merge from dev

CMakeLists.txt

@@ -131,6 +131,11 @@ if(CMAKE_C_COMPILER_ID MATCHES "AppleClang|Clang|GNU|Intel")
   endif()
 endif()
 
+# Architecture flags
+if(${CMAKE_HOST_SYSTEM_PROCESSOR} MATCHES "arm")
+    list(APPEND mi_cflags -march=native)
+endif()
+
 # extra needed libraries
 if(WIN32)
   list(APPEND mi_libraries psapi shell32 user32 bcrypt)
@@ -216,7 +221,7 @@ if(NOT WIN32)
   # install a symlink in the /usr/local/lib to the versioned library
   set(mi_symlink "${CMAKE_SHARED_MODULE_PREFIX}${mi_basename}${CMAKE_SHARED_LIBRARY_SUFFIX}")
   set(mi_soname "mimalloc-${mi_version}/${mi_symlink}.${mi_version}")
-  install(CODE "execute_process(COMMAND ${CMAKE_COMMAND} -E create_symlink ${mi_soname} ${mi_symlink} WORKING_DIRECTORY ${CMAKE_INSTALL_PREFIX}/${mi_install_dir}/..)")
+  install(CODE "execute_process(COMMAND ${CMAKE_COMMAND} -E create_symlink ${mi_soname} ${mi_symlink} WORKING_DIRECTORY ${mi_install_dir}/..)")
   install(CODE "MESSAGE(\"-- Symbolic link: ${CMAKE_INSTALL_PREFIX}/lib/${mi_symlink} -> ${mi_soname}\")")
 endif()
 

include/mimalloc-types.h

@@ -332,6 +332,8 @@ struct mi_heap_s {
   uintptr_t             keys[2];                             // two random keys used to encode the `thread_delayed_free` list
   mi_random_ctx_t       random;                              // random number context used for secure allocation
   size_t                page_count;                          // total number of pages in the `pages` queues.
+  size_t                page_retired_min;                    // smallest retired index (retired pages are fully free, but still in the page queues)
+  size_t                page_retired_max;                    // largest retired index into the `pages` array.
   mi_heap_t*            next;                                // list of heaps per thread
   bool                  no_reclaim;                          // `true` if this heap should not reclaim abandoned pages
 };

readme.md

@@ -11,7 +11,7 @@ mimalloc (pronounced "me-malloc")
 is a general purpose allocator with excellent [performance](#performance) characteristics.
 Initially developed by Daan Leijen for the run-time systems of the
 [Koka](https://github.com/koka-lang/koka) and [Lean](https://github.com/leanprover/lean) languages.
-Latest release:`v1.6.0` (2020-02-09).
+Latest release:`v1.6.1` (2020-02-17).
 
 It is a drop-in replacement for `malloc` and can be used in other programs
 without code changes, for example, on dynamically linked ELF-based systems (Linux, BSD, etc.) you can use it as:
@@ -57,6 +57,7 @@ Enjoy!
 
 ### Releases
 
+* 2020-02-17, `v1.6.1`: stable release 1.6: minor updates (build with clang-cl, fix alignment issue for small objects).
 * 2020-02-09, `v1.6.0`: stable release 1.6: fixed potential memory leak, improved overriding
   and thread local support on FreeBSD, NetBSD, DragonFly, and macOSX. New byte-precise
   heap block overflow detection in debug mode (besides the double-free detection and free-list
@@ -275,8 +276,7 @@ resolved to the _mimalloc_ library.
 Note that certain security restrictions may apply when doing this from
 the [shell](https://stackoverflow.com/questions/43941322/dyld-insert-libraries-ignored-when-calling-application-through-bash).
 
-Note: unfortunately, at this time, dynamic overriding on macOS seems broken but it is
+(Note: macOS support for dynamic overriding is recent, please report any issues.)
-actively worked on to fix this (see issue [`#50`](https://github.com/microsoft/mimalloc/issues/50)).
 
