Merge pull request #1284 from microsoft/dev-main

Release v3.3.2

Author: daanx

.github/workflows/release.yaml

@@ -9,7 +9,7 @@ permissions:
   contents: write
 
 env:
-  RELEASE: Release v3.3.1
+  RELEASE: Release v3.3.2
   FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: true
 
 name: Release
@@ -19,7 +19,7 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        branch: [v1.9.9,v2.3.1,v3.3.1]  # [dev,dev2,dev3]
+        branch: [v1.9.10,v2.3.2,v3.3.2]  # [dev,dev2,dev3]
         # we build on the oldest ubuntu version for better binary compatibility.
         os: [windows-latest, macOS-latest, macos-15-intel, ubuntu-22.04, ubuntu-22.04-arm]  
 

.github/workflows/stale.yaml

@@ -0,0 +1,27 @@
+on:
+  workflow_dispatch:       # allow running the workflow manually   
+  schedule:
+    - cron: "15 21 * * *"  # minute, hour, day (1-31), month (1-12), day of the week (0 - 6 or SUN-SAT)
+
+name: Close inactive issues
+jobs:
+  close-issues:
+    runs-on: ubuntu-latest
+    permissions:
+      issues: write
+      pull-requests: write
+    steps:
+      - uses: actions/stale@v10
+        with:
+          days-before-issue-stale: 360
+          days-before-issue-close: 14
+          stale-issue-label: "stale"
+          stale-issue-message: "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs in the next 14 days. Thank you for your contributions!"
+          close-issue-message: "This issue was closed because it has been inactive for 14 days since being marked as stale. Please feel free to reopen if this is still an active issue."
+          days-before-pr-stale: -1
+          days-before-pr-close: -1
+          stale-pr-label: "stale"
+          stale-pr-message: "This PR has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs in the next 14 days. Thank you for your contributions!"
+          close-pr-message: "This PR was closed because it has been inactive for 14 days since being marked as stale. Please feel free to reopen if you think this PR should still be considered. Thank you again for your help."
+          operations-per-run: 32
+          repo-token: ${{ secrets.GITHUB_TOKEN }}

cmake/mimalloc-config-version.cmake

@@ -1,6 +1,6 @@
 set(mi_version_major 2)
 set(mi_version_minor 3)
-set(mi_version_patch 1)
+set(mi_version_patch 2)
 set(mi_version ${mi_version_major}.${mi_version_minor})
 
 set(PACKAGE_VERSION ${mi_version})

contrib/vcpkg/portfile.cmake

@@ -3,7 +3,7 @@ vcpkg_from_github(
   REPO microsoft/mimalloc
   HEAD_REF master
 
-  # The "REF" can be a commit hash, branch name (dev3), or a version (v3.3.1).
+  # The "REF" can be a commit hash, branch name (dev3), or a version (v3.3.2).
   REF "v${VERSION}"
   # REF e2db21e9ba9fb9172b7b0aa0fe9b8742525e8774
 

contrib/vcpkg/vcpkg.json

@@ -1,6 +1,6 @@
 {
   "name": "mimalloc",
-  "version": "3.3.0",
+  "version": "3.3.2",
   "port-version": 0,
   "description": "Compact general purpose allocator with excellent performance",
   "homepage": "https://github.com/microsoft/mimalloc",

doc/release-notes.md

@@ -10,6 +10,6 @@ Notes:
 - Generally it is recommended to download sources (or use `vcpkg` etc.) and build mimalloc as 
   part of your project.
 - Source releases can also be downloaded directly from github by the tag.  
-  For example <https://github.com/microsoft/mimalloc/archive/v3.3.0.tar.gz>.  
+  For example <https://github.com/microsoft/mimalloc/archive/v3.3.2.tar.gz>.  
 - Binary releases include a release-, debug-, and secure build.
 - Linux binaries are built on Ubuntu 22.

include/mimalloc.h

@@ -8,7 +8,7 @@ terms of the MIT license. A copy of the license can be found in the file
 #ifndef MIMALLOC_H
 #define MIMALLOC_H
 
-#define MI_MALLOC_VERSION 20301  // major + 2 digits minor + 2 digits patch
+#define MI_MALLOC_VERSION 20302  // major + 2 digits minor + 2 digits patch
 
