Swisstable Hash implementation for dict

Include/internal/pycore_dict.h

@@ -52,7 +52,6 @@ Py_ssize_t _Py_dict_lookup(PyDictObject *mp, PyObject *key, Py_hash_t hash, PyOb
 int _PyDict_SetItem_Take2(PyDictObject *op, PyObject *key, PyObject *value);
 
 #define DKIX_EMPTY (-1)
-#define DKIX_DUMMY (-2) /* Used internally */
 #define DKIX_ERROR (-3)
 
 typedef enum {
@@ -61,6 +60,39 @@ typedef enum {
  DICT_KEYS_SPLIT = 2
 } DictKeysKind;
 
+// Currently, we support only 8-wide.
+// But it is possible to use 16-wide when SSE2 is available, but NEON can not optimize it.
+#define GROUP_WIDTH (8)
+
+typedef union {
+ char c[GROUP_WIDTH];
+ uint64_t u64;
+} group_control;
+
+typedef struct {
+ group_control control;
+ uint8_t index[GROUP_WIDTH];
+} group8; // 16byte
+
+typedef struct {
+ group_control control;
+ uint16_t index[GROUP_WIDTH];
+} group16; // 24byte
+
+typedef struct {
+ group_control control;
+ uint32_t index[GROUP_WIDTH];
+} group32; // 40byte
+
+typedef struct {
+ group_control control;
+ uint64_t index[GROUP_WIDTH];
+} group64; // 72byte
+
+// todo: group64 can use uint8_t index[7*GROUP_WIDTH] instead.
+// Then sizeof(group64) become 64byte. It is friendly to cache line.
+
+
 /* See dictobject.c for actual layout of DictKeysObject */
 struct _dictkeysobject {
  Py_ssize_t dk_refcnt;
@@ -80,20 +112,7 @@ struct _dictkeysobject {
  /* Number of used entries in dk_entries. */
  Py_ssize_t dk_nentries;
 
- /* Actual hash table of dk_size entries. It holds indices in dk_entries,
- or DKIX_EMPTY(-1) or DKIX_DUMMY(-2).
-
- Indices must be: 0 <= indice < USABLE_FRACTION(dk_size).
-
- The size in bytes of an indice depends on dk_size:
-
- - 1 byte if dk_size <= 0xff (char*)
- - 2 bytes if dk_size <= 0xffff (int16_t*)
- - 4 bytes if dk_size <= 0xffffffff (int32_t*)
- - 8 bytes otherwise (int64_t*)
-
- Dynamically sized, SIZEOF_VOID_P is minimum. */
- char dk_indices[]; /* char is required to avoid strict aliasing. */
+ unsigned char dk_groups[]; /* char is required to avoid strict aliasing. */
 
  /* "PyDictKeyEntry dk_entries[dk_usable];" array follows:
  see the DK_ENTRIES() macro */
@@ -113,23 +132,19 @@ struct _dictvalues {
  PyObject *values[1];
 };
 
-#define DK_LOG_SIZE(dk) ((dk)->dk_log2_size)
-#if SIZEOF_VOID_P > 4
+#define DK_LOG_SIZE(dk) ((dk)->dk_log2_size+3)
+
 #define DK_SIZE(dk) (((int64_t)1)<<DK_LOG_SIZE(dk))
-#define DK_IXSIZE(dk) \
- (DK_LOG_SIZE(dk) <= 7 ? \
- 1 : DK_LOG_SIZE(dk) <= 15 ? \
- 2 : DK_LOG_SIZE(dk) <= 31 ? \
- 4 : sizeof(int64_t))
-#else
-#define DK_SIZE(dk) (1<<DK_LOG_SIZE(dk))
-#define DK_IXSIZE(dk) \
- (DK_LOG_SIZE(dk) <= 7 ? \
- 1 : DK_LOG_SIZE(dk) <= 15 ? \
- 2 : sizeof(int32_t))
-#endif
+#define DK_GROUPS(dk) (((int64_t)1)<<((dk)->dk_log2_size))
+
+#define DK_GROUP_SIZE(dk) \
+ (DK_LOG_SIZE(dk) <= 8 ? sizeof(group8) \
+ : DK_LOG_SIZE(dk) <= 16 ? sizeof(group16) \
+ : DK_LOG_SIZE(dk) <= 32 ? sizeof(group32) \
+ : sizeof(group64))
+
 #define DK_ENTRIES(dk) \
- ((PyDictKeyEntry*)(&((int8_t*)((dk)->dk_indices))[DK_SIZE(dk) * DK_IXSIZE(dk)]))
+ ((PyDictKeyEntry*)(&((dk)->dk_groups)[DK_GROUP_SIZE(dk) << ((dk)->dk_log2_size)]))
 
 extern uint64_t _pydict_global_version;