Skip to content

TODO.md

Latest commit

 

History

History
114 lines (96 loc) · 3.62 KB

TODO.md

File metadata and controls

114 lines (96 loc) · 3.62 KB

Immediate

  • Refactor scc_ir_parser messages into generic mixin.
  1. Document scc/lexer.h and scc/feed.h.
  2. Finish intermediate representation lexer.
  3. Parse intermediate representation.
  4. Generate intermediate representation using a builder.
  5. Implement a crap tonne of sanity checks in builder.
  6. Document builder.
  7. Setup testing infrastructure.
  • Build a relatively large test suite to cover IR.
  • Then (re)write vigilant and use it to manage test suites.
  • C/C++, not Ruby!
  • Seperate process to protect against faults.
  1. Publish.
  2. Improve build infrastructure.
  • AppVeyor and Travis builds.
  • Refactor Ryb a lot.

Nebulous

TODO:

  • Generate DOT files (for Graphviz).
    • Layout engines are terrible; write my own?
  • Optimize compilation by making "modules" first class?
  • "Asynchronous" pipeline a la Jai.
  • Unicode
    • Implement UTF-8 encoding and decoding.
    • Implement Unicode canonicalization.
    • Fix assumptions about ASCII.
  • Optimizations
    • Strength reduction.
    • Instruction selection pass.
  • Analysis
    • Poisoning.
      • Yell at user for undefined behavior.
  • Big ol' list to make classification of errors easier.
    • Use a substitution macro!
  • Smart console.
    • Look to Ninja for technical specifics.
    • Look to pull requests against Ninja for UI/UX ideas.

PERF:

  • File/feed caching to improve pre-processor #include performance.
    • Ideally, we would compile all shaders with one invocation of the driver.
      • The driver could take hints about "ONE" or "MULTIPLE." If more than one, but hinted otherwise, yell at the user and then setup caching.

REFACTOR:

  • Replace SCC_TRUE and SCC_FALSE with true and false?
  • Refactor number parsing in scc_ir_lexer_handle_number into two shared functions...
    • One that sanity checks the number and determines its components.
      • Indicate if a binary, octal, decimal, or hexadecimal number by checking significand.
      • Determine width.
      • Extract sign.
      • Extract exponent, if specified.
      • Indicate relevant spans...
        • Sign
        • Significand
        • Integer part
        • Fractional part
        • Exponent
    • Another that parses it into a fitting type.
      • A signed or unsigned 64-bit integer.
      • A signed 64-bit floating-point number.
      • Check for ERANGE when parsing.
  • Handle old control characters.
  • Only accept UTF-8 in comments.
  • Generate and use a radix tree to make matching faster.
    • Finite stage recognizers are cool, but overkill. Future work?

Custom graph visualization tool.

https://en.wikipedia.org/wiki/Graph_drawing

Shader Cross-Compiler

Translator Maps intermediate representation to graph. Reflects modifications?

Tool

Parser Constitutes graph from a language like DOT.

Builder Aids in building a graph representation. Tries to avoid pointer chasing; indexes rather than pointers.

Selector/Selection Refers to a portion of the graph. In our case, perhaps a function and optionally dependencies. Visual selection may differ from logical selection, e.g. "zoom."

LayoutEngine Lays out graph, ideally planar.

Library Various graph components. Components may be defined in a language like DOT.

Presenter Caching generates flat list, a mapping from source to flattened span. Makes drawing fast! Presents a graph on canvas. Can rely on transparency to allow layers!

Canvas Immediate mode rendering of two-dimensional primitives, like rectangles and lines. Save to image, offline. Draw in real-time through OpenGL.