When is posix_memalign/memalign used? [duplicate]

This question already has answers here: Closed 11 years ago.

Possible Dupl开发者_如何学运维icate:

When do we need to use posix_memalign instead of malloc ?

When would you need "size bytes of memory such that the allocation's base address is an exact multiple of alignment" over just a simple malloc()?

---

I believe it is memalign() under Linux and posix_memalign() under Mac OS X/BSD.

---

One of the most common use cases is when working with SIMD code (SSE, AltiVec, AVX, Neon, etc) where you typically need all buffers to be aligned on a SIMD vector boundary (16 byte or 32 byte alignment).

Another common alignment requirement is virtual memory page alignment (4 kB typically).

---

One reason that popped up recently is to ensure that data buffer structs/objects do not share cache lines. This reduces the need for cache flushes.

Rgds, Martin

---

if you need strict memory alignment than what malloc offers.

malloc usually would return a pointer aligned in such away that it may be used with any of the primitive types (usually 8 bytes).

However, sometimes you need memory aligned on other boundaries, In this case, you would need memalign.

---

Some special instructions (eg. SSE instructions) require correct alignment of data in order to work. If you're developing in kernel mode, some DMA controllers also require correct alignment.