aboutsummaryrefslogtreecommitdiffstats
path: root/erts/emulator/beam/erl_thr_progress.h
diff options
context:
space:
mode:
authorRickard Green <[email protected]>2010-09-15 22:14:51 +0200
committerRickard Green <[email protected]>2011-11-13 20:39:30 +0100
commita67e91e658bdbba24fcc3c79b06fdf10ff830bc9 (patch)
tree07f9e6b1fd715d516d2571521307fe1b9d7c3948 /erts/emulator/beam/erl_thr_progress.h
parent55358c54778ead444e51f565d00175ba887ef182 (diff)
downloadotp-a67e91e658bdbba24fcc3c79b06fdf10ff830bc9.tar.gz
otp-a67e91e658bdbba24fcc3c79b06fdf10ff830bc9.tar.bz2
otp-a67e91e658bdbba24fcc3c79b06fdf10ff830bc9.zip
Optimize memory allocation
A number of memory allocation optimizations have been implemented. Most optimizations reduce contention caused by synchronization between threads during allocation and deallocation of memory. Most notably: * Synchronization of memory management in scheduler specific allocator instances has been rewritten to use lock-free synchronization. * Synchronization of memory management in scheduler specific pre-allocators has been rewritten to use lock-free synchronization. * The 'mseg_alloc' memory segment allocator now use scheduler specific instances instead of one instance. Apart from reducing contention this also ensures that memory allocators always create memory segments on the local NUMA node on a NUMA system.
Diffstat (limited to 'erts/emulator/beam/erl_thr_progress.h')
-rw-r--r--erts/emulator/beam/erl_thr_progress.h210
1 files changed, 210 insertions, 0 deletions
diff --git a/erts/emulator/beam/erl_thr_progress.h b/erts/emulator/beam/erl_thr_progress.h
new file mode 100644
index 0000000000..36880c10f0
--- /dev/null
+++ b/erts/emulator/beam/erl_thr_progress.h
@@ -0,0 +1,210 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson AB 2011. All Rights Reserved.
+ *
+ * The contents of this file are subject to the Erlang Public License,
+ * Version 1.1, (the "License"); you may not use this file except in
+ * compliance with the License. You should have received a copy of the
+ * Erlang Public License along with this software. If not, it can be
+ * retrieved online at http://www.erlang.org/.
+ *
+ * Software distributed under the License is distributed on an "AS IS"
+ * basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
+ * the License for the specific language governing rights and limitations
+ * under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+/*
+ * Description: Thread progress information. Used by lock free algorithms
+ * to determine when all involved threads are guaranteed to
+ * have passed a specific point of execution.
+ *
+ * Usage instructions can be found in ert_thr_progress.c
+ *
+ * Author: Rickard Green
+ */
+
+#if !defined(ERL_THR_PROGRESS_H__TSD_TYPE__)
+#define ERL_THR_PROGRESS_H__TSD_TYPE__
+
+#include "sys.h"
+
+#ifdef ERTS_SMP
+
+typedef Uint64 ErtsThrPrgrVal;
+
+#define ERTS_THR_PRGR_WAKEUP_DATA_SIZE 4 /* Need to be an even power of 2. */
+
+typedef struct {
+ int id;
+ int is_managed;
+ ErtsThrPrgrVal wakeup_request[ERTS_THR_PRGR_WAKEUP_DATA_SIZE];
+
+ /* --- Part below only for managed threads */
+
+ int leader; /* Needs to be first in the managed threads part */
+ int active;
+ struct {
+ ErtsThrPrgrVal local;
+ ErtsThrPrgrVal next;
+ ErtsThrPrgrVal current;
+ } previous;
+} ErtsThrPrgrData;
+#endif /* ERTS_SMP */
+
+#endif
+
+#if !defined(ERL_THR_PROGRESS_H__) && !defined(ERL_THR_PROGRESS_TSD_TYPE_ONLY)
+#define ERL_THR_PROGRESS_H__
+
+#include "erl_threads.h"
+#include "erl_process.h"
+
+#ifdef ERTS_SMP
+
+#define ERTS_THR_PRGR_VAL_WAITING (~((ErtsThrPrgrVal) 0))
+
+extern erts_tsd_key_t erts_thr_prgr_data_key__;
+
+#ifdef ARCH_64
+# define ERTS_THR_PRGR_ATOMIC erts_atomic_t
+#else /* ARCH_32 */
+# define ERTS_THR_PRGR_ATOMIC erts_dw_atomic_t
+#endif
+
+typedef struct {
+ void *arg;
+ void (*wakeup)(void *arg);
