Genesys
ToolGenesys is a test-program generator used in hardware functional verification. Evidence from a 1999 DSP verification paper describes Genesys as known technology previously used for large-scale processor families such as PowerPC and x86, and as the core technology used to build a generator for IBM's C54XDSP.
First seen 5/24/2026
Last seen 7/6/2026
Evidence 63 chunks
Wiki v3
WIKI
Genesys
Overview
Genesys is a test-program generator used in hardware functional verification. A 1999 paper, Fast Construction of Test-Program Generators for Digital Signal Processors, states that test-program generators play a key role in functional verification of large-scale processors and identifies Genesys as the core technology used in its DSP test-generator construction model.
NEIGHBORHOOD
No graph connections found for this entity yet. It may appear in future ingestion runs.
explore full graph →RELATIONSHIPS
50 connectionsThe paper mentions Genesys as a typical tool for pseudo-random test generation.
GENESYS investigates and uses various restart policies to escape local minima.
Genesys evaluates program state-space coverage by comparing GEN vs REAL training sets using SpreadRatio.
Genesys introduces the SpreadRatio metric to quantify state-space coverage.
Genesys is evaluated on its ability to improve ML-based power prediction accuracy.
Genesys implements the concept of automatic workload generation for creating representative training sets.
GENESYS uses an error function to evaluate candidate programs against the specification.
Genesys is evaluated on its ability to improve ML-based performance prediction accuracy.
GENESYS is compared with NetSyn in synthesis ability experiments.
GENESYS is compared with NPO in terms of synthesis rate and stability.
GENESYS is compared with RobustFill in synthesis ability experiments.
GENESYS is compared with PCCoder in synthesis ability experiments.
Genesys PE is compared to its predecessor Genesys in terms of verification productivity, quality and cost.
GENESYS is compared with PushGP in synthesis ability experiments.
Functional Verification Methodology for Microprocessors Using the Genesys Test-Program Generator ← uses 100% 2e
The paper uses Genesys as the primary test-program generator to demonstrate the methodology.
GENESYS can use token probabilities from a neural network model to bias bin widths.
GENESYS uses CMA-ES as its core optimization algorithm to solve the continuous program synthesis problem.
GENESYS uses a domain specific language with 41 tokens to define valid programs.
GENESYS uses an LSTM-based neural network model to predict token probabilities.
GENESYS uses bin mapping as its primary mapping scheme to convert continuous parameters into DSL tokens.
GENESYS uses a multivariate normal distribution to sample candidate programs via CMA-ES.
GENESYS uses input-output examples as the specification for program synthesis.
GENESYS explored multi-group mapping as an alternative mapping scheme.
GENESYS explored dynamic bin mapping as an alternative mapping scheme.
Genesys controls temporal locality using 8 bins.
GENESYS is compared with DeepCoder in synthesis ability experiments.
Genesys is designed to support predictive benchmarking by generating representative training sets.
Genesys uses convex hull area to define the SpreadRatio metric.
Genesys generates synthetic benchmarks covering a broad range of program behavior.
Genesys generates X86 assembly code using C with embedded assembly instructions.
Genesys uses data footprint to control the range of data addresses accessed.
Genesys models regular data memory accesses using simple strided stream classes.
Genesys: Automatically Generating Representative Training Sets for Predictive Benchmarking ← introduces 100% 2e
The paper proposes and introduces the Genesys automatic workload generation framework.
Genesys PE was developed as a successor and enhancement of Genesys with AI-based improvements.
Genesys is compared against SPEC CPU2006 benchmarks for state-space coverage.
Genesys is compared against MiBench benchmarks for state-space coverage.
Genesys is compared against MediaBench benchmarks for state-space coverage.
Genesys is compared against TPC-H benchmarks for state-space coverage.
Genesys uses instruction mix as one of its core controllable metrics.
The paper presents and introduces the GENESYS program synthesis framework.
Genesys uses branch transition rate as a core metric for controlling control-flow behavior.
GENESYS formulates program synthesis as a continuous optimization problem.
Genesys models branch misprediction rate using transition probability of branch instructions.
Genesys uses PCA to reduce dimensionality of performance data for state-space comparison.
Genesys uses KMeans clustering to visualize program behavior differences.
GENESYS adapts the covariance matrix during CMA-ES evolution.
Genesys models L1/L2 data cache miss rates as core metrics.
Genesys is a tool for pseudo-random directed test generation.
Genesys controls spatial locality through stride bins.
Genesys is a pseudo-random test-program generator.
CITATIONS
6 sources6 citations — click to expand
[1] Genesys is identified as a known test-program generator and as the core technology used in the DSP test-generator construction model. Fast Construction of Test-Program Generators for Digital Signal Processors
[2] Test-program generators play a key role in hardware functional verification of large-scale processors. Fast Construction of Test-Program Generators for Digital Signal Processors
[3] Genesys had been used for verification of large-scale processor families such as PowerPC and x86. Fast Construction of Test-Program Generators for Digital Signal Processors
[4] The DSP paper used Genesys for verification of the IBM C54XDSP, a recently announced fixed-point DSP. Fast Construction of Test-Program Generators for Digital Signal Processors
[5] The DSP paper says full-blown test-program generators were less popular in the DSP domain mainly because of limited time and money, and reports that a full test-program generator could be built quickly to improve verification coverage despite shorter development time. Fast Construction of Test-Program Generators for Digital Signal Processors
[6] The public GitHub repository allenai/genesys is a Python repository for a distributed language-model architecture discovery system. allenai/genesys