@shd101wyy/yo 0.1.24 → 0.1.25

package/vendor/mimalloc/src/stats.c

@@ -117,7 +117,7 @@ static void mi_stat_counter_add_mt(mi_stat_counter_t* stat, const mi_stat_counte
 
 // must be thread safe as it is called from stats_merge
 static void mi_stats_add(mi_stats_t* stats, const mi_stats_t* src) {
-  if (stats==src) return;
+  if (stats==NULL || src==NULL || stats==src) return;
 
   // copy all fields
   MI_STAT_FIELDS()
@@ -214,10 +214,9 @@ static void mi_stat_print_ex(const mi_stat_count_t* stat, const char* msg, int64
     }
   }
   else {
-    mi_print_amount(stat->peak, 1, out, arg);
-    mi_print_amount(stat->total, 1, out, arg);
-    _mi_fprintf(out, arg, "%11s", " ");  // no freed
-    mi_print_amount(stat->current, 1, out, arg);
+    mi_print_amount(stat->peak, 0, out, arg);
+    mi_print_amount(stat->total, 0, out, arg);
+    mi_print_amount(stat->current, 0, out, arg);
     _mi_fprintf(out, arg, "\n");
   }
 }
@@ -237,16 +236,21 @@ static void mi_stat_total_print(const mi_stat_count_t* stat, const char* msg, in
 
 static void mi_stat_counter_print(const mi_stat_counter_t* stat, const char* msg, mi_output_fun* out, void* arg ) {
   _mi_fprintf(out, arg, "  %-10s:", msg);
-  mi_print_amount(stat->total, -1, out, arg);
+  mi_print_amount(stat->total, 0, out, arg);
   _mi_fprintf(out, arg, "\n");
 }
 
+static void mi_stat_counter_print_size(const mi_stat_counter_t* stat, const char* msg, mi_output_fun* out, void* arg ) {
+  _mi_fprintf(out, arg, "  %-10s:", msg);
+  mi_print_amount(stat->total, 1, out, arg);
+  _mi_fprintf(out, arg, "\n");
+}
 
-static void mi_stat_average_print(size_t count, size_t total, const char* msg, mi_output_fun* out, void* arg) {
+static void mi_stat_average_print(int64_t count, int64_t total, const char* msg, mi_output_fun* out, void* arg) {
   const int64_t avg_tens = (count == 0 ? 0 : (total*10 / count));
-  const long avg_whole = (long)(avg_tens/10);
-  const long avg_frac1 = (long)(avg_tens%10);
-  _mi_fprintf(out, arg, "  %-10s: %5ld.%ld avg\n", msg, avg_whole, avg_frac1);
+  const int64_t avg_whole = avg_tens/10;
+  const int64_t avg_frac1 = avg_tens%10;
+  _mi_fprintf(out, arg, "  %-10s: %5lld.%lld avg\n", msg, avg_whole, avg_frac1);
 }
 
 
@@ -335,7 +339,7 @@ mi_decl_export void mi_process_info_print_out(mi_output_fun* out, void* arg) mi_
   _mi_fprintf(out, arg, "\n");
 }
 
