Hi Ben, Thanks for your reply. Having 1GB huge-page makes PA to VA conversion somewhat easier than with 2MB huge-pages. The VPP has its own buffer management and cacheing, and the DPDK rte mempool functionality is not used by VPP at all. Is that statement correct? I am developing a (DPDK) PMD with VPP network stack. The driver needs to replenish rx-buffers and tx-buffers need to be returned to the pool/cache. Currently I am using DPDK rte_mempool APIs to allocate and free mbuf. I am wondering why not call vlib_buffer_alloc/free instead of rte_pkmbuf_free/alloc_bulk since mbuf is a vlib_buffer and VPP does all the buffer and cache-management work? Does vlib_buffer_t have any dependency on rte_mbuf contents? I keep hearing that VPP can operate without DPDK. Is there any example code/driver that does not use DPDK rte APIs for tx/rx and for buffer management ? The VPP (default)creates a single buffer pool with contiguous VA address space with buffer-index thats point to a buffer which make buffer recycling much easier for the PMD.
Thank you, - P K Das
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#27221): https://lists.fd.io/g/vpp-dev/message/27221 Mute This Topic: https://lists.fd.io/mt/121458452/21656 Group Owner: [email protected] Unsubscribe: https://lists.fd.io/g/vpp-dev/leave/14379924/21656/631435203/xyzzy [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
