Is there a non-blocking memcpy i.e. "software DMA"?

[histrungalot]

I want to know if there is a non-blocking memcpy? I have a program that delivers 2MB of data into a chunk of memory every ~7 ms. I need to copy that data out into a larger circular buffer. If I could not call memcpy and use a non-blocking "software DMA", that would greatly increase my though-put.

[kdg1955]

As far as I know, DMA can only be used with I/O (not sure, sorry if I'm wrong).

You may try with a file-socket as workaround (fork a new process and write/read to/from that socket in O_NONBLOKING mode). I'm not sure it meet your (performance) needs and there is a lot of development.

[Rubberman]

What do mean specifically by blocking/non-blocking? Those terms have a lot of nuances, depending upon the situation and environment. I think I understand your main problem - to move large chunks of data from one location in memory to another quickly and without interfering with system operation. So, do you mean non-blocking of other processes attempting to access memory, other threads in the same process that may want to access the same buffers, or something else?