-void _mi_stats_print(const char* name, size_t id, mi_stats_t* stats, mi_output_fun* out0, void* arg0) mi_attr_noexcept {
+void _mi_stats_print(const char* name, size_t id, const mi_stats_t* stats, mi_output_fun* out0, void* arg0) mi_attr_noexcept {
   // wrap the output function to be line buffered
   char buf[256]; _mi_memzero_var(buf);
   buffered_t buffer = { out0, arg0, NULL, 0, 255 };
@@ -352,8 +356,8 @@ void _mi_stats_print(const char* name, size_t id, mi_stats_t* stats, mi_output_f
     mi_stats_print_bins(stats->malloc_bins, MI_BIN_HUGE, out, arg);
     #endif
     #if MI_STAT
-    mi_stat_print(&stats->malloc_normal, "binned", (stats->malloc_normal_count.total == 0 ? 1 : -1), out, arg);
-    mi_stat_print(&stats->malloc_huge, "huge", (stats->malloc_huge_count.total == 0 ? 1 : -1), out, arg);
+    mi_stat_print(&stats->malloc_normal, "binned", (stats->malloc_normal_count.total == 0 ? -1 : 1), out, arg);
+    mi_stat_print(&stats->malloc_huge, "huge", (stats->malloc_huge_count.total == 0 ? -1 : 1), out, arg);
     mi_stat_count_t total = { 0,0,0 };
     mi_stat_count_add_mt(&total, &stats->malloc_normal);
     mi_stat_count_add_mt(&total, &stats->malloc_huge);
@@ -371,8 +375,8 @@ void _mi_stats_print(const char* name, size_t id, mi_stats_t* stats, mi_output_f
     // mi_stat_print(&stats->segments, "segments", -1, out, arg);
     // mi_stat_print(&stats->segments_abandoned, "-abandoned", -1, out, arg);
     // mi_stat_print(&stats->segments_cache, "-cached", -1, out, arg);
-    mi_stat_print(&stats->pages, "pages", -1, out, arg);
-    mi_stat_print(&stats->pages_abandoned, "abandoned", -1, out, arg);
+    mi_stat_print(&stats->pages, "pages", 0, out, arg);
+    mi_stat_print(&stats->pages_abandoned, "abandoned", 0, out, arg);
     mi_stat_counter_print(&stats->pages_reclaim_on_alloc, "reclaima", out, arg);
     mi_stat_counter_print(&stats->pages_reclaim_on_free, "reclaimf", out, arg);
     mi_stat_counter_print(&stats->pages_reabandon_full, "reabandon", out, arg);
@@ -387,8 +391,8 @@ void _mi_stats_print(const char* name, size_t id, mi_stats_t* stats, mi_output_f
     mi_print_header("arenas", out, arg);
     mi_stat_print_ex(&stats->reserved, "reserved", 1, out, arg, "");
     mi_stat_print_ex(&stats->committed, "committed", 1, out, arg, "");
-    mi_stat_counter_print(&stats->reset, "reset", out, arg);
-    mi_stat_counter_print(&stats->purged, "purged", out, arg);
+    mi_stat_counter_print_size(&stats->reset, "reset", out, arg);
+    mi_stat_counter_print_size(&stats->purged, "purged", out, arg);
 
     mi_stat_counter_print(&stats->arena_count, "arenas", out, arg);
     mi_stat_counter_print(&stats->arena_rollback_count, "rollback", out, arg);
@@ -397,11 +401,13 @@ void _mi_stats_print(const char* name, size_t id, mi_stats_t* stats, mi_output_f
     mi_stat_counter_print(&stats->reset_calls, "resets", out, arg);
     mi_stat_counter_print(&stats->purge_calls, "purges", out, arg);
     mi_stat_counter_print(&stats->malloc_guarded_count, "guarded", out, arg);
-    mi_stat_print_ex(&stats->heaps, "heaps", -1, out, arg, "");
+    mi_stat_print_ex(&stats->theaps, "theaps", 0, out, arg, "");
+    mi_stat_print_ex(&stats->heaps, "heaps", 0, out, arg, "");
+    mi_stat_counter_print(&stats->heaps_delete_wait, "heap waits", out, arg);
     _mi_fprintf(out, arg, "\n");
 
     mi_print_header("process", out, arg);
-    mi_stat_print_ex(&stats->threads, "threads", -1, out, arg, "");
+    mi_stat_print_ex(&stats->threads, "threads", 0, out, arg, "");
     _mi_fprintf(out, arg, "  %-10s: %5i\n", "numa nodes", _mi_os_numa_node_count());
     mi_process_info_print_out(out, arg);
   }
@@ -416,7 +422,7 @@ void _mi_stats_init(void) {
   if (mi_process_start == 0) { mi_process_start = _mi_clock_start(); };
 }
 
-static void mi_stats_add_into(mi_stats_t* to, mi_stats_t* from) {
+static void mi_stats_add_into(mi_stats_t* to, const mi_stats_t* from) {
   mi_assert_internal(to != NULL && from != NULL);
   if (to == from) return;
   mi_stats_add(to, from);
@@ -429,14 +435,14 @@ void _mi_stats_merge_into(mi_stats_t* to, mi_stats_t* from) {
   _mi_memzero(from, sizeof(mi_stats_t));
 }
 
-static mi_stats_t* mi_stats_merge_theap_to_heap(mi_theap_t* theap) mi_attr_noexcept {
+static const mi_stats_t* mi_stats_merge_theap_to_heap(mi_theap_t* theap) mi_attr_noexcept {
   mi_stats_t* stats = &theap->stats;
-  mi_stats_t* heap_stats = &theap->heap->stats;
+  mi_stats_t* heap_stats = &_mi_theap_heap(theap)->stats;
   _mi_stats_merge_into( heap_stats, stats );
   return heap_stats;
 }
 