+} ErtsThrPrgrWakeupCallback;
+
+typedef struct {
+ ERTS_THR_PRGR_ATOMIC current;
+} ErtsThrPrgr;
+
+extern ErtsThrPrgr erts_thr_prgr__;
+
+void erts_thr_progress_pre_init(void);
+void erts_thr_progress_init(int no_schedulers, int managed, int unmanaged);
+void erts_thr_progress_register_managed_thread(ErtsSchedulerData *esdp,
+ ErtsThrPrgrWakeupCallback *,
+ int);
+void erts_thr_progress_register_unmanaged_thread(ErtsThrPrgrWakeupCallback *);
+void erts_thr_progress_active(ErtsSchedulerData *esdp, int on);
+void erts_thr_progress_wakeup(ErtsSchedulerData *esdp,
+ ErtsThrPrgrVal value);
+int erts_thr_progress_update(ErtsSchedulerData *esdp);
+int erts_thr_progress_leader_update(ErtsSchedulerData *esdp);
+void erts_thr_progress_prepare_wait(ErtsSchedulerData *esdp);
+void erts_thr_progress_finalize_wait(ErtsSchedulerData *esdp);
+
+void erts_thr_progress_dbg_print_state(void);
+
+#ifdef ARCH_32
+#define ERTS_THR_PRGR_ATOMIC erts_dw_atomic_t
+ERTS_GLB_INLINE ErtsThrPrgrVal erts_thr_prgr_dw_sint_to_val__(ethr_dw_sint_t *dw_sint);
+#endif
+ERTS_GLB_INLINE ErtsThrPrgrVal erts_thr_prgr_read_acqb__(ERTS_THR_PRGR_ATOMIC *atmc);
+
+ERTS_GLB_INLINE int erts_thr_progress_is_managed_thread(void);
+ERTS_GLB_INLINE ErtsThrPrgrVal erts_thr_progress_later(void);
+ERTS_GLB_INLINE ErtsThrPrgrVal erts_thr_progress_current(void);
+ERTS_GLB_INLINE int erts_thr_progress_has_passed__(ErtsThrPrgrVal val1, ErtsThrPrgrVal val2);
+ERTS_GLB_INLINE int erts_thr_progress_has_reached(ErtsThrPrgrVal val);
+
+#if ERTS_GLB_INLINE_INCL_FUNC_DEF
+
+#ifdef ARCH_64
+
+ERTS_GLB_INLINE ErtsThrPrgrVal
+erts_thr_prgr_read_acqb__(ERTS_THR_PRGR_ATOMIC *atmc)
+{
+ return (ErtsThrPrgrVal) erts_atomic_read_acqb(atmc);
+}
+
+#else /* ARCH_32 */
+
+ERTS_GLB_INLINE ErtsThrPrgrVal
+erts_thr_prgr_dw_sint_to_val__(ethr_dw_sint_t *dw_sint)
+{
+#ifdef ETHR_SU_DW_NAINT_T__
+ return (ErtsThrPrgrVal) dw_sint->dw_sint;
+#else
+ ErtsThrPrgrVal res;
+ res = (ErtsThrPrgrVal) ((Uint32) dw_sint->sint[ETHR_DW_SINT_HIGH_WORD]);
+ res <<= 32;
+ res |= (ErtsThrPrgrVal) ((Uint32) dw_sint->sint[ETHR_DW_SINT_LOW_WORD]);
+ return res;
+#endif
+}
+
+ERTS_GLB_INLINE ErtsThrPrgrVal
+erts_thr_prgr_read_acqb__(ERTS_THR_PRGR_ATOMIC *atmc)
+{
+ ethr_dw_sint_t dw_sint;
+ erts_dw_atomic_read_acqb(atmc, &dw_sint);
+ return erts_thr_prgr_dw_sint_to_val__(&dw_sint);
+}
+
+#endif
+
+ERTS_GLB_INLINE int
+erts_thr_progress_is_managed_thread(void)
+{
+ ErtsThrPrgrData *tpd = erts_tsd_get(erts_thr_prgr_data_key__);
+ return tpd && tpd->is_managed;
+}
+
+ERTS_GLB_INLINE ErtsThrPrgrVal
+erts_thr_progress_later(void)
+{
+ ErtsThrPrgrVal val = erts_thr_prgr_read_acqb__(&erts_thr_prgr__.current);
+ if (val == (ERTS_THR_PRGR_VAL_WAITING-((ErtsThrPrgrVal)2)))
+ return ((ErtsThrPrgrVal) 0);
+ else if (val == (ERTS_THR_PRGR_VAL_WAITING-((ErtsThrPrgrVal)1)))
+ return ((ErtsThrPrgrVal) 1);
+ else
+ return val + ((ErtsThrPrgrVal) 2);
+}
+
+ERTS_GLB_INLINE ErtsThrPrgrVal
+erts_thr_progress_current(void)
+{
+ return erts_thr_prgr_read_acqb__(&erts_thr_prgr__.current);
+}
+
+ERTS_GLB_INLINE int
+erts_thr_progress_has_passed__(ErtsThrPrgrVal val1, ErtsThrPrgrVal val0)
+{
+ if ((((((ErtsThrPrgrVal) 1) << 63) & val1)
+ ^ ((((ErtsThrPrgrVal) 1) << 63) & val0)) != 0) {
+ /* May have wrapped... */
+ if (val1 < (((ErtsThrPrgrVal) 1) << 62)
+ && val0 > (((ErtsThrPrgrVal) 3) << 62)) {
+ /*
+ * 'val1' has wrapped but 'val0' has not yet wrapped. While in
+ * these ranges 'current' is considered later than 'val0'.
+ */
+ return 1;
+ }
+ }
+ return val1 > val0;
+}
+
+ERTS_GLB_INLINE int
+erts_thr_progress_has_reached(ErtsThrPrgrVal val)
+{
+ ErtsThrPrgrVal current;
+ current = erts_thr_prgr_read_acqb__(&erts_thr_prgr__.current);
+ if (current == val)
+ return 1;
+ return erts_thr_progress_has_passed__(current, val);
+}
+
+#endif
+
+#endif /* ERTS_SMP */
+
+#endif