Sign Up
Log In
Log In
or
Sign Up
Places
All Projects
Status Monitor
Collapse sidebar
home:Ledest:erlang:23
erlang
4495-Fixes-due-to-comments-by-sverker.patch
Overview
Repositories
Revisions
Requests
Users
Attributes
Meta
File 4495-Fixes-due-to-comments-by-sverker.patch of Package erlang
From 49c2445ade21abf1dc5c5ee557418e56aa87af50 Mon Sep 17 00:00:00 2001 From: Kjell Winblad <kjellwinblad@gmail.com> Date: Wed, 6 Oct 2021 15:37:23 +0200 Subject: [PATCH 5/8] Fixes due to comments by @sverker --- erts/emulator/beam/erl_db.c | 29 +++++++++-------- erts/emulator/beam/erl_db_hash.c | 54 +++++++++++++++++--------------- erts/emulator/beam/erl_db_hash.h | 2 +- lib/stdlib/test/ets_SUITE.erl | 12 +++++-- 4 files changed, 55 insertions(+), 42 deletions(-) diff --git a/erts/emulator/beam/erl_db.c b/erts/emulator/beam/erl_db.c index a047c6ecb5..9cfc89c8e9 100644 --- a/erts/emulator/beam/erl_db.c +++ b/erts/emulator/beam/erl_db.c @@ -58,6 +58,9 @@ #include "bif.h" +#define DB_WRITE_CONCURRENCY_MIN_LOCKS 1 +#define DB_WRITE_CONCURRENCY_MAX_LOCKS 32768 + erts_atomic_t erts_ets_misc_mem_size; /* @@ -815,7 +818,7 @@ DbTable* db_get_table_aux(Process *p, tb = tid2tab(id); if (tb) { - erl_db_hash_adapt_no_locks(tb); + erl_db_hash_adapt_number_of_locks(tb); db_lock(tb, kind); #ifdef ETS_DBG_FORCE_TRAP if (erts_atomic_read_nob(&tb->common.dbg_force_trap)) { @@ -2255,7 +2258,7 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2) Sint keypos; int is_named, is_compressed; int is_fine_locked, frequent_read; - int no_locks; + int number_of_locks; int is_decentralized_counters; int is_decentralized_counters_option; int is_explicit_lock_granularity; @@ -2280,7 +2283,7 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2) heir = am_none; heir_data = (UWord) am_undefined; is_compressed = erts_ets_always_compress; - no_locks = -1; + number_of_locks = -1; is_explicit_lock_granularity = 0; is_write_concurrency_auto = 0; @@ -2308,22 +2311,22 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2) keypos = signed_val(tp[2]); } else if (tp[1] == am_write_concurrency) { - Sint no_locks_param; + Sint number_of_locks_param; if (is_integer(tp[2]) && - term_to_Sint(tp[2], &no_locks_param) && - no_locks_param >= 1 && - no_locks_param <= 32768) { + term_to_Sint(tp[2], &number_of_locks_param) && + number_of_locks_param >= DB_WRITE_CONCURRENCY_MIN_LOCKS && + number_of_locks_param <= DB_WRITE_CONCURRENCY_MAX_LOCKS) { is_decentralized_counters = 1; is_fine_locked = 1; is_explicit_lock_granularity = 1; is_write_concurrency_auto = 0; - no_locks = no_locks_param; + number_of_locks = number_of_locks_param; } else if (tp[2] == am_auto) { is_decentralized_counters = 1; is_write_concurrency_auto = 1; is_fine_locked = 1; is_explicit_lock_granularity = 0; - no_locks = -1; + number_of_locks = -1; } else if (tp[2] == am_true) { if (!