If using the POSIX memory allocator (rather than _mm_malloc), the code originally accepted an int to store the size of the allocation, which fails if the allocation size exceeds about 2 GB. (Empirically, this caused problems for me when allocating a $48^3 \times 96$ gauge field.) Changing this to type size_t, as is done in the analogous version using _mm_malloc, allows much larger memory allocations, including my test $48^3$ field.
If using the POSIX memory allocator (rather than
_mm_malloc), the code originally accepted anintto store the size of the allocation, which fails if the allocation size exceeds about 2 GB. (Empirically, this caused problems for me when allocating a $48^3 \times 96$ gauge field.) Changing this to typesize_t, as is done in the analogous version using_mm_malloc, allows much larger memory allocations, including my test $48^3$ field.