Age | Commit message (Collapse) | Author |
|
* maint:
Handle addition of bignum + variable in unoptimized code
|
|
* bjorn/erts/unoptimized-plus:
Handle addition of bignum + variable in unoptimized code
|
|
|
|
|
|
Seems the right thing to do as the compiled code is not necessarily
executed by a single CPU thread.
Only warning if old gcc and will work as good/bad as it did before.
|
|
|
|
|
|
Wake AIX up from coma
|
|
* kjell/stdlib/ets_ordered_set_slow_react/OTP-15906:
ETS ordered_set: Improvements to the CA tree implementation
|
|
This commit only affects the implementation of ETS `ordered_set`
tables with the `write_concurrency` option enabled. Such tables are
implemented with a data structure that is called the contention
adapting search tree (CA tree). This commit introduces the following
changes:
* This commit causes a join to be triggered in one randomly selected
base node in about one of 1000 read unlock calls for base node
locks. No such joins happened before this commit. Before this
commit, operations that only acquired looks in read-mode never
triggered any contention adaptation. Therefore, the CA tree could
get stuck in a sub-optimal state in certain scenarios. This could
happen, for example, when a CA tree is first populated with parallel
inserts (which will cause splits of base nodes) and then only
read-only operations are applied to the data structure. Benchmark
results from the
`ets_SUITE:lookup_catree_par_vs_seq_init_benchmark/0` benchmark
function (which is included in this commit) shows that this change
can improve the throughput of the CA tree in the scenario described
above.
* Read-only operations will now also increase values of statistics
counters when they detect that they need to wait for other
operations. Only write operation changed statistics counters before
this commit. This improves the statistics that the adaptation
heuristics is based on.
* Additionally, this commit adds an upper and lower limit to the
contention statistics variables in the base nodes. Such limits did
not exist before this commit. This should, for example, make the CA
tree more responsive to contention after long periods of low
contention.
|
|
|
|
Return real pend value in erlang:dist_get_stat/1
|
|
|
|
garazdawi/lukas/erts/fix_active_n_close_win32/ERL-960/OTP-15901
Fix {active,N} close race condition on windows
|
|
* maint:
erts: Relax the constraint introduced in OTP-15871
|
|
* john/erts/fix-conservative/OTP-15881:
erts: Relax the constraint introduced in OTP-15871
|
|
* maint:
Don't disconnect on remote exit/2 with old incarnation as recipient
|
|
* rickard/dist-exit2/22/OTP-15867:
Don't disconnect on remote exit/2 with old incarnation as recipient
|
|
* bjorn/better-fun-info/OTP-15837:
Create a shared wrapper function for all occurrences of 'fun F/A'
Support sharing of fun entries in the runtime system
erl_fun.c: Remove unused struct definition
hipe: Use the new index when translating funs
Stop supporting decoding of old funs in the external term format
genop.tab: Insert an "OTP 23" comment for clarity
|
|
|
|
ingela/merge-294
* ingela/raimo/udp-send-TOS/ERIERL-294/OTP-15747:
Introduce udp send ancillary data argument down to inet_drv
Fix old warnings
|
|
rickard/dist-exit2/22/OTP-15867
* rickard/dist-exit2/21/OTP-15867:
Don't disconnect on remote exit/2 with old incarnation as recipient
|
|
rickard/dist-exit2/21/OTP-15867
* rickard/dist-exit2/20/OTP-15867:
Don't disconnect on remote exit/2 with old incarnation as recipient
|
|
|
|
|
|
|
|
|
|
Before OTP 23, there had to be a one-to-one correspodence between
`make_fun2` instructions and fun entries.
In the OTP 23 runtime system, start keying the fun entries with the
explicit ("new") index included in the fun entry. That will allow
multiple `make_fun2` instructions to refer to the same entry in
the fun table.
This change is safe for code compiled by OTP R15B or later, because
the old and new indices are identical. The loader will refuse to
load code where the old and new indices are not equal unless the
code is known to be compiled with an OTP 23 compiler.
|
|
|
|
The new external format for funs (NEW_FUN_EXT) was introduced in OTP
R8 along with the `make_fun2` instruction. Therefore, it should be
safe to stop decoding the old FUN_EXT tag.
|
|
|
|
|
|
When geting the local address, make sure we don't accept
the loopback address.
|
|
|
|
into maint
|
|
on non-Linux with gcc < 4.3
hipe/hipe_arm.c:65: #error "Don't know how to flush instruction cache"
This is benign because "make depend" runs preprocessor on more files
than actually needed.
|
|
|
|
Also handles code compiled by OTP 21 and earlier.
|
|
|
|
* raimo/incomplete-socket-close/ERIERL-353/OTP-15370:
Copy linger zero flag from listen socket
Test linger zero flag from listen socket
|
|
'sverker/erts/process-info-reductions-idle-proc/ERL-964/OTP-15865' into maint-22
* sverker/erts/process-info-reductions-idle-proc/ERL-964/OTP-15865:
erts: Improve test of process_info(reductions)
Revert "erts: Force process_info(reductions) as signal"
|
|
* sverker/erts/break-p-segv/ERL-965/OTP-15873:
erts: Fix SEGV crash on shell break (p)roc info
|
|
This commit only affects the implementation of ETS `ordered_set`
tables with the `write_concurrency` option enabled. Such tables are
implemented with a data structure that is called the contention
adapting search tree (CA tree). This commit introduces the following
changes:
* This commit causes a join to be triggered in one randomly selected
base node in about one of 1000 read unlock calls for base node
locks. No such joins happened before this commit. Before this
commit, operations that only acquired looks in read-mode never
triggered any contention adaptation. Therefore, the CA tree could
get stuck in a sub-optimal state in certain scenarios. This could
happen, for example, when a CA tree is first populated with parallel
inserts (which will cause splits of base nodes) and then only
read-only operations are applied to the data structure. Benchmark
results from the
`ets_SUITE:lookup_catree_par_vs_seq_init_benchmark/0` benchmark
function (which is included in this commit) shows that this change
can improve the throughput of the CA tree in the scenario described
above.
* Read-only operations will now also increase values of statistics
counters when they detect that they need to wait for other
operations. Only write operation changed statistics counters before
this commit. This improves the statistics that the adaptation
heuristics is based on.
* Additionally, this commit adds an upper and lower limit to the
contention statistics variables in the base nodes. Such limits did
not exist before this commit. This should, for example, make the CA
tree more responsive to contention after long periods of low
contention.
|
|
* maint:
erts: Fix bad loader optimization of get_tuple_element
beam_ssa_type: Fix incorrect bitstring unit determination
|
|
into maint
* john/erts/fix-bad-get_tuple_element-opt/OTP-15871/ERIERL-374:
erts: Fix bad loader optimization of get_tuple_element
|
|
|
|
OTP-15874
PR-2266
* sverker/erts/hipe-arm-freebsd-clear-cache/ERL-958:
erts: Fix hipe_flush_icache_range for FreeBSD on ARM
|
|
Avoid old Linux specific syscall assembler if we can.
|
|
|
|
'sverker/erts/process-info-reductions-idle-proc/ERL-964/OTP-15865' into maint
* sverker/erts/process-info-reductions-idle-proc/ERL-964/OTP-15865:
erts: Improve test of process_info(reductions)
Revert "erts: Force process_info(reductions) as signal"
|