors that use two levels of branch history | 377 | n/a | 0.9032 | 0.1017 |
| 3 | 1994 | Improving the accuracy of static branch prediction using branch correlation | 88 | n/a | 0.8384 | 0.0781 |
| 4 | 1992 | Alternative implementations of two level adaptive branch prediction | 502 | n/a | 0.8351 | 0.1421 |
| 5 | 1991 | Two level adaptive training branch prediction | 542 | 20 | 0.7697 | 0.3641 |

Top-1 note: Improving the accuracy of dynamic branch prediction using branch correlation is ranked highly because it combines cluster impact 0.908, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.164, so it is treated as a canonical family member within this cluster.

Top-2 note: A comparison of dynamic branch predictors that use two levels of branch history is ranked highly because it combines cluster impact 0.903, downstream semantic support 0.909, and citation metadata status `ok`. Its novelty score is 0.102, so it is treated as a canonical family member within this cluster.

Top-3 note: Improving the accuracy of static branch prediction using branch correlation is ranked highly because it combines cluster impact 0.838, downstream semantic support 0.930, and citation metadata status `ok`. Its novelty score is 0.078, so it is treated as a canonical family member within this cluster.

## core_architecture:2 - Dynamic / Static Predictor Characterization

- Scope: Core Predictor Architecture
- Size: 31
- Year span: 1996 to 2025
- Top terms: history, dynamic, instruction, static, instructions, taken, execution, accuracy

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2013 | Dynamic branch prediction for high level synthesis | 9 | n/a | 0.7283 | 0.3837 |
| 2 | 1996 | An analysis of dynamic branch prediction schemes on system workloads | 67 | n/a | 0.6976 | 0.6041 |
| 3 | 2019 | Exploring deep learning based branch prediction for computer devices | 0 | n/a | 0.6095 | 0.1639 |
| 4 | 2010 | An analysis of hard to predict branches | 8 | n/a | 0.5538 | 0.4445 |
| 5 | 2011 | Improving branch prediction related wcet abstract interpretation | 1 | n/a | 0.5274 | 0.4635 |

Top-1 note: Dynamic branch prediction for high level synthesis is ranked highly because it combines cluster impact 0.728, downstream semantic support 0.922, and citation metadata status `ok`. Its novelty score is 0.384, so it is treated as a canonical family member within this cluster.

Top-2 note: An analysis of dynamic branch prediction schemes on system workloads is ranked highly because it combines cluster impact 0.698, downstream semantic support 0.585, and citation metadata status `ok`. Its novelty score is 0.604, so it is treated as a canonical family member within this cluster.

Top-3 note: Exploring deep learning based branch prediction for computer devices is ranked highly because it combines cluster impact 0.610, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.164, so it is treated as a canonical family member within this cluster.

## core_architecture:3 - Perceptron / Neural

- Scope: Core Predictor Architecture
- Size: 34
- Year span: 2002 to 2023
- Top terms: neural, perceptron, history, accuracy, university, more, section, one

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2002 | Dynamic branch prediction with perceptrons | 406 | n/a | 0.9879 | 0.0628 |
| 2 | 2002 | Neural methods for dynamic branch prediction | 197 | 19 | 0.8821 | 0.0000 |
| 3 | 2004 | Fast path based neural branch prediction | 71 | n/a | 0.5708 | 0.1957 |
| 4 | 2005 | Improved latency and accuracy for neural branch prediction | 19 | 3 | 0.5319 | 0.0786 |
| 5 | 2005 | Merging path and gshare indexing in perceptron branch prediction | 40 | 3 | 0.4541 | 0.2682 |

Top-1 note: Dynamic branch prediction with perceptrons is ranked highly because it combines cluster impact 0.988, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.063, so it is treated as a canonical family member within this cluster.

Top-2 note: Neural methods for dynamic branch prediction is ranked highly because it combines cluster impact 0.882, downstream semantic support 0.759, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

Top-3 note: Fast path based neural branch prediction is ranked highly because it combines cluster impact 0.571, downstream semantic support 0.358, and citation metadata status `ok`. Its novelty score is 0.196, so it is treated as a canonical family member within this cluster.

