romenskiy2012/jemalloc
1
0
Fork 0
mirror of https://github.com/jemalloc/jemalloc.git synced 2026-07-27 02:57:21 +03:00
Code Issues Projects Releases Packages Wiki Activity

Remove hpa_sec_batch_fill_extra and calculate nallocs automatically.

Browse source 
This change includes the following improvements:

- Remove the hpa_sec_batch_fill_extra parameter.
- Refactor the hpa_alloc() code and helper functions to be able to
  allocate more than one extent out of a single pageslab. This way
  we can amortize the per-pageslab costs (active bitmap iteration,
  pageslab metadata updates) across multiple extents.
- Decide on a min and max number of extents that will be allocated
  in hpa_alloc(). The code will try to allocate at least the min
  and allocate up to the max as long as we can allocate additional
  ones from the pageslab we already have, as additional allocations
  are relatively cheap.
- Add extent allocation distribution stats.
- Amend hpa_sec_integration.c unit test.
This commit is contained in:
Tony Printezis 2026-03-02 11:11:09 -08:00 • committed by Guangli Dai
parent 639e70fcfb
commit f008ce9fe1
15 changed files with 675 additions and 105 deletions
Download patch file Download diff file Expand all files Collapse all files

81
src/stats.c
View file

@ -902,9 +902,7 @@ stats_arena_hpa_shard_counters_print(
" / sec)\n"
" Hugify failures: %" FMTu64 " (%" FMTu64
" / sec)\n"
" Dehugifies: %" FMTu64 " (%" FMTu64
" / sec)\n"
"\n",
" Dehugifies: %" FMTu64 " (%" FMTu64 " / sec)\n",
npageslabs, npageslabs_huge, npageslabs_nonhuge, nactive,
nactive_huge, nactive_nonhuge, ndirty, ndirty_huge, ndirty_nonhuge,
nretained_nonhuge, npurge_passes,
@ -944,6 +942,82 @@ stats_arena_hpa_shard_counters_print(
emitter_json_kv(
emitter, "ndirty_huge", emitter_type_size, &ndirty_huge);
emitter_json_object_end(emitter); /* End "slabs" */
/* alloc_batch stats */
uint64_t hpa_alloc_min_extents[SEC_MAX_NALLOCS + 1];
uint64_t hpa_alloc_max_extents[SEC_MAX_NALLOCS + 1];
uint64_t hpa_alloc_extents[SEC_MAX_NALLOCS + 1];
uint64_t hpa_alloc_ps[SEC_MAX_NALLOCS + 1];
uint64_t hpa_alloc_pages_per_ps[SEC_MAX_NALLOCS + 1];
uint64_t hpa_alloc_extents_per_ps[SEC_MAX_NALLOCS + 1];
uint64_t hpa_alloc_total_elapsed_ns_per_ps[SEC_MAX_NALLOCS + 1];
size_t alloc_mib[CTL_MAX_DEPTH];
CTL_LEAF_PREPARE(alloc_mib, 0, "stats.arenas");
alloc_mib[2] = i;
CTL_LEAF_PREPARE(alloc_mib, 3, "hpa_shard.alloc");
for (size_t j = 0; j <= SEC_MAX_NALLOCS; j += 1) {
alloc_mib[5] = j;
CTL_LEAF(alloc_mib, 6, "min_extents", &hpa_alloc_min_extents[j],
uint64_t);
CTL_LEAF(alloc_mib, 6, "max_extents", &hpa_alloc_max_extents[j],
uint64_t);
CTL_LEAF(
alloc_mib, 6, "extents", &hpa_alloc_extents[j], uint64_t);
CTL_LEAF(alloc_mib, 6, "ps", &hpa_alloc_ps[j], uint64_t);
CTL_LEAF(alloc_mib, 6, "pages_per_ps",
&hpa_alloc_pages_per_ps[j], uint64_t);
CTL_LEAF(alloc_mib, 6, "extents_per_ps",
&hpa_alloc_extents_per_ps[j], uint64_t);
CTL_LEAF(alloc_mib, 6, "total_elapsed_ns_per_ps",
&hpa_alloc_total_elapsed_ns_per_ps[j], uint64_t);
}
emitter_table_printf(emitter, " extent allocation distribution:\n");
emitter_table_printf(emitter,
" %4s %20s %20s %20s %20s %20s %20s %24s %24s\n", "",
"min_extents", "max_extents",
"extents", "ps", "pages_per_ps", "extents_per_ps",
"total_elapsed_ns_per_ps", "elapsed_ns_per_ps");
for (size_t j = 0; j <= SEC_MAX_NALLOCS; j += 1) {
const uint64_t extents_per_ps = hpa_alloc_extents_per_ps[j];
const uint64_t total_elapsed_ns_per_ps =
hpa_alloc_total_elapsed_ns_per_ps[j];
const uint64_t elapsed_ns_per_ps = (extents_per_ps != 0)
? (total_elapsed_ns_per_ps / extents_per_ps)
: 0;
emitter_table_printf(emitter,
" %4zu %20" FMTu64 " %20" FMTu64 " %20" FMTu64
" %20" FMTu64 " %20" FMTu64 " %20" FMTu64 " %24" FMTu64
" %24" FMTu64 "\n",
j, hpa_alloc_min_extents[j], hpa_alloc_max_extents[j],
hpa_alloc_extents[j],
hpa_alloc_ps[j], hpa_alloc_pages_per_ps[j], extents_per_ps,
total_elapsed_ns_per_ps, elapsed_ns_per_ps);
}
emitter_table_printf(emitter, "\n");
emitter_json_array_kv_begin(emitter, "extent_allocation_distribution");
for (size_t j = 0; j <= SEC_MAX_NALLOCS; j += 1) {
emitter_json_object_begin(emitter);
emitter_json_kv(emitter, "min_extents", emitter_type_uint64,
&hpa_alloc_min_extents[j]);
emitter_json_kv(emitter, "max_extents", emitter_type_uint64,
&hpa_alloc_max_extents[j]);
emitter_json_kv(emitter, "extents", emitter_type_uint64,
&hpa_alloc_extents[j]);
emitter_json_kv(
emitter, "ps", emitter_type_uint64, &hpa_alloc_ps[j]);
emitter_json_kv(emitter, "pages_per_ps", emitter_type_uint64,
&hpa_alloc_pages_per_ps[j]);
emitter_json_kv(emitter, "extents_per_ps", emitter_type_uint64,
&hpa_alloc_extents_per_ps[j]);
emitter_json_kv(emitter, "total_elapsed_ns_per_ps",
emitter_type_uint64, &hpa_alloc_total_elapsed_ns_per_ps[j]);
emitter_json_object_end(emitter);
}
emitter_json_array_end(emitter); /* End "alloc_batch" */
}
static void
@ -1687,7 +1761,6 @@ stats_general_print(emitter_t *emitter) {
OPT_WRITE_SIZE_T("hpa_sec_nshards")
OPT_WRITE_SIZE_T("hpa_sec_max_alloc")
OPT_WRITE_SIZE_T("hpa_sec_max_bytes")
OPT_WRITE_SIZE_T("hpa_sec_batch_fill_extra")
OPT_WRITE_BOOL("huge_arena_pac_thp")
OPT_WRITE_CHAR_P("metadata_thp")
OPT_WRITE_INT64("mutex_max_spin")