AMD did not have useful implementation of PDEP/PEXT instructions until Zen 3, which is barely more than 3 years old. All prior AMD implementations were microcoded for compatibility purposes only. If you cared about performance it was faster to do it in software for most applications. The Zen 2 implementation was something like 6-7x slower than the original Haswell implementation.

For all practical purposes, you had to write performance code like AMD did not support PDEP/PEXT.