Merge branch 'develop' into wip-isildur-g-cppcheck-47-48-58

2025-11-17 01:41:51 +03:00 · 2024-05-01 10:59:59 +03:00
parent 98243a6a08 46aba61156
commit 583e1d576a
89 changed files with 1608 additions and 588 deletions
--- a/src/nfa/castle.c
+++ b/src/nfa/castle.c
@@ -400,7 +400,7 @@ char castleFindMatch(const struct Castle *c, const u64a begin, const u64a end,
 }

 static really_inline
-u64a subCastleNextMatch(const struct Castle *c, void *full_state,
+u64a subCastleNextMatch(const struct Castle *c, const void *full_state,
                        void *stream_state, const u64a loc,
                        const u32 subIdx) {
    DEBUG_PRINTF("subcastle %u\n", subIdx);
@@ -489,7 +489,6 @@ char castleMatchLoop(const struct Castle *c, const u64a begin, const u64a end,
        // full_state (scratch).

        u64a offset = end; // min offset of next match
-        u32 activeIdx = 0;
        mmbit_clear(matching, c->numRepeats);
        if (c->exclusive) {
            u8 *active = (u8 *)stream_state;
@@ -497,7 +496,7 @@ char castleMatchLoop(const struct Castle *c, const u64a begin, const u64a end,
            for (u32 i = mmbit_iterate(groups, c->numGroups, MMB_INVALID);
                 i != MMB_INVALID; i = mmbit_iterate(groups, c->numGroups, i)) {
                u8 *cur = active + i * c->activeIdxSize;
-                activeIdx = partial_load_u32(cur, c->activeIdxSize);
+                u32 activeIdx = partial_load_u32(cur, c->activeIdxSize);
                u64a match = subCastleNextMatch(c, full_state, stream_state,
                                                loc, activeIdx);
                set_matching(c, match, groups, matching, c->numGroups, i,
@@ -907,7 +906,6 @@ s64a castleLastKillLoc(const struct Castle *c, struct mq *q) {
        if (castleRevScan(c, q->history, sp + hlen, ep + hlen, &loc)) {
            return (s64a)loc - hlen;
        }
-        ep = 0;
    }

    return sp - 1; /* the repeats are never killed */
--- a/src/nfa/castlecompile.cpp
+++ b/src/nfa/castlecompile.cpp
@@ -655,7 +655,8 @@ buildCastle(const CastleProto &proto,
    if (!stale_iter.empty()) {
        c->staleIterOffset = verify_u32(ptr - base_ptr);
        copy_bytes(ptr, stale_iter);
-        ptr += byte_length(stale_iter);
+        // Removed unused increment operation
+        // ptr += byte_length(stale_iter);
    }

    return nfa;
--- a/src/nfa/limex_common_impl.h
+++ b/src/nfa/limex_common_impl.h
@@ -332,7 +332,7 @@ void EXPIRE_ESTATE_FN(const IMPL_NFA_T *limex, struct CONTEXT_T *ctx,
 // UE-1636) need to guard cyclic tug-accepts as well.
 static really_inline
 char LIMEX_INACCEPT_FN(const IMPL_NFA_T *limex, STATE_T state,
-                       union RepeatControl *repeat_ctrl, char *repeat_state,
+                       const union RepeatControl *repeat_ctrl, const char *repeat_state,
                       u64a offset, ReportID report) {
    assert(limex);

@@ -382,7 +382,7 @@ char LIMEX_INACCEPT_FN(const IMPL_NFA_T *limex, STATE_T state,

 static really_inline
 char LIMEX_INANYACCEPT_FN(const IMPL_NFA_T *limex, STATE_T state,
-                          union RepeatControl *repeat_ctrl, char *repeat_state,
+                          const union RepeatControl *repeat_ctrl, const char *repeat_state,
                          u64a offset) {
    assert(limex);