-static mi_stats_t* mi_heap_get_stats(mi_heap_t* heap) {
+static const mi_stats_t* mi_heap_get_stats(mi_heap_t* heap) {
   if (heap==NULL) { heap = mi_heap_main(); }
   mi_theap_t* theap = _mi_heap_theap_peek(heap);
   if (theap==NULL) return &heap->stats;
@@ -473,7 +479,7 @@ void mi_subproc_heap_stats_print_out(mi_subproc_id_t subproc_id, mi_output_fun*
   mi_subproc_t* subproc = _mi_subproc_from_id(subproc_id);
   if (subproc==NULL) return;
   mi_heap_print_visit_info_t vinfo = { out, arg };
-  mi_subproc_visit_heaps(subproc, &mi_heap_print_visitor, &vinfo);
+  mi_subproc_visit_heaps(subproc_id, &mi_heap_print_visitor, &vinfo);
   _mi_stats_print("subproc", subproc->subproc_seq, &subproc->stats, out, arg);
 }
 
@@ -482,7 +488,7 @@ void mi_subproc_heap_stats_print_out(mi_subproc_id_t subproc_id, mi_output_fun*
 void mi_subproc_stats_print_out(mi_subproc_id_t subproc_id, mi_output_fun* out, void* arg) mi_attr_noexcept {
   mi_subproc_t* subproc = _mi_subproc_from_id(subproc_id);
   if (subproc==NULL) return;
-  mi_stats_t_decl(stats); 
+  mi_stats_t_decl(stats);
   if (mi_subproc_stats_get(subproc_id, &stats)) {
     _mi_stats_print("subproc", subproc->subproc_seq, &stats, out, arg);
   }
@@ -502,7 +508,7 @@ void mi_stats_print(void* out) mi_attr_noexcept {
 void mi_thread_stats_print_out(mi_output_fun* out, void* arg) mi_attr_noexcept {
   mi_theap_t* theap = _mi_theap_default();
   if (theap==NULL || !mi_theap_is_initialized(theap)) return;
-  _mi_stats_print("heap", theap->heap->heap_seq, &theap->stats, out, arg);
+  _mi_stats_print("heap", _mi_theap_heap(theap)->heap_seq, &theap->stats, out, arg);
   mi_stats_merge_theap_to_heap(_mi_theap_default());
 }
 
@@ -537,12 +543,15 @@ mi_msecs_t _mi_clock_end(mi_msecs_t start) {
 
 mi_decl_export void mi_process_info(size_t* elapsed_msecs, size_t* user_msecs, size_t* system_msecs, size_t* current_rss, size_t* peak_rss, size_t* current_commit, size_t* peak_commit, size_t* page_faults) mi_attr_noexcept
 {
-  mi_subproc_t* subproc = _mi_subproc_main();
   mi_process_info_t pinfo;
   _mi_memzero_var(pinfo);
   pinfo.elapsed        = _mi_clock_end(mi_process_start);
-  pinfo.current_commit = (size_t)(mi_atomic_loadi64_relaxed((_Atomic(int64_t)*)(&subproc->stats.committed.current)));
-  pinfo.peak_commit    = (size_t)(mi_atomic_loadi64_relaxed((_Atomic(int64_t)*)(&subproc->stats.committed.peak)));
+  { const mi_subproc_t* subproc = _mi_subproc_main();
+    if (subproc!=NULL) {
+      pinfo.current_commit = (size_t)(mi_atomic_loadi64_relaxed((_Atomic(int64_t)*)(&subproc->stats.committed.current)));
+      pinfo.peak_commit    = (size_t)(mi_atomic_loadi64_relaxed((_Atomic(int64_t)*)(&subproc->stats.committed.peak)));
+    }  
+  }
   pinfo.current_rss    = pinfo.current_commit;
   pinfo.peak_rss       = pinfo.peak_commit;
   pinfo.utime          = 0;
@@ -575,34 +584,36 @@ size_t mi_stats_get_bin_size(size_t bin) mi_attr_noexcept {
   return _mi_bin_size(bin);
 }
 
