Merge remote-tracking branch 'upstream/main' into gh-144278

Author: zooba

Android/testbed/app/build.gradle.kts

@@ -92,7 +92,12 @@ android {
             }
             throw GradleException("Failed to find API level in $androidEnvFile")
         }
-        targetSdk = 35
+
+        // This controls the API level of the maxVersion managed emulator, which is used
+        // by CI and cibuildwheel. 34 takes up too much disk space (#142289), 35 has
+        // issues connecting to the internet (#142387), and 36 and later are not
+        // available as aosp_atd images yet.
+        targetSdk = 33
 
         versionCode = 1
         versionName = "1.0"

Doc/c-api/memory.rst

@@ -677,7 +677,11 @@ The pymalloc allocator
 Python has a *pymalloc* allocator optimized for small objects (smaller or equal
 to 512 bytes) with a short lifetime. It uses memory mappings called "arenas"
 with a fixed size of either 256 KiB on 32-bit platforms or 1 MiB on 64-bit
-platforms. It falls back to :c:func:`PyMem_RawMalloc` and
+platforms. When Python is configured with :option:`--with-pymalloc-hugepages`,
+the arena size on 64-bit platforms is increased to 2 MiB to match the huge page
+size, and arena allocation will attempt to use huge pages (``MAP_HUGETLB`` on
+Linux, ``MEM_LARGE_PAGES`` on Windows) with automatic fallback to regular pages.
+It falls back to :c:func:`PyMem_RawMalloc` and
 :c:func:`PyMem_RawRealloc` for allocations larger than 512 bytes.
 
 *pymalloc* is the :ref:`default allocator <default-memory-allocators>` of the

Doc/using/configure.rst

@@ -783,6 +783,21 @@ also be used to improve performance.
 
    See also :envvar:`PYTHONMALLOC` environment variable.
 
+.. option:: --with-pymalloc-hugepages
+
+   Enable huge page support for :ref:`pymalloc <pymalloc>` arenas (disabled by
+   default). When enabled, the arena size on 64-bit platforms is increased to
+   2 MiB and arena allocation uses ``MAP_HUGETLB`` (Linux) or
+   ``MEM_LARGE_PAGES`` (Windows) with automatic fallback to regular pages.
+
+   The configure script checks that the platform supports ``MAP_HUGETLB``
+   and emits a warning if it is not available.
+
+   On Windows, use the ``--pymalloc-hugepages`` flag with ``build.bat`` or
+   set the ``UsePymallocHugepages`` MSBuild property.
+
+   .. versionadded:: 3.15
+
 .. option:: --without-doc-strings
 
    Disable static documentation strings to reduce the memory footprint (enabled

Doc/whatsnew/3.15.rst

@@ -1477,6 +1477,12 @@ Build changes
   modules that are missing or packaged separately.
   (Contributed by Stan Ulbrych and Petr Viktorin in :gh:`139707`.)
 
+* The new configure option :option:`--with-pymalloc-hugepages` enables huge
+  page support for :ref:`pymalloc <pymalloc>` arenas. When enabled, arena size
+  increases to 2 MiB and allocation uses ``MAP_HUGETLB`` (Linux) or
+  ``MEM_LARGE_PAGES`` (Windows) with automatic fallback to regular pages.
+  On Windows, use ``build.bat --pymalloc-hugepages``.
+
 * Annotating anonymous mmap usage is now supported if Linux kernel supports
   :manpage:`PR_SET_VMA_ANON_NAME <PR_SET_VMA(2const)>` (Linux 5.17 or newer).
   Annotations are visible in ``/proc/<pid>/maps`` if the kernel supports the feature

Include/internal/pycore_obmalloc.h

@@ -208,7 +208,11 @@ typedef unsigned int pymem_uint;  /* assuming >= 16 bits */
  * mappings to reduce heap fragmentation.
  */
 #ifdef USE_LARGE_ARENAS
-#define ARENA_BITS              20                    /* 1 MiB */
+#  ifdef PYMALLOC_USE_HUGEPAGES
+#    define ARENA_BITS            21                    /* 2 MiB */
+#  else
+#    define ARENA_BITS            20                    /* 1 MiB */
+#  endif
 #else
 #define ARENA_BITS              18                    /* 256 KiB */
 #endif
@@ -469,7 +473,7 @@ nfp free pools in usable_arenas.
 */
 
 /* How many arena_objects do we initially allocate?
- * 16 = can allocate 16 arenas = 16 * ARENA_SIZE = 4MB before growing the
+ * 16 = can allocate 16 arenas = 16 * ARENA_SIZE before growing the
  * `arenas` vector.
  */
 #define INITIAL_ARENA_OBJECTS 16
@@ -512,14 +516,26 @@ struct _obmalloc_mgmt {
 
    memory address bit allocation for keys
 
-   64-bit pointers, IGNORE_BITS=0 and 2^20 arena size:
+   ARENA_BITS is configurable: 20 (1 MiB) by default on 64-bit, or
+   21 (2 MiB) when PYMALLOC_USE_HUGEPAGES is enabled.  All bit widths
+   below are derived from ARENA_BITS automatically.
+
+   64-bit pointers, IGNORE_BITS=0 and 2^20 arena size (default):
      15 -> MAP_TOP_BITS
      15 -> MAP_MID_BITS
      14 -> MAP_BOT_BITS
      20 -> ideal aligned arena
    ----
      64
 
