diff options
author | Kjell Winblad <[email protected]> | 2019-02-22 16:49:37 +0100 |
---|---|---|
committer | Kjell Winblad <[email protected]> | 2019-04-10 15:42:42 +0200 |
commit | c5e9766712436bea2b91bccd062f66a3ad1841bb (patch) | |
tree | 25ce2b16b9f847088997b0d4e77af46dbb6d2d66 /erts/emulator/beam/erl_db_tree_util.h | |
parent | 326c3cb70c1b37c794b781a42c50725766098810 (diff) | |
download | otp-c5e9766712436bea2b91bccd062f66a3ad1841bb.tar.gz otp-c5e9766712436bea2b91bccd062f66a3ad1841bb.tar.bz2 otp-c5e9766712436bea2b91bccd062f66a3ad1841bb.zip |
Decentralized counters for ETS ordered_set with write_concurrency
Previously, all ETS tables used centralized counter variables to keep
track of the number of items stored and the amount of memory
consumed. These counters can cause scalability problems (especially on
big NUMA systems). This commit adds an implementation of a
decentralized counter and modifies the implementation of ETS so that
ETS tables of type ordered_set with write_concurrency enabled use the
decentralized counter. [Experiments][1] indicate that this change
substantially improves the scalability of ETS ordered_set tables with
write_concurrency enabled in scenarios with frequent `ets:insert/2`
and `ets:delete/2` calls.
The new counter is implemented in the module erts_flxctr
(`erts_flxctr.h` and `erts_flxctr.c`). The module has the suffix
flxctr as it contains the implementation of a flexible counter (i.e.,
counter instances can be configured to be either centralized or
decentralized). Counters that are configured to be centralized are
implemented with a single counter variable which is modified with
atomic operations. Decentralized counters are spread over several
cache lines (how many can be configured with the parameter
`+dcg`). The scheduler threads are mapped to cache lines so that there
is no single point of contention when decentralized counters are
updated. The thread progress functionality of the Erlang VM is
utilized to implement support for linearizable snapshots of
decentralized counters. The snapshot functionality is used by the
`ets:info/1` and `ets:info/2` functions.
[1]: http://winsh.me/ets_catree_benchmark/flxctr_res.html
Diffstat (limited to 'erts/emulator/beam/erl_db_tree_util.h')
-rw-r--r-- | erts/emulator/beam/erl_db_tree_util.h | 18 |
1 files changed, 17 insertions, 1 deletions
diff --git a/erts/emulator/beam/erl_db_tree_util.h b/erts/emulator/beam/erl_db_tree_util.h index 02df74678d..ba4a8f79e5 100644 --- a/erts/emulator/beam/erl_db_tree_util.h +++ b/erts/emulator/beam/erl_db_tree_util.h @@ -25,6 +25,8 @@ ** Internal functions and macros used by both the CA tree and the AVL tree */ + +#if defined(ARCH_32) /* ** A stack of this size is enough for an AVL tree with more than ** 0xFFFFFFFF elements. May be subject to change if @@ -34,8 +36,19 @@ ** Where n denotes the number of nodes, h(n) the height of the tree ** with n nodes and log is the binary logarithm. */ - #define STACK_NEED 50 +#elif defined(ARCH_64) +/* +** A stack of this size is enough for an AVL tree with more than +** 2^61 elements. +** The Maximal height of an AVL tree is calculated as above. +*/ +#define STACK_NEED 90 +#else +#error "Unsported architecture" +#endif + + #define PUSH_NODE(Dtt, Tdt) \ ((Dtt)->array[(Dtt)->pos++] = Tdt) @@ -50,6 +63,9 @@ #define EMPTY_NODE(Dtt) (TOP_NODE(Dtt) == NULL) +#define DEC_NITEMS(DB) \ + erts_flxctr_dec(&(DB)->common.counters, ERTS_DB_TABLE_NITEMS_COUNTER_ID) + static ERTS_INLINE void free_term(DbTable *tb, TreeDbTerm* p) { db_free_term(tb, p, offsetof(TreeDbTerm, dbterm)); |