## core_architecture:4 - BTB / Fetch / Target

- Scope: Core Predictor Architecture
- Size: 40
- Year span: 1991 to 2024
- Top terms: instruction, btb, instructions, target, cache, address, indirect, fetch

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 1993 | Branch target buffer design and optimization | 146 | n/a | 0.9663 | 0.1440 |
| 2 | 1993 | Increasing the instruction fetch rate via multiple branch prediction and a branch address cache | 159 | 11 | 0.8452 | 0.3081 |
| 3 | 1991 | Strategies for branch target buffers | 36 | 1 | 0.7217 | 0.0000 |
| 4 | 1994 | Fast and accurate instruction fetch and branch prediction | 92 | n/a | 0.6069 | 0.4539 |
| 5 | 1997 | Instruction prefetching using branch prediction information | 52 | n/a | 0.5357 | 0.5152 |

Top-1 note: Branch target buffer design and optimization is ranked highly because it combines cluster impact 0.966, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.144, so it is treated as a canonical family member within this cluster.

Top-2 note: Increasing the instruction fetch rate via multiple branch prediction and a branch address cache is ranked highly because it combines cluster impact 0.845, downstream semantic support 0.690, and citation metadata status `ok`. Its novelty score is 0.308, so it is treated as a canonical family member within this cluster.

Top-3 note: Strategies for branch target buffers is ranked highly because it combines cluster impact 0.722, downstream semantic support 0.889, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

## core_architecture:5 - Long Global-History / GEHL-TAGE Evolution

- Scope: Core Predictor Architecture
- Size: 28
- Year span: 2002 to 2020
- Top terms: history, accuracy, global, tage, more, section, global history, instruction

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2005 | Analysis of the o geometric history length branch predictor | 125 | 11 | 0.9910 | 0.0000 |
| 2 | 2002 | Design tradeoffs for the alpha ev8 conditional branch predictor | 147 | n/a | 0.6376 | 0.5226 |
| 3 | 2011 | A new case for the tage branch predictor | 101 | n/a | 0.5381 | 0.4415 |
| 4 | 2009 | Creating artificial global history to improve branch prediction accuracy | 14 | n/a | 0.4923 | 0.2045 |
| 5 | 2012 | Energy efficient branch prediction with compiler guided history stack | n/a | n/a | 0.4557 | 0.2270 |

Top-1 note: Analysis of the o geometric history length branch predictor is ranked highly because it combines cluster impact 0.991, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

Top-2 note: Design tradeoffs for the alpha ev8 conditional branch predictor is ranked highly because it combines cluster impact 0.638, downstream semantic support 0.416, and citation metadata status `ok`. Its novelty score is 0.523, so it is treated as a canonical family member within this cluster.

Top-3 note: A new case for the tage branch predictor is ranked highly because it combines cluster impact 0.538, downstream semantic support 0.219, and citation metadata status `ok`. Its novelty score is 0.441, so it is treated as a canonical family member within this cluster.

## core_architecture:6 - Energy-Aware Predictor and Fetch Design

- Scope: Core Predictor Architecture
- Size: 23
- Year span: 2003 to 2024
- Top terms: power, energy, dynamic, consumption, instruction, btb, instructions, reduce

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2003 | Branch prediction on demand an energy efficient solution microprocessor architecture | 15 | n/a | 0.8861 | 0.0000 |
| 2 | 2004 | Power aware branch prediction characterization and design | 57 | 10 | 0.8860 | 0.0557 |
| 3 | 2004 | Power aware branch prediction techniques | n/a | n/a | 0.8343 | 0.1126 |
| 4 | 2003 | Customizing the branch predictor to reduce complexity and energy consumption | 27 | n/a | 0.7537 | 0.1218 |
| 5 | 2003 | Branch prediction on demand | 49 | 2 | 0.6949 | 0.1403 |

Top-1 note: Branch prediction on demand an energy efficient solution microprocessor architecture is ranked highly because it combines cluster impact 0.886, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

Top-2 note: Power aware branch prediction characterization and design is ranked highly because it combines cluster impact 0.886, downstream semantic support 0.698, and citation metadata status `ok`. Its novelty score is 0.056, so it is treated as a canonical family member within this cluster.

