A fast AVX memcpy macro which copies the content of a 64 byte source buffer into a 64 byte destination buffer. Buffers must be 32byte aligned.

fast_copy.hpp

#pragma once

#include <immintrin.h>

#define I32B __m256i
#define STORE_I32B(ptrD, ptrS) _mm256_store_si256(ptrD, *ptrS)

/**
 * Copies the content of one 32 byte aligned 
 * 64 byte buffer into another
 *
 * @param dest
 *		A pointer to the destination buffer
 *
 * @param src
 *		A pointer to the source buffer
 */
#define A_memcpy_64(dest,src)	{															\
					I32B *A_memcpy_64_bufferS = (I32B*) (src);		\
					I32B *A_memcpy_64_bufferD = (I32B*) (dest);		\
					STORE_I32B(A_memcpy_64_bufferD, A_memcpy_64_bufferS);   \
					A_memcpy_64_bufferS++;					\
					A_memcpy_64_bufferD++;					\
					STORE_I32B(A_memcpy_64_bufferD, A_memcpy_64_bufferS);	\
				}