Hello, The patchset drops idle compression streams list from zcomp, so zram now is using per-cpu compression streams. This removes two spin_lock()/spin_unlock() calls from the write path and prevents write OP from being preempted with active compression stream, which improves performance (according to the tests, see commit messages) and simplifies the code.
against next-20160428 Sergey Senozhatsky (2): zsmalloc: require GFP in zs_malloc() zram: user per-cpu compression streams drivers/block/zram/zcomp.c | 293 ++++++++++++------------------------------ drivers/block/zram/zcomp.h | 12 +- drivers/block/zram/zram_drv.c | 34 ++++- include/linux/zsmalloc.h | 2 +- mm/zsmalloc.c | 15 +-- 5 files changed, 119 insertions(+), 237 deletions(-) -- 2.8.0
