Add SIMD optimization for int_to_float conversion#580

Merged

veluca93 merged 14 commits into

libjxl:mainfrom

hjanuschka:simd-int-to-float

Jan 21, 2026

Commits on Jan 21, 2026

Add SIMD optimization for int_to_float conversion

hjanuschka
authored and
veluca93
committed
Fix clippy excessive precision warnings in f16 tests

hjanuschka
authored and
veluca93
committed
Refactor f16 conversion to use F32Vec::load_f16_bits/store_f16

hjanuschka
authored and
veluca93
committed
Add SAFETY comments to unsafe blocks

hjanuschka
authored and
veluca93
committed
Address PR review comments for SIMD f16 conversion

hjanuschka
authored and
veluca93
committed
Address reviewer feedback: add I32Vec::store_u16 and remove scalar remainder

hjanuschka
authored and
veluca93
committed
Add TODO for SIMD optimization of generic float format conversion

hjanuschka
authored and
veluca93
committed
Move assert inside inner function for load_f16_bits/store_f16 in AVX

hjanuschka
authored and
veluca93
committed
Remove unnecessary f16c comments in AVX load/store

hjanuschka
authored and
veluca93
committed
Rename store_f16 to store_f16_bits for consistency

hjanuschka
authored and
veluca93
committed
Fix SIMD int_to_float to handle non-SIMD-aligned sizes

hjanuschka
authored and
veluca93
committed
Add f16c to target features defined on dispatch functions.
veluca93
committed
Remove scalar fallback and fix tests & pipeline code
veluca93
committed
Improve store_u16 impl & some safety comments.
veluca93
committed