mempool cache zero-copy get/put bulk ------------------------------------ Zero-copy access to mempool caches is beneficial for PMD performance, and must be provided by the mempool library to fix [Bug 1052] without a performance regression.
[Bug 1052]: https://bugs.dpdk.org/show_bug.cgi?id=1052 Refer to patch v2 and RFC discussions: http://inbox.dpdk.org/dev/20221116180419.98937-1...@smartsharesystems.com/ http://inbox.dpdk.org/dev/98cbd80474fa8b44bf855df32c47dc35d87...@smartserver.smartshare.dk/ Non-temporal memory copy ------------------------ The primary purpose of this feature is to avoid polluting the CPU's data cache when copying packets using the CPU, e.g. when capturing packets. E.g. copying a 32 packet burst of 1514 B packets reads 48 KB and writes 48 KB, and thus pollutes the entire L1 cache of a typical x86 CPU core with 64 KB data cache. The API for this feature is generic, but we will initially only implement the feature for x86 architecture. This implementation can be used as inspiration for implementations for other architectures. Refer to patch v4 and RFC discussions: http://inbox.dpdk.org/dev/20221010064600.16495-1...@smartsharesystems.com/ http://inbox.dpdk.org/dev/98cbd80474fa8b44bf855df32c47dc35d87...@smartserver.smartshare.dk/ http://inbox.dpdk.org/dev/98cbd80474fa8b44bf855df32c47dc35d87...@smartserver.smartshare.dk/ Med venlig hilsen / Kind regards, -Morten Brørup