remove rb_binding_new_with_cfp, and use rb_vm_make_binding instead.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48001 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Remove some unnecessary VALUE/struct conversions and aggressively inline
functions used during fiber_switch. Either of these changes alone does
not yield significant performance increase, but in combination they
improve performance by ~6%.
Arguably, removal of separate VALUE/rb_fiber_t* variables also makes the
code more readable in a few places.
* vm_core.h: declare rb_fiber_t typedef
(rb_thread_t): fiber and root_fiber become rb_fiber_t * (from VALUE)
* vm.c (rb_thread_mark): use rb_fiber_mark_self
* cont.c (rb_fiber_t): prev becomes rb_fiber_t * (from VALUE)
(cont_mark, cont_free): simplify conditions
(rb_fiber_mark_self): new function
(fiber_mark): use rb_fiber_mark_self
(cont_save_thread, cont_restore_thread): inline
(cont_restore_thread): simplify
(fiber_setcontext): simplify conditions
(rb_cont_call): remove dereference
(fiber_t_alloc): update for rb_fiber_t->prev type change
(rb_fiber_start): ditto
(fiber_current): extract from rb_fiber_current
(return_fiber): move, simplify type checks
(rb_fiber_current): use fiber_current
(fiber_store): simplify type checks
(fiber_switch): ditto, simplify call to fiber_setcontext,
use fiber_current
(rb_fiber_transfer): update for type changes
(rb_fiber_terminate): move, use fiber_switch
(rb_fiber_resume): update for type changes
(rb_fiber_reset_root_local_storage): ditto
(rb_fiber_yield): use rb_fiber_switch instead of rb_fiber_transfer
(rb_fiber_m_transfer): ditto
[ruby-core:65518] [Feature #10341]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47964 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* proc.c (rb_proc_alloc): inline and move to vm.c
(rb_proc_wrap): new wrapper function used by rb_proc_alloc
(proc_dup): simplify alloc + copy + wrap operation
[ruby-core:64994]
* vm.c (rb_proc_alloc): new inline function
(rb_vm_make_proc): call rb_proc_alloc
* vm_core.h: remove rb_proc_alloc, add rb_proc_wrap
* benchmark/bm_vm2_newlambda.rb: short test to show difference
First we allocate and populate an rb_proc_t struct inline to avoid
unnecessary zeroing of the large struct. Inlining speeds up callers as
this takes many parameters to ensure correctness. We then call the new
rb_proc_wrap function to create the object.
rb_proc_wrap - wraps a rb_proc_t pointer as a Ruby object, but
we only use it inside rb_proc_alloc. We must call this before
the compiler may clobber VALUE parameters passed to rb_proc_alloc.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47562 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
We may tag the running_thread pointer to avoid making the "once" struct
bigger than "struct iseq_inline_cache_entry".
This only saves a small amount with "valgrind ruby -e exit"
before:
total heap usage: 48,122 allocs, 19,248 frees, 8,110,149 bytes allocated
after:
total heap usage: 48,122 allocs, 19,253 frees, 8,099,197 bytes allocated
* insns.def (once): define and use fake RUNNING_THREAD_ONCE_DONE
pointer to indicate is->once.running_thread is done.
* vm_core.h (iseq_inline_storage_entry): remove done field,
allowing the union to be reduced from 24=>16 bytes on 64-bit
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47542 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
make and register special exceptions.
* vm.c (rb_vm_mark): do not need to mark special exceptions
because they are registerd by rb_gc_register_mark_object().
* eval.c (Init_eval): use rb_vm_register_special_exception().
* gc.c (Init_GC): ditto.
* proc.c (Init_Proc): ditto.
* thread.c (Init_Thread): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47534 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This keeps ci->flag and ci->aux.index consistent across 32-bit
and 64-bit platforms.
ci->flag: VM_CALL_* flags only use 9 bits, currently
ci->aux.index: 2 billion ivars per class should be enough for anybody
This saves around 50K allocations on "valgrind ruby -e exit" on x86-64
before:
total heap usage: 48,122 allocs, 19,253 frees, 8,099,197 bytes allocated
after:
total heap usage: 48,069 allocs, 19,214 frees, 8,047,266 bytes allocated
* vm_core.h (rb_call_info_t): ci->flag becomes 32-bit unsigned int
ci->index becomes a 32-bit signed int (from signed long).
Reorder for better packing on 64-bit, giving an 8 byte reduction
from 104 to 96 bytes for each ci.
* compile.c (new_callinfo, setup_args, iseq_compile_each,
iseq_build_from_ary_body): adjust for type changes
* vm_insnhelper.c (vm_getivar): ditto
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47509 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Nearly all current uses of iseq->location.first_lineno are as a
VALUE, not a size_t. The only exception was the experimental
(and currently unused) rb_iseq_build_for_ruby2cext function.
