| Age | Commit message (Collapse) | Author |
|---|
|
Introduce a 'rename' instruction that can be used to optimize
simple renaming with unchanged operands such as:
get_tuple_element Reg P Dst => i_get_tuple_element Reg P Dst
By allowing it to lower the arity of instruction, transformations
such as the following can be handled:
trim N Remaining => i_trim N
All in all, currently 67 transformations can be optimized in this
way, including some commonly used ones.
|
|
Generic instructions have a min_window field. Its purpose is to
avoid calling transform_engine() when there are too few instructions
in the current "transformation window" for a transformation to
succeed.
Currently it does not do much good since the window size will be
decremented by one before being used. The reason for the subtraction
is probably that in some circumstances in the past, the loader could
read past the end of the BEAM module while attempting to fetch
instructions to increase the window size. Therefore, it would not
be safe to just remove the subtraction by one.
The simplest and safest solution seems to always ensure that there
are always at least TWO instructions when calling transform_engine().
That will be safe, as long as a BEAM module is always finished with
an int_code_end/0 that is not involved in any transformation.
|
|
When an instruction with a variable number operands (such as
select_val) is seen of the left side of a transformation, the
'next_arg' instruction will allocate a buffer to fit all variables and
all operands will be copied into the buffer. Very often, the 'commit'
instruction will never be reached because of a test or predicate
failing or because of a short window; in that case, the variable
buffer will be deallocated.
Note that originally there were only few instructions with a variable
number of operands, but now common operations such as tuple building
also have a variable number of operands.
To avoid those frequent allocations and deallocations, modify the
'next_arg' instruction to only save a pointer to the first of the
"rest" arguments. Also move the deallocation of the instructions
on the left side from the 'commit' instruction to the 'end'
instruction to ensure that 'store_rest_args' will still work.
|
|
This is an optimization for reducing the number of heap fragments
allocated when sending a message where the majority of the
message payload is on the sending process' heap.
|
|
* egil/erts/tracing-beam-lttng/OTP-10282:
erts: Don't use ratio in carrier lttng tracepoints
Add lttng testcases
erts: Extend erlang:system_info/1 with lttng
Refactor and fix dtrace define in erl_message
erts: Add lttng tracepoints for async pool queue
erts: Add lttng tracepoints for drivers
erts: Add lttng tracepoints for scheduler events
erts: Add lttng tracepoints for memory carriers
erts: Update lttng-wrapper with mfa conversion
erts: Teach lttng to configure and build system
|
|
|
|
Let erlang:system_info(dynamic_trace) be able to return 'lttng' if enabled.
|
|
|
|
* aio_pool_get
* aio_pool_add
|
|
* driver_event
* driver_flush
* driver_finish
* driver_init
* driver_output
* driver_outputv
* driver_process_exit
* driver_ready_async
* driver_ready_input
* driver_ready_output
* driver_start
* driver_stop
* driver_stop_select
* driver_timeout
|
|
* scheduler_poll
|
|
* carrier_create
* carrier_destroy
* carrier_pool_put
* carrier_pool_get
|
|
|
|
Introduce a wrapper API for lttng.
|
|
We used to set last_op_next and last_op to NULL just in case.
Setting last_op_next to causes a rescan of the instructions
to find the last instruction in the chain, so we would want
to avoid that unless really necessary.
|
|
The transformation on the following line will do the job.