+   64-bit pointers, IGNORE_BITS=0 and 2^21 arena size (hugepages):
+     15 -> MAP_TOP_BITS
+     15 -> MAP_MID_BITS
+     13 -> MAP_BOT_BITS
+     21 -> ideal aligned arena
+   ----
+     64
+
    64-bit pointers, IGNORE_BITS=16, and 2^20 arena size:
      16 -> IGNORE_BITS
      10 -> MAP_TOP_BITS

Include/internal/pycore_optimizer.h

@@ -37,10 +37,15 @@ typedef struct _JitOptContext {
     // Arena for the symbolic types.
     ty_arena t_arena;
 
+    // Arena for the descriptor mappings.
+    descr_arena d_arena;
+
     JitOptRef *n_consumed;
     JitOptRef *limit;
     JitOptRef locals_and_stack[MAX_ABSTRACT_INTERP_SIZE];
     _PyJitUopBuffer out_buffer;
+    // Index of the last escaped uop in out_buffer.
+    int last_escape_index;
 } JitOptContext;
 
 
@@ -295,6 +300,9 @@ extern JitOptRef _Py_uop_sym_new_truthiness(JitOptContext *ctx, JitOptRef value,
 extern bool _Py_uop_sym_is_compact_int(JitOptRef sym);
 extern JitOptRef _Py_uop_sym_new_compact_int(JitOptContext *ctx);
 extern void _Py_uop_sym_set_compact_int(JitOptContext *ctx,  JitOptRef sym);
+extern JitOptRef _Py_uop_sym_new_descr_object(JitOptContext *ctx, unsigned int type_version);
+extern JitOptRef _Py_uop_sym_get_attr(JitOptContext *ctx, JitOptRef ref, uint16_t slot_index);
+extern JitOptRef _Py_uop_sym_set_attr(JitOptContext *ctx, JitOptRef ref, uint16_t slot_index, JitOptRef value);
 extern JitOptRef _Py_uop_sym_new_predicate(JitOptContext *ctx, JitOptRef lhs_ref, JitOptRef rhs_ref, JitOptPredicateKind kind);
 extern void _Py_uop_sym_apply_predicate_narrowing(JitOptContext *ctx, JitOptRef sym, bool branch_is_true);
 

Include/internal/pycore_optimizer_types.h

@@ -16,6 +16,10 @@ extern "C" {
 
 #define TY_ARENA_SIZE (UOP_MAX_TRACE_LENGTH * 5)
 
+// Maximum descriptor mappings per object tracked symbolically
+#define MAX_SYMBOLIC_DESCR_SIZE 16
+#define DESCR_ARENA_SIZE (MAX_SYMBOLIC_DESCR_SIZE * 100)
+
 // Need extras for root frame and for overflow frame (see TRACE_STACK_PUSH())
 #define MAX_ABSTRACT_FRAME_DEPTH (16)
 
@@ -41,6 +45,7 @@ typedef enum _JitSymType {
     JIT_SYM_TRUTHINESS_TAG = 9,
     JIT_SYM_COMPACT_INT = 10,
     JIT_SYM_PREDICATE_TAG = 11,
+    JIT_SYM_DESCR_TAG = 12,
 } JitSymType;
 
 typedef struct _jit_opt_known_class {
@@ -91,6 +96,31 @@ typedef struct {
     uint8_t tag;
 } JitOptCompactInt;
 
+/*
+Mapping from slot index or attribute offset to its symbolic value.
+SAFETY:
+This structure is used for both STORE_ATTR_SLOT and STORE_ATTR_INSTANCE_VALUE.
+These two never appear on the same object type because:
+__slots__ classes don't have Py_TPFLAGS_INLINE_VALUES
+Therefore, there is no index collision between slot offsets and inline value offsets.
+Note:
+STORE_ATTR_WITH_HINT is NOT currently tracked.
+If we want to track it in the future, we need to be careful about
+potential index collisions with STORE_ATTR_INSTANCE_VALUE.
+*/
+typedef struct {
+    uint16_t slot_index;
+    uint16_t symbol;
+} JitOptDescrMapping;
+
+typedef struct _jit_opt_descr {
+    uint8_t tag;
+    uint8_t num_descrs;
+    uint16_t last_modified_index;  // Index in out_buffer when this object was last modified
+    uint32_t type_version;
+    JitOptDescrMapping *descrs;
+} JitOptDescrObject;
+
 typedef union _jit_opt_symbol {
     uint8_t tag;
     JitOptKnownClass cls;
@@ -99,6 +129,7 @@ typedef union _jit_opt_symbol {
     JitOptTuple tuple;
     JitOptTruthiness truthiness;
     JitOptCompactInt compact;
+    JitOptDescrObject descr;
     JitOptPredicate predicate;
 } JitOptSymbol;
 
@@ -128,6 +159,11 @@ typedef struct ty_arena {
     JitOptSymbol arena[TY_ARENA_SIZE];
 } ty_arena;
 
+typedef struct descr_arena {
+    int descr_curr_number;
+    int descr_max_number;
+    JitOptDescrMapping arena[DESCR_ARENA_SIZE];
+} descr_arena;
 
 #ifdef __cplusplus
 }