[libc++] Instantiate hash function externally (#127040)

This has multiple benefits:
- There is a single instance of our hash function, reducing object file
size
- The hash implementation isn't instantiated in every TU anymore,
reducing compile times
- Behind an ABI configuration macro it would be possible to salt the
hash

libcxx/include/__configuration/availability.h

@@ -78,6 +78,9 @@
 // in all versions of the library are available.
 #if !_LIBCPP_HAS_VENDOR_AVAILABILITY_ANNOTATIONS
 
+#  define _LIBCPP_INTRODUCED_IN_LLVM_21 1
+#  define _LIBCPP_INTRODUCED_IN_LLVM_21_ATTRIBUTE /* nothing */
+
 #  define _LIBCPP_INTRODUCED_IN_LLVM_20 1
 #  define _LIBCPP_INTRODUCED_IN_LLVM_20_ATTRIBUTE /* nothing */
 
@@ -114,6 +117,11 @@
 
 // clang-format off
 
+// LLVM 21
+// TODO: Fill this in
+#  define _LIBCPP_INTRODUCED_IN_LLVM_21 0
+#  define _LIBCPP_INTRODUCED_IN_LLVM_21_ATTRIBUTE __attribute__((unavailable))
+
 // LLVM 20
 // TODO: Fill this in
 #  define _LIBCPP_INTRODUCED_IN_LLVM_20 0
@@ -359,6 +367,11 @@
 #define _LIBCPP_AVAILABILITY_HAS_FROM_CHARS_FLOATING_POINT _LIBCPP_INTRODUCED_IN_LLVM_20
 #define _LIBCPP_AVAILABILITY_FROM_CHARS_FLOATING_POINT _LIBCPP_INTRODUCED_IN_LLVM_20_ATTRIBUTE
 
+// This controls whether `std::__hash_memory` is available in the dylib, which
+// is used for some `std::hash` specializations.
+#define _LIBCPP_AVAILABILITY_HAS_HASH_MEMORY _LIBCPP_INTRODUCED_IN_LLVM_21
+// No attribute, since we've had hash in the headers before
+
 // Define availability attributes that depend on _LIBCPP_HAS_EXCEPTIONS.
 // Those are defined in terms of the availability attributes above, and
 // should not be vendor-specific.

libcxx/include/__functional/hash.h

@@ -238,6 +238,14 @@ private:
   }
 };
 
+#if _LIBCPP_AVAILABILITY_HAS_HASH_MEMORY
+[[__gnu__::__pure__]] _LIBCPP_EXPORTED_FROM_ABI size_t __hash_memory(_LIBCPP_NOESCAPE const void*, size_t) _NOEXCEPT;
+#else
+_LIBCPP_HIDE_FROM_ABI inline size_t __hash_memory(const void* __ptr, size_t __size) _NOEXCEPT {
+  return __murmur2_or_cityhash<size_t>()(__ptr, __size);
+}
+#endif
+
 template <class _Tp, size_t = sizeof(_Tp) / sizeof(size_t)>
 struct __scalar_hash;
 
@@ -277,7 +285,7 @@ struct __scalar_hash<_Tp, 2> : public __unary_function<_Tp, size_t> {
       } __s;
     } __u;
     __u.__t = __v;
-    return __murmur2_or_cityhash<size_t>()(std::addressof(__u), sizeof(__u));
+    return std::__hash_memory(std::addressof(__u), sizeof(__u));
   }
 };
 
@@ -293,7 +301,7 @@ struct __scalar_hash<_Tp, 3> : public __unary_function<_Tp, size_t> {
       } __s;
     } __u;
     __u.__t = __v;
-    return __murmur2_or_cityhash<size_t>()(std::addressof(__u), sizeof(__u));
+    return std::__hash_memory(std::addressof(__u), sizeof(__u));
   }
 };
 
@@ -310,7 +318,7 @@ struct __scalar_hash<_Tp, 4> : public __unary_function<_Tp, size_t> {
       } __s;
     } __u;
     __u.__t = __v;
-    return __murmur2_or_cityhash<size_t>()(std::addressof(__u), sizeof(__u));
+    return std::__hash_memory(std::addressof(__u), sizeof(__u));
   }
 };
 
@@ -333,7 +341,7 @@ struct _LIBCPP_TEMPLATE_VIS hash<_Tp*> : public __unary_function<_Tp*, size_t> {
       size_t __a;
     } __u;
     __u.__t = __v;
-    return __murmur2_or_cityhash<size_t>()(std::addressof(__u), sizeof(__u));
+    return std::__hash_memory(std::addressof(__u), sizeof(__u));
   }
 };
 

libcxx/include/__string/char_traits.h

@@ -534,7 +534,7 @@ __str_find_last_not_of(const _CharT* __p, _SizeT __sz, _CharT __c, _SizeT __pos)
 template <class _Ptr>
 inline _LIBCPP_HIDE_FROM_ABI size_t __do_string_hash(_Ptr __p, _Ptr __e) {
   typedef typename iterator_traits<_Ptr>::value_type value_type;
-  return __murmur2_or_cityhash<size_t>()(__p, (__e - __p) * sizeof(value_type));
+  return std::__hash_memory(__p, (__e - __p) * sizeof(value_type));
 }
 
 _LIBCPP_END_NAMESPACE_STD

libcxx/lib/abi/CHANGELOG.TXT

@@ -12,6 +12,21 @@ To generate a summary, re-generate the new ABI list using the
 
 New entries should be added directly below the "Version" header.
 
