For this series: Reviewed-and-tested-by: Bill Fischofer <bill.fischo...@linaro.org>
On Thu, Nov 10, 2016 at 5:07 AM, Petri Savolainen < petri.savolai...@nokia.com> wrote: > Pool performance is optimized by using a ring as the global buffer storage. > IPC build is disabled, since it needs large modifications due to > dependency to > pool internals. Old pool implementation was based on locks and linked list > of > buffer headers. New implementation maintain a ring of buffer handles, which > enable fast, burst based allocs and frees. Also ring scales better with > number > of cpus than a list (enq and deq operations update opposite ends of the > pool). > > L2fwd link rate (%), 2 x 40GE, 64 byte packets > > direct- parallel- atomic- > cpus orig direct diff orig parall diff orig atomic > diff > 1 7 % 8 % 1 % 6 % 6 % 2 % 5.4 % 5.6 % 4 % > 2 14 % 15 % 7 % 9 % 9 % 5 % 8 % 9 % 8 % > 4 28 % 30 % 6 % 13 % 14 % 13 % 12 % 15 % 19 > % > 6 42 % 44 % 6 % 16 % 19 % 19 % 8 % 20 % > 150 % > 8 46 % 59 % 28 % 19 % 23 % 26 % 18 % 24 % 34 > % > 10 55 % 57 % 3 % 20 % 27 % 37 % 8 % 28 % > 264 % > 12 56 % 56 % -1 % 22 % 31 % 43 % 7 % 32 % > 357 % > > Max packet rate of NICs are reached with 10-12 cpu on direct mode. > Otherwise, > all cases were improved. Especially, scheduler driven cases suffered on bad > pool scalability. > > changed in v3: > * rebased > * ipc disabled with #ifdef > * added support for multi-segment packets > * API: added explicit limits for packet length in alloc calls > * Corrected validation test and example application bugs found during > segmentation implementation > > changed in v2: > * rebased to api-next branch > * added a comment that ring size must be larger than number of items in it > * fixed clang build issue > * added parens in align macro > > v1 reviews: > Reviewed-by: Brian Brooks <brian.bro...@linaro.org> > > > > Petri Savolainen (19): > linux-gen: ipc: disable build of ipc pktio > linux-gen: pktio: do not free zero packets > linux-gen: ring: created common ring implementation > linux-gen: align: added round up power of two > linux-gen: pool: reimplement pool with ring > linux-gen: ring: added multi enq and deq > linux-gen: pool: use ring multi enq and deq operations > linux-gen: pool: optimize buffer alloc > linux-gen: pool: clean up pool inlines functions > linux-gen: pool: ptr instead of hdl in buffer_alloc_multi > test: validation: buf: test alignment > test: performance: crypto: use capability to select max packet > test: correctly initialize pool parameters > test: validation: packet: fix bugs in tailroom and concat tests > linux-gen: packet: added support for segmented packets > test: validation: packet: improved multi-segment alloc test > api: packet: added limits for packet len on alloc > linux-gen: packet: remove zero len support from alloc > linux-gen: packet: enable multi-segment packets > > example/generator/odp_generator.c | 2 +- > include/odp/api/spec/packet.h | 9 +- > include/odp/api/spec/pool.h | 6 + > platform/linux-generic/Makefile.am | 1 + > .../include/odp/api/plat/packet_types.h | 6 +- > .../include/odp/api/plat/pool_types.h | 6 - > .../linux-generic/include/odp_align_internal.h | 34 +- > .../linux-generic/include/odp_buffer_inlines.h | 167 +-- > .../linux-generic/include/odp_buffer_internal.h | 120 +- > .../include/odp_classification_datamodel.h | 2 +- > .../linux-generic/include/odp_config_internal.h | 55 +- > .../linux-generic/include/odp_packet_internal.h | 87 +- > platform/linux-generic/include/odp_pool_internal.h | 289 +--- > platform/linux-generic/include/odp_ring_internal.h | 176 +++ > .../linux-generic/include/odp_timer_internal.h | 4 - > platform/linux-generic/odp_buffer.c | 22 +- > platform/linux-generic/odp_classification.c | 25 +- > platform/linux-generic/odp_crypto.c | 12 +- > platform/linux-generic/odp_packet.c | 717 ++++++++-- > platform/linux-generic/odp_packet_io.c | 2 +- > platform/linux-generic/odp_pool.c | 1440 > ++++++++------------ > platform/linux-generic/odp_queue.c | 4 +- > platform/linux-generic/odp_schedule.c | 102 +- > platform/linux-generic/odp_schedule_ordered.c | 4 +- > platform/linux-generic/odp_timer.c | 3 +- > platform/linux-generic/pktio/dpdk.c | 10 +- > platform/linux-generic/pktio/ipc.c | 3 +- > platform/linux-generic/pktio/loop.c | 2 +- > platform/linux-generic/pktio/netmap.c | 14 +- > platform/linux-generic/pktio/socket.c | 17 +- > platform/linux-generic/pktio/socket_mmap.c | 10 +- > test/common_plat/performance/odp_crypto.c | 47 +- > test/common_plat/performance/odp_pktio_perf.c | 2 +- > test/common_plat/performance/odp_scheduling.c | 8 +- > test/common_plat/validation/api/buffer/buffer.c | 113 +- > test/common_plat/validation/api/crypto/crypto.c | 2 +- > test/common_plat/validation/api/packet/packet.c | 96 +- > test/common_plat/validation/api/pktio/pktio.c | 21 +- > 38 files changed, 1745 insertions(+), 1895 deletions(-) > create mode 100644 platform/linux-generic/include/odp_ring_internal.h > > -- > 2.8.1 > >
