> -----Original Message----- > From: Trahe, Fiona > Sent: Tuesday, May 9, 2017 5:14 PM > To: dev@dpdk.org; Mrozowicz, SlawomirX <slawomirx.mrozow...@intel.com> > Cc: Doherty, Declan <declan.dohe...@intel.com>; Griffin, John > <john.grif...@intel.com>; De Lara Guarch, > Pablo <pablo.de.lara.gua...@intel.com>; Trahe, Fiona <fiona.tr...@intel.com> > Subject: [PATCH] app/crypto-perf: add minimise-offload-cost flag > > The throughput test enqueues and dequeues bursts of operations > to the device. For software devices the full burst size will > usually be successfully en/dequeued, on hardware devices > however the CPU can call the API more frequently than necessary, > as it has nothing else to do. > Minimum offload cost is achieved when the specified > burst_size is en/dequeued. So rather than > wasting CPU cycles continually retrying, with a > fraction of the burst being en/dequeued each time, > fewer CPU cycles are used by backing off until a full > burst can be enqueued. > > This patch adds a --minimise-offload-cost flag. > When set the test backs off until full bursts are > en/dequeued and counts the cycles while waiting. > These cycles represent cycles saved by > offloading, which in a real application are > available for other work. Hence these cycles are > deducted from the total cycle-count to show the > minimum offload-cost which can be acheived. > > Signed-off-by: Fiona Trahe <fiona.tr...@intel.com> Self-nack As we're intending to push a better way to measure HW offload cost in next release.
