github_mirrors/vectorscan
3
0
Fork 1
mirror of https://github.com/VectorCamp/vectorscan.git synced 2025-06-28 16:41:01 +03:00
Code Issues Packages Projects Releases Wiki Activity

Make shufti's OR reduce size agnostic

Browse Source 
Signed-off-by: Yoan Picchi <yoan.picchi@arm.com>
This commit is contained in:
Yoan Picchi 2025-06-06 13:55:15 +00:00
parent 9ecd4b53bc
commit d98e67aad7
1 changed files with 3 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
src/nfa/shufti_simd.hpp
View File

@ -208,18 +208,9 @@ const u8 *check_last_byte(SuperVector<S> mask2_lo, SuperVector<S> mask2_hi,
uint8_t last_elem = mask.u.u8[mask_len - 1]; uint8_t last_elem = mask.u.u8[mask_len - 1];
SuperVector<S> reduce = mask2_lo | mask2_hi; SuperVector<S> reduce = mask2_lo | mask2_hi;
#if defined(HAVE_SIMD_512_BITS) for(uint16_t i = S; i > 2; i/=2) {
if constexpr (S >= 64) reduce = reduce | reduce.vshr(i/2);
reduce = reduce | reduce.vshr_512(32); }
#endif
#if defined(HAVE_SIMD_256_BITS)
if constexpr (S >= 32)
reduce = reduce | reduce.vshr_256(16);
#endif
reduce = reduce | reduce.vshr_128(8);
reduce = reduce | reduce.vshr_64(32);
reduce = reduce | reduce.vshr_32(16);
reduce = reduce | reduce.vshr_16(8);
uint8_t match_inverted = reduce.u.u8[0] | last_elem; uint8_t match_inverted = reduce.u.u8[0] | last_elem;
// if 0xff, then no match // if 0xff, then no match