-static bool _mi_stats_get(mi_stats_t* stats_in, mi_stats_t* stats_out) mi_attr_noexcept {
-  if (stats_out == NULL || stats_out->size != sizeof(mi_stats_t) || stats_out->version != MI_STAT_VERSION) return false;
-  if (stats_in == NULL || stats_in->size != stats_out->size) return false;
-  _mi_memcpy(stats_out, stats_in, stats_out->size);
+static bool mi_stats_copy(mi_stats_t* stats_to, const mi_stats_t* stats_from) mi_attr_noexcept {
+  if (stats_to == NULL || stats_to->size != sizeof(mi_stats_t) || stats_to->version != MI_STAT_VERSION) return false;
+  if (stats_from == NULL || stats_from->size != stats_to->size) return false;
+  _mi_memcpy(stats_to, stats_from, stats_to->size);
   return true;
 }
 
 bool mi_subproc_stats_get_exclusive(mi_subproc_id_t subproc_id, mi_stats_t* stats) mi_attr_noexcept {
-  return _mi_stats_get(&_mi_subproc_from_id(subproc_id)->stats, stats);
+  const mi_subproc_t* subproc = _mi_subproc_from_id(subproc_id);
+  if (subproc==NULL) return false;
+  return mi_stats_copy(stats,&subproc->stats);
 }
 
 bool mi_heap_stats_get(mi_heap_t* heap, mi_stats_t* stats) mi_attr_noexcept {
-  return _mi_stats_get(mi_heap_get_stats(heap), stats);
+  return mi_stats_copy(stats, mi_heap_get_stats(heap));
 }
 
 
 static bool mi_cdecl mi_heap_aggregate_visitor(mi_heap_t* heap, void* arg) {
   mi_stats_t* stats = (mi_stats_t*)arg;
-  mi_stats_add_into(stats, &heap->stats);
+  mi_stats_add_into(stats, mi_heap_get_stats(heap));
   return true;
 }
 
 bool mi_subproc_stats_get(mi_subproc_id_t subproc_id, mi_stats_t* stats) mi_attr_noexcept {
+  if (stats==NULL) return false;
   mi_subproc_t* subproc = _mi_subproc_from_id(subproc_id);
-  if (stats == NULL || stats->size != sizeof(mi_stats_t) || stats->version != MI_STAT_VERSION) return false;
-  _mi_memzero(stats,stats->size);  
-  mi_subproc_visit_heaps(subproc, &mi_heap_aggregate_visitor, stats);
-  mi_stats_add_into(stats, &subproc->stats);
+  if (subproc == NULL) return false;
+  if (!mi_stats_copy(stats, &subproc->stats)) return false;
+  mi_subproc_visit_heaps(subproc_id, &mi_heap_aggregate_visitor, stats);
   return true;
 }
 
@@ -651,7 +662,7 @@ static void mi_json_buf_print(mi_json_buf_t* hbuf, const char* msg) {
   hbuf->buf[hbuf->used] = 0;
 }
 
-static void mi_json_buf_print_count_bin(mi_json_buf_t* hbuf, const char* prefix, mi_stat_count_t* stat, size_t bin, bool add_comma) {
+static void mi_json_buf_print_count_bin(mi_json_buf_t* hbuf, const char* prefix, const mi_stat_count_t* stat, size_t bin, bool add_comma) {
   const size_t binsize = mi_stats_get_bin_size(bin);
   const size_t pagesize = (binsize <= MI_SMALL_MAX_OBJ_SIZE ? MI_SMALL_PAGE_SIZE :
                             (binsize <= MI_MEDIUM_MAX_OBJ_SIZE ? MI_MEDIUM_PAGE_SIZE :
@@ -662,7 +673,7 @@ static void mi_json_buf_print_count_bin(mi_json_buf_t* hbuf, const char* prefix,
   mi_json_buf_print(hbuf, buf);
 }
 
-static void mi_json_buf_print_count_cbin(mi_json_buf_t* hbuf, const char* prefix, mi_stat_count_t* stat, mi_chunkbin_t bin, bool add_comma) {
+static void mi_json_buf_print_count_cbin(mi_json_buf_t* hbuf, const char* prefix, const mi_stat_count_t* stat, mi_chunkbin_t bin, bool add_comma) {
   const char* cbin = " ";
   switch(bin) {
     case MI_CBIN_SMALL:  cbin = "S"; break;
@@ -678,14 +689,14 @@ static void mi_json_buf_print_count_cbin(mi_json_buf_t* hbuf, const char* prefix
   mi_json_buf_print(hbuf, buf);
 }
 
-static void mi_json_buf_print_count(mi_json_buf_t* hbuf, const char* prefix, mi_stat_count_t* stat, bool add_comma) {
+static void mi_json_buf_print_count(mi_json_buf_t* hbuf, const char* prefix, const mi_stat_count_t* stat, bool add_comma) {
   char buf[128];
   _mi_snprintf(buf, 128, "%s{ \"total\": %lld, \"peak\": %lld, \"current\": %lld }%s\n", prefix, stat->total, stat->peak, stat->current, (add_comma ? "," : ""));
   buf[127] = 0;
   mi_json_buf_print(hbuf, buf);
 }
 
