-
Vikram Narayanan authored
With a single hctx, multi-threaded fio bottlenecks on accessing an atomic variable. To avoid accessing a single global variable, add support for running null_blk with as many hctx as the number of CPUs.
ed1a5fbd
With a single hctx, multi-threaded fio bottlenecks on accessing an atomic variable. To avoid accessing a single global variable, add support for running null_blk with as many hctx as the number of CPUs.