projects / platform / kernel / linux-exynos.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: a5a9e8c) | patch
staging/rdma/hfi1: convert buffers allocated atomic to per cpu
authorMike Marciniszyn <mike.marciniszyn@intel.com>
Mon, 7 Dec 2015 20:39:22 +0000 (15:39 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 21 Dec 2015 22:00:17 +0000 (14:00 -0800)
commita054374f15428cbc1d9cb9cba17ce870eaa7d60f
tree07dd3a9d52cb4c409fd799c65ab93e699a6e4c2ftree | snapshot
parenta5a9e8ccab4d24c7d9e1da8222f373688745ca6acommit | diff
staging/rdma/hfi1: convert buffers allocated atomic to per cpu

Profiling has shown the the atomic is a performance issue
for the pio hot path.

If multiple cpus allocated an sc's buffer, the cacheline
containing the atomic will bounce from L0 to L0.

Convert the atomic to a percpu variable.

Reviewed-by: Jubin John <jubin.john@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/rdma/hfi1/pio.c diff | blob | history
drivers/staging/rdma/hfi1/pio.h diff | blob | history
drivers/staging/rdma/hfi1/pio_copy.c diff | blob | history
Domain: System / Kernel;