 // ------------------------------------------------------
 // Compiler specific attributes
@@ -380,7 +380,7 @@ typedef mi_heap_t  mi_theap_t;
 #define mi_theap_collect(hp,force)        mi_heap_collect(hp,force)
 #define mi_theap_malloc(hp,sz)            mi_heap_malloc(hp,sz)
 #define mi_theap_zalloc(hp,sz)            mi_heap_zalloc(hp,sz)
-#define mi_theap_calloc(hp,cnt,sz)        mi_heap_malloc(hp,cnt,sz)
+#define mi_theap_calloc(hp,cnt,sz)        mi_heap_calloc(hp,cnt,sz)
 #define mi_theap_malloc_small(hp,sz)      mi_heap_malloc_small(hp,sz)
 #define mi_theap_malloc_aligned(hp,sz,a)  mi_heap_malloc_aligned(hp,sz,a)
 #define mi_theap_realloc(hp,p,newsz)      mi_heap_realloc(hp,p,newsz)

include/mimalloc/atomic.h

@@ -155,15 +155,17 @@ static inline void mi_atomic_maxi64_relaxed(volatile int64_t* p, int64_t x) {
 #elif defined(_MSC_VER)
 
 // Deprecated: MSVC plain C compilation wrapper that uses Interlocked operations to model C11 atomics.
-// It is recommended to always compile as C++ when using MSVC
+// It is recommended to always compile as C++ when using MSVC.
 
 #include <intrin.h>
 #ifdef _WIN64
-typedef LONG64   msc_intptr_t;
-#define MI_64(f) f##64
+typedef LONG64        msc_intptr_t;
+#define MI_MSC_64(f)  f##64
+#define MI_MSC_XX(f)  f##64
 #else
-typedef LONG     msc_intptr_t;
-#define MI_64(f) f
+typedef LONG          msc_intptr_t;
+#define MI_MSC_64(f)  f
+#define MI_MSC_XX(f)  f##32
 #endif
 
 typedef enum mi_memory_order_e {
@@ -177,23 +179,23 @@ typedef enum mi_memory_order_e {
 