--- a/src/nfa/limex_compile.cpp
+++ b/src/nfa/limex_compile.cpp
@@ -1572,7 +1572,7 @@ u32 findMaxVarShift(const build_info &args, u32 nShifts) {
 static
 int getLimexScore(const build_info &args, u32 nShifts) {
    const NGHolder &h = args.h;
-    u32 maxVarShift = nShifts;
+    u32 maxVarShift;
    int score = 0;

    score += SHIFT_COST * nShifts;
--- a/src/nfa/mpv.c
+++ b/src/nfa/mpv.c
@@ -512,7 +512,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,

    verm_restart:;
        assert(buf[curr] == kp->u.verm.c);
-        size_t test = curr;
+        size_t test;
        if (curr + min_rep < length) {
            test = curr + min_rep;
        } else {
@@ -534,7 +534,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,
        m128 hi = kp->u.shuf.mask_hi;
    shuf_restart:
        assert(do_single_shufti(lo, hi, buf[curr]));
-        size_t test = curr;
+        size_t test;
        if (curr + min_rep < length) {
            test = curr + min_rep;
        } else {
@@ -556,7 +556,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,
        const m128 mask1 = kp->u.truffle.mask1;
        const m128 mask2 = kp->u.truffle.mask2;
    truffle_restart:;
-        size_t test = curr;
+        size_t test;
        if (curr + min_rep < length) {
            test = curr + min_rep;
        } else {
@@ -582,7 +582,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,

    nverm_restart:;
        assert(buf[curr] != kp->u.verm.c);
-        size_t test = curr;
+        size_t test;
        if (curr + min_rep < length) {
            test = curr + min_rep;
        } else {
@@ -607,7 +607,7 @@ size_t find_last_bad(const struct mpv_kilopuff *kp, const u8 *buf,
 }

 static really_inline
-void restartKilo(const struct mpv *m, UNUSED u8 *active, u8 *reporters,
+void restartKilo(const struct mpv *m, UNUSED const u8 *active, u8 *reporters,
                 struct mpv_decomp_state *dstate, struct mpv_pq_item *pq,
                 const u8 *buf, u64a prev_limit, size_t buf_length, u32 i) {
    const struct mpv_kilopuff *kp = (const void *)(m + 1);
--- a/src/nfa/repeatcompile.cpp
+++ b/src/nfa/repeatcompile.cpp
@@ -94,9 +94,6 @@ u32 repeatRecurTable(struct RepeatStateInfo *info, const depth &repeatMax,
 static
 u32 findOptimalPatchSize(struct RepeatStateInfo *info, const depth &repeatMax,
                         const u32 minPeriod, u32 rv) {
-    u32 cnt = 0;
-    u32 patch_bits = 0;
-    u32 total_size = 0;
    u32 min = ~0U;
    u32 patch_len = 0;

@@ -105,11 +102,11 @@ u32 findOptimalPatchSize(struct RepeatStateInfo *info, const depth &repeatMax,
    }

    for (u32 i = minPeriod; i <= rv; i++) {
-        cnt = ((u32)repeatMax + (i - 1)) / i + 1;
+        u32 cnt = ((u32)repeatMax + (i - 1)) / i + 1;

        // no bit packing version
-        patch_bits = calcPackedBits(info->table[i]);
-        total_size = (patch_bits + 7U) / 8U * cnt;
+        u32 patch_bits = calcPackedBits(info->table[i]);
+        u32 total_size = (patch_bits + 7U) / 8U * cnt;

        if (total_size < min) {
            patch_len = i;
--- a/src/nfa/sheng.c
+++ b/src/nfa/sheng.c
@@ -154,7 +154,7 @@ char fireReports(const struct sheng *sh, NfaCallback cb, void *ctxt,
    return MO_CONTINUE_MATCHING; /* continue execution */
 }

