When loading Ruby source files, we can save the result of
successful opens as open(2)/openat(2) are a fairly expensive
syscalls. This also avoids a time-of-check-to-time-of-use
(TOCTTOU) problem.
This reduces open(2) syscalls during `require'; but should be
most apparent when users have a small $LOAD_PATH. Users with
large $LOAD_PATH will benefit less since there'll be more
open(2) failures due to ENOENT.
With `strace -c -e openat ruby -e exit' under Linux, this
results in a ~14% reduction of openat(2) syscalls
(glibc uses openat(2) to implement open(2)).
% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
0.00 0.000000 0 296 110 openat
0.00 0.000000 0 254 110 openat
Additionally, the introduction of `struct ruby_file_load_state'
may make future optimizations more apparent.
This change cannot benefit binary (.so) loading since the
dlopen(3) API requires a filename and I'm not aware of an
alternative that takes a pre-existing FD. In typical
situations, Ruby source files outnumber the mount of .so
files.
If the previous instruction is not a leaf instruction, then the PC was
incremented before the instruction was ran (meaning the currently
executing instruction is actually the previous instruction), so we
should not increment the PC otherwise we will calculate the source
line for the next instruction.
This bug can be reproduced in the following script:
```
require "objspace"
ObjectSpace.trace_object_allocations_start
a =
1.0 / 0.0
p [ObjectSpace.allocation_sourceline(a), ObjectSpace.allocation_sourcefile(a)]
```
Which outputs: [4, "test.rb"]
This is incorrect because the object was allocated on line 10 and not
line 4. The behaviour is correct when we use a leaf instruction (e.g.
if we replaced `1.0 / 0.0` with `"hello"`), then the output is:
[10, "test.rb"].
[Bug #19456]
Given that signleton classes don't have an allocator,
we can re-use these bytes to store the attached object
in `rb_classext_struct` without making it larger.
Right now the attached object is stored as an instance variable
and all the call sites that either get or set it have to know how it's
stored.
It's preferable to hide this implementation detail behind accessors
so that it is easier to change how it's stored.
This patch is follo-up of 0a82bfe.
Without this patch, if env is escaped (Proc'ed), strange svar
can be touched.
This patch tracks escaped env and use it.
[Bug #19415]
If multiple threads attemps to load the same file concurrently
it's not a circular dependency issue.
So we check that the existing ThreadShield is owner by the current
fiber before warning about circular dependencies.
[Bug #19415]
If multiple threads attemps to load the same file concurrently
it's not a circular dependency issue.
So we check that the existing ThreadShield is owner by the current
fiber before warning about circular dependencies.
* use correct svar
Without this patch, svar location is used "nearest Ruby frame".
It is almost correct but it doesn't correct when the `each` method
is written in Ruby.
```ruby
class C
include Enumerable
def each
%w(bar baz).each{|e| yield e}
end
end
C.new.grep(/(b.)/){|e| p [$1, e]}
```
This patch fix this issue by traversing ifunc's cfp.
Note that if cfp doesn't specify this Thread's cfp stack, reserved
svar location (`ec->root_svar`) is used.
* make yjit-bindgen
---------
Co-authored-by: Takashi Kokubun <takashikkbn@gmail.com>
Instance variables held in gen_ivtbl are marked with rb_gc_mark. It
prevents the referenced objects from moving, which is bad for copying
garbage collectors.
This commit allows those instance variables to be updated during
gc_update_object_references.
This commit adds rb_gc_mark_and_move which takes a pointer to an object
and marks it during marking phase and updates references during compaction.
This allows for marking and reference updating to be combined into a
single function, which reduces code duplication and prevents bugs if
marking and reference updating goes out of sync.
This commit also implements rb_gc_mark_and_move on iseq as an example.
This commit moves the classpath (and tmp_classpath) from instance
variables to the rb_classext_t. This improves performance as we no
longer need to set an instance variable when assigning a classpath to
a class.