-static void mi_json_buf_print_count_value(mi_json_buf_t* hbuf, const char* name, mi_stat_count_t* stat) {
+static void mi_json_buf_print_count_value(mi_json_buf_t* hbuf, const char* name, const mi_stat_count_t* stat) {
   char buf[128];
   _mi_snprintf(buf, 128, "  \"%s\": ", name);
   buf[127] = 0;
@@ -707,14 +718,14 @@ static void mi_json_buf_print_size(mi_json_buf_t* hbuf, const char* name, size_t
   mi_json_buf_print(hbuf, buf);
 }
 
-static void mi_json_buf_print_counter_value(mi_json_buf_t* hbuf, const char* name, mi_stat_counter_t* stat) {
+static void mi_json_buf_print_counter_value(mi_json_buf_t* hbuf, const char* name, const mi_stat_counter_t* stat) {
   mi_json_buf_print_value(hbuf, name, stat->total);
 }
 
 #define MI_STAT_COUNT(stat)    mi_json_buf_print_count_value(&hbuf, #stat, &stats->stat);
 #define MI_STAT_COUNTER(stat)  mi_json_buf_print_counter_value(&hbuf, #stat, &stats->stat);
 
-static char* mi_stats_get_json_from(mi_stats_t* stats, size_t output_size, char* output_buf) mi_attr_noexcept {
+static char* mi_stats_get_json_from(const mi_stats_t* stats, size_t output_size, char* output_buf) mi_attr_noexcept {
   if (stats==NULL || stats->size!=sizeof(mi_stats_t) || stats->version!=MI_STAT_VERSION) return NULL;
   mi_json_buf_t hbuf = { NULL, 0, 0, true };
   if (output_size > 0 && output_buf != NULL) {
@@ -769,7 +780,7 @@ static char* mi_stats_get_json_from(mi_stats_t* stats, size_t output_size, char*
   for (size_t i = 0; i < MI_CBIN_COUNT; i++) {
     mi_json_buf_print_count_cbin(&hbuf, "    ", &stats->chunk_bins[i], (mi_chunkbin_t)i, i!=MI_CBIN_COUNT-1);
   }
-  mi_json_buf_print(&hbuf, "  ]\n");  
+  mi_json_buf_print(&hbuf, "  ]\n");
   mi_json_buf_print(&hbuf, "}\n");
   if (hbuf.used >= hbuf.size) {
     // failed
@@ -786,7 +797,7 @@ char* mi_subproc_stats_get_json(mi_subproc_id_t subproc_id, size_t buf_size, cha
   if (subproc==NULL) return NULL;
   mi_stats_t_decl(stats);
   if (!mi_subproc_stats_get(subproc_id,&stats)) return NULL;
-  return mi_stats_get_json_from(&subproc->stats, buf_size, buf);  
+  return mi_stats_get_json_from(&stats, buf_size, buf);
 }
 
 char* mi_heap_stats_get_json(mi_heap_t* heap, size_t buf_size, char* buf) mi_attr_noexcept {

package/vendor/mimalloc/src/theap.c

@@ -21,7 +21,7 @@ terms of the MIT license. A copy of the license can be found in the file
 typedef bool (theap_page_visitor_fun)(mi_theap_t* theap, mi_page_queue_t* pq, mi_page_t* page, void* arg1, void* arg2);
 
 // Visit all pages in a theap; returns `false` if break was called.
-static bool mi_theap_visit_pages(mi_theap_t* theap, theap_page_visitor_fun* fn, void* arg1, void* arg2)
+static bool mi_theap_visit_pages(mi_theap_t* theap, theap_page_visitor_fun* fn, bool include_full, void* arg1, void* arg2)
 {
   if (theap==NULL || theap->page_count==0) return 0;
 
@@ -31,7 +31,8 @@ static bool mi_theap_visit_pages(mi_theap_t* theap, theap_page_visitor_fun* fn,
   size_t count = 0;
   #endif
 
-  for (size_t i = 0; i <= MI_BIN_FULL; i++) {
+  const size_t max_bin = (include_full ? MI_BIN_FULL : MI_BIN_FULL - 1);
+  for (size_t i = 0; i <= max_bin; i++) {
     mi_page_queue_t* pq = &theap->pages[i];
     mi_page_t* page = pq->first;
     while(page != NULL) {
@@ -44,7 +45,7 @@ static bool mi_theap_visit_pages(mi_theap_t* theap, theap_page_visitor_fun* fn,
       page = next; // and continue
     }
   }
-  mi_assert_internal(count == total);
+  mi_assert_internal(!include_full || count == total);
   return true;
 }
 