|
|
=== OTP-18.3.1 ===
Changed Applications:
- erts-7.3.1
- inets-6.2.1
- mnesia-4.13.4
Unchanged Applications:
- asn1-4.0.2
- common_test-1.12
- compiler-6.0.3
- cosEvent-2.2
- cosEventDomain-1.2
- cosFileTransfer-1.2
- cosNotification-1.2.1
- cosProperty-1.2
- cosTime-1.2.1
- cosTransactions-1.3.1
- crypto-3.6.3
- debugger-4.1.2
- dialyzer-2.9
- diameter-1.11.2
- edoc-0.7.18
- eldap-1.2.1
- erl_docgen-0.4.2
- erl_interface-3.8.2
- et-1.5.1
- eunit-2.2.13
- gs-1.6
- hipe-3.15
- ic-4.4
- jinterface-1.6.1
- kernel-4.2
- megaco-3.18
- observer-2.1.2
- odbc-2.11.1
- orber-3.8.1
- os_mon-2.4
- ose-1.1
- otp_mibs-1.1
- parsetools-2.1.1
- percept-0.8.11
- public_key-1.1.1
- reltool-0.7
- runtime_tools-1.9.3
- sasl-2.7
- snmp-5.2.2
- ssh-4.2.2
- ssl-7.3
- stdlib-2.8
- syntax_tools-1.7
- test_server-3.10
- tools-2.8.3
- typer-0.9.10
- webtool-0.9.1
- wx-1.6.1
- xmerl-1.3.10
Conflicts:
OTP_VERSION
erts/emulator/test/save_calls_SUITE.erl
erts/vsn.mk
|
|
* bjorn/erts/clang-opt:
Fix unsafe transformation of apply/3 with fixed arguments
|
|
* egil/erts/fix-flatmap-get/OTP-13459:
erts: Don't search for non-existing Map keys twice
|
|
* sverker/erts/trap_exit-race/OTP-13452:
erts: Fix race for process_flag(trap_exit,true)
|
|
* rickard/proc-free-fix/OTP-13446:
Fix bad refc management of process struct
# Conflicts:
# erts/emulator/beam/erl_process.c
|
|
* rickard/port-sig-dropped-fix/OTP-13424:
Fix implementation of dropped signal to port
|
|
* rickard/last_calls/OTP-13418:
Unbreak process_info(Pid,last_calls)
|
|
and a concurrent exit signal.
We now actually guarantee that the process will not die
from exit signal *after* the call to process_flag(trap_exit,true)
has returned.
The race is narrow and probably quite hard to observe even if you
manage to provoke it. Has only been confirmed with the help of
return trace and a sleep in send_exit_signal().
Solution:
Seize status lock to prevent send_exit_signal() from reading
an old status (without TRAP_EXIT) and then writing PENDING_EXIT
after TRAP_EXIT has been set by process_flag_2().
|
|
* For maps:get/2,3 and maps:find/2, searching for an immediate key, e.g. an atom,
the search was performed twice if the key did not exist in the map.
|
|
* rickard/proc-free-fix/OTP-13446:
Fix bad refc management of process struct
Conflicts:
erts/emulator/beam/erl_process.c
|
|
|
|
62473daf introduced an unsafe optimization in the loader.
See the comments in the test case for an explanation of
the problem.
|
|
* lukas/erts/enif_trace_functions/OTP-13442:
erts: Fix bug in enif_send
erts: Fix windows nif port tests
erts: Fix bug in enif_term_to_binary
erts: Remove printout when dec_term fails in DEBUG
erts: Polish erl_nif docs
erts: Improve enif_binary_to_term
erts: Add enif_port_command
erts: Add enif_term_to_binary and enif_binary_to_term
erts: Add enif_is_process/port_alive
erts: Add enif_cpu/now_time and enif_make_unique_integer
|
|
Let cache_env() set env->heap_frag to same as MBUF(p)
as it is in any other case.
|
|
Wait until after dec_term and factory_close to do cache_env(),
otherwise we will cache the wrong state.
|
|
|
|
* Accept a raw data buffer instead of ErlNifBinary
* Accept option ERL_NIF_BIN2TERM_SAFE
* Return number of read bytes
|
|
|
|
|
|
|
|
|
|
* rickard/improve-psd/OTP-13428:
Improve process/port specific data management
|
|
* rickard/delayed-gc-fix:
Allow delayed gc while scheduled out
|
|
* rickard/port-sig-dropped-fix/OTP-13424:
Fix implementation of dropped signal to port
|
|
* lukas/erts/inline_tagvaldef/OTP-13440:
erts: inline tag_val_def
|
|
The tag_val_def function was called and multiple switch statements
had to be traversed in term.c, and then a big switch in the calling
code to branch on the term types. By inlining the switches are
merged by the compiler and a lot fewer branches have to be taken.
Benchmarks show that this increases performance of enc_term by as
much as 10%.
|
|
* lukas/erts/low_write_freq_darwin/OTP-12345:
erts: Fix LOW_WRITE section for non llvm os x compilation
|
|
|
|
|
|
|
|
* lukas/erts/fix_scheduler_suspend/ERL-94/PR-978/OTP-13425:
erts: Fix install of suspend handler
Conflicts:
erts/emulator/sys/unix/erl_unix_sys.h
erts/emulator/sys/unix/sys.c
|
|
|
|
|
|
|