* vm_core.h (rb_iseq_location_t): change first_lineno type to VALUE
* iseq.c (rb_iseq_build_for_ruby2cext): update based on argument
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47269 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
iseq_size and line_info_size may be 32-bit.
4GB instruction sequences should be big enough for anyone.
Other existing line info counters are 32-bit, and nobody
should need all that.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46957 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_core.h (rb_proc_t): reduce to 64 bytes from 72 on 64-bit
This allows rb_proc_t to fit inside a single cache line on x86-64
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46886 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* enum.c (enum_any): optimize object allocations for Array and
Hash when `each` is not redefined, always false if empty and the
case without a block. [fix GH-617]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46859 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
because all calculations related to stack_max in compile.c
(iseq_set_sequence) and vm_insnhelper.c (vm_push_frame) are
conducted by using int. This partly reverts r23945.
* vm_insnhelper.c (vm_push_frame): ditto. This reverts r42401.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46839 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
rb_vm_t::trap_list_cmds (an array) and
rb_vm_t::trap_list_safes[RUBY_NSIG]
(separate to two different array).
This modification reduce root objects.
* signal.c: ditto.
* vm.c (rb_vm_mark): remove marking code for rb_vm_t::trap_list.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46834 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_core.h (struct rb_iseq_struct): reduce to 288 bytes on 64-bit
(found with pahole, from the dwarves package)
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46814 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Most iseq do not have a catch_table, so avoid needlessly adding
4-8 bytes to the struct for the common case.
Changes from v2:
- iseq_catch_table_size removed, use if (...) for (;...;)
Changes from v1:
- renamed iseq->_catch_table to iseq->catch_table
- iseq_catch_table_bytes: made a static inline function
- iseq_catch_table_size: new function replaces the
iseq_catch_table_each iterator macro
* iseq.h (struct iseq_catch_table): new flexible array struct
(iseq_catch_table_bytes): allocated size function
* vm_core.h (struct rb_iseq_struct): uupdate catch_table member
* compile.c (iseq_set_exception_table): update for struct changes
* iseq.c (iseq_free): ditto
* iseq.c (iseq_memsize): ditto
* iseq.c (rb_iseq_disasm): ditto
* iseq.c (iseq_data_to_ary): ditto
* iseq.c (rb_iseq_build_for_ruby2cext): ditto (untested)
* vm.c (vm_exec): ditto
* vm_core.h (struct rb_iseq_struct): ditto
* vm_insnhelper.c (vm_throw): ditto
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46811 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_core.h (struct rb_call_info_struct): improve packing
On x86-64, it goes from 112 to 104 bytes. This results in
~20K of savings from just "./ruby -e exit"
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46800 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_core.h (struct rb_iseq_struct): stack_max is uint32_t
No program will ever need more than a few megabytes of stack,
so there's no sense in using a 64-bit counter for stack
accounting. Packing this with the 32-bit type enum reduces
rb_iseq_struct from 312 to 304 bytes on 64-bit systems.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46798 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm.c (rb_vm_env_local_variables): returns array of local
variable name symbols in the environment by envval.
* proc.c (bind_local_variables): use rb_vm_env_local_variables.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46648 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* configure.in (rb_cv_scalar_pthread_t): pthread_t is not required
to be a scalar type.
* thread.c (fill_thread_id_string, thread_id_str): dump pthread_t
in hexadecimal form if it is not a scalar type, assume it can be
represented in a pointer form otherwise. based on the patch by
Rei Odaira at [ruby-core:62867]. [ruby-core:62857] [Bug #9884]
* thread_pthread.c (Init_native_thread, thread_start_func_1),
(native_thread_create): set thread_id_str if needed.
* vm_core.h (rb_thread_t): add thread_id_string if needed.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46406 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* error.c (rb_bug_context): new function to report bug with
context.
* vm_dump.c (rb_vm_bugreport): accepts `ucontext_t` argument to
dump machine regisiters. based on [GH-584].
* signal.c (sigbus, sigsegv): dump machine regisiters if available.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46106 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
A doubly-linked list for tracking living threads guarantees
constant-time insert/delete performance with no corner cases of a
hash table. I chose this ccan implementation of doubly-linked
lists over the BSD sys/queue.h implementation since:
1) insertion and removal are both branchless
2) locality is improved if a struct may be a member of multiple lists
(0002 patch in Feature 9632 will introduce a secondary list
for waiting FDs)
This also increases cache locality during iteration: improving
performance in a new IO#close benchmark with many sleeping threads
while still scanning the same number of threads.