I benchmarked with the following script:
```ruby
name = :MyClass
puts(Benchmark.measure do
10_000_000.times do |i|
Object.const_set(name, Class.new)
Object.send(:remove_const, name)
end
end)
```
Before this patch:
```
5.440119 0.025264 5.465383 ( 5.467105)
```
After this patch:
```
4.889646 0.028325 4.917971 ( 4.942678)
```
On a hash miss we need to call default if it is redefined in order to
return the default value to be used. Previously we checked this with
rb_method_basic_definition_p, which avoids the method call but requires
a method lookup.
This commit replaces the previous check with BASIC_OP_UNREDEFINED_P and
a new BOP_DEFAULT. We still need to fall back to
rb_method_basic_definition_p when called on a subclasss of hash.
| |compare-ruby|built-ruby|
|:---------------|-----------:|---------:|
|hash_aref_miss | 2.692| 3.531|
| | -| 1.31x|
Co-authored-by: Daniel Colson <danieljamescolson@gmail.com>
Co-authored-by: "Ian C. Anderson" <ian@iancanderson.com>
Co-authored-by: Jack McCracken <me@jackmc.xyz>
SIZE_POOL_COUNT is a GC macro, it should belong in gc.h and not shape.h.
SIZE_POOL_COUNT doesn't depend on shape.h so we can have shape.h depend
on gc.h.
Co-Authored-By: Matt Valentine-House <matt@eightbitraptor.com>
When an object becomes "too complex" (in other words it has too many
variations in the shape tree), we transition it to use a "too complex"
shape and use a hash for storing instance variables.
Without this patch, there were rare cases where shape tree growth could
"explode" and cause performance degradation on what would otherwise have
been cached fast paths.
This patch puts a limit on shape tree growth, and gracefully degrades in
the rare case where there could be a factorial growth in the shape tree.
For example:
```ruby
class NG; end
HUGE_NUMBER.times do
NG.new.instance_variable_set(:"@unique_ivar_#{_1}", 1)
end
```
We consider objects to be "too complex" when the object's class has more
than SHAPE_MAX_VARIATIONS (currently 8) leaf nodes in the shape tree and
the object introduces a new variation (a new leaf node) associated with
that class.
For example, new variations on instances of the following class would be
considered "too complex" because those instances create more than 8
leaves in the shape tree:
```ruby
class Foo; end
9.times { Foo.new.instance_variable_set(":@uniq_#{_1}", 1) }
```
However, the following class is *not* too complex because it only has
one leaf in the shape tree:
```ruby
class Foo
def initialize
@a = @b = @c = @d = @e = @f = @g = @h = @i = nil
end
end
9.times { Foo.new }
``
This case is rare, so we don't expect this change to impact performance
of most applications, but it needs to be handled.
Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org>
Count how many "variations" each class creates. A "variation" is a a
unique ordering of instance variables on a particular class. This can
also be thought of as a branch in the shape tree.
For example, the following Foo class will have 2 variations:
```ruby
class Foo ; end
Foo.new.instance_variable_set(:@a, 1) # case 1: creates one variation
Foo.new.instance_variable_set(:@b, 1) # case 2: creates another variation
foo = Foo.new
foo.instance_variable_set(:@a, 1) # does not create a new variation
foo.instance_variable_set(:@b, 1) # does not create a new variation (a continuation of the variation in case 1)
```
We will use this number to limit the amount of shapes that a class can
create and fallback to using a hash iv lookup.
Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org>
We want to hint to the compiler that it's likely that the BOP is
unredefined (the bit is 0). Previously we were accidentally hinting to
the compiler that it was non-zero due to a misplaced parenthesis.
Prior to this commit the `OPTIMIZED_CMP` macro relied on a method lookup
to determine whether `<=>` was overridden. The result of the lookup was
cached, but only for the duration of the specific method that
initialized the cmp_opt_data cache structure.
With this method lookup, `[x,y].max` is slower than doing `x > y ?
x : y` even though there's an optimized instruction for "new array max".
