Bugfix/fix avx512vbmi regressions (#335)

Multiple AVX512VBMI-related fixes:

src/nfa/mcsheng_compile.cpp: No need for an assert here, impl_id can be set to 0
src/nfa/nfa_api_queue.h: Make sure this compiles on both C++ and C
src/nfagraph/ng_fuzzy.cpp: Fix compilation error when DEBUG_OUTPUT=on
src/runtime.c: Fix crash when data == NULL
unit/internal/sheng.cpp: Unit test has to enable AVX512VBMI manually as autodetection does not get trigger, this causes test to fail
src/fdr/teddy_fat.cpp: AVX512 loads need to be 64-bit aligned, caused a crash on clang-18

src/fdr/teddy_fat.cpp

@@ -100,7 +100,7 @@ const m512 *getDupMaskBase(const struct Teddy *teddy, u8 numMask) {
 }
 
 
-const u8 ALIGN_AVX_DIRECTIVE p_mask_interleave[64] = {
+const u8 ALIGN_CL_DIRECTIVE p_mask_interleave[64] = {
     0, 32, 1, 33, 2, 34, 3, 35, 4, 36, 5, 37, 6, 38, 7, 39,
     8, 40, 9, 41, 10, 42, 11, 43, 12, 44, 13, 45, 14, 46, 15, 47,
     16, 48, 17, 49, 18, 50, 19, 51, 20, 52, 21, 53, 22, 54, 23, 55,

src/nfa/mcsheng_compile.cpp

@@ -610,7 +610,9 @@ dstate_id_t find_sheng_states(dfa_info &info,
     for (auto v : sheng_states) {
         dstate_id_t s = g[v].index;
         if (contains(accel_escape_info, s)) {
-            assert(!info.states[s].impl_id);
+            if (info.states[s].impl_id == 0) {
+                DEBUG_PRINTF("impl_id == 0!\n");
+            }
             info.states[s].impl_id = sheng_end++;
             info.extra[s].sheng_id = info.states[s].impl_id - 1;
         }

src/nfa/nfa_api_queue.h

@@ -251,7 +251,11 @@ void q_skip_forward_to(struct mq *q, s64a min_loc) {
 // Dump the contents of the given queue.
 static never_inline UNUSED
 void debugQueue(const struct mq *q) {
+#ifdef __cplusplus
     if (q == nullptr) {
+#else
+    if (q == NULL) {
+#endif
        DEBUG_PRINTF("q=NULL!\n");
        return;
     }

src/nfagraph/ng_fuzzy.cpp

@@ -76,7 +76,7 @@ vector<flat_set<NFAVertex>> gatherSuccessorsByDepth(const NGHolder &g,
                 if (v == succr) {
                     continue;
                 }
-                DEBUG_PRINTF("Node %zu depth %u\n", g[succ].index, d + 1);
+                DEBUG_PRINTF("Node %zu depth %u\n", g[succr].index, d + 1);
                 next.insert(succr);
             }
         }
@@ -118,7 +118,7 @@ vector<flat_set<NFAVertex>> gatherPredecessorsByDepth(const NGHolder &g,
                 if (v == predc) {
                     continue;
                 }
-                DEBUG_PRINTF("Node %zu depth %u\n", g[pred].index, d + 1);
+                DEBUG_PRINTF("Node %zu depth %u\n", g[predc].index, d + 1);
                 next.insert(predc);
             }
         }

src/rose/rose_build_long_lit.cpp

@@ -207,8 +207,7 @@ vector<RoseLongLitHashEntry> buildHashTable(
             u32 lit_id = lit_offset.first;
             u32 offset = lit_offset.second;
 
-            DEBUG_PRINTF("hash 0x%08x lit_id %u offset %u bucket %u\n", hash,
+            DEBUG_PRINTF("hash 0x%08x lit_id %u offset %u bucket %u\n", m.first, lit_id, offset, bucket);
-                         lit_id, offset, bucket);
 
             auto &entry = tab[bucket];
             entry.str_offset = verify_u32(litToOffsetVal.at(lit_id));

src/runtime.c

@@ -1089,6 +1089,7 @@ hs_error_t HS_CDECL hs_stream_size(const hs_database_t *db,
 // perusal.
 static UNUSED
 void dumpData(const char *data, size_t len) {
+    if (!data) return;
     DEBUG_PRINTF("BUFFER:");
     for (size_t i = 0; i < len; i++) {
         u8 c = data[i];

unit/internal/sheng.cpp

@@ -198,7 +198,11 @@ typedef void (*init_raw_dfa_ptr)(struct ue2::raw_dfa*, const ReportID);
 
 static inline void init_nfa(struct NFA **out_nfa, sheng_compile_ptr compile_function, init_raw_dfa_ptr init_dfa_function) {
     ue2::Grey *g = new ue2::Grey();
+#if defined(HAVE_AVX512VBMI)
+    hs_platform_info plat_info = {0, HS_CPU_FEATURES_AVX512VBMI, 0, 0};
+#else
     hs_platform_info plat_info = {0, 0, 0, 0};
+#endif
     ue2::CompileContext *cc = new ue2::CompileContext(false, false, ue2::target_t(plat_info), *g);
     ue2::ReportManager *rm = new ue2::ReportManager(*g);
     ue2::Report *report = new ue2::Report(ue2::EXTERNAL_CALLBACK, 0);
@@ -522,7 +526,7 @@ TEST(Sheng32, std_compile_header) {
     }
 #endif
     ue2::Grey *g = new ue2::Grey();
-    hs_platform_info plat_info = {0, 0, 0, 0};
+    hs_platform_info plat_info = {0, HS_CPU_FEATURES_AVX512VBMI, 0, 0};
     ue2::CompileContext *cc = new ue2::CompileContext(false, false, ue2::target_t(plat_info), *g);
     ue2::ReportManager *rm = new ue2::ReportManager(*g);
     ue2::Report *report = new ue2::Report(ue2::EXTERNAL_CALLBACK, 0);