feat(rust): port compression-parameter selection

Move the context-free compression-parameter logic of zstd_compress.c to
rust/src/zstd_compress_params.rs: the compression-level tables (formerly
clevels.h), parameter bounds/checking/clamping, cycle log, level-table
selection, source/dictionary parameter adjustment, default frame
parameters, and the match-state/CDict size estimators.

The boundary follows the module's design notes: Rust owns only leaves
whose behavior is independent of C preprocessor configuration.  C keeps
the public ZSTD_* symbols and feeds the leaves everything that is
configuration-owned as explicit scalars:

- The ZSTD_EXCLUDE_*_BLOCK_COMPRESSOR strategy cascade stays in
  ZSTD_adjustCParams_internal() ahead of the Rust adjustment leaf, so
  reduced builds keep their fallback policy.
- Workspace estimation receives struct sizes (ZSTD_CDict, ZSTD_match_t,
  ZSTD_optimal_t), HUF workspace size, and the sanitizer redzone size,
  because those depend on private layouts and ASAN configuration.
- ZSTD_cParam_getBounds() forwards only the compression level and the
  seven core parameters; all other parameter bounds remain C.
- Frozen private constants the leaves hardcode (short-cache and row-hash
  tag widths, MaxML/MaxLL/MaxOff/Litbits, ZSTD_OPT_SIZE, cwksp alignment)
  are pinned by ZSTD_STATIC_ASSERTs at the C call sites.

Two latent 32-bit bugs in the previously unwired module were fixed
before integration: dictAndWindowLog's max window size and the
window-resize threshold were hardcoded to the 64-bit constants
(1<<31, 1<<30) instead of deriving from ZSTD_WINDOWLOG_MAX, which is
30 on 32-bit targets.

clevels.h is no longer included anywhere but stays in-tree as the
reference for mechanical comparison against the Rust table.

Test plan:
- cd rust && cargo fmt --check && cargo clippy --all-targets -- -D
  warnings && cargo test --all-targets (125 tests)
- make -C tests fuzzer && ./tests/fuzzer -i1 --no-big-tests
- make -C tests test-rust-lib-smoke
- Byte-identity: zstd CLI frames for COPYING, a 250 KB C source, and a
  5 MB datagen sample at levels 1/3/9/19/--fast=5 are identical between
  this change and its parent commit.
This commit is contained in:
2026-07-11 09:32:49 +02:00
parent 959e485201
commit 529cd297e2
4 changed files with 1162 additions and 271 deletions
+6
View File
@@ -34,6 +34,12 @@ zstd ABI:
- `zstd_compress_frame` serializes frame headers, skippable frames, and the
last empty block; it takes scalar frame parameters so the C-owned
`ZSTD_CCtx_params` layout never crosses the language boundary.
- `zstd_compress_params` owns the compression-level tables (formerly
`clevels.h`), parameter bounds, clamping, validation, table selection,
source/dictionary adjustment, and match-state/CDict size estimation.
The C integration layer keeps the public `ZSTD_*` symbols and feeds the
leaves configuration-owned scalars: the excluded-block-compressor
strategy cascade, struct sizes, and sanitizer redzone policy.
- `zstd_fast` and `zstd_double_fast` implement the single- and two-table
fast block match finders, including attached and external dictionary paths.
- `zstd_lazy` implements greedy, lazy, lazy2, and binary-tree matching,