Add aarch64 defines

With atomic operations and system call support in place, enable
with __aarch64__ defines Aarch64 support in other files around
the google-perftools header files. After these, google-perftools
testsuite (make check) results:

8 of 46 tests failed.

FAIL: sampling_test.sh
FAIL: heap-profiler_unittest.sh
FAIL: heap-checker_unittest.sh
FAIL: heap-checker-death_unittest.sh
FAIL: sampling_debug_test.sh
FAIL: heap-profiler_debug_unittest.sh
FAIL: heap-checker_debug_unittest.sh
FAIL: profiler_unittest.sh

While it indicates that there is still work to do, This is still
better than the result I get on ARMv7:

12 of 46 tests failed.

4 files changed, 8 insertions, 5 deletions

diff --git a/src/base/basictypes.h b/src/base/basictypes.h
index 677c3cd..2862c11 100644
--- a/src/base/basictypes.h
+++ b/src/base/basictypes.h
@@ -348,6 +348,9 @@ class AssignAttributeStartEnd {
     // some ARMs have shorter cache lines (ARM1176JZF-S is 32 bytes for example) but obviously 64-byte aligned implies 32-byte aligned
 # elif (defined(__mips__))
 #   define CACHELINE_ALIGNED __attribute__((aligned(128)))
+# elif (defined(__aarch64__))
+#   define CACHELINE_ALIGNED __attribute__((aligned(64)))
+    // implementation specific, Cortex-A53 and 57 should have 64 bytes
 # else
 #   error Could not determine cache line length - unknown architecture
 # endif
diff --git a/src/base/cycleclock.h b/src/base/cycleclock.h
index d6a6662..dc2d569 100644
--- a/src/base/cycleclock.h
+++ b/src/base/cycleclock.h
@@ -50,7 +50,7 @@
 #include "base/arm_instruction_set_select.h"
 // base/sysinfo.h is really big and we don't want to include it unless
 // it is necessary.
-#if defined(__arm__) || defined(__mips__)
+#if defined(__arm__) || defined(__mips__) || defined(__aarch64__)
 # include "base/sysinfo.h"
 #endif
 #if defined(__MACH__) && defined(__APPLE__)
@@ -67,7 +67,7 @@
 extern "C" uint64 __rdtsc();
 #pragma intrinsic(__rdtsc)
 #endif
-#if defined(ARMV3) || defined(__mips__)
+#if defined(ARMV3) || defined(__mips__) || defined(__aarch64__)
 #include <sys/time.h>
 #endif
 
@@ -133,7 +133,7 @@ struct CycleClock {
     _asm rdtsc
 #elif defined(_MSC_VER)
     return __rdtsc();
-#elif defined(ARMV3)
+#elif defined(ARMV3) || defined(__aarch64__)
 #if defined(ARMV7)  // V7 is the earliest arch that has a standard cyclecount
     uint32 pmccntr;
     uint32 pmuseren;
diff --git a/src/base/linuxthreads.h b/src/base/linuxthreads.h
index 5c318fe..16bc8c6 100644
--- a/src/base/linuxthreads.h
+++ b/src/base/linuxthreads.h
@@ -41,7 +41,7 @@
  * related platforms should not be difficult.
  */
 #if (defined(__i386__) || defined(__x86_64__) || defined(__ARM_ARCH_3__) || \
-     defined(__mips__) || defined(__PPC__)) && defined(__linux)
+     defined(__mips__) || defined(__PPC__) || defined(__aarch64__)) && defined(__linux)
 
 /* Define the THREADS symbol to make sure that there is exactly one core dumper
  * built into the library.
diff --git a/src/malloc_hook_mmap_linux.h b/src/malloc_hook_mmap_linux.h
index 54e5d87..8e5a3b0 100755
--- a/src/malloc_hook_mmap_linux.h
+++ b/src/malloc_hook_mmap_linux.h
@@ -52,7 +52,7 @@
 
 // I test for 64-bit first so I don't have to do things like
 // '#if (defined(__mips__) && !defined(__MIPS64__))' as a mips32 check.
-#if defined(__x86_64__) || defined(__PPC64__) || (defined(_MIPS_SIM) && _MIPS_SIM == _ABI64)
+#if defined(__x86_64__) || defined(__PPC64__) || defined(__aarch64__) || (defined(_MIPS_SIM) && _MIPS_SIM == _ABI64)
 
 static inline void* do_mmap64(void *start, size_t length,
                               int prot, int flags,