ARC-0044: Building an EpochProgram of Similar Computational Loads

[OzielLa]

---

arc: ARC-0044
title: Building an EpochProgram of Similar Computational Loads
authors: OzielLa
topic: Protocol
status: Draft
created: 10/17/2024

Abstract

This proposal aims to address the issue of 1 to 4 times difference in puzzle- solving workload across different epochs by limiting the number of num_private instead of opcode . When the overall computational power of provers across the network stabilizes and significant fluctuations in the Proof Target occur, Puzzle rewards may vary by 1 to 4 times. These frequent fluctuations impact the estimated output model of the project.

Background

After a period of operation and technical analysis, it has been observed that the workload of puzzles has a significant correlation with the quantity of num_private generated during program execution.

Proposal

This proposal advises to replace the current limit on opcode with that on the quantity of num_private.

1. Evaluate a reasonable number of leaves as MAX_LEAVES_COUNT.
2. Build an InstructionSet by adding a data relationship between puzzle workloads and the quantity of num_private.
3. Enhance the building process of instruction_string.

Technical Considerations

1. When determining MAX_LEAVES_COUNT, it needs to ensure a sufficient number of Operation Instructions to guarantee sufficient randomness in the EpochProgram.
2. After introducing num_private into InstructionSet, how to reasonably adjust the original weight values needs to be considered .

Benefits

1. Proof Target adjustment can align with the computational power of provers.
2. No unusual fluctuations in Puzzle Rewards and Stake Rewards.

Risks and Mitigations

TBD

Conclusion

By building an EpochProgram of similar computational loads, that can facilitate the assessment of current network computational power change and ensure fairer returns for provers.

[xiongqiumin]

May I ask if there is sample code on how to achieve this goal