Salsa 3.0 overall plan

• Working towards an 1.0alpha1 release

  • The too

• Big projects we are using to drive progress:

  • Red-knot
  • Rust-analyzer

• Hassle-free interface

• Hashtable-free tracked functions

• Lightweight tracked structs

• Documentation for current setup

• Ez Parallel execution – David Barsky

  • Ruff works around the lack of parallelism through cloning: https://github.com/astral-sh/ruff/blob/ca0ae0a484906dd29a55a97572df527667330507/crates/red_knot_workspace/src/workspace.rs#L194-L222
  • basically ready to merge. David ported tests, but Miri is unhappy. Should we just treat them as warnings?

• Garbage collection for interned values – has mentioned instructions, needs owner

• Coarse-grained dependencies

• Remove the global hashtable for tracked fns with >1 argument

• Port rust-analyzer to salsa 3.0: not a salsa task, but requires entries above it

  • 01/06/2025 meeting notes between David and Niko: https://hackmd.io/A_-2uaXOSTuiYIUs6bbpXA
  • rust-analyzer runs on new Salsa! David's been using it. Concerns are that memory is too high. Working on fixing it.

• Fixed-point solving

• Speculative execution

• New trait solver integration

• WASM compatibility and/or async

  • Immediate resolution: don't use Result, but evaluate async/await in ~3 months.
  • Or maybe we don't need this because of cycle resolution.

• Persistent caching/deserialization/serialization

• Addressing papercuts

Adjacent documents

• Salsa patterns
  • Dada using new Salsa: https://github.com/nikomatsakis/dada/tree/checked-dada
  • Salsa Patterns and FAQs: https://hackmd.io/7f8isuLzQ7CA1GiPqNDYfA
• Salsa 1.0 versus 3.0: https://hackmd.io/z5M6WYHFQ6qKxYF8Wwjv_w
  • David is working on a query_group macro: https://salsa.zulipchat.com/#narrow/channel/333573-Contributing-to-Salsa/topic/OG.20Salsa-Esque.20Macro

Per meeting notes

• 2025-06-11
• 2025-04-16
• 2025-03-19
• 2025-03-05
• 2025-02-19
• 2025-01-22
• Meeting 05-02-2025