Top-3 note: Power aware branch prediction techniques is ranked highly because it combines cluster impact 0.834, downstream semantic support 0.812, and citation metadata status `unmatched`. Its novelty score is 0.113, so it is treated as a canonical family member within this cluster.

## core_architecture:7 - TAGE and Soft-Processor Implementations

- Scope: Core Predictor Architecture
- Size: 34
- Year span: 2003 to 2026
- Top terms: instruction, pipeline, accuracy, instructions, tage, history, used, execution

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2022 | An optimized risc v processor with five stage pipelining using tournament branch predictor for e | n/a | n/a | 0.7842 | 0.0026 |
| 2 | 2017 | Design space exploration of tage branch predictor with ultra small ram | 9 | n/a | 0.7609 | 0.1939 |
| 3 | 2003 | Branch predictor design and performance estimation for a high performance embedded microprocesso | 1 | 0 | 0.7368 | 0.4273 |
| 4 | 2017 | A high performance branch predictor design considering memory capacity limitations | 8 | n/a | 0.7237 | 0.2685 |
| 5 | 2019 | An efficient implementation of a tage branch predictor for soft processors on fpga | 5 | n/a | 0.6746 | 0.0319 |

Top-1 note: An optimized risc v processor with five stage pipelining using tournament branch predictor for e is ranked highly because it combines cluster impact 0.784, downstream semantic support 0.693, and citation metadata status `unmatched`. Its novelty score is 0.003, so it is treated as a canonical family member within this cluster.

Top-2 note: Design space exploration of tage branch predictor with ultra small ram is ranked highly because it combines cluster impact 0.761, downstream semantic support 0.582, and citation metadata status `ok`. Its novelty score is 0.194, so it is treated as a canonical family member within this cluster.

Top-3 note: Branch predictor design and performance estimation for a high performance embedded microprocesso is ranked highly because it combines cluster impact 0.737, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.427, so it is treated as a canonical family member within this cluster.

## power_embedded:0 - Embedded BTB / Branch-Address Power Reduction

- Scope: Power / Embedded / Energy
- Size: 6
- Year span: 2000 to 2011
- Top terms: power, embedded, btb, consumption, power consumption, instruction, information, university

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2006 | Power efficient branch prediction through early identification of branch addresses | 20 | 3 | 0.8916 | 0.2221 |
| 2 | 2003 | Low power branch target buffer for application specific embedded processors | n/a | n/a | 0.7462 | 0.4347 |
| 3 | 2010 | Low power branch prediction for embedded application processors | 13 | n/a | 0.6051 | 0.0000 |
| 4 | 2010 | Low power branch predictor for embedded processors | 9 | n/a | 0.5274 | 0.1951 |
| 5 | 2000 | Offline program re mapping to improve branch prediction efficiency in embedded systems | 1 | 0 | 0.4137 | 1.0000 |

Top-1 note: Power efficient branch prediction through early identification of branch addresses is ranked highly because it combines cluster impact 0.892, downstream semantic support 0.786, and citation metadata status `ok`. Its novelty score is 0.222, so it is treated as a canonical family member within this cluster.

Top-2 note: Low power branch target buffer for application specific embedded processors is ranked highly because it combines cluster impact 0.746, downstream semantic support 0.824, and citation metadata status `unmatched`. Its novelty score is 0.435, so it is treated as a canonical family member within this cluster.