-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 // Sheng32
 static really_inline
 const struct sheng32 *get_sheng32(const struct NFA *n) {
@@ -351,7 +351,7 @@ char fireReports64(const struct sheng64 *sh, NfaCallback cb, void *ctxt,
    }
    return MO_CONTINUE_MATCHING; /* continue execution */
 }
-#endif // end of HAVE_AVX512VBMI
+#endif // end of HAVE_AVX512VBMI || HAVE_SVE

 /* include Sheng function definitions */
 #include "sheng_defs.h"
@@ -871,7 +871,7 @@ char nfaExecSheng_expandState(UNUSED const struct NFA *nfa, void *dest,
    return 0;
 }

-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 // Sheng32
 static really_inline
 char runSheng32Cb(const struct sheng32 *sh, NfaCallback cb, void *ctxt,
@@ -1874,4 +1874,4 @@ char nfaExecSheng64_expandState(UNUSED const struct NFA *nfa, void *dest,
    *(u8 *)dest = *(const u8 *)src;
    return 0;
 }
-#endif // end of HAVE_AVX512VBMI
+#endif // end of HAVE_AVX512VBMI || HAVE_SVE
--- a/src/nfa/sheng.h
+++ b/src/nfa/sheng.h
@@ -58,7 +58,7 @@ char nfaExecSheng_reportCurrent(const struct NFA *n, struct mq *q);
 char nfaExecSheng_B(const struct NFA *n, u64a offset, const u8 *buffer,
                    size_t length, NfaCallback cb, void *context);

-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define nfaExecSheng32_B_Reverse NFA_API_NO_IMPL
 #define nfaExecSheng32_zombie_status NFA_API_ZOMBIE_NO_IMPL

@@ -106,8 +106,7 @@ char nfaExecSheng64_reportCurrent(const struct NFA *n, struct mq *q);

 char nfaExecSheng64_B(const struct NFA *n, u64a offset, const u8 *buffer,
                      size_t length, NfaCallback cb, void *context);
-
-#else // !HAVE_AVX512VBMI
+#else // !HAVE_AVX512VBMI && !HAVE_SVE

 #define nfaExecSheng32_B_Reverse NFA_API_NO_IMPL
 #define nfaExecSheng32_zombie_status NFA_API_ZOMBIE_NO_IMPL
@@ -138,6 +137,7 @@ char nfaExecSheng64_B(const struct NFA *n, u64a offset, const u8 *buffer,
 #define nfaExecSheng64_testEOD NFA_API_NO_IMPL
 #define nfaExecSheng64_reportCurrent NFA_API_NO_IMPL
 #define nfaExecSheng64_B NFA_API_NO_IMPL
-#endif // end of HAVE_AVX512VBMI
+#endif // end of HAVE_AVX512VBMI || defined(HAVE_SVE)
+

 #endif /* SHENG_H_ */
--- a/src/nfa/sheng_defs.h
+++ b/src/nfa/sheng_defs.h
@@ -52,7 +52,7 @@ u8 hasInterestingStates(const u8 a, const u8 b, const u8 c, const u8 d) {
    return (a | b | c | d) & (SHENG_STATE_FLAG_MASK);
 }