vm_thread_close 1.762
* vm_core.h (rb_vm_t): list_head and counter for living_threads
(rb_thread_t): vmlt_node for living_threads linkage
(rb_vm_living_threads_init): new function wrapper
(rb_vm_living_threads_insert): ditto
(rb_vm_living_threads_remove): ditto
* vm.c (rb_vm_living_threads_foreach): new function wrapper
* thread.c (terminate_i, thread_start_func_2, thread_create_core,
thread_fd_close_i, thread_fd_close): update to use new APIs
* vm.c (vm_mark_each_thread_func, rb_vm_mark, ruby_vm_destruct,
vm_memsize, vm_init2, Init_VM): ditto
* vm_trace.c (clear_trace_func_i, rb_clear_trace_func): ditto
* benchmark/bm_vm_thread_close.rb: added to show improvement
* ccan/build_assert/build_assert.h: added as a dependency of list.h
* ccan/check_type/check_type.h: ditto
* ccan/container_of/container_of.h: ditto
* ccan/licenses/BSD-MIT: ditto
* ccan/licenses/CC0: ditto
* ccan/str/str.h: ditto (stripped of unused macros)
* ccan/list/list.h: ditto
* common.mk: add CCAN_LIST_INCLUDES
[ruby-core:61871][Feature 9632 (part 1)]
Apologies for the size of this commit, but I think a good
doubly-linked list will be useful for future features, too.
This may be used to add ordering to a container_of-based hash
table to preserve compatibility if required (e.g. feature 9614).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45913 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
it is bmethod frame.
* vm.c (vm_exec): invoke RUBY_EVENT_RETURN event if rollbacked frame
is VM_FRAME_FLAG_BMETHOD.
[Bug #9759]
* test/ruby/test_settracefunc.rb: add a test for TracePoint/set_trace_func.
* vm_core.h: renmae rb_thread_t::passed_me to
rb_thread_t::passed_bmethod_me to clarify the usage.
* vm_insnhelper.c (vm_call_bmethod_body): use renamed member.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45758 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_core.h (rb_thread_struct): aggregate cpu stuff into a struct,
so that a debugger can show its content at once.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44722 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
vm_insnhelper.c, vm_insnhelper.h, vm_method.c: Rename method_serial
to global_method_state and constant_serial to global_constant_state
after discussion with ko1.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44097 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
PC is cached into local stack and cfp->pc is incorrent at next of
branch or jump.
* vm_exec.h (DEBUG_ENTER_INSN): catch up this change.
* vm_core.h: update signature of rb_vmdebug_debug_print_pre.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43916 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* internal.h (ruby_register_rollback_func_for_ensure): catch up above change.
Add rollback mechanism API.
* vm_core.h (typedef struct rb_vm_struct): catch up above change.
Introdule ensure-rollback relation table.
* vm_core.h (typedef struct rb_thread_struct): catch up above change.
Introdule ensure stack.
* eval.c (rb_ensure): catch up above change.
Introdule ensure stack.
* hash.c : New function for rollback ensure, and register it to
ensure-rollback relation table. [ruby-dev:47803] [Bug #9105]
Ensure Rollback Mechanism:
A rollback's function is a function to rollback a state before ensure's
function execution.
When the jump of callcc is across the scope of rb_ensure,
ensure's functions and rollback's functions are executed appropriately
for keeping consistency.
Current API is unstable, and only internal use.
ruby_register_rollback_func_for_ensure(ensure_func,rollback_func)
This API create relation ensure's function to rollback's function.
By registered rollback's function, it is executed When jumpping into
corresponding rb_ensure scope.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43688 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
ic_serial. This is possible because these fields are only ever used
exclusively with each other.
* insns.def: ditto
* vm_core.h: ditto
* vm_insnhelper.c: ditto
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43596 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
ruby_vm_global_state_version into two separate counters - one for the
global method state and one for the global constant state. This means
changes to constants do not affect method caches, and changes to
methods do not affect constant caches. In particular, this means
inclusions of modules containing constants no longer globally
invalidate the method cache.
* class.c, eval.c, include/ruby/intern.h, insns.def, vm.c, vm_method.c:
rename rb_clear_cache_by_class to rb_clear_method_cache_by_class
* class.c, include/ruby/intern.h, variable.c, vm_method.c: add
rb_clear_constant_cache
* compile.c, vm_core.h, vm_insnhelper.c: rename vmstat field in
rb_call_info_struct to method_state
* vm_method.c: rename vmstat field in struct cache_entry to method_state
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43455 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm.c, vm_core.h (rb_vm_add_root_module): added to register as a
defined root module or class.
This guard helps mark miss from defined classes/modules they are
only refered from C's global variables in C-exts.
Basically, it is extension's bug.
Register to hash object VM has.
Marking a hash objects allows generational GC supports.
* gc.c (RGENGC_PRINT_TICK): disable (revert).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43263 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
ktsj
akr
normal
ko1
nobu
ngoto
charliesome
tarui
shyouhei