Top-3 note: Low power branch prediction for embedded application processors is ranked highly because it combines cluster impact 0.605, downstream semantic support 0.227, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

## power_embedded:1 - Loop-Based Leakage Energy Control

- Scope: Power / Embedded / Energy
- Size: 2
- Year span: 2004 to 2007
- Top terms: leakage, energy, leakage energy, carbondale, illinois, entries, zhang, southern

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2007 | Reducing branch predictor leakage energy by exploiting loops | 9 | 1 | 0.5000 | 1.0000 |
| 2 | 2004 | Loop based leakage control for branch predictors | 0 | 0 | 0.3500 | 1.0000 |

Top-1 note: Reducing branch predictor leakage energy by exploiting loops is ranked highly because it combines cluster impact 0.500, downstream semantic support 0.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

Top-2 note: Loop based leakage control for branch predictors is ranked highly because it combines cluster impact 0.350, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

## power_embedded:2 - Embedded Forth Core Outlier

- Scope: Power / Embedded / Energy
- Size: 3
- Year span: 2013 to 2014
- Top terms: cache, locking, embedded, forth, instructions, path, cache locking, core

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2013 | An embedded forth core with floating point and branch prediction | 4 | n/a | 0.7463 | 1.0000 |
| 2 | 2013 | Pseudo dual path processing to reduce the branch misprediction penalty in embedded processors | 0 | n/a | 0.6410 | 0.0000 |
| 3 | 2014 | Branch prediction directed dynamic instruction cache locking for embedded systems | 2 | n/a | 0.3664 | 0.3691 |

Top-1 note: An embedded forth core with floating point and branch prediction is ranked highly because it combines cluster impact 0.746, downstream semantic support 0.863, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

Top-2 note: Pseudo dual path processing to reduce the branch misprediction penalty in embedded processors is ranked highly because it combines cluster impact 0.641, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

Top-3 note: Branch prediction directed dynamic instruction cache locking for embedded systems is ranked highly because it combines cluster impact 0.366, downstream semantic support 0.000, and citation metadata status `ok`. Its novelty score is 0.369, so it is treated as a canonical family member within this cluster.

## security:0 - Branch Predictor Side Channels and Mitigations

- Scope: Security / Side-Channel
- Size: 41
- Year span: 2007 to 2026
- Top terms: attacks, execution, security, attack, such, target, hardware, transient

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2018 | Branchscope | 179 | n/a | 0.7179 | 0.2404 |
| 2 | 2021 | Leaking secrets through modern branch predictor in the speculative world | 1 | n/a | 0.6690 | 0.0812 |
| 3 | 2007 | On the power of simple branch prediction analysis | 301 | 16 | 0.6546 | 0.5345 |
| 4 | 2020 | Exploring branch predictors for constructing transient execution trojans | 31 | n/a | 0.5913 | 0.2068 |
| 5 | 2014 | The micro architectural support countermeasures against the branch prediction analysis attack | 13 | n/a | 0.5199 | 0.3171 |

Top-1 note: Branchscope is ranked highly because it combines cluster impact 0.718, downstream semantic support 0.547, and citation metadata status `ok`. Its novelty score is 0.240, so it is treated as a canonical family member within this cluster.

Top-2 note: Leaking secrets through modern branch predictor in the speculative world is ranked highly because it combines cluster impact 0.669, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.081, so it is treated as a canonical family member within this cluster.

Top-3 note: On the power of simple branch prediction analysis is ranked highly because it combines cluster impact 0.655, downstream semantic support 0.242, and citation metadata status `ok`. Its novelty score is 0.534, so it is treated as a canonical family member within this cluster.

## security:1 - Return Stack Buffer Attacks and Defenses

- Scope: Security / Side-Channel
- Size: 8
- Year span: 2004 to 2024
- Top terms: return, attack, address, stack, attacks, execution, return address, buffer

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2018 | Spectre returns speculation attacks using the return stack buffer | 97 | n/a | 0.8512 | 0.0284 |
| 2 | 2018 | Ret2spec | 185 | n/a | 0.8330 | 0.2766 |
| 3 | 2004 | Repairing return address stack for buffer overflow protection | 22 | 0 | 0.7553 | 0.8847 |
| 4 | 2024 | Spectre returns speculation attacks using the return stack buffer | 97 | n/a | 0.5073 | 0.0000 |
| 5 | 2019 | Reinforcing meltdown attack by using a return stack buffer | 7 | n/a | 0.4695 | 0.1667 |

Top-1 note: Spectre returns speculation attacks using the return stack buffer is ranked highly because it combines cluster impact 0.851, downstream semantic support 0.779, and citation metadata status `ok`. Its novelty score is 0.028, so it is treated as a canonical family member within this cluster.

