|
| 1 | +# Roadmap |
| 2 | + |
| 3 | +This project is research- and education-oriented, but there are several important improvements planned to make the codebase cleaner, faster, and more extensible. |
| 4 | + |
| 5 | +Contributions and discussion are very welcome. |
| 6 | + |
| 7 | +--- |
| 8 | + |
| 9 | +## 1. Decouple `GF2Matrix` from `Matrix<u8>` |
| 10 | + |
| 11 | +Currently, `GF2Matrix` is closely tied to the generic `Matrix<u8>` type. Conceptually, this is not ideal: a matrix over GF(2) is a distinct mathematical object with different semantics and invariants. |
| 12 | + |
| 13 | +Goals: |
| 14 | +- Make `GF2Matrix` a first-class type |
| 15 | +- Avoid treating `GF2Matrix` as an alias or thin wrapper over `Matrix<u8>` |
| 16 | +- Clarify invariants (entries are always in {0,1}) |
| 17 | +- Improve API clarity and correctness |
| 18 | +- Prepare the codebase for alternative internal representations |
| 19 | + |
| 20 | +This change is primarily architectural and should preserve existing functionality and behavior. |
| 21 | + |
| 22 | +--- |
| 23 | + |
| 24 | +## 2. Optimized storage and processing for GF(2) matrices |
| 25 | + |
| 26 | +Introduce a bit-packed representation for `GF2Matrix` to improve performance and memory efficiency. |
| 27 | + |
| 28 | +Goals: |
| 29 | +- Add a bit-packed backend (e.g. using `u64` blocks) |
| 30 | +- Keep the current non-bit-packed representation for clarity, |
| 31 | + experimentation, and teaching |
| 32 | +- Allow both representations to coexist |
| 33 | +- Avoid premature optimization or loss of readability |
| 34 | +- Enable future performance comparisons and benchmarks |
| 35 | + |
| 36 | +This project explicitly aims to remain suitable for research and education, so clarity and correctness are more important than maximal optimization. |
| 37 | + |
| 38 | +--- |
| 39 | + |
| 40 | +## 3. Generalization to linear algebra over arbitrary fields |
| 41 | + |
| 42 | +The long-term goal of `lin_algebra` is to support linear algebra over general fields, not only GF(2). |
| 43 | + |
| 44 | +GF(2) currently serves as the primary reference case: it is simple, well-defined, and particularly relevant for cryptography and coding theory. Design decisions and abstractions will be validated on GF(2) before being generalized. |
| 45 | + |
| 46 | +Goals: |
| 47 | +- Support linear algebra over arbitrary fields (e.g. GF(p), rationals) |
| 48 | +- Introduce field abstractions only when they are justified |
| 49 | +- Avoid over-engineering and unnecessary trait complexity |
| 50 | +- Keep the codebase readable and suitable for research and teaching |
| 51 | +- Allow experimentation with different field implementations |
| 52 | + |
| 53 | +Generalization will be approached incrementally, starting from GF(2) and extending to other fields once the design is stable. |
| 54 | + |
| 55 | +--- |
| 56 | + |
| 57 | +## Getting involved |
| 58 | + |
| 59 | +If you are interested in working on any of these topics: |
| 60 | +- comment on an existing issue, |
| 61 | +- open a new issue to discuss design ideas, |
| 62 | +- or submit a draft pull request. |
| 63 | + |
| 64 | +Even partial contributions, experiments, or design discussions are welcome. |
0 commit comments