Refactor FFI guard and SIMD scalar implementations#649
Conversation
The repeated handle/lock/init matches in fff-c collapse into try_inst!/try_cstr!/read_picker! macros, and the free/accessor bodies share free_array/elem_at helpers. The NEON and AVX2 memmem kernels now share one packed_pair_tail and verify_scalar_from. Also drops the stale fff-core copy of parse_bench.rs left behind by the query-parser crate extraction. Exported C symbols and header unchanged.
seuros
force-pushed
the
refactor/dedup-round-2
branch
from
6830af1 to
8287aec
Compare
|
seem like flacky windows tests (i dont test in windows). But the PR is just code restructured so the failing test are order dependant. fixed in #644 |
|
I appreciate your initiative but I am not a fan of custom propitiatory macros - it is hard to understand what the code is doing when it is hidden behind the macro, so for this kind of 5-10 lines duplications I prefer copy-pasting the code especially given the fact that AI today does this pretty well |
1 similar comment
|
I appreciate your initiative but I am not a fan of custom propitiatory macros - it is hard to understand what the code is doing when it is hidden behind the macro, so for this kind of 5-10 lines duplications I prefer copy-pasting the code especially given the fact that AI today does this pretty well |
2 participants
The repeated handle/lock/init matches in fff-c collapse into try_inst!/try_cstr!/read_picker! macros, and the free/accessor bodies share free_array/elem_at helpers.
The NEON and AVX2 memmem kernels now share one packed_pair_tail and verify_scalar_from.
Also i drops the stale fff-core copy of parse_bench.rs left behind by the query-parser crate extraction.
1 duplicate file removed.
No logic was modified