Top-2 note: Ret2spec is ranked highly because it combines cluster impact 0.833, downstream semantic support 0.746, and citation metadata status `ok`. Its novelty score is 0.277, so it is treated as a canonical family member within this cluster.

Top-3 note: Repairing return address stack for buffer overflow protection is ranked highly because it combines cluster impact 0.755, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.885, so it is treated as an atypical deviation within this cluster.

## simulator_fpga:0 - Pipeline Simulator and FPGA Implementation

- Scope: Simulator / FPGA / Soft Processor
- Size: 2
- Year span: 2019 to 2026
- Top terms: instruction, simulator, execution, pipeline, pipelined, mars, hazards, computer

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2019 | Pipelined mips simulation a plug in to mars simulator for supporting pipeline simulation and bra | n/a | n/a | 0.7000 | 1.0000 |
| 2 | 2026 | Five stage pipeline simulator in c forwarding and branch prediction | 0 | n/a | 0.3590 | 1.0000 |

Top-1 note: Pipelined mips simulation a plug in to mars simulator for supporting pipeline simulation and bra is ranked highly because it combines cluster impact 0.700, downstream semantic support 1.000, and citation metadata status `unmatched`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

Top-2 note: Five stage pipeline simulator in c forwarding and branch prediction is ranked highly because it combines cluster impact 0.359, downstream semantic support 0.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

## simulator_fpga:1 - RISC-V / FPGA Predictor Validation

- Scope: Simulator / FPGA / Soft Processor
- Size: 6
- Year span: 1998 to 2026
- Top terms: risc-v, instruction, instructions, core, validation, convolution, computer, processing

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 1998 | Retrospective a study of branch prediction strategies | 738 | 34 | 0.8500 | 1.0000 |
| 2 | 2024 | Optimizing convolutional image processing on risc v using branch prediction on fpga | 2 | n/a | 0.5166 | 0.4466 |
| 3 | 2025 | Fpga implementation of a pipelined rv32im processor with dynamic branch prediction | 0 | n/a | 0.2833 | 0.0000 |
| 4 | 2025 | Design of a dual issue superscalar risc v core with dynamic branch prediction | 0 | n/a | 0.2735 | 0.0488 |
| 5 | 2026 | Ml guided branch predictor optimization for embedded risc v socs | 0 | n/a | 0.1672 | 0.1305 |

Top-1 note: Retrospective a study of branch prediction strategies is ranked highly because it combines cluster impact 0.850, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

Top-2 note: Optimizing convolutional image processing on risc v using branch prediction on fpga is ranked highly because it combines cluster impact 0.517, downstream semantic support 0.781, and citation metadata status `ok`. Its novelty score is 0.447, so it is treated as a canonical family member within this cluster.

Top-3 note: Fpga implementation of a pipelined rv32im processor with dynamic branch prediction is ranked highly because it combines cluster impact 0.283, downstream semantic support 0.203, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

## software_static_wcet:0 - Program Trace Compression and Debugging

- Scope: Software / Static / WCET
- Size: 2
- Year span: 2010 to 2014
- Top terms: debugging, software, program, embedded, trace, systems, tracing, execution

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2010 | Real time unobtrusive program execution trace compression using branch predictor events | 13 | n/a | 0.8077 | 1.0000 |
| 2 | 2014 | Using branch predictors and variable encoding for on the fly program tracing | 8 | n/a | 0.0000 | 1.0000 |

Top-1 note: Real time unobtrusive program execution trace compression using branch predictor events is ranked highly because it combines cluster impact 0.808, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

Top-2 note: Using branch predictors and variable encoding for on the fly program tracing is ranked highly because it combines cluster impact 0.000, downstream semantic support 0.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

