With this 'struct map' uses a bit over 3 cachelines:
$ pahole -C map ~/bin/perf
<SNIP>
/* --- cacheline 2 boundary (128 bytes) --- */
u64 (*unmap_ip)(struct map *, u64); /* 128 8 */
struct dso * dso; /* 136 8 */
refcount_t refcnt; /* 144 4 */
/* size: 152, cachelines: 3, members: 18 */
/* sum members: 145, holes: 1, sum holes: 3 */
/* padding: 4 */
/* forced alignments: 2 */
/* last cacheline: 24 bytes */
} __attribute__((__aligned__(8)));
$
We probably can move map->map/unmap_ip() moved to 'struct map_groups',
that will shave more 16 bytes, getting this almost to two cachelines.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-ymlv3nzpofv2fugnjnizkrwy@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
map->map_ip = map__map_ip;
map->unmap_ip = map__unmap_ip;
RB_CLEAR_NODE(&map->rb_node);
- map->groups = NULL;
map->erange_warned = false;
refcount_set(&map->refcnt, 1);
}
refcount_set(&map->refcnt, 1);
RB_CLEAR_NODE(&map->rb_node);
dso__get(map->dso);
- map->groups = NULL;
}
return map;
void map_groups__insert(struct map_groups *mg, struct map *map)
{
maps__insert(&mg->maps, map);
- map->groups = mg;
}
static void __maps__purge(struct maps *maps)
{
__maps__insert(&mg->maps, map);
__maps__insert_name(&mg->maps, map);
- map->groups = mg;
}
int map_groups__fixup_overlappings(struct map_groups *mg, struct map *map, FILE *fp)
u64 (*unmap_ip)(struct map *, u64);
struct dso *dso;
- struct map_groups *groups;
refcount_t refcnt;
};