-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 static really_inline
 u8 isDeadState32(const u8 a) {
    return a & SHENG32_STATE_DEAD;
@@ -108,7 +108,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define SHENG_IMPL sheng_cod
 #define DEAD_FUNC isDeadState
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_cod
 #define DEAD_FUNC32 isDeadState32
 #define ACCEPT_FUNC32 isAcceptState32
@@ -121,7 +121,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef SHENG_IMPL
 #undef DEAD_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef DEAD_FUNC32
 #undef ACCEPT_FUNC32
@@ -135,7 +135,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define SHENG_IMPL sheng_co
 #define DEAD_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_co
 #define DEAD_FUNC32 dummyFunc
 #define ACCEPT_FUNC32 isAcceptState32
@@ -148,7 +148,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef SHENG_IMPL
 #undef DEAD_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef DEAD_FUNC32
 #undef ACCEPT_FUNC32
@@ -162,7 +162,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define SHENG_IMPL sheng_samd
 #define DEAD_FUNC isDeadState
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_samd
 #define DEAD_FUNC32 isDeadState32
 #define ACCEPT_FUNC32 isAcceptState32
@@ -175,7 +175,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef SHENG_IMPL
 #undef DEAD_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef DEAD_FUNC32
 #undef ACCEPT_FUNC32
@@ -189,7 +189,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define SHENG_IMPL sheng_sam
 #define DEAD_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_sam
 #define DEAD_FUNC32 dummyFunc
 #define ACCEPT_FUNC32 isAcceptState32
@@ -202,7 +202,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef SHENG_IMPL
 #undef DEAD_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef DEAD_FUNC32
 #undef ACCEPT_FUNC32
@@ -216,7 +216,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define SHENG_IMPL sheng_nmd
 #define DEAD_FUNC isDeadState
 #define ACCEPT_FUNC dummyFunc
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_nmd
 #define DEAD_FUNC32 isDeadState32
 #define ACCEPT_FUNC32 dummyFunc
@@ -229,7 +229,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef SHENG_IMPL
 #undef DEAD_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef DEAD_FUNC32
 #undef ACCEPT_FUNC32
@@ -243,7 +243,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define SHENG_IMPL sheng_nm
 #define DEAD_FUNC dummyFunc
 #define ACCEPT_FUNC dummyFunc
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_nm
 #define DEAD_FUNC32 dummyFunc
 #define ACCEPT_FUNC32 dummyFunc
@@ -256,7 +256,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef SHENG_IMPL
 #undef DEAD_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef DEAD_FUNC32
 #undef ACCEPT_FUNC32
@@ -277,7 +277,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC isAccelState
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_coda
 #define INTERESTING_FUNC32 hasInterestingStates32
 #define INNER_DEAD_FUNC32 isDeadState32
@@ -296,7 +296,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -316,7 +316,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC dummyFunc
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_cod
 #define INTERESTING_FUNC32 hasInterestingStates32
 #define INNER_DEAD_FUNC32 isDeadState32
@@ -339,7 +339,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -363,7 +363,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC isAccelState
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_coa
 #define INTERESTING_FUNC32 hasInterestingStates32
 #define INNER_DEAD_FUNC32 dummyFunc
@@ -382,7 +382,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -402,7 +402,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC dummyFunc
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_co
 #define INTERESTING_FUNC32 hasInterestingStates32
 #define INNER_DEAD_FUNC32 dummyFunc
@@ -425,7 +425,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -449,7 +449,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC isAccelState
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_samda
 #define INTERESTING_FUNC32 hasInterestingStates32
 #define INNER_DEAD_FUNC32 isDeadState32
@@ -468,7 +468,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -488,7 +488,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC dummyFunc
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_samd
 #define INTERESTING_FUNC32 hasInterestingStates32
 #define INNER_DEAD_FUNC32 isDeadState32
@@ -511,7 +511,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -535,7 +535,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC isAccelState
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_sama
 #define INTERESTING_FUNC32 hasInterestingStates32
 #define INNER_DEAD_FUNC32 dummyFunc
@@ -554,7 +554,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -574,7 +574,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC dummyFunc
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC isAcceptState
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_sam
 #define INTERESTING_FUNC32 hasInterestingStates32
 #define INNER_DEAD_FUNC32 dummyFunc
@@ -597,7 +597,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -623,7 +623,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC dummyFunc
 #define OUTER_ACCEL_FUNC isAccelState
 #define ACCEPT_FUNC dummyFunc
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_nmda
 #define INTERESTING_FUNC32 dummyFunc4
 #define INNER_DEAD_FUNC32 dummyFunc
@@ -642,7 +642,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -662,7 +662,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC dummyFunc
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC dummyFunc
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_nmd
 #define INTERESTING_FUNC32 dummyFunc4
 #define INNER_DEAD_FUNC32 dummyFunc
@@ -685,7 +685,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
@@ -712,7 +712,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #define INNER_ACCEL_FUNC dummyFunc
 #define OUTER_ACCEL_FUNC dummyFunc
 #define ACCEPT_FUNC dummyFunc
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #define SHENG32_IMPL sheng32_4_nm
 #define INTERESTING_FUNC32 dummyFunc4
 #define INNER_DEAD_FUNC32 dummyFunc
@@ -735,7 +735,7 @@ u8 dummyFunc(UNUSED const u8 a) {
 #undef INNER_ACCEL_FUNC
 #undef OUTER_ACCEL_FUNC
 #undef ACCEPT_FUNC
-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 #undef SHENG32_IMPL
 #undef INTERESTING_FUNC32
 #undef INNER_DEAD_FUNC32
--- a/src/nfa/sheng_impl.h
+++ b/src/nfa/sheng_impl.h
@@ -96,7 +96,7 @@ char SHENG_IMPL(u8 *state, NfaCallback cb, void *ctxt, const struct sheng *s,
    return MO_CONTINUE_MATCHING;
 }