 static inline uintptr_t mi_atomic_fetch_add_explicit(_Atomic(uintptr_t)*p, uintptr_t add, mi_memory_order mo) {
   (void)(mo);
-  return (uintptr_t)MI_64(_InterlockedExchangeAdd)((volatile msc_intptr_t*)p, (msc_intptr_t)add);
+  return (uintptr_t)MI_MSC_64(_InterlockedExchangeAdd)((volatile msc_intptr_t*)p, (msc_intptr_t)add);
 }
 static inline uintptr_t mi_atomic_fetch_sub_explicit(_Atomic(uintptr_t)*p, uintptr_t sub, mi_memory_order mo) {
   (void)(mo);
-  return (uintptr_t)MI_64(_InterlockedExchangeAdd)((volatile msc_intptr_t*)p, -((msc_intptr_t)sub));
+  return (uintptr_t)MI_MSC_64(_InterlockedExchangeAdd)((volatile msc_intptr_t*)p, -((msc_intptr_t)sub));
 }
 static inline uintptr_t mi_atomic_fetch_and_explicit(_Atomic(uintptr_t)*p, uintptr_t x, mi_memory_order mo) {
   (void)(mo);
-  return (uintptr_t)MI_64(_InterlockedAnd)((volatile msc_intptr_t*)p, (msc_intptr_t)x);
+  return (uintptr_t)MI_MSC_64(_InterlockedAnd)((volatile msc_intptr_t*)p, (msc_intptr_t)x);
 }
 static inline uintptr_t mi_atomic_fetch_or_explicit(_Atomic(uintptr_t)*p, uintptr_t x, mi_memory_order mo) {
   (void)(mo);
-  return (uintptr_t)MI_64(_InterlockedOr)((volatile msc_intptr_t*)p, (msc_intptr_t)x);
+  return (uintptr_t)MI_MSC_64(_InterlockedOr)((volatile msc_intptr_t*)p, (msc_intptr_t)x);
 }
 static inline bool mi_atomic_compare_exchange_strong_explicit(_Atomic(uintptr_t)*p, uintptr_t* expected, uintptr_t desired, mi_memory_order mo1, mi_memory_order mo2) {
   (void)(mo1); (void)(mo2);
-  uintptr_t read = (uintptr_t)MI_64(_InterlockedCompareExchange)((volatile msc_intptr_t*)p, (msc_intptr_t)desired, (msc_intptr_t)(*expected));
+  const uintptr_t read = (uintptr_t)MI_MSC_64(_InterlockedCompareExchange)((volatile msc_intptr_t*)p, (msc_intptr_t)desired, (msc_intptr_t)(*expected));
   if (read == *expected) {
     return true;
   }
@@ -207,68 +209,119 @@ static inline bool mi_atomic_compare_exchange_weak_explicit(_Atomic(uintptr_t)*p
 }
 static inline uintptr_t mi_atomic_exchange_explicit(_Atomic(uintptr_t)*p, uintptr_t exchange, mi_memory_order mo) {
   (void)(mo);
-  return (uintptr_t)MI_64(_InterlockedExchange)((volatile msc_intptr_t*)p, (msc_intptr_t)exchange);
+  return (uintptr_t)MI_MSC_64(_InterlockedExchange)((volatile msc_intptr_t*)p, (msc_intptr_t)exchange);
 }
 static inline void mi_atomic_thread_fence(mi_memory_order mo) {
   (void)(mo);
   _Atomic(uintptr_t) x = 0;
   mi_atomic_exchange_explicit(&x, 1, mo);
 }
+
 static inline uintptr_t mi_atomic_load_explicit(_Atomic(uintptr_t) const* p, mi_memory_order mo) {
   (void)(mo);
-#if defined(_M_IX86) || defined(_M_X64)
-  return *p;
-#else
-  uintptr_t x = *p;
-  if (mo > mi_memory_order_relaxed) {
-    while (!mi_atomic_compare_exchange_weak_explicit((_Atomic(uintptr_t)*)p, &x, x, mo, mi_memory_order_relaxed)) { /* nothing */ };
-  }
-  return x;
-#endif
+  // assert(mo<=mi_memory_order_acquire); // others are not used by mimalloc
+  #if defined(_M_IX86) || defined(_M_X64)
+    return (uintptr_t)MI_MSC_XX(__iso_volatile_load)((volatile const intptr_t*)p);
+  #elif defined(_M_ARM) || defined(_M_ARM64)
+    if (mo == mi_memory_order_relaxed) {
+      return (uintptr_t)MI_MSC_XX(__iso_volatile_load)((volatile const intptr_t*)p);
+    }
+    else if (mo <= mi_memory_order_acquire) {
+      return MI_MSC_XX(__ldar)((volatile const uintptr_t*)p);
+    }
+    else {
+      const uintptr_t u = (uintptr_t)MI_MSC_XX(__iso_volatile_load)((volatile const intptr_t*)p);
+      __dmb(15);  // _ARM(64)_BARRIER_SY
+      return u;
+    }
+  #else
+    #warning "define mi_atomic_load_explicit for MSVC C compilation on this platform (which should be readonly, see issue #1277)"
+    return MI_MSC_XX(__iso_volatile_load)((volatile const intptr_t*)p);
+  #endif
 }
 static inline void mi_atomic_store_explicit(_Atomic(uintptr_t)*p, uintptr_t x, mi_memory_order mo) {
   (void)(mo);
-#if defined(_M_IX86) || defined(_M_X64)
-  *p = x;
-#else
-  mi_atomic_exchange_explicit(p, x, mo);
-#endif
+  // assert(mo<=mi_memory_order_release); // others are not used by mimalloc
+  #if defined(_M_IX86) || defined(_M_X64)
+    MI_MSC_XX(__iso_volatile_store)((volatile intptr_t*)p, x);
+  #elif defined(_M_ARM) || defined(_M_ARM64)
+    if (mo == mi_memory_order_relaxed) {
+      MI_MSC_XX(__iso_volatile_store)((volatile intptr_t*)p, x);
+    }
+    else if (mo <= mi_memory_order_release) {
+      MI_MSC_XX(__stlr)((volatile uintptr_t*)p,x);
+    }
+    else {
+      mi_atomic_exchange_explicit(p, x, mo);
+    }
+  #else
+    mi_atomic_exchange_explicit(p, x, mo);
+  #endif
 }
+
 static inline int64_t mi_atomic_loadi64_explicit(_Atomic(int64_t)*p, mi_memory_order mo) {
   (void)(mo);
-#if defined(_M_X64)
-  return *p;
-#else
-  int64_t old = *p;
-  int64_t x = old;
-  while ((old = InterlockedCompareExchange64(p, x, old)) != x) {
-    x = old;
-  }
-  return x;
-#endif
+  // assert(mo<=mi_memory_order_acquire); // others are not used by mimalloc
+  #if defined(_M_IX86) || defined(_M_X64)
+    return __iso_volatile_load64((volatile const int64_t*)p);
+  #elif defined(_M_ARM) || defined(_M_ARM64)
+    if (mo == mi_memory_order_relaxed) {
+      return __iso_volatile_load64((volatile const int64_t*)p);
+    }
+    #if defined(_M_ARM64)
+    else if (mo <= mi_memory_order_acquire) {
+      return __ldar64((volatile const uintptr_t*)p);
+    }
+    #endif
+    else {
+      const int64_t i = __iso_volatile_load64((volatile const int64_t*)p);
+      __dmb(15);  // _ARM(64)_BARRIER_SY
+      return i;
+    }
+  #else
+    #warning "define mi_atomic_loadi64_explicit for MSVC C compilation on this platform (which should be readonly, see issue #1277)"
+    return __iso_volatile_load64((volatile const int64_t*)p);
+  #endif
 }
+
 static inline void mi_atomic_storei64_explicit(_Atomic(int64_t)*p, int64_t x, mi_memory_order mo) {
   (void)(mo);
-#if defined(_M_X64)
-  *p = x;
-#else
-  InterlockedExchange64(p, x);
-#endif
+  // assert(mo<=mi_memory_order_release); // others are not used by mimalloc
+  #if defined(_M_IX86) || defined(_M_X64)
+    __iso_volatile_store64((volatile int64_t*)p,x);
+  #elif defined(_M_ARM) || defined(_M_ARM64)
+    if (mo == mi_memory_order_relaxed) {
+      __iso_volatile_store64((volatile int64_t*)p,x);
+    }
+    #if defined(_M_ARM64)
+    else if (mo == mi_memory_order_release) {
+      __stlr64((volatile uint64_t*)p, (uint64_t)x);
+    }
+    #endif
+    else {
+      InterlockedExchange64(p, x);
+    }
+  #else
+    InterlockedExchange64(p, x);
+  #endif
 }
 