## software_static_wcet:1 - Historical MIPS / Branch Prediction Retrospective

- Scope: Software / Static / WCET
- Size: 2
- Year span: 2016 to 2023
- Top terms: mips, hardware, loop, instruction, ldb, hps, simple, outcomes

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2016 | Common bonds mips hps two level branch prediction and compressed code risc processor | 5 | n/a | 0.8077 | 1.0000 |
| 2 | 2023 | By software branch prediction in loops | 2 | n/a | 0.0000 | 1.0000 |

Top-1 note: Common bonds mips hps two level branch prediction and compressed code risc processor is ranked highly because it combines cluster impact 0.808, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

Top-2 note: By software branch prediction in loops is ranked highly because it combines cluster impact 0.000, downstream semantic support 0.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

## software_static_wcet:2 - WCET and Timing Predictability

- Scope: Software / Static / WCET
- Size: 4
- Year span: 2009 to 2019
- Top terms: wcet, analysis, bounds, systems, static, execution, time, dynamic

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2009 | Branch target buffers wcet analysis framework and timing predictability | 16 | n/a | 0.9154 | 0.4397 |
| 2 | 2011 | Wcet driven branch prediction aware code positioning | 5 | n/a | 0.4913 | 0.7245 |
| 3 | 2015 | Persistence based branch misprediction bounds for wcet analysis | 2 | n/a | 0.3368 | 0.0000 |
| 4 | 2019 | Evaluating software diversity in branch prediction analyses for static wcet estimation | 2 | n/a | 0.0000 | 1.0000 |

Top-1 note: Branch target buffers wcet analysis framework and timing predictability is ranked highly because it combines cluster impact 0.915, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.440, so it is treated as a canonical family member within this cluster.

Top-2 note: Wcet driven branch prediction aware code positioning is ranked highly because it combines cluster impact 0.491, downstream semantic support 0.657, and citation metadata status `ok`. Its novelty score is 0.725, so it is treated as an atypical deviation within this cluster.

Top-3 note: Persistence based branch misprediction bounds for wcet analysis is ranked highly because it combines cluster impact 0.337, downstream semantic support 0.322, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

## software_static_wcet:3 - Static Prediction and Compiler Control-Flow Techniques

- Scope: Software / Static / WCET
- Size: 8
- Year span: 1992 to 2009
- Top terms: execution, instructions, instruction, program, value, guarded, conditional, dynamic

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 1992 | Predicting conditional branch directions from previous runs of a program | 270 | 13 | 0.9714 | 0.1709 |
| 2 | 1993 | Branch prediction for free | 287 | n/a | 0.8732 | 0.0000 |
| 3 | 1993 | Experiments with low level speculative computation based on multiple branch prediction | 20 | n/a | 0.5501 | 0.3763 |
| 4 | 1995 | Accurate static branch prediction by value range propagation | 122 | n/a | 0.3470 | 0.7816 |
| 5 | 1994 | Guarded execution and branch prediction in dynamic ilp processors | 65 | n/a | 0.3017 | 0.8123 |

Top-1 note: Predicting conditional branch directions from previous runs of a program is ranked highly because it combines cluster impact 0.971, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.171, so it is treated as a canonical family member within this cluster.

Top-2 note: Branch prediction for free is ranked highly because it combines cluster impact 0.873, downstream semantic support 0.717, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

Top-3 note: Experiments with low level speculative computation based on multiple branch prediction is ranked highly because it combines cluster impact 0.550, downstream semantic support 0.528, and citation metadata status `ok`. Its novelty score is 0.376, so it is treated as a canonical family member within this cluster.

## software_static_wcet:4 - Interpreter Indirect-Branch Prediction

- Scope: Software / Static / WCET
- Size: 3
- Year span: 2003 to 2015
- Top terms: interpreters, indirect, interpreter, accuracy, code, techniques, section, execution

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2003 | Optimizing indirect branch prediction accuracy in virtual machine interpreters | 61 | n/a | 0.9456 | 0.0964 |
| 2 | 2007 | Optimizing indirect branch prediction accuracy in virtual machine interpreters | 97 | 9 | 0.8153 | 0.0000 |
| 3 | 2015 | Branch prediction and the performance of interpreters x2014 don t trust folklore | n/a | n/a | 0.0000 | 1.0000 |

Top-1 note: Optimizing indirect branch prediction accuracy in virtual machine interpreters is ranked highly because it combines cluster impact 0.946, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.096, so it is treated as a canonical family member within this cluster.