(John noticed somebody a proposed micro-optimization based on this fact
in https://github.com/mastodon/mastodon/pull/19903.)
```rb
a, b = 1, 2
Benchmark.ips do |bm|
bm.report('conditional') { a > b ? a : b }
bm.report('method') { [a, b].max }
bm.compare!
end
```
Before:
```
Comparison:
conditional: 22603733.2 i/s
method: 19820412.7 i/s - 1.14x (± 0.00) slower
```
This commit replaces the method lookup with a new CMP basic op, which
gives the examples above equivalent performance.
After:
```
Comparison:
method: 24022466.5 i/s
conditional: 23851094.2 i/s - same-ish: difference falls within
error
```
Relevant benchmarks show an improvement to Array#max and Array#min when
not using the optimized newarray_max instruction as well. They are
noticeably faster for small arrays with the relevant types, and the same
or maybe a touch faster on larger arrays.
```
$ make benchmark COMPARE_RUBY=<master@5958c305> ITEM=array_min
$ make benchmark COMPARE_RUBY=<master@5958c305> ITEM=array_max
```
The benchmarks added in this commit also look generally improved.
Co-authored-by: John Hawthorn <jhawthorn@github.com>
This commit moves ruby_basic_operators and the unredefined macros out of
vm_core.h and into basic_operators.h so that we can use them more
broadly in places where we currently use a method look up via
`rb_method_basic_definition_p` (e.g. object.c, numeric.c, complex.c,
enum.c, but also in internal/compar.h after introducing BOP_CMP and
elsewhere if we introduce more BOPs)
The most controversial part of this change is probably moving
redefined_flag out of rb_vm_t. [vm_opt_method_def_table and
vm_opt_mid_table](9da2a5204f/vm.c)
are not part of rb_vm_t either, and I think this fits well with those.
But more significantly it seems to result in one fewer instruction. For
example:
Before:
```
(lldb) disassemble -n vm_opt_str_freeze
miniruby`vm_exec_core:
miniruby[0x10028233e] <+14558>: movq 0x11a86b(%rip), %rax ; ruby_current_vm_ptr
miniruby[0x100282345] <+14565>: testb $0x4, 0x242c(%rax)
```
After:
```
(lldb) disassemble -n vm_opt_str_freeze
ruby`vm_exec_core:
ruby[0x100280ebe] <+14510>: testb $0x4, 0x120147(%rip) ; ruby_vm_redefined_flag + 43
```
Co-authored-by: John Hawthorn <jhawthorn@github.com>
obj_ivar_set and vm_setivar_slowpath is essentially doing the same thing,
but the code is duplicated and not quite implemented in the same way,
which could cause bugs. This commit refactors vm_setivar_slowpath to use
obj_ivar_set.
Implementation for Language Server Protocol (LSP) sometimes needs token information.
For example both `m(1)` and `m(1, )` has same AST structure other than node locations
then it's impossible to check the existence of `,` from AST. However in later case,
it might be better to suggest variables list for the second argument.
Token information is important for such case.
This commit adds these methods.
* Add `keep_tokens` option for `RubyVM::AbstractSyntaxTree.parse`, `.parse_file` and `.of`
* Add `RubyVM::AbstractSyntaxTree::Node#tokens` which returns tokens for the node including tokens for descendants nodes.
* Add `RubyVM::AbstractSyntaxTree::Node#all_tokens` which returns all tokens for the input script regardless the receiver node.