 ### Override on Windows
 

src/alloc.c

@@ -67,6 +67,11 @@ MI_ALLOC_API1(inline mi_decl_restrict void*, malloc_small, mi_heap_t*, heap, siz
   mi_assert(heap!=NULL);
   mi_assert(heap->thread_id == 0 || heap->thread_id == _mi_thread_id()); // heaps are thread local
   mi_assert(size <= MI_SMALL_SIZE_MAX);
+  #if (MI_PADDING)
+  if (size == 0) {
+    size = sizeof(void*);
+  }
+  #endif
   mi_page_t* page = _mi_heap_get_free_small_page(heap,size + MI_PADDING_SIZE);
   void* p = _mi_page_malloc(heap, page, size + MI_PADDING_SIZE  MI_SOURCE_XARG);
   mi_assert_internal(p==NULL || mi_usable_size(p) >= size);

src/init.c

@@ -97,6 +97,7 @@ const mi_heap_t _mi_heap_empty = {
   { 0, 0 },         // keys
   { {0}, {0}, 0 },
   0,                // page count
+  MI_BIN_FULL, 0,   // page retired min/max
   NULL,             // next
   false
 };
@@ -131,6 +132,7 @@ mi_heap_t _mi_heap_main = {
   { 0, 0 },         // the key of the main heap can be fixed (unlike page keys that need to be secure!)
   { {0x846ca68b}, {0}, 0 },  // random
   0,                // page count
+  MI_BIN_FULL, 0,   // page retired min/max
   NULL,             // next heap
   false             // can reclaim
 };
@@ -241,7 +243,9 @@ static bool _mi_heap_done(mi_heap_t* heap) {
     mi_assert_internal(heap->tld->segments.count == 0);
     _mi_os_free(heap, sizeof(mi_thread_data_t), &_mi_stats_main);
   }
-#if (MI_DEBUG > 0)
+#if 0  
+  // never free the main thread even in debug mode; if a dll is linked statically with mimalloc,
+  // there may still be delete/free calls after the mi_fls_done is called. Issue #207
   else {
     _mi_heap_destroy_pages(heap);
     mi_assert_internal(heap->tld->heap_backing == &_mi_heap_main);
@@ -483,6 +487,10 @@ static void mi_process_done(void) {
   if (process_done) return;
   process_done = true;
 
+  #if defined(_WIN32) && !defined(MI_SHARED_LIB)
+  FlsSetValue(mi_fls_key, NULL);  // don't call main-thread callback
+  FlsFree(mi_fls_key);            // call thread-done on all threads to prevent dangling callback pointer if statically linked with a DLL; Issue #208
+  #endif
   #ifndef NDEBUG
   mi_collect(true);
   #endif
@@ -490,7 +498,7 @@ static void mi_process_done(void) {
       mi_option_is_enabled(mi_option_verbose)) {
     mi_stats_print(NULL);
   }
-  mi_allocator_done();
+  mi_allocator_done();  
   _mi_verbose_message("process done: 0x%zx\n", _mi_heap_main.thread_id);
   os_preloading = true; // don't call the C runtime anymore
 }

src/os.c

@@ -209,7 +209,12 @@ static void* mi_win_virtual_allocx(void* addr, size_t size, size_t try_alignment
   // on 64-bit systems, try to use the virtual address area after 4TiB for 4MiB aligned allocations
   void* hint;
   if (addr == NULL && (hint = mi_os_get_aligned_hint(try_alignment,size)) != NULL) {
-    return VirtualAlloc(hint, size, flags, PAGE_READWRITE);
+    void* p = VirtualAlloc(hint, size, flags, PAGE_READWRITE);
+    if (p != NULL) return p;
+    DWORD err = GetLastError();
+    if (err != ERROR_INVALID_ADDRESS) { // if linked with multiple instances, we may have tried to allocate at an already allocated area
+      return NULL;
+    }
   }
 #endif
 #if defined(MEM_EXTENDED_PARAMETER_TYPE_BITS)

src/page.c

@@ -380,7 +380,8 @@ void _mi_page_free(mi_page_t* page, mi_page_queue_t* pq, bool force) {
   _mi_segment_page_free(page, force, segments_tld);
 }
 
