I've implemented optimizations to rte_memcpy targeting RISC-V architectures, 
<br />achieving&nbsp;an&nbsp;average&nbsp;10%~15%&nbsp;reduction&nbsp;in&nbsp;execution&nbsp;time&nbsp;for&nbsp;data&nbsp;sizes&nbsp;between<br />129&nbsp;to&nbsp;1024&nbsp;bytes(&nbsp;1025~1600&nbsp;gains&nbsp;little).<br />These&nbsp;enhancements&nbsp;draw&nbsp;inspiration&nbsp;from&nbsp;x86&nbsp;implementations,<br />&nbsp;specifically&nbsp;focusing&nbsp;on:<br />1)Alignment&nbsp;Handling&nbsp;for&nbsp;Unaligned&nbsp;Scenarios<br />2)Vector&nbsp;Configuration&nbsp;Tuning<br />3)Strategic&nbsp;Prefetching<br />
<br />-&nbsp;Patch&nbsp;1:&nbsp;Cover&nbsp;letter<br />-&nbsp;Patch&nbsp;2:&nbsp;Base&nbsp;implementation<br />-&nbsp;Patch&nbsp;3:&nbsp;Benchmark&nbsp;report<br />
<br />
<br />Tested&nbsp;on&nbsp;Tested&nbsp;on&nbsp;SG2044&nbsp;(VLEN=128)<br />
<br />
<br />Qiguo&nbsp;Chen&nbsp;(2):<br />&nbsp;&nbsp;riscv&nbsp;support&nbsp;rte_memcpy&nbsp;in&nbsp;vector<br />&nbsp;&nbsp;benchmark&nbsp;report&nbsp;for&nbsp;rte_memcpy<br />
<br />&nbsp;.mailmap&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;|&nbsp;&nbsp;&nbsp;1&nbsp;+<br />&nbsp;benchmark_report.txt&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;|&nbsp;149&nbsp;++++++++++++++<br />&nbsp;config/riscv/meson.build&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;|&nbsp;&nbsp;14&nbsp;++<br />&nbsp;lib/eal/riscv/include/rte_memcpy.h&nbsp;|&nbsp;310&nbsp;++++++++++++++++++++++++++++-<br />&nbsp;4&nbsp;files&nbsp;changed,&nbsp;472&nbsp;insertions(+),&nbsp;2&nbsp;deletions(-)<br />&nbsp;create&nbsp;mode&nbsp;100644&nbsp;benchmark_report.txt<br />
<br />--&nbsp;
<br />2.21.0.windows.1<br />