get_string slower than it needs to be.

[ikeydoherty]

Going by the general rule of thumb that gets are quicker than sets (read vs write),
and the single get_string clearly being quicker than set_string, it seems odd that
get_string4k is slower than set_string4k in bxt_timing

set_string                    32.665us      86.291us           0
get_string                    29.088us     115.403us           0
set_unset_string              62.137us      81.381us           0
set_string4k                  33.275us      20.620us           0
get_string4k                  40.344us      27.054us           0
set_unset_string4k            61.957us      23.804us           0

[sofar]

Won't surprise me if there's an extra copy involved during "get" here...

[ikeydoherty]

Thinking the same thing -- hoping it exposes itself post hashmap merge.