[libcxx-commits] [libcxx] 20f39bf - Lower std::string's alignment requirement from 16 to 8. (#68807)

[via libcxx-commits]

Author: Eric
Date: 2023-10-13T11:56:24-04:00
New Revision: 20f39bf48218515e05126d02f26cec73ac655b0f

URL: https://github.com/llvm/llvm-project/commit/20f39bf48218515e05126d02f26cec73ac655b0f
DIFF: https://github.com/llvm/llvm-project/commit/20f39bf48218515e05126d02f26cec73ac655b0f.diff

LOG: Lower std::string's alignment requirement from 16 to 8. (#68807)

This allows smaller allocations to occur, closer to the actual
std::string's required size. This is particularly effective in
decreasing the allocation size upon initial construction (where
__recommend is called to determine the size).

Although the memory savings per-string are never more than 8 bytes per
string initially, this quickly adds up. And has lead to not insigficant
memory savings at Google.

Unfortunately, this change is ABI breaking because it changes the value
returned by max_size. So it has to be guarded.

Added: 
    libcxx/test/libcxx/strings/basic.string/string.capacity/allocation_size.pass.cpp

Modified: 
    libcxx/docs/ReleaseNotes/18.rst
    libcxx/include/__config
    libcxx/include/string
    libcxx/test/libcxx/strings/basic.string/string.capacity/max_size.pass.cpp

Removed: 
    


################################################################################
diff  --git a/libcxx/docs/ReleaseNotes/18.rst b/libcxx/docs/ReleaseNotes/18.rst
index 5f43d2f2afe22d3..ac78563aa73848f 100644
--- a/libcxx/docs/ReleaseNotes/18.rst
+++ b/libcxx/docs/ReleaseNotes/18.rst
@@ -133,6 +133,13 @@ ABI Affecting Changes
   results in an ABI break, however in practice we expect uses of ``std::projected`` in ABI-sensitive places to be
   extremely rare. Any error resulting from this change should result in a link-time error.
 
+- Under the unstable ABI, the internal alignment requirements for heap allocations
+  inside ``std::string`` has decreased from 16 to 8 This save memory since string requests fewer additional
+  bytes than it did previously. However, this also changes the return value of ``std::string::max_size``
+  and can cause code compiled against older libc++ versions but linked at runtime to a new version
+  to throw a 
diff erent exception when attempting allocations that are too large
+  (``std::bad_alloc`` vs ``std::length_error``).
+
 Build System Changes
 --------------------
 

diff  --git a/libcxx/include/__config b/libcxx/include/__config
index 55d9f1c737652e6..65ce6d6a27f8326 100644
--- a/libcxx/include/__config
+++ b/libcxx/include/__config
@@ -167,6 +167,11 @@
 // The implementation moved to the header, but we still export the symbols from
 // the dylib for backwards compatibility.
 #    define _LIBCPP_ABI_DO_NOT_EXPORT_TO_CHARS_BASE_10
+// Save memory by providing the allocator more freedom to allocate the most
+// efficient size class by dropping the alignment requirements for std::string's
+// pointer from 16 to 8. This changes the output of std::string::max_size,
+// which makes it ABI breaking
+#    define _LIBCPP_ABI_STRING_8_BYTE_ALIGNMENT
 #  elif _LIBCPP_ABI_VERSION == 1
 #    if !(defined(_LIBCPP_OBJECT_FORMAT_COFF) || defined(_LIBCPP_OBJECT_FORMAT_XCOFF))
 // Enable compiling copies of now inline methods into the dylib to support

diff  --git a/libcxx/include/string b/libcxx/include/string
index 33e87406a1156a6..3078715e02b358a 100644
--- a/libcxx/include/string
+++ b/libcxx/include/string
@@ -1851,7 +1851,14 @@ private:
         _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20
         size_type __align_it(size_type __s) _NOEXCEPT
             {return (__s + (__a-1)) & ~(__a-1);}
-    enum {__alignment = 16};
+    enum {
+      __alignment =
+#ifdef _LIBCPP_ABI_STRING_8_BYTE_ALIGNMENT
+      8
+#else
+      16
+#endif
+    };
     static _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20
     size_type __recommend(size_type __s) _NOEXCEPT
     {

diff  --git a/libcxx/test/libcxx/strings/basic.string/string.capacity/allocation_size.pass.cpp b/libcxx/test/libcxx/strings/basic.string/string.capacity/allocation_size.pass.cpp
new file mode 100644
index 000000000000000..c7df56c815a8054
--- /dev/null
+++ b/libcxx/test/libcxx/strings/basic.string/string.capacity/allocation_size.pass.cpp
@@ -0,0 +1,45 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// <string>
+
+// This test demonstrates the smaller allocation sizes when the alignment
+// requirements of std::string are dropped from 16 to 8.
+#include <algorithm>
+#include <cassert>
+#include <cstddef>
+#include <string>
+
+#include "test_macros.h"
+
+// alignment of the string heap buffer is hardcoded to either 16 or 8
+
+const std::size_t alignment =
+#ifdef _LIBCPP_ABI_STRING_8_BYTE_ALIGNMENT
+    8;
+#else
+    16;
+#endif
+
+int main(int, char**) {
+  std::string input_string;
+  input_string.resize(64, 'a');
+
+  // Call a constructor which selects its size using __recommend.
+  std::string test_string(input_string.data());
+  const std::size_t expected_align8_size = 71;
+
+  // Demonstrate the lesser capacity/allocation size when the alignment requirement is 8.
+  if (alignment == 8) {
+    assert(test_string.capacity() == expected_align8_size);
+  } else {
+    assert(test_string.capacity() == expected_align8_size + 8);
+  }
+
+  return 0;
+}

diff  --git a/libcxx/test/libcxx/strings/basic.string/string.capacity/max_size.pass.cpp b/libcxx/test/libcxx/strings/basic.string/string.capacity/max_size.pass.cpp
index 5af9cab0be4e80a..a3cb79522f2e1eb 100644
--- a/libcxx/test/libcxx/strings/basic.string/string.capacity/max_size.pass.cpp
+++ b/libcxx/test/libcxx/strings/basic.string/string.capacity/max_size.pass.cpp
@@ -18,7 +18,13 @@
 #include "test_macros.h"
 
 // alignment of the string heap buffer is hardcoded to 16
-static const std::size_t alignment = 16;
+
+static const std::size_t alignment =
+#ifdef _LIBCPP_ABI_STRING_8_BYTE_ALIGNMENT
+    8;
+#else
+    16;
+#endif
 
 template <class = int>
 TEST_CONSTEXPR_CXX20 void full_size() {