 // These are used by the statistics
 static inline int64_t mi_atomic_addi64_relaxed(volatile _Atomic(int64_t)*p, int64_t add) {
-#ifdef _WIN64
-  return (int64_t)mi_atomic_addi((int64_t*)p, add);
-#else
-  int64_t current;
-  int64_t sum;
-  do {
-    current = *p;
-    sum = current + add;
-  } while (_InterlockedCompareExchange64(p, sum, current) != current);
-  return current;
-#endif
+  #ifdef _WIN64
+    return (int64_t)mi_atomic_addi((int64_t*)p, add);
+  #elif defined(_M_ARM)
+    return _InterlockedExchangeAdd64(p, add);
+  #else
+    // x86
+    int64_t current;
+    int64_t sum;
+    do {
+      current = __iso_volatile_load64((volatile const int64_t*)p);
+      sum = current + add;
+    } while (_InterlockedCompareExchange64(p, sum, current) != current);
+    return current;
+  #endif
 }
 static inline void mi_atomic_void_addi64_relaxed(volatile int64_t* p, const volatile int64_t* padd) {
   const int64_t add = *padd;
@@ -289,7 +342,7 @@ static inline void mi_atomic_addi64_acq_rel(volatile _Atomic(int64_t*)p, int64_t
 }
 
 static inline bool mi_atomic_casi64_strong_acq_rel(volatile _Atomic(int64_t*)p, int64_t* exp, int64_t des) {
-  int64_t read = _InterlockedCompareExchange64(p, des, *exp);
+  const int64_t read = _InterlockedCompareExchange64(p, des, *exp);
   if (read == *exp) {
     return true;
   }

include/mimalloc/internal.h

@@ -352,7 +352,7 @@ mi_decl_noreturn mi_decl_cold void _mi_assert_fail(const char* assertion, const
   Inlined definitions
 ----------------------------------------------------------- */
 #define MI_UNUSED(x)     (void)(x)
-#if (MI_DEBUG>0)
+#if (MI_DEBUG>1)
 #define MI_UNUSED_RELEASE(x)
 #else
 #define MI_UNUSED_RELEASE(x)  MI_UNUSED(x)
@@ -378,8 +378,7 @@ static inline bool _mi_is_power_of_two(uintptr_t x) {
 
 // Is a pointer aligned?
 static inline bool _mi_is_aligned(void* p, size_t alignment) {
-  mi_assert_internal(alignment != 0);
-  return (((uintptr_t)p % alignment) == 0);
+  return (alignment==0 || ((uintptr_t)p % alignment) == 0);
 }
 
 // Align upwards

include/mimalloc/prim.h

@@ -282,7 +282,9 @@ static inline mi_threadid_t _mi_prim_thread_id(void) mi_attr_noexcept;
 #if defined(MI_PRIM_THREAD_ID)
 
 static inline mi_threadid_t _mi_prim_thread_id(void) mi_attr_noexcept {
-  return MI_PRIM_THREAD_ID();  // used for example by CPython for a free threaded build (see python/cpython#115488)
+  const mi_threadid_t tid = MI_PRIM_THREAD_ID();  // used for example by CPython for a free threaded build (see python/cpython#115488)
+  mi_assert_internal( (tid & 0x03) == 0 );        // mimalloc reserves the bottom 2 bits
+  return tid;
 }
 
 #elif defined(_WIN32)