-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 static really_inline
 char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
                  const struct sheng32 *s,
@@ -114,14 +114,28 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
    }
    DEBUG_PRINTF("Scanning %lli bytes\n", (s64a)(end - start));

+#if defined(HAVE_SVE)
+    const svbool_t lane_pred_32 = svwhilelt_b8(0, 32);
+    svuint8_t cur_state = svdup_u8(*state);
+    svuint8_t tbl_mask = svdup_u8((unsigned char)0x1F);
+    const m512 *masks = s->succ_masks;
+#else
    m512 cur_state = set1_64x8(*state);
    const m512 *masks = s->succ_masks;
+#endif

    while (likely(cur_buf != end)) {
        const u8 c = *cur_buf;
+
+#if defined(HAVE_SVE)
+        svuint8_t succ_mask = svld1(lane_pred_32, (const u8*)(masks + c));
+        cur_state = svtbl(succ_mask, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 tmp = svlastb(lane_pred_32, cur_state);
+#else
        const m512 succ_mask = masks[c];
        cur_state = vpermb512(cur_state, succ_mask);
        const u8 tmp = movd512(cur_state);
+#endif

        DEBUG_PRINTF("c: %02hhx '%c'\n", c, ourisprint(c) ? c : '?');
        DEBUG_PRINTF("s: %u (flag: %u)\n", tmp & SHENG32_STATE_MASK,
@@ -153,7 +167,11 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        }
        cur_buf++;
    }
+#if defined(HAVE_SVE)
+    *state = svlastb(lane_pred_32, cur_state);
+#else
    *state = movd512(cur_state);
+#endif
    *scan_end = cur_buf;
    return MO_CONTINUE_MATCHING;
 }
@@ -175,14 +193,28 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
    }
    DEBUG_PRINTF("Scanning %lli bytes\n", (s64a)(end - start));

+#if defined(HAVE_SVE)
+    const svbool_t lane_pred_64 = svwhilelt_b8(0, 64);
+    svuint8_t cur_state = svdup_u8(*state);
+    svuint8_t tbl_mask = svdup_u8((unsigned char)0x3F);
+    const m512 *masks = s->succ_masks;
+#else
    m512 cur_state = set1_64x8(*state);
    const m512 *masks = s->succ_masks;
+#endif

    while (likely(cur_buf != end)) {
        const u8 c = *cur_buf;
+
+#if defined(HAVE_SVE)
+        svuint8_t succ_mask = svld1(lane_pred_64, (const u8*)(masks + c));
+        cur_state = svtbl(succ_mask, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 tmp = svlastb(lane_pred_64, cur_state);
+#else
        const m512 succ_mask = masks[c];
        cur_state = vpermb512(cur_state, succ_mask);
        const u8 tmp = movd512(cur_state);
+#endif