-#define MI_MAX_RETIRE_SIZE    (4*MI_SMALL_SIZE_MAX)
+#define MI_MAX_RETIRE_SIZE    MI_LARGE_OBJ_SIZE_MAX  
+#define MI_RETIRE_CYCLES      (16)
 
 // Retire a page with no more used blocks
 // Important to not retire too quickly though as new
@@ -405,7 +406,13 @@ void _mi_page_retire(mi_page_t* page) {
   if (mi_likely(page->xblock_size <= MI_MAX_RETIRE_SIZE && !mi_page_is_in_full(page))) {
     if (pq->last==page && pq->first==page) { // the only page in the queue?
       mi_stat_counter_increase(_mi_stats_main.page_no_retire,1);
-      page->retire_expire = 16;
+      page->retire_expire = MI_RETIRE_CYCLES;
+      mi_heap_t* heap = mi_page_heap(page);
+      mi_assert_internal(pq >= heap->pages);
+      const size_t index = pq - heap->pages;
+      mi_assert_internal(index < MI_BIN_FULL && index < MI_BIN_HUGE);
+      if (index < heap->page_retired_min) heap->page_retired_min = index;
+      if (index > heap->page_retired_max) heap->page_retired_max = index;
       mi_assert_internal(mi_page_all_free(page));
       return; // dont't free after all
     }
@@ -415,22 +422,32 @@ void _mi_page_retire(mi_page_t* page) {
 }
 
 // free retired pages: we don't need to look at the entire queues
-// since we only retire pages that are the last one in a queue.
+// since we only retire pages that are at the head position in a queue.
 void _mi_heap_collect_retired(mi_heap_t* heap, bool force) {
-  for(mi_page_queue_t* pq = heap->pages; pq->block_size <= MI_MAX_RETIRE_SIZE; pq++) {
+  size_t min = MI_BIN_FULL;
-    mi_page_t* page = pq->first;
+  size_t max = 0;
+  for(size_t bin = heap->page_retired_min; bin <= heap->page_retired_max; bin++) {
+    mi_page_queue_t* pq   = &heap->pages[bin];
+    mi_page_t*       page = pq->first;
     if (page != NULL && page->retire_expire != 0) {
       if (mi_page_all_free(page)) {
         page->retire_expire--;
         if (force || page->retire_expire == 0) {
           _mi_page_free(pq->first, pq, force);
         }
+        else {
+          // keep retired, update min/max
+          if (bin < min) min = bin;
+          if (bin > max) max = bin;
+        }
       }
       else {
         page->retire_expire = 0;
       }
     }
   }
+  heap->page_retired_min = min;
+  heap->page_retired_max = max;
 }
 
 

test/main-override.cpp

@@ -7,7 +7,7 @@
 #include <vector>
 #include <thread>
 #include <mimalloc.h>
-// #include <mimalloc-new-delete.h>
+#include <mimalloc-new-delete.h>
 #include <mimalloc-override.h>
 
 #ifdef _WIN32
@@ -28,6 +28,7 @@ int main() {
   // heap_no_delete();  // issue #202
   // heap_late_free();  // issue #204
   // dangling_ptr_write();
+  // padding_shrink();  // issue #209
   various_tests();
   mi_stats_print(NULL);
   return 0;
@@ -159,3 +160,18 @@ static void heap_late_free() {
 
   t1.join();
 }
+
+// issue  #209
+static void* shared_p;
+static void alloc0(/* void* arg */)
+{
+  shared_p = mi_malloc(8);
+}
+
+void padding_shrink(void)
+{
+  auto t1 = std::thread(alloc0);
+  t1.join();
+  mi_free(shared_p);
+}
+

test/test-api.c

@@ -123,7 +123,7 @@ int main() {
   });
   CHECK_BODY("posix_memalign_nomem", {
     void* p = &p;
-    int err = mi_posix_memalign(&p, sizeof(void*), SIZE_MAX);
+    volatile int err = mi_posix_memalign(&p, sizeof(void*), SIZE_MAX);
     result = (err==ENOMEM && p==&p);
   });