@@ -62,7 +63,7 @@ static bool mi_theap_page_is_valid(mi_theap_t* theap, mi_page_queue_t* pq, mi_pa
 #if MI_DEBUG>=3
 static bool mi_theap_is_valid(mi_theap_t* theap) {
   mi_assert_internal(theap!=NULL);
-  mi_theap_visit_pages(theap, &mi_theap_page_is_valid, NULL, NULL);
+  mi_theap_visit_pages(theap, &mi_theap_page_is_valid, true, NULL, NULL);
   for (size_t bin = 0; bin < MI_BIN_COUNT; bin++) {
     mi_assert_internal(_mi_page_queue_is_valid(theap, &theap->pages[bin]));
   }
@@ -109,7 +110,7 @@ static bool mi_theap_page_collect(mi_theap_t* theap, mi_page_queue_t* pq, mi_pag
 
 static void mi_theap_merge_stats(mi_theap_t* theap) {
   mi_assert_internal(mi_theap_is_initialized(theap));
-  _mi_stats_merge_into(&theap->heap->stats, &theap->stats);
+  _mi_stats_merge_into(&_mi_theap_heap(theap)->stats, &theap->stats);
 }
 
 static void mi_theap_collect_ex(mi_theap_t* theap, mi_collect_t collect)
@@ -127,7 +128,7 @@ static void mi_theap_collect_ex(mi_theap_t* theap, mi_collect_t collect)
   _mi_theap_collect_retired(theap, force);
 
   // collect all pages owned by this thread
-  mi_theap_visit_pages(theap, &mi_theap_page_collect, &collect, NULL);
+  mi_theap_visit_pages(theap, &mi_theap_page_collect, (collect!=MI_NORMAL), &collect, NULL);  // dont normally visit full pages, see issue #1220
 
   // collect arenas (this is program wide so don't force purges on abandonment of threads)
   //mi_atomic_storei64_release(&theap->tld->subproc->purge_expire, 1);
@@ -173,11 +174,14 @@ mi_theap_t* mi_theap_get_default(void) {
 void _mi_theap_init(mi_theap_t* theap, mi_heap_t* heap, mi_tld_t* tld)
 {
   mi_assert_internal(theap!=NULL);
+  mi_assert_internal(heap!=NULL);
   mi_memid_t memid = theap->memid;
   _mi_memcpy_aligned(theap, &_mi_theap_empty, sizeof(mi_theap_t));
   theap->memid = memid;
-  theap->heap  = heap;
+  theap->refcount = 1;
   theap->tld   = tld;  // avoid reading the thread-local tld during initialization
+  mi_atomic_store_ptr_relaxed(mi_heap_t,&theap->heap,heap);
+  mi_assert_internal(theap->stats.size == sizeof(mi_stats_t));
   
   _mi_theap_options_init(theap);
   if (theap->tld->is_in_threadpool) {
@@ -191,11 +195,14 @@ void _mi_theap_init(mi_theap_t* theap, mi_heap_t* heap, mi_tld_t* tld)
   }
 
   // push on the thread local theaps list
-  mi_theap_t* head = theap->tld->theaps;
-  theap->tprev = NULL;
-  theap->tnext = head;
-  if (head!=NULL) { head->tprev = theap; }
-  theap->tld->theaps = theap;
+  mi_theap_t* head = NULL;
+  mi_lock(&theap->tld->theaps_lock) {
+    head = theap->tld->theaps;
+    theap->tprev = NULL;
+    theap->tnext = head;
+    if (head!=NULL) { head->tprev = theap; }
+    theap->tld->theaps = theap;
+  }
 
   // initialize random
   if (head == NULL) {  // first theap in this thread?
@@ -210,6 +217,7 @@ void _mi_theap_init(mi_theap_t* theap, mi_heap_t* heap, mi_tld_t* tld)
   }
   theap->cookie  = _mi_theap_random_next(theap) | 1;
   _mi_theap_guarded_init(theap);
+  mi_subproc_stat_increase(_mi_subproc(),theaps,1);
 
   // push on the heap's theap list
   mi_lock(&heap->theaps_lock) {
@@ -256,37 +264,70 @@ uintptr_t _mi_theap_random_next(mi_theap_t* theap) {
   return _mi_random_next(&theap->random);
 }
 
