v0.5.0: Unified Benchmarks and Faster DFT Paths

[0.5.0] - 2026-03-31

poulpy-bench (new crate)

• Consolidate all benchmark suites into a single poulpy-bench crate; remove bench_suite modules from poulpy-hal, poulpy-core, and poulpy-schemes.
• Organize bench suite under three namespaces: bench_suite::hal, bench_suite::core, bench_suite::schemes.
• Add standard binary: one representative run across all layers with fixed parameters, intended for version-to-version regression tracking.
• Add JSON-configurable benchmark parameters via the POULPY_BENCH_PARAMS environment variable (file path or inline JSON). All sweep ranges and layout constants are overridable; any omitted field falls back to its built-in default.
  • hal.sweeps — [log_n, cols, size] points for vec_znx_big, vec_znx_dft, svp
  • cnv.sweeps — [log_n, size] points for convolution
  • vmp.sweeps — [log_n, rows, cols_in, cols_out, size] points for vmp
  • svp_prepare.log_n — ring degrees for SVP prepare
  • core.{n, base2k, k, rank, dsize} — layout for all core / scheme / standard benchmarks
• Add run JSON field: list of bench binary names or function names to execute; binary names run the whole binary, function names are applied as a Criterion regex filter across the default binary set.
• Add backends JSON field: list of backend labels (fft64-ref, ntt120-ref, fft64-avx, ntt120-avx) to restrict which backends are benchmarked; listing an AVX backend automatically enables --features enable-avx and sets RUSTFLAGS="-C target-feature=+avx2,+fma".
• Replace per-group measurement_time overrides with a shared criterion_config() (100 samples, 5 s measurement budget).
• Add examples/custom_params.json and examples/run_custom_params.sh: runnable example demonstrating JSON-configurable parameters, backend selection, operation filtering, and baseline comparison.

poulpy-hal

• Remove VmpApplyDftToDftAdd and SvpApplyDftToDftAdd traits; merge additive variant into VmpApplyDftToDft / SvpApplyDftToDft via a new limb_offset parameter.
  These traits accumulated VMP results directly into a scattered output buffer, causing severe cache misses. Writing into a contiguous temporary buffer and folding with VecZnxDftAddInplace is ~2× faster.
• Remove all associated OEP (VmpApplyDftToDftAddImpl, VmpApplyDftToDftAddTmpBytesImpl, SvpApplyDftToDftAddImpl), delegate, and bench-suite plumbing.

poulpy-cpu-ref / poulpy-cpu-avx

• Update FFT64 and NTT120 vmp_apply_dft_to_dft implementations to accept limb_offset directly, replacing the separate _add codepath.
• NTT120 AVX2 (arithmetic_avx.rs): add reduce_b_and_apply_crt that fuses the CRT multiply into the Barrett reduction pass, using new compile-time constants POW32_CRT and POW16_CRT; apply to compact_all_blocks to reduce instruction count by a factor of ~2x.

poulpy-core

• Rewrite external product (glwe_external_product_internal) and GLWE keyswitching inner loops to write intermediate per-digit VMP results into a dedicated temporary buffer before accumulating with VecZnxDftAddInplace, avoiding scattered-write cache thrashing. where bounds updated accordingly.
• Add bench_suite::keyswitch::gglwe module and keyswitch_glwe criterion benchmark targeting the NTT120 backend; remove the old FFT64-specific keyswitch_glwe_fft64 benchmark.