Hi, The 1st patch removes memory footprint of percpu_ref in fast path from 7 words to 2 words, since it is often used in fast path and embedded in user struct. The 2nd patch moves .q_usage_counter to 1st cacheline of 'request_queue'. Simple test on null_blk shows ~2% IOPS boost on one 16cores(two threads per core) machine, dual socket/numa. V2: - pass 'gfp' to kzalloc() for fixing block/027 failure reported by kernel test robot - protect percpu_ref_is_zero() with destroying percpu-refcount by spin lock Ming Lei (2): percpu_ref: reduce memory footprint of percpu_ref in fast path block: move 'q_usage_counter' into front of 'request_queue' drivers/infiniband/sw/rdmavt/mr.c | 2 +- include/linux/blkdev.h | 3 +- include/linux/percpu-refcount.h | 45 ++++------- lib/percpu-refcount.c | 128 ++++++++++++++++++++++-------- 4 files changed, 115 insertions(+), 63 deletions(-) Cc: Sagi Grimberg <sagi@xxxxxxxxxxx> Cc: Tejun Heo <tj@xxxxxxxxxx> Cc: Christoph Hellwig <hch@xxxxxx> Cc: Jens Axboe <axboe@xxxxxxxxx> Cc: Bart Van Assche <bvanassche@xxxxxxx> -- 2.25.2
