Provide EIGEN_ALIGNOF macro, and give handmade_aligned_malloc the possibility for alignments larger than the standard alignment.

2 files changed, 9 insertions, 4 deletions

diff --git a/Eigen/src/Core/util/ConfigureVectorization.h b/Eigen/src/Core/util/ConfigureVectorization.h
index f30503b33..e75c7d89e 100644
--- a/Eigen/src/Core/util/ConfigureVectorization.h
+++ b/Eigen/src/Core/util/ConfigureVectorization.h
@@ -39,15 +39,19 @@
  */
 #if (defined EIGEN_CUDACC)
   #define EIGEN_ALIGN_TO_BOUNDARY(n) __align__(n)
+  #define EIGEN_ALIGNOF(x) __alignof(x)
 #elif EIGEN_COMP_GNUC || EIGEN_COMP_PGI || EIGEN_COMP_IBM || EIGEN_COMP_ARM
   #define EIGEN_ALIGN_TO_BOUNDARY(n) __attribute__((aligned(n)))
+  #define EIGEN_ALIGNOF(x) __alignof(x)
 #elif EIGEN_COMP_MSVC
   #define EIGEN_ALIGN_TO_BOUNDARY(n) __declspec(align(n))
+  #define EIGEN_ALIGNOF(x) __alignof(x)
 #elif EIGEN_COMP_SUNCC
   // FIXME not sure about this one:
   #define EIGEN_ALIGN_TO_BOUNDARY(n) __attribute__((aligned(n)))
+  #define EIGEN_ALIGNOF(x) __alignof(x)
 #else
-  #error Please tell me what is the equivalent of __attribute__((aligned(n))) for your compiler
+  #error Please tell me what is the equivalent of __attribute__((aligned(n))) and __alignof(x) for your compiler
 #endif
 
 // If the user explicitly disable vectorization, then we also disable alignment
diff --git a/Eigen/src/Core/util/Memory.h b/Eigen/src/Core/util/Memory.h
index 056070159..76472703e 100644
--- a/Eigen/src/Core/util/Memory.h
+++ b/Eigen/src/Core/util/Memory.h
@@ -96,11 +96,12 @@ inline void throw_std_bad_alloc()
 /** \internal Like malloc, but the returned pointer is guaranteed to be 16-byte aligned.
   * Fast, but wastes 16 additional bytes of memory. Does not throw any exception.
   */
-inline void* handmade_aligned_malloc(std::size_t size)
+inline void* handmade_aligned_malloc(std::size_t size, std::size_t alignment = EIGEN_DEFAULT_ALIGN_BYTES)
 {
-  void *original = std::malloc(size+EIGEN_DEFAULT_ALIGN_BYTES);
+  eigen_assert(alignment >= sizeof(void*) && (alignment & -alignment) == alignment && "Alignment must be at least sizeof(void*) and a power of 2");
+  void *original = std::malloc(size+alignment);
   if (original == 0) return 0;
-  void *aligned = reinterpret_cast<void*>((reinterpret_cast<std::size_t>(original) & ~(std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1))) + EIGEN_DEFAULT_ALIGN_BYTES);
+  void *aligned = reinterpret_cast<void*>((reinterpret_cast<std::size_t>(original) & ~(std::size_t(alignment-1))) + alignment);
   *(reinterpret_cast<void**>(aligned) - 1) = original;
   return aligned;
 }