-// called from `mi_theap_delete` to free the internal theap resources.
-void _mi_theap_free(mi_theap_t* theap) {
-  mi_assert(theap != NULL);
-  mi_assert_internal(mi_theap_is_initialized(theap));
-  if (theap==NULL || !mi_theap_is_initialized(theap)) return;
+static void mi_theap_free_mem(mi_theap_t* theap) {
+  if (theap!=NULL) {
+    mi_subproc_stat_decrease(_mi_subproc(),theaps,1);
+    // free the used memory
+    if (theap->memid.memkind == MI_MEM_HEAP_MAIN) {  // note: for now unused as it would access theap_default stats in mi_free of the current theap
+      mi_assert_internal(_mi_is_heap_main(mi_heap_of(theap)));
+      mi_free(theap);
+    }
+    else if (theap->memid.memkind == MI_MEM_META) {
+      _mi_meta_free(theap, sizeof(*theap), theap->memid);
+    }
+    else {
+      _mi_arenas_free(theap, _mi_align_up(sizeof(*theap),MI_ARENA_MIN_OBJ_SIZE), theap->memid ); // issue #1168, avoid assertion failure
+    }
+  }
+}
 
-  // merge stats to the owning heap
-  mi_theap_merge_stats(theap);
+void _mi_theap_incref(mi_theap_t* theap) {
+  if (theap!=NULL && theap->memid.memkind > MI_MEM_STATIC) {
+    mi_atomic_increment_acq_rel(&theap->refcount);
+  }
+}
 
-  // remove ourselves from the heap theaps list
-  mi_lock(&theap->heap->theaps_lock) {
-    if (theap->hnext != NULL) { theap->hnext->hprev = theap->hprev; }
-    if (theap->hprev != NULL) { theap->hprev->hnext = theap->hnext; }
-                         else { mi_assert_internal(theap->heap->theaps == theap); theap->heap->theaps = theap->hnext; }
+void _mi_theap_decref(mi_theap_t* theap) {
+  if (theap!=NULL && theap->memid.memkind > MI_MEM_STATIC) {
+    if (mi_atomic_decrement_acq_rel(&theap->refcount) == 1) {
+      mi_theap_free_mem(theap);
+    }
   }
+}
 
-  // remove ourselves from the thread local theaps list
-  if (theap->tnext != NULL) { theap->tnext->tprev = theap->tprev;  }
-  if (theap->tprev != NULL) { theap->tprev->tnext = theap->tnext;  }
-                       else { mi_assert_internal(theap->tld->theaps == theap); theap->tld->theaps = theap->tnext; }
 
-  // and free the used memory
-  if (theap->memid.memkind == MI_MEM_HEAP_MAIN) {  // note: for now unused as it would access theap_default stats in mi_free of the current theap
-    mi_assert_internal(_mi_is_heap_main(mi_heap_of(theap)));
-    mi_free(theap);
-  }
-  else if (theap->memid.memkind == MI_MEM_META) {
-    _mi_meta_free(theap, sizeof(*theap), theap->memid);
+// called from `mi_theap_delete` to free the internal theap resources.
+bool _mi_theap_free(mi_theap_t* theap, bool acquire_heap_theaps_lock, bool acquire_tld_theaps_lock) {
+  mi_assert(theap != NULL);
+  if (theap==NULL) return true;
+
+  mi_heap_t* const heap = mi_atomic_exchange_ptr_acq_rel(mi_heap_t, &theap->heap, NULL);
+  if (heap==NULL) {
+    // concurrent interaction, retry in an outer loop (as the other thread may be blocked on our lock)
+    return false;
   }
   else {
-    _mi_arenas_free(theap, _mi_align_up(sizeof(*theap),MI_ARENA_MIN_OBJ_SIZE), theap->memid ); // issue #1168, avoid assertion failure
+    // merge stats to the owning heap
+    _mi_stats_merge_into(&heap->stats, &theap->stats);
+
+    // remove ourselves from the heap theaps list
+    mi_lock_maybe(&heap->theaps_lock, acquire_heap_theaps_lock) {
+      if (theap->hnext != NULL) { theap->hnext->hprev = theap->hprev; }
+      if (theap->hprev != NULL) { theap->hprev->hnext = theap->hnext; }
+                          else { mi_assert_internal(heap->theaps == theap); heap->theaps = theap->hnext; }
+      theap->hnext = theap->hprev = NULL;
+    }
+
+    // remove ourselves from the thread local theaps list
+    mi_lock_maybe(&theap->tld->theaps_lock, acquire_tld_theaps_lock) {
+      if (theap->tnext != NULL) { theap->tnext->tprev = theap->tprev;  }
+      if (theap->tprev != NULL) { theap->tprev->tnext = theap->tnext;  }
+                          else { mi_assert_internal(theap->tld->theaps == theap); theap->tld->theaps = theap->tnext; }
+      theap->tnext = theap->tprev = NULL;                        
+    }
+    theap->tld = NULL;
+    _mi_theap_decref(theap);
+    return true;
   }
 }
 
