https://devzone.nordicsemi.com/f/nordic-q-a/83501/copy-instruction-cyclesI tested various copy methods in terms of instruction cycles, that I determined with DWT. For that I copied an uint8_t array size 70 by for Loop --&gt; 63 Cycles memcpy --&gt; 514 Cycles std::array copy --&gt; 514 Cycles Are theses result valid?en-USTelligent Community 13Mon, 10 Jan 2022 13:41:43 GMThttps://devzone.nordicsemi.com/thread/346959?ContentTypeID=1Mon, 10 Jan 2022 13:41:43 GMT137ad170-7792-4731-bb38-c0d22fbe4515:5444eb73-244e-4884-9f05-abbff9717440stackGreen<p>so as far as I understood its about the context and optimization. &nbsp;If I disable optimization with -O0, the &ldquo;copy by for loop&rdquo; cycles increased to 2129 (memcpy 828)</p><div style="clear:both;"></div>https://devzone.nordicsemi.com/thread/346869?ContentTypeID=1Mon, 10 Jan 2022 09:51:47 GMT137ad170-7792-4731-bb38-c0d22fbe4515:fc4d7f42-16e7-4ea3-a14f-27e08bd16483Vidar Berg<p>Hello,</p> <p>I would also have expected memcpy to be faster than a for loop, and I certainly did not expect it to be that much slower. Have you tried to optimize for speed as discussed in this blog post by memfault: <a href="https://interrupt.memfault.com/blog/memcpy-newlib-nano">https://interrupt.memfault.com/blog/memcpy-newlib-nano</a> ?</p><div style="clear:both;"></div>