lihao/jemalloc
1
0
Fork 0
mirror of https://github.com/jemalloc/jemalloc.git synced 2026-07-23 21:23:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

Implement per-CPU arena.

Browse Source 
The new feature, opt.percpu_arena, determines thread-arena association
dynamically based CPU id. Three modes are supported: "percpu", "phycpu"
and disabled.

"percpu" uses the current core id (with help from sched_getcpu())
directly as the arena index, while "phycpu" will assign threads on the
same physical CPU to the same arena. In other words, "percpu" means # of
arenas == # of CPUs, while "phycpu" has # of arenas == 1/2 * (# of
CPUs). Note that no runtime check on whether hyper threading is enabled
is added yet.

When enabled, threads will be migrated between arenas when a CPU change
is detected. In the current design, to reduce overhead from reading CPU
id, each arena tracks the thread accessed most recently. When a new
thread comes in, we will read CPU id and update arena if necessary.
This commit is contained in:
Qi Wang
2017-02-02 17:02:05 -08:00
committed by Qi Wang
parent 8721e19c04
commit ec532e2c5c
16 changed files with 415 additions and 119 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
src/ctl.c
View File

@@ -72,6 +72,7 @@ CTL_PROTO(config_xmalloc)
CTL_PROTO(opt_abort)
CTL_PROTO(opt_dss)
CTL_PROTO(opt_narenas)
CTL_PROTO(opt_percpu_arena)
CTL_PROTO(opt_decay_time)
CTL_PROTO(opt_stats_print)
CTL_PROTO(opt_junk)
@@ -229,6 +230,7 @@ static const ctl_named_node_t opt_node[] = {
{NAME("abort"), CTL(opt_abort)},
{NAME("dss"), CTL(opt_dss)},
{NAME("narenas"), CTL(opt_narenas)},
{NAME("percpu_arena"), CTL(opt_percpu_arena)},
{NAME("decay_time"), CTL(opt_decay_time)},
{NAME("stats_print"), CTL(opt_stats_print)},
{NAME("junk"), CTL(opt_junk)},
@@ -1284,6 +1286,7 @@ CTL_RO_CONFIG_GEN(config_xmalloc, bool)
CTL_RO_NL_GEN(opt_abort, opt_abort, bool)
CTL_RO_NL_GEN(opt_dss, opt_dss, const char *)
CTL_RO_NL_GEN(opt_narenas, opt_narenas, unsigned)
CTL_RO_NL_GEN(opt_percpu_arena, opt_percpu_arena, const char *)
CTL_RO_NL_GEN(opt_decay_time, opt_decay_time, ssize_t)
CTL_RO_NL_GEN(opt_stats_print, opt_stats_print, bool)
CTL_RO_NL_CGEN(config_fill, opt_junk, opt_junk, const char *)
@@ -1317,10 +1320,10 @@ thread_arena_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
if (oldarena == NULL) {
return EAGAIN;
}
newind = oldind = arena_ind_get(oldarena);
WRITE(newind, unsigned);
READ(oldind, unsigned);
if (newind != oldind) {
arena_t *newarena;
@@ -1330,6 +1333,19 @@ thread_arena_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
goto label_return;
}
if (have_percpu_arena &&
(percpu_arena_mode != percpu_arena_disabled)) {
if (newind < percpu_arena_ind_limit()) {
/*
* If perCPU arena is enabled, thread_arena
* control is not allowed for the auto arena
* range.
*/
ret = EPERM;
goto label_return;
}
}
/* Initialize arena if necessary. */
newarena = arena_get(tsd_tsdn(tsd), newind, true);
if (newarena == NULL) {