[Feature #19070]
Impacts on memory usage and performance are below:
Memory usage:
```
$ cat test.rb
root = RubyVM::AbstractSyntaxTree.parse_file(File.expand_path('../test/ruby/test_keyword.rb', __FILE__), keep_tokens: true)
$ /usr/bin/time -f %Mkb /usr/local/bin/ruby -v
ruby 3.2.0dev (2022-11-19T09:41:54Z 19070-keep_tokens d3af1b8057) [x86_64-linux]
11408kb
# keep_tokens :false
$ /usr/bin/time -f %Mkb /usr/local/bin/ruby test.rb
17508kb
# keep_tokens :true
$ /usr/bin/time -f %Mkb /usr/local/bin/ruby test.rb
30960kb
```
Performance:
```
$ cat ../ast_keep_tokens.yml
prelude: |
src = <<~SRC
module M
class C
def m1(a, b)
1 + a + b
end
end
end
SRC
benchmark:
without_keep_tokens: |
RubyVM::AbstractSyntaxTree.parse(src, keep_tokens: false)
with_keep_tokens: |
RubyVM::AbstractSyntaxTree.parse(src, keep_tokens: true)
$ make benchmark COMPARE_RUBY="./ruby" ARGS=../ast_keep_tokens.yml
/home/kaneko.y/.rbenv/shims/ruby --disable=gems -rrubygems -I../benchmark/lib ../benchmark/benchmark-driver/exe/benchmark-driver \
--executables="compare-ruby::./ruby -I.ext/common --disable-gem" \
--executables="built-ruby::./miniruby -I../lib -I. -I.ext/common ../tool/runruby.rb --extout=.ext -- --disable-gems --disable-gem" \
--output=markdown --output-compare -v ../ast_keep_tokens.yml
compare-ruby: ruby 3.2.0dev (2022-11-19T09:41:54Z 19070-keep_tokens d3af1b8057) [x86_64-linux]
built-ruby: ruby 3.2.0dev (2022-11-19T09:41:54Z 19070-keep_tokens d3af1b8057) [x86_64-linux]
warming up..
| |compare-ruby|built-ruby|
|:--------------------|-----------:|---------:|
|without_keep_tokens | 21.659k| 21.303k|
| | 1.02x| -|
|with_keep_tokens | 6.220k| 5.691k|
| | 1.09x| -|
```
This commit adds a `capacity` field to shapes, and adds shape
transitions whenever an object's capacity changes. Objects which are
allocated out of a bigger size pool will also make a transition from the
root shape to the shape with the correct capacity for their size pool
when they are allocated.
This commit will allow us to remove numiv from objects completely, and
will also mean we can guarantee that if two objects share shapes, their
IVs are in the same positions (an embedded and extended object cannot
share shapes). This will enable us to implement ivar sets in YJIT using
object shapes.
Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org>
* Avoid RCLASS_IV_TBL in marshal.c
* Avoid RCLASS_IV_TBL for class names
* Avoid RCLASS_IV_TBL for autoload
* Avoid RCLASS_IV_TBL for class variables
* Avoid copying RCLASS_IV_TBL onto ICLASSes
* Use object shapes for Class and Module IVs
Before object shapes, we were using class serial to invalidate
inline caches. Now that we use shape_id for inline cache keys,
the class serial is unnecessary.
Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org>
This patch pushes dummy frames when loading code for the
profiling purpose.
The following methods push a dummy frame:
* `Kernel#require`
* `Kernel#load`
* `RubyVM::InstructionSequence.compile_file`
* `RubyVM::InstructionSequence.load_from_binary`
https://bugs.ruby-lang.org/issues/18559
This solves multiple problems.
First, RB_VM_LOCK_ENTER/LEAVE is a barrier. We could at least use the
_NO_BARRIER variant.
Second, this doesn't need to interfere with GC or other GVL users when
multiple Ractors are used. This needs to be used in very few places, so
the benefit of fine-grained locking would outweigh its small maintenance
cost.
Third, it fixes a crash for YJIT. Because YJIT is never disabled until a
process exits unlike MJIT that finishes earlier, we could call jit_cont_free
when EC no longer exists, which crashes RB_VM_LOCK_ENTER.
Eric Wong
Peter Zhu
Jean Boussier
Koichi Sasada
Matt Valentine-House
Jean byroot Boussier
Kunshan Wang
Nobuyoshi Nakada
John Hawthorn
Jemma Issroff
Daniel Colson
S-H-GAMELINKS
Samuel Williams
yui-knk
Takashi Kokubun
Aaron Patterson