@@ -411,7 +452,7 @@ void _mi_theap_unsafe_destroy_all(mi_theap_t* theap) {
 ----------------------------------------------------------- */
 
 // Safe delete a theap without freeing any still allocated blocks in that theap.
-void _mi_theap_delete(mi_theap_t* theap)
+void _mi_theap_delete(mi_theap_t* theap, bool acquire_tld_theaps_lock)
 {
   mi_assert(theap != NULL);
   mi_assert(mi_theap_is_initialized(theap));
@@ -422,7 +463,7 @@ void _mi_theap_delete(mi_theap_t* theap)
   _mi_theap_collect_abandon(theap);
 
   mi_assert_internal(theap->page_count==0);
-  _mi_theap_free(theap);
+  _mi_theap_free(theap, true /* acquire heap->theaps_lock */, acquire_tld_theaps_lock);
 }
 
 
@@ -435,7 +476,7 @@ void mi_theap_unload(mi_theap_t* theap) {
   mi_assert(mi_theap_is_initialized(theap));
   mi_assert_expensive(mi_theap_is_valid(theap));
   if (theap==NULL || !mi_theap_is_initialized(theap)) return;
-  if (theap->heap->exclusive_arena == NULL) {
+  if (_mi_theap_heap(theap)->exclusive_arena == NULL) {
     _mi_warning_message("cannot unload theaps that are not associated with an exclusive arena\n");
     return;
   }
@@ -455,7 +496,7 @@ void mi_theap_unload(mi_theap_t* theap) {
 bool mi_theap_reload(mi_theap_t* theap, mi_arena_id_t arena_id) {
   mi_assert(mi_theap_is_initialized(theap));
   if (theap==NULL || !mi_theap_is_initialized(theap)) return false;
-  if (theap->heap->exclusive_arena == NULL) {
+  if (_mi_theap_heap(theap)->exclusive_arena == NULL) {
     _mi_warning_message("cannot reload theaps that were not associated with an exclusive arena\n");
     return false;
   }
@@ -464,8 +505,8 @@ bool mi_theap_reload(mi_theap_t* theap, mi_arena_id_t arena_id) {
     return false;
   }
   mi_arena_t* arena = _mi_arena_from_id(arena_id);
-  if (theap->heap->exclusive_arena != arena) {
-    _mi_warning_message("trying to reload a theap at a different arena address: %p vs %p\n", theap->heap->exclusive_arena, arena);
+  if (_mi_theap_heap(theap)->exclusive_arena != arena) {
+    _mi_warning_message("trying to reload a theap at a different arena address: %p vs %p\n", _mi_theap_heap(theap)->exclusive_arena, arena);
     return false;
   }
 
@@ -645,7 +686,7 @@ static bool mi_theap_visit_areas_page(mi_theap_t* theap, mi_page_queue_t* pq, mi
 // Visit all theap pages as areas
 static bool mi_theap_visit_areas(const mi_theap_t* theap, mi_theap_area_visit_fun* visitor, void* arg) {
   if (visitor == NULL) return false;
-  return mi_theap_visit_pages((mi_theap_t*)theap, &mi_theap_visit_areas_page, (void*)(visitor), arg); // note: function pointer to void* :-{
+  return mi_theap_visit_pages((mi_theap_t*)theap, &mi_theap_visit_areas_page, true, (void*)(visitor), arg); // note: function pointer to void* :-{
 }
 
 // Just to pass arguments
@@ -657,7 +698,7 @@ typedef struct mi_visit_blocks_args_s {
 
 static bool mi_theap_area_visitor(const mi_theap_t* theap, const mi_theap_area_ex_t* xarea, void* arg) {
   mi_visit_blocks_args_t* args = (mi_visit_blocks_args_t*)arg;
-  if (!args->visitor(theap->heap, &xarea->area, NULL, xarea->area.block_size, args->arg)) return false;
+  if (!args->visitor(_mi_theap_heap(theap), &xarea->area, NULL, xarea->area.block_size, args->arg)) return false;
   if (args->visit_blocks) {
     return _mi_theap_area_visit_blocks(&xarea->area, xarea->page, args->visitor, args->arg);
   }