Perform scalar simple match lookup in AVX512 DPIF by reusing the simple match lookup functions that are used in the scalar DPIF. This allows feature parity with the scalar DPIF and better lookup performance in the AVX512 DPIF for supported flows over existing lookup methods (6% faster than EMC for a simple Phy-to-Phy when compiling with "-Ofast -march=native". Link to Harry's performance tests on the v1: https://mail.openvswitch.org/pipermail/ovs-dev/2022-June/395374.html
The performance of simple match lookup in the scalar DPIF is not affected as measured in a simple Phy-to-Phy scenario. Cian Ferriter (2): dpif-netdev: Refactor simple match lookup functions. dpif-avx512: Add support for simple match lookup. lib/dpif-netdev-avx512.c | 53 ++++++++++++++++++++++++++++++++-- lib/dpif-netdev-private-dpif.h | 12 ++++++++ lib/dpif-netdev.c | 11 ++----- 3 files changed, 65 insertions(+), 11 deletions(-) -- 2.25.1 _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
