Memory mapped IO and files

The lower-level parts of computers General: Computer power consumption · Computer noises Memory: Some understanding of memory hardware · CPU cache · Flash memory · Virtual memory · Memory mapped IO and files · RAM disk · Memory limits on 32-bit and 64-bit machines Related: Network wiring notes - Power over Ethernet · 19" rack sizes Unsorted: Interrupts · Computer booting · GPU, GPGPU, OpenCL, CUDA notes

✎ This article/section is a stub — some half-sorted notes, not necessarily checked, not necessarily correct. Feel free to ignore, or tell me about it.

Note that

memory mapped IO is a hardware-level construction, while

memory mapped files are a software construction -- because files are.

Memory mapped files

Memory mapping of files is a technique (OS feature, system call) that pretends a file is accessible at some address in memory.

When the process accesses those memory locations, the OS will scramble for the actual contents from disk.

...it can also be cached -- whether the OS does that at all depends a little on the OS and details(verify), whether it is in RAM depends on whether it was recently accessed by something.

For caching

In e.g. linux you get that interaction with the page cache, and the data stays cached as long as the RAM isn't used by other things.

This can also save memory - in that without memory mapping, compared to the easy choice of manually cacheing the entire thing in your process.

With mmap you may cache only the parts you use, and if multiple processes want this file, you may avoid a little duplication.

The fact that the OS can flush most or all of this data can be seen as a limitation or a feature - it's not always predictable, but it does mean you can deal with large data sets without having to think about very large allocations, and how those aren't nice to other apps.

shared memory via memory mapped files

Most kernel implementations allow multiple processes to mmap the same file -- which effectively shares memory, and probably one of the simplest in a protected mode system. (Some methods of Inter-Process communication work via mmapping)

Not clobbering each other's memory is still something you need to do yourself.

The implementation, limitations, and method of use varies per OS / kernel.

Often relies on demand paging to work.

Memory mapped IO

Map devices into memory space (statically or dynamically), meaning that memory accesses to those areas will actually be served via IO accesses (...that you can typically also do directly).

This mapping is made and resolved at hardware-level, and only works for DMA-capable devices (which is many).

It seems to often be done to have a simple generic interface (verify) - it means drivers and software can avoid many hardware-specific details.

See also:

• http://en.wikipedia.org/wiki/Memory-mapped_I/O

DMA

✎ This article/section is a stub — some half-sorted notes, not necessarily checked, not necessarily correct. Feel free to ignore, or tell me about it.

In contrast with occupying the CPU with copy instructions on CPU (called PIO when it's done to do IO via MMIO or PMIO), DMA refers to code deciding to delegate the copy to a peripheral (which can only do copying).

DMA implies that that there is at least one device beyond the CPU that can also take can take control of the bus.

This way, the CPU can do other things, so from a time perspective DMA transfers are often nearly free.

DMA is also independent enough at hardware level that its transfers can often work at high clock rates, meaning it might also be faster than CPU-based copies, though it takes a little setup so only really makes sense when there is a range to be copied.

Broadly, there is either:

• First party DMA, a.k.a. bus mastering

means any device on the bus could potentially initiate a DMA transaction.

• Third party DMA

means there is a single system DMA controller actually does the transfer

Aside from memory-to-memory uses, DMA also allows memory-to-peripheral copies (if a specific supporting device is memory mapped(verify)).

While first party DMA could theoretically have any device directly talk to any other, it is much more generic (and thereby more compatible) to have the OS drivers deal with just one device, meaning that in practice they'll typically use OS-managed RAM as an intermediate.

Some terms:

These terms are coined to discuss use of a single shared system bus.

The details look intimidating at first, but mostly because they are low-level.

The 'you say copy, it does copy' idea is actually relatively simple.

And yes, around DMA you do start having to worry about avoiding race conditions, though there are some standard-ish tricks.