(status & DB_ORDERED_SET)) { is_decentralized_counters = 0; @@ -2331,12 +2334,12 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2) is_fine_locked = 1; is_explicit_lock_granularity = 0; is_write_concurrency_auto = 0; - no_locks = -1; + number_of_locks = -1; } else if (tp[2] == am_false) { is_fine_locked = 0; is_explicit_lock_granularity = 0; is_write_concurrency_auto = 0; - no_locks = -1; + number_of_locks = -1; } else break; if (DB_LOCK_FREE(NULL)) is_fine_locked = 0; @@ -2411,7 +2414,7 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2) status |= DB_FINE_LOCKED_AUTO; } } else { - no_locks = -1; + number_of_locks = -1; } } else if (IS_TREE_TABLE(status)) { @@ -2464,7 +2467,7 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2) if (IS_HASH_TABLE(status)) { DbTableHash* hash_db = (DbTableHash*) tb; - hash_db->nlocks = no_locks; + hash_db->nlocks = number_of_locks; } cret = meth->db_create(BIF_P, tb); ASSERT(cret == DB_ERROR_NONE); (void)cret; diff --git a/erts/emulator/beam/erl_db_hash.c b/erts/emulator/beam/erl_db_hash.c index fd79dd3344..6b160eb97a 100644 --- a/erts/emulator/beam/erl_db_hash.c +++ b/erts/emulator/beam/erl_db_hash.c @@ -138,6 +138,8 @@ #define GROW_LIMIT(NACTIVE) ((NACTIVE)*1) #define SHRINK_LIMIT(TB) erts_atomic_read_nob(&(TB)->shrink_limit) +#define IS_POW2(x) ((x) && !((x) & ((x)-1))) + /* ** We want the first mandatory segment to be small (to reduce minimal footprint) ** and larger extra segments (to reduce number of alloc/free calls). @@ -264,7 +266,7 @@ static ERTS_INLINE int is_pseudo_deleted(HashDbTerm* p) ((is_atom(term) ? (atom_tab(atom_val(term))->slot.bucket.hvalue) : \ make_internal_hash(term, 0)) & MAX_HASH_MASK) -# define GET_LOCK_MASK(NO_LOCKS) ((NO_LOCKS)-1) +# define GET_LOCK_MASK(NUMBER_OF_LOCKS) ((NUMBER_OF_LOCKS)-1) # define GET_LOCK(tb,hval) (&(tb)->locks[(hval) & GET_LOCK_MASK(tb->nlocks)].u.lck_ctr.lck) # define GET_LOCK_AND_CTR(tb,hval) (&(tb)->locks[(hval) & GET_LOCK_MASK(tb->nlocks)].u.lck_ctr) @@ -276,24 +278,24 @@ static ERTS_INLINE int is_pseudo_deleted(HashDbTerm* p) # define LCK_AUTO_SHRINK_LIMIT -10000000 # define LCK_AUTO_MAX_LOCKS 8192 # define LCK_AUTO_MIN_LOCKS 4 -# define LCK_AUTO_DEFAULT_NO_LOCKS 64 +# define LCK_AUTO_DEFAULT_NUMBER_OF_LOCKS 64 # define LCK_AUTO_MAX_LOCKS_FREQ_READ_RW_LOCKS 128 static Sint get_lock_nitems_form_prev_lock_array(int index, - int new_no_locks, - int old_no_locks, + int new_number_of_locks, + int old_number_of_locks, DbTableHashFineLockSlot* old_locks) { - if (new_no_locks > old_no_locks) { + if (new_number_of_locks > old_number_of_locks) { Sint nitems = 0; - Sint in_source = old_locks[index % old_no_locks].u.lck_ctr.nitems; + Sint in_source = old_locks[index % old_number_of_locks].u.lck_ctr.nitems; nitems += in_source / 2; - if (index >= old_no_locks) { + if (index >= old_number_of_locks) { nitems += in_source % 2; } return nitems; } else { Sint in_source_1 = old_locks[index].u.lck_ctr.nitems; - Sint in_source_2 = old_locks[index + new_no_locks].u.lck_ctr.nitems; + Sint in_source_2 = old_locks[index + new_number_of_locks].u.lck_ctr.nitems; return in_source_1 + in_source_2; } @@ -301,10 +303,10 @@ static Sint get_lock_nitems_form_prev_lock_array(int index, static void calc_shrink_limit(DbTableHash* tb); -void erl_db_hash_adapt_no_locks(DbTable* tb) { +void erl_db_hash_adapt_number_of_locks(DbTable* tb) { db_hash_lock_array_resize_state current_state; DbTableHash* tbl; - int new_no_locks; + int new_number_of_locks; if(!(tb->common.type & DB_FINE_LOCKED_AUTO)) { return; } @@ -335,9 +337,9 @@ void erl_db_hash_adapt_no_locks(DbTable* tb) { } if (current_state == DB_HASH_LOCK_ARRAY_RESIZE_STATUS_GROW && erts_atomic_read_nob(&tbl->nactive) >= (2*tbl->nlocks)) { - new_no_locks = 2*tbl->nlocks; + new_number_of_locks = 2*tbl->nlocks; } else if (current_state == DB_HASH_LOCK_ARRAY_RESIZE_STATUS_SHRINK) { - new_no_locks = tbl->nlocks / 2; + new_number_of_locks = tbl->nlocks / 2; } else { /* Do not do any adaptation if the number of active buckets is @@ -359,11 +361,11 @@ void erl_db_hash_adapt_no_locks(DbTable* tb) { erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER; int i; DbTableHashFineLockSlot* old_locks = tbl->locks; - Uint old_no_locks = tbl->nlocks; - ASSERT(new_no_locks != 0); - tbl->nlocks = new_no_locks; + Uint old_number_of_locks = tbl->nlocks; + ASSERT(new_number_of_locks != 0); + tbl->nlocks = new_number_of_locks; if (tb->common.type & DB_FREQ_READ && - new_no_locks <= LCK_AUTO_MAX_LOCKS_FREQ_READ_RW_LOCKS) { + new_number_of_locks <= LCK_AUTO_MAX_LOCKS_FREQ_READ_RW_LOCKS) { rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ; } if (erts_ets_rwmtx_spin_count >= 0) { @@ -376,7 +378,7 @@ void erl_db_hash_adapt_no_locks(DbTable* tb) { erts_rwmtx_init_opt(&tbl->locks[i].u.lck_ctr.lck, &rwmtx_opt, "db_hash_slot", tb->common.the_name, ERTS_LOCK_FLAGS_CATEGORY_DB); tbl->locks[i].u.lck_ctr.nitems = - get_lock_nitems_form_prev_lock_array(i, tbl->nlocks, old_no_locks, old_locks); + get_lock_nitems_form_prev_lock_array(i, tbl->nlocks, old_number_of_locks, old_locks); tbl->locks[i].u.lck_ctr.lck_stat = 0; } /* #define HARD_DEBUG_ITEM_CNT_LOCK_CHANGE 1 */ @@ -385,7 +387,7 @@ void erl_db_hash_adapt_no_locks(DbTable* tb) { Sint total_old = 0; Sint total_new = 0; int i; - for (i=0; i < old_no_locks; i++) { + for (i=0; i < old_number_of_locks; i++) { total_old += old_locks[i].u.lck_ctr.nitems; } for (i=0; i < tbl->nlocks; i++) { @@ -400,10 +402,10 @@ void erl_db_hash_adapt_no_locks(DbTable* tb) { erts_atomic_set_nob(&tbl->lock_array_resize_state, DB_HASH_LOCK_ARRAY_RESIZE_STATUS_NORMAL); erts_rwmtx_rwunlock(&tb->common.rwlock); - for (i = 0; i < old_no_locks; i++) { + for (i = 0; i < old_number_of_locks; i++) { erts_rwmtx_destroy(&old_locks[i].u.lck_ctr.lck); } - erts_db_free(ERTS_ALC_T_DB_SEG, tb, old_locks, sizeof(DbTableHashFineLockSlot) * old_no_locks); + erts_db_free(ERTS_ALC_T_DB_SEG, tb, old_locks, sizeof(DbTableHashFineLockSlot) * old_number_of_locks); } } @@ -997,12 +999,14 @@ int db_create_hash(Process *p, DbTable *tbl) erts_atomic_init_nob(&tb->is_resizing, 0); erts_atomic_init_nob(&tb->lock_array_resize_state, (erts_aint_t)DB_HASH_LOCK_ARRAY_RESIZE_STATUS_NORMAL); - if (tb->nlocks == -1 || !