Performance problem with memmove

Hi all,

while I was profiling and optimizing my DRM transmitter flow graph, I
noticed that __memmove_sse3_back.S is the dominating part in my program,
although I do not explicitly call memmove() once in my own code. I am
using a vector based approach which leads to quite large buffers (and
lots of gr-buffer warnings).

Does memmove get called in the buffer context and is it possible that
this in combination with large buffers leads to a high CPU load?

Best regards,
Felix W.