align fast-path functions only if compiler supports that

Apparently gcc only supports __attribute__((aligned(N))) on functions
only since version 4.3. So lets test it in configure script and only
use when possible. We now use CACHELINE_ALIGNED_FN macro for aligning
functions.

3 files changed, 29 insertions, 7 deletions

diff --git a/configure.ac b/configure.ac
index e0599ff..59b2467 100644
--- a/configure.ac
+++ b/configure.ac
@@ -165,6 +165,22 @@ AC_PROG_LIBTOOL
 AC_C_INLINE
 AX_C___ATTRIBUTE__
 
+AC_MSG_CHECKING(for __attribute__((aligned(N))) on functions)
+AC_CACHE_VAL(ac_cv___attribute__aligned_fn, [
+  AC_TRY_COMPILE(
+    [#include <stdlib.h>
+     void foo(void) __attribute__((aligned(128)));
+     void foo(void) { exit(1); }],
+    [],
+    ac_cv___attribute__aligned_fn=yes,
+    ac_cv___attribute__aligned_fn=no
+  )])
+if test "$ac_cv___attribute__aligned_fn" = "yes"; then
+  AC_DEFINE(HAVE___ATTRIBUTE__ALIGNED_FN, 1, [define if your compiler supports alignment of functions])
+fi
+AC_MSG_RESULT($ac_cv___attribute__aligned_fn)
+
+
 # Check whether some low-level functions/files are available
 AC_HEADER_STDC
 
diff --git a/src/base/basictypes.h b/src/base/basictypes.h
index 1a7b73a..42dbe5c 100644
--- a/src/base/basictypes.h
+++ b/src/base/basictypes.h
@@ -388,7 +388,13 @@ class AssignAttributeStartEnd {
 # endif
 #else
 # define CACHELINE_ALIGNED
-#endif  // defined(HAVE___ATTRIBUTE__) && (__i386__ || __x86_64__)
+#endif  // defined(HAVE___ATTRIBUTE__)
+
+#if defined(HAVE___ATTRIBUTE__ALIGNED_FN)
+#  define CACHELINE_ALIGNED_FN CACHELINE_ALIGNED
+#else
+#  define CACHELINE_ALIGNED_FN
+#endif
 
 // Structure for discovering alignment
 union MemoryAligner {
diff --git a/src/tcmalloc.cc b/src/tcmalloc.cc
index fde8a8a..c6edefd 100644
--- a/src/tcmalloc.cc
+++ b/src/tcmalloc.cc
@@ -1394,7 +1394,7 @@ void do_free_with_callback(void* ptr,
     // cache could return bogus result (cl = 0 as of this
     // writing). But since there is no way we could be dealing with
     // ptr we've allocated, since successfull malloc implies IsInited,
-    // we can just call invalid_free_fn here.
+    // we can just call "invalid free" handling code.
     free_null_or_invalid(ptr, invalid_free_fn);
     return;
   }
@@ -1749,12 +1749,12 @@ static void * malloc_fast_path(size_t size) {
   return CheckedMallocResult(cache->Allocate(allocated_size, cl));
 }
 
-extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED
+extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED_FN
 void* tc_malloc(size_t size) PERFTOOLS_THROW {
   return malloc_fast_path<tcmalloc::allocate_full_malloc_oom>(size);
 }
 
-extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED
+extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED_FN
 void tc_free(void* ptr) PERFTOOLS_THROW {
   if (PREDICT_FALSE(!base::internal::delete_hooks_.empty())) {
     tcmalloc::invoke_hooks_and_free(ptr);
@@ -1763,7 +1763,7 @@ void tc_free(void* ptr) PERFTOOLS_THROW {
   do_free(ptr);
 }
 
-extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED
+extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED_FN
 void tc_free_sized(void *ptr, size_t size) PERFTOOLS_THROW {
   if (PREDICT_FALSE(!base::internal::delete_hooks_.empty())) {
     tcmalloc::invoke_hooks_and_free(ptr);
@@ -1844,12 +1844,12 @@ extern "C" PERFTOOLS_DLL_DECL void* tc_realloc(void* old_ptr,
   return do_realloc(old_ptr, new_size);
 }
 
-extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED
+extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED_FN
 void* tc_new(size_t size) {
   return malloc_fast_path<tcmalloc::allocate_full_cpp_throw_oom>(size);
 }
 
-extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED
+extern "C" PERFTOOLS_DLL_DECL CACHELINE_ALIGNED_FN
 void* tc_new_nothrow(size_t size, const std::nothrow_t&) PERFTOOLS_THROW {
   return malloc_fast_path<tcmalloc::allocate_full_cpp_nothrow_oom>(size);
 }