        DEBUG_PRINTF("c: %02hhx '%c'\n", c, ourisprint(c) ? c : '?');
        DEBUG_PRINTF("s: %u (flag: %u)\n", tmp & SHENG64_STATE_MASK,
@@ -214,7 +246,11 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        }
        cur_buf++;
    }
+#if defined(HAVE_SVE)
+    *state = svlastb(lane_pred_64, cur_state);
+#else
    *state = movd512(cur_state);
+#endif
    *scan_end = cur_buf;
    return MO_CONTINUE_MATCHING;
 }
--- a/src/nfa/sheng_impl4.h
+++ b/src/nfa/sheng_impl4.h
@@ -283,7 +283,7 @@ char SHENG_IMPL(u8 *state, NfaCallback cb, void *ctxt, const struct sheng *s,
    return MO_CONTINUE_MATCHING;
 }

-#if defined(HAVE_AVX512VBMI)
+#if defined(HAVE_AVX512VBMI) || defined(HAVE_SVE)
 static really_inline
 char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
                  const struct sheng32 *s,
@@ -320,8 +320,15 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        return MO_CONTINUE_MATCHING;
    }

+#if defined(HAVE_SVE)
+    const svbool_t lane_pred_32 = svwhilelt_b8(0, 32);
+    svuint8_t cur_state = svdup_u8(*state);
+    svuint8_t tbl_mask = svdup_u8((unsigned char)0x1F);
+    const m512 *masks = s->succ_masks;
+#else
    m512 cur_state = set1_64x8(*state);
    const m512 *masks = s->succ_masks;
+#endif

    while (likely(end - cur_buf >= 4)) {
        const u8 *b1 = cur_buf;
@@ -333,6 +340,23 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        const u8 c3 = *b3;
        const u8 c4 = *b4;

+#if defined(HAVE_SVE)
+        svuint8_t succ_mask1 = svld1(lane_pred_32, (const u8*)(masks+c1));
+        cur_state = svtbl(succ_mask1, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 a1 = svlastb(lane_pred_32, cur_state);
+
+        svuint8_t succ_mask2 = svld1(lane_pred_32, (const u8*)(masks+c2));
+        cur_state = svtbl(succ_mask2, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 a2 = svlastb(lane_pred_32, cur_state);
+
+        svuint8_t succ_mask3 = svld1(lane_pred_32, (const u8*)(masks+c3));
+        cur_state = svtbl(succ_mask3, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 a3 = svlastb(lane_pred_32, cur_state);
+
+        svuint8_t succ_mask4 = svld1(lane_pred_32, (const u8*)(masks+c4));
+        cur_state = svtbl(succ_mask4, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 a4 = svlastb(lane_pred_32, cur_state);
+#else
        const m512 succ_mask1 = masks[c1];
        cur_state = vpermb512(cur_state, succ_mask1);
        const u8 a1 = movd512(cur_state);
@@ -348,6 +372,7 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        const m512 succ_mask4 = masks[c4];
        cur_state = vpermb512(cur_state, succ_mask4);
        const u8 a4 = movd512(cur_state);
+#endif

        DEBUG_PRINTF("c: %02hhx '%c'\n", c1, ourisprint(c1) ? c1 : '?');
        DEBUG_PRINTF("s: %u (flag: %u)\n", a1 & SHENG32_STATE_MASK,
@@ -517,7 +542,11 @@ char SHENG32_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        };
        cur_buf += 4;
    }
+#if defined(HAVE_SVE)
+    *state = svlastb(lane_pred_32, cur_state);
+#else
    *state = movd512(cur_state);
+#endif
    *scan_end = cur_buf;
    return MO_CONTINUE_MATCHING;
 }
@@ -541,9 +570,15 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        *scan_end = end;
        return MO_CONTINUE_MATCHING;
    }
-
+#if defined(HAVE_SVE)
+    const svbool_t lane_pred_64 = svwhilelt_b8(0, 64);
+    svuint8_t cur_state = svdup_u8(*state);
+    svuint8_t tbl_mask = svdup_u8((unsigned char)0x3F);
+    const m512 *masks = s->succ_masks;
+#else
    m512 cur_state = set1_64x8(*state);
    const m512 *masks = s->succ_masks;
+#endif

    while (likely(end - cur_buf >= 4)) {
        const u8 *b1 = cur_buf;
@@ -555,6 +590,23 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        const u8 c3 = *b3;
        const u8 c4 = *b4;

