Ivan PleshkovandClaude Fable 5 5959e7c421 Fix TurboQuantizer::quantized_size double-padding for Bits1_5 (#10390) ...

`quantized_size_for` pads its `dim` argument, so passing the already
padded `self.padded_dim` applied the x1.5 expansion of `Bits1_5` a
second time.  Nothing on disk depends on the value for that width: the
quantization path sizes its records with `quantized_size_for` from the
raw dim, and the Turbo datatype storages, which do use
`quantized_size()` as their record size, are fixed at Bits4, where the
padding is idempotent.  The wrong value only over-reserved the
`quantize` output buffer.

Compute the packed size from `padded_dim` directly and cover `Bits1_5`
in the byte-length test.

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

2026-09-03 12:45:52 +02:00

..

benches

Replace permutation_iterator with rand's index sampler (#10217)

2026-09-03 12:41:00 +02:00

cpp

Hotfix for Windows ARM64 builds, disable some Neon features (#7690)

2025-12-18 17:26:37 +01:00

src

Fix TurboQuantizer::quantized_size double-padding for Bits1_5 (#10390)

2026-09-03 12:45:52 +02:00

tests/integration

[UpdateOnly] implement the appendable quantized-vector overlay (dense, Binary/Turbo) (#10161)

2026-09-03 12:38:58 +02:00

build.rs

fix(quantization): skip cc compile when no SIMD sources were added (#10209)

2026-09-03 12:39:03 +02:00

Cargo.toml

Replace permutation_iterator with rand's index sampler (#10217)

2026-09-03 12:41:00 +02:00