+/// Wrappers for systems where the c++17 implementation doesn't guarantee the availability of aligned_alloc.
+/// Memory allocated with std_aligned_alloc must be freed with std_aligned_free.
+///
+
+void* std_aligned_alloc(size_t alignment, size_t size) {
+#if defined(__APPLE__)
+ return aligned_alloc(alignment, size);
+#elif defined(_WIN32)
+ return _mm_malloc(size, alignment);
+#else
+ return std::aligned_alloc(alignment, size);
+#endif
+}
+
+void std_aligned_free(void* ptr) {
+#if defined(__APPLE__)
+ free(ptr);
+#elif defined(_WIN32)
+ _mm_free(ptr);
+#else
+ free(ptr);
+#endif
+}
+
+/// aligned_ttmem_alloc() will return suitably aligned memory, and if possible use large pages.
+/// The returned pointer is the aligned one, while the mem argument is the one that needs
+/// to be passed to free. With c++17 some of this functionality could be simplified.
+
+#if defined(__linux__) && !defined(__ANDROID__)
+
+void* aligned_ttmem_alloc(size_t allocSize, void*& mem) {
+
+ constexpr size_t alignment = 2 * 1024 * 1024; // assumed 2MB page sizes
+ size_t size = ((allocSize + alignment - 1) / alignment) * alignment; // multiple of alignment
+ if (posix_memalign(&mem, alignment, size))
+ mem = nullptr;
+ madvise(mem, allocSize, MADV_HUGEPAGE);
+ return mem;
+}
+
+#elif defined(_WIN64)
+
+static void* aligned_ttmem_alloc_large_pages(size_t allocSize) {
+
+ HANDLE hProcessToken { };
+ LUID luid { };
+ void* mem = nullptr;
+
+ const size_t largePageSize = GetLargePageMinimum();
+ if (!largePageSize)
+ return nullptr;
+
+ // We need SeLockMemoryPrivilege, so try to enable it for the process
+ if (!OpenProcessToken(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &hProcessToken))
+ return nullptr;
+
+ if (LookupPrivilegeValue(NULL, SE_LOCK_MEMORY_NAME, &luid))
+ {
+ TOKEN_PRIVILEGES tp { };
+ TOKEN_PRIVILEGES prevTp { };
+ DWORD prevTpLen = 0;
+
+ tp.PrivilegeCount = 1;
+ tp.Privileges[0].Luid = luid;
+ tp.Privileges[0].Attributes = SE_PRIVILEGE_ENABLED;
+
+ // Try to enable SeLockMemoryPrivilege. Note that even if AdjustTokenPrivileges() succeeds,
+ // we still need to query GetLastError() to ensure that the privileges were actually obtained.
+ if (AdjustTokenPrivileges(
+ hProcessToken, FALSE, &tp, sizeof(TOKEN_PRIVILEGES), &prevTp, &prevTpLen) &&
+ GetLastError() == ERROR_SUCCESS)
+ {
+ // Round up size to full pages and allocate
+ allocSize = (allocSize + largePageSize - 1) & ~size_t(largePageSize - 1);
+ mem = VirtualAlloc(
+ NULL, allocSize, MEM_RESERVE | MEM_COMMIT | MEM_LARGE_PAGES, PAGE_READWRITE);
+
+ // Privilege no longer needed, restore previous state
+ AdjustTokenPrivileges(hProcessToken, FALSE, &prevTp, 0, NULL, NULL);
+ }
+ }
+
+ CloseHandle(hProcessToken);