+#if defined(HAVE_SVE)
+        svuint8_t succ_mask1 = svld1(lane_pred_64, (const u8*)(masks+c1));
+        cur_state = svtbl(succ_mask1, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 a1 = svlastb(lane_pred_64, cur_state);
+
+        svuint8_t succ_mask2 = svld1(lane_pred_64, (const u8*)(masks+c2));
+        cur_state = svtbl(succ_mask2, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 a2 = svlastb(lane_pred_64, cur_state);
+
+        svuint8_t succ_mask3 = svld1(lane_pred_64, (const u8*)(masks+c3));
+        cur_state = svtbl(succ_mask3, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 a3 = svlastb(lane_pred_64, cur_state);
+
+        svuint8_t succ_mask4 = svld1(lane_pred_64, (const u8*)(masks+c4));
+        cur_state = svtbl(succ_mask4, svand_x(svptrue_b8(), tbl_mask, cur_state));
+        const u8 a4 = svlastb(lane_pred_64, cur_state);
+#else
        const m512 succ_mask1 = masks[c1];
        cur_state = vpermb512(cur_state, succ_mask1);
        const u8 a1 = movd512(cur_state);
@@ -570,6 +622,7 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        const m512 succ_mask4 = masks[c4];
        cur_state = vpermb512(cur_state, succ_mask4);
        const u8 a4 = movd512(cur_state);
+#endif

        DEBUG_PRINTF("c: %02hhx '%c'\n", c1, ourisprint(c1) ? c1 : '?');
        DEBUG_PRINTF("s: %u (flag: %u)\n", a1 & SHENG64_STATE_MASK,
@@ -703,7 +756,11 @@ char SHENG64_IMPL(u8 *state, NfaCallback cb, void *ctxt,
        }
        cur_buf += 4;
    }
+#if defined(HAVE_SVE)
+    *state = svlastb(lane_pred_64, cur_state);
+#else
    *state = movd512(cur_state);
+#endif
    *scan_end = cur_buf;
    return MO_CONTINUE_MATCHING;
 }
--- a/src/nfa/shengcompile.cpp
+++ b/src/nfa/shengcompile.cpp
@@ -730,10 +730,17 @@ bytecode_ptr<NFA> sheng32Compile(raw_dfa &raw, const CompileContext &cc,
        return nullptr;
    }

+#ifdef HAVE_SVE
+    if (svcntb()<32) {
+        DEBUG_PRINTF("Sheng32 failed, SVE width is too small!\n");
+        return nullptr;
+    }
+#else
    if (!cc.target_info.has_avx512vbmi()) {
        DEBUG_PRINTF("Sheng32 failed, no HS_CPU_FEATURES_AVX512VBMI!\n");
        return nullptr;
    }
+#endif

    sheng_build_strat strat(raw, rm, only_accel_init);
    dfa_info info(strat);
@@ -762,10 +769,17 @@ bytecode_ptr<NFA> sheng64Compile(raw_dfa &raw, const CompileContext &cc,
        return nullptr;
    }

+#ifdef HAVE_SVE
+    if (svcntb()<64) {
+        DEBUG_PRINTF("Sheng64 failed, SVE width is too small!\n");
+        return nullptr;
+    }
+#else
    if (!cc.target_info.has_avx512vbmi()) {
        DEBUG_PRINTF("Sheng64 failed, no HS_CPU_FEATURES_AVX512VBMI!\n");
        return nullptr;
    }
+#endif

    sheng_build_strat strat(raw, rm, only_accel_init);
    dfa_info info(strat);