[PATCH v6] eal/x86: optimize memcpy of small sizes

[Varghese, Vipin]

[Public]

Hi Wang,

We use dpdk dma-perf application for testing memcpy via cpu.
Please find the ini configuration shared below

```
[GLOBAL]
eal_args=--in-memory --file-prefix=test -l 34,35
cache_flush=0
test_seconds=10

[case1]
type=CPU_MEM_COPY
mem_size=100
buf_size=1,17,1,ADD
src_numa_node=0
dst_numa_node=0
lcore=35
```

Please adapt eal_args and lcore for machine on which you are running.
Command used is `dpdk-vanilla/build-no512/app/dpdk-test-dma-perf --config ../config.ini --result /tmp/memcpy.log`

After running we rebuild dpdk to use avx512 MEMCPY using `-DRTE_MEMCPY_AVX512` and then rerun the dma-perf again.

Regards
Vipin Varghese