(tb->common.type & DB_FINE_LOCKED)) { + if (!(tb->common.type & DB_FINE_LOCKED)) { /* The number of locks needs to be set even if fine grained locking is not used as this variable is used when iterating - over the table + over the table. */ + tb->nlocks = 1; + } else if(tb->nlocks == -1) { tb->nlocks = DB_HASH_LOCK_CNT; } @@ -1010,7 +1014,7 @@ int db_create_hash(Process *p, DbTable *tbl) erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER; int i; if (tb->common.type & DB_FINE_LOCKED_AUTO) { - tb->nlocks = LCK_AUTO_DEFAULT_NO_LOCKS; + tb->nlocks = LCK_AUTO_DEFAULT_NUMBER_OF_LOCKS; } /* nlocks needs to be a power of two so we round down to @@ -1047,8 +1051,8 @@ int db_create_hash(Process *p, DbTable *tbl) ASSERT(tb->nlocks <= erts_atomic_read_nob(&tb->nactive)); ASSERT(erts_atomic_read_nob(&tb->nactive) <= tb->nslots); ASSERT(tb->nslots <= (erts_atomic_read_nob(&tb->szm) + 1)); - ASSERT((tb->nlocks % 2) == 0); - ASSERT((erts_atomic_read_nob(&tb->szm) + 1) % 2 == 0); + ASSERT(IS_POW2(tb->nlocks)); + ASSERT(IS_POW2(erts_atomic_read_nob(&tb->szm) + 1)); } else { /* coarse locking */ tb->locks = NULL; diff --git a/erts/emulator/beam/erl_db_hash.h b/erts/emulator/beam/erl_db_hash.h index 4154a2aa2b..086ca0837b 100644 --- a/erts/emulator/beam/erl_db_hash.h +++ b/erts/emulator/beam/erl_db_hash.h @@ -95,7 +95,7 @@ typedef enum { } db_hash_lock_array_resize_state; /* To adapt number of locks if hash table with {write_concurrency, auto} */ -void erl_db_hash_adapt_no_locks(DbTable* tb); +void erl_db_hash_adapt_number_of_locks(DbTable* tb); /* ** Function prototypes, looks the same (except the suffix) for all diff --git a/lib/stdlib/test/ets_SUITE.erl b/lib/stdlib/test/ets_SUITE.erl index b44d67aec4..c7a4b38c51 100644 --- a/lib/stdlib/test/ets_SUITE.erl +++ b/lib/stdlib/test/ets_SUITE.erl @@ -7427,9 +7427,15 @@ whereis_table(Config) when is_list(Config) -> Tid = ets:whereis(whereis_test), ets:insert(whereis_test, [{hello}, {there}]), - - [[{hello}],[{there}]] = ets:match(whereis_test, '$1'), - [[{hello}],[{there}]] = ets:match(Tid, '$1'), + CheckMatch = + fun(MatchRes) -> + case MatchRes of + [[{there}],[{hello}]] -> ok; + [[{hello}],[{there}]] -> ok + end + end, + CheckMatch(ets:match(whereis_test, '$1')), + CheckMatch(ets:match(Tid, '$1')), true = ets:delete_all_objects(Tid), -- 2.31.1
Locations
Projects
Search
Status Monitor
Help
OpenBuildService.org
Documentation
API Documentation
Code of Conduct
Contact
Support
@OBShq
Terms
openSUSE Build Service is sponsored by
The Open Build Service is an
openSUSE project
.
Sign Up
Log In
Places
Places
All Projects
Status Monitor