Top-2 note: Optimizing indirect branch prediction accuracy in virtual machine interpreters is ranked highly because it combines cluster impact 0.815, downstream semantic support 0.472, and citation metadata status `ok`. Its novelty score is 0.000, so it is treated as a canonical family member within this cluster.

Top-3 note: Branch prediction and the performance of interpreters x2014 don t trust folklore is ranked highly because it combines cluster impact 0.000, downstream semantic support 0.000, and citation metadata status `unmatched`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

## software_static_wcet:5 - Branch-History-Guided Symbolic Execution

- Scope: Software / Static / WCET
- Size: 2
- Year span: 2017 to 2018
- Top terms: software, testing, execution, program, test, search, symbolic, dynamic

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2017 | A branch history directed heuristic search for effective binary level dynamic symbolic execution | 2 | n/a | 0.8077 | 1.0000 |
| 2 | 2018 | Perdice towards discovering software inefficiencies leading to cache misses and branch mispredic | n/a | n/a | 0.0000 | 1.0000 |

Top-1 note: A branch history directed heuristic search for effective binary level dynamic symbolic execution is ranked highly because it combines cluster impact 0.808, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

Top-2 note: Perdice towards discovering software inefficiencies leading to cache misses and branch mispredic is ranked highly because it combines cluster impact 0.000, downstream semantic support 0.000, and citation metadata status `unmatched`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

## testing_verification:0 - Functional BPU / BTB Fault Testing

- Scope: Testing / Verification / Faults
- Size: 7
- Year span: 2000 to 2015
- Top terms: test, functional, approach, units, faults, bpu, memory, instruction

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2011 | On the functional test of branch prediction units based on branch history table | 9 | n/a | 0.7877 | 0.1399 |
| 2 | 2000 | Formal verification of superscale microprocessors with multicycle functional units exception and | 116 | 5 | 0.6943 | 1.0000 |
| 3 | 2012 | A sbst strategy to test microprocessors branch target buffer | 9 | n/a | 0.6814 | 0.1635 |
| 4 | 2013 | On the functional test of the btb logic in pipelined and superscalar processors | 10 | n/a | 0.5963 | 0.1453 |
| 5 | 2015 | On the functional test of branch prediction units | 26 | n/a | 0.4407 | 0.0000 |

Top-1 note: On the functional test of branch prediction units based on branch history table is ranked highly because it combines cluster impact 0.788, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 0.140, so it is treated as a canonical family member within this cluster.

Top-2 note: Formal verification of superscale microprocessors with multicycle functional units exception and is ranked highly because it combines cluster impact 0.694, downstream semantic support 0.555, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.

Top-3 note: A sbst strategy to test microprocessors branch target buffer is ranked highly because it combines cluster impact 0.681, downstream semantic support 0.773, and citation metadata status `ok`. Its novelty score is 0.164, so it is treated as a canonical family member within this cluster.

## testing_verification:1 - Return-Stack Hardware Trojan Outlier

- Scope: Testing / Verification / Faults
- Size: 1
- Year span: 2018 to 2018
- Top terms: hts, trojan, ras, stuck-at, hardware, effect, logic, payload

| Rank | Year | Paper | Citations | Influential | Impact | Novelty |
| ---: | ---: | --- | ---: | ---: | ---: | ---: |
| 1 | 2018 | Stuck at 0 1 trojans on return address stack | 0 | n/a | 0.8077 | 1.0000 |

Top-1 note: Stuck at 0 1 trojans on return address stack is ranked highly because it combines cluster impact 0.808, downstream semantic support 1.000, and citation metadata status `ok`. Its novelty score is 1.000, so it is treated as an atypical deviation within this cluster.