+------------
+Version 21.0
+------------
+
+* [libc++] Instantiate hash function externally
+
+  This has multiple benefits:
+  - There is a single instance of our hash function, reducing object file size
+  - The hash implementation isn't instantiated in every TU anymore, reducing compile times
+  - Behind an ABI configuration macro, it would be possible to salt the hash
+
+  All platforms
+  -------------
+  Symbol added: _ZNSt3__113__hash_memoryEPKvm
+
 ------------
 Version 20.0
 ------------

libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist

@@ -1305,6 +1305,7 @@
 {'is_defined': True, 'name': '__ZNSt3__113shared_futureIvED1Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNSt3__113shared_futureIvED2Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNSt3__113shared_futureIvEaSERKS1_', 'type': 'FUNC'}
+{'is_defined': True, 'name': '__ZNSt3__113__hash_memoryEPKvm', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNSt3__114__num_get_base10__get_baseERNS_8ios_baseE', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNSt3__114__num_get_base5__srcE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZNSt3__114__num_put_base12__format_intEPcPKcbj', 'type': 'FUNC'}

libcxx/lib/abi/i686-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist

@@ -941,6 +941,7 @@
 {'is_defined': True, 'name': '_ZNSt6__ndk113shared_futureIvED1Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt6__ndk113shared_futureIvED2Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt6__ndk113shared_futureIvEaSERKS1_', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNSt6__ndk113__hash_memoryEPKvj', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt6__ndk114__num_get_base10__get_baseERNS_8ios_baseE', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt6__ndk114__num_get_base5__srcE', 'size': 33, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZNSt6__ndk114__num_put_base12__format_intEPcPKcbj', 'type': 'FUNC'}

libcxx/lib/abi/powerpc-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist

@@ -418,6 +418,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED1Ev', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED2Ev', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__113shared_futureIvEaSERKS1_', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__113__hash_memoryEPKvm', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__114__num_get_base10__get_baseERNS_8ios_baseE', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__114__num_get_base5__srcE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__114__num_put_base12__format_intEPcPKcbj', 'storage_mapping_class': 'DS', 'type': 'FUNC'}

libcxx/lib/abi/powerpc64-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist

@@ -418,6 +418,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED1Ev', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED2Ev', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__113shared_futureIvEaSERKS1_', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__113__hash_memoryEPKvm', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__114__num_get_base10__get_baseERNS_8ios_baseE', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__114__num_get_base5__srcE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNSt3__114__num_put_base12__format_intEPcPKcbj', 'storage_mapping_class': 'DS', 'type': 'FUNC'}

libcxx/lib/abi/x86_64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist

@@ -1305,6 +1305,7 @@
 {'is_defined': True, 'name': '__ZNSt3__113shared_futureIvED1Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNSt3__113shared_futureIvED2Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNSt3__113shared_futureIvEaSERKS1_', 'type': 'FUNC'}
+{'is_defined': True, 'name': '__ZNSt3__113__hash_memoryEPKvm', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNSt3__114__num_get_base10__get_baseERNS_8ios_baseE', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNSt3__114__num_get_base5__srcE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZNSt3__114__num_put_base12__format_intEPcPKcbj', 'type': 'FUNC'}

libcxx/lib/abi/x86_64-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist

@@ -941,6 +941,7 @@
 {'is_defined': True, 'name': '_ZNSt6__ndk113shared_futureIvED1Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt6__ndk113shared_futureIvED2Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt6__ndk113shared_futureIvEaSERKS1_', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNSt6__ndk113__hash_memoryEPKvm', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt6__ndk114__num_get_base10__get_baseERNS_8ios_baseE', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt6__ndk114__num_get_base5__srcE', 'size': 33, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZNSt6__ndk114__num_put_base12__format_intEPcPKcbj', 'type': 'FUNC'}

libcxx/lib/abi/x86_64-unknown-freebsd.libcxxabi.v1.stable.exceptions.nonew.abilist

@@ -956,6 +956,7 @@
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED1Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED2Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvEaSERKS1_', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNSt3__113__hash_memoryEPKvm', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__114__num_get_base10__get_baseERNS_8ios_baseE', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__114__num_get_base5__srcE', 'size': 33, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZNSt3__114__num_put_base12__format_intEPcPKcbj', 'type': 'FUNC'}

libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.exceptions.nonew.abilist

@@ -954,6 +954,7 @@
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED1Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED2Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvEaSERKS1_', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNSt3__113__hash_memoryEPKvm', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__114__num_get_base10__get_baseERNS_8ios_baseE', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__114__num_get_base5__srcE', 'size': 33, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZNSt3__114__num_put_base12__format_intEPcPKcbj', 'type': 'FUNC'}

libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.noexceptions.nonew.abilist

@@ -924,6 +924,7 @@
 {'is_defined': True, 'name': '_ZNSt3__113random_deviceclEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED1Ev', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvED2Ev', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNSt3__113__hash_memoryEPKvm', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__113shared_futureIvEaSERKS1_', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__114__num_get_base10__get_baseERNS_8ios_baseE', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNSt3__114__num_get_base5__srcE', 'size': 33, 'type': 'OBJECT'}

libcxx/src/functional.cpp

@@ -16,4 +16,8 @@ bad_function_call::~bad_function_call() noexcept {}
 const char* bad_function_call::what() const noexcept { return "std::bad_function_call"; }
 #endif
 
+size_t __hash_memory(_LIBCPP_NOESCAPE const void* ptr, size_t size) noexcept {
+  return __murmur2_or_cityhash<size_t>()(ptr, size);
+}
+
 _LIBCPP_END_NAMESPACE_STD