With this change, we're storing the iv name on an inline cache on
setinstancevariable instructions. This allows us to check the inline
cache to count instance variables set in initialize and give us an
estimate of iv capacity for an object.
For the purpose of estimating the number of instance variables required
for an object, we're assuming that all initialize methods will call
`super`.
This change allows us to estimate the number of instance variables
required without disassembling instruction sequences.
Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org>
This was introduced by b609bdeb53
to suppress warnings. However these warngins were deleted by
beae6cbf0f. Therefore these codes
are not needed anymore.
`throw TAG_BREAK` instruction makes a jump only if the continuation of
catch of TAG_BREAK exactly matches the instruction immediately following
the "send" instruction that is currently being executed. Otherwise, it
seems to determine break from proc-closure.
Branch coverage may insert some recording instructions after "send"
instruction, which broke the conditions for TAG_BREAK to work properly.
This change forces to set the continuation of catch of TAG_BREAK
immediately after "send" (or "invokesuper") instruction.
[Bug #18991]
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
Object Shapes is used for accessing instance variables and representing the
"frozenness" of objects. Object instances have a "shape" and the shape
represents some attributes of the object (currently which instance variables are
set and the "frozenness"). Shapes form a tree data structure, and when a new
instance variable is set on an object, that object "transitions" to a new shape
in the shape tree. Each shape has an ID that is used for caching. The shape
structure is independent of class, so objects of different types can have the
same shape.
For example:
```ruby
class Foo
def initialize
# Starts with shape id 0
@a = 1 # transitions to shape id 1
@b = 1 # transitions to shape id 2
end
end
class Bar
def initialize
# Starts with shape id 0
@a = 1 # transitions to shape id 1
@b = 1 # transitions to shape id 2
end
end
foo = Foo.new # `foo` has shape id 2
bar = Bar.new # `bar` has shape id 2
```
Both `foo` and `bar` instances have the same shape because they both set
instance variables of the same name in the same order.
This technique can help to improve inline cache hits as well as generate more
efficient machine code in JIT compilers.
This commit also adds some methods for debugging shapes on objects. See
`RubyVM::Shape` for more details.
For more context on Object Shapes, see [Feature: #18776]
Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org>
Co-Authored-By: Eileen M. Uchitelle <eileencodes@gmail.com>
Co-Authored-By: John Hawthorn <john@hawthorn.email>
Tabs were expanded because the file did not have any tab indentation in unedited lines.
Please update your editor config, and use misc/expand_tabs.rb in the pre-commit hook.
Object Shapes is used for accessing instance variables and representing the
"frozenness" of objects. Object instances have a "shape" and the shape
represents some attributes of the object (currently which instance variables are
set and the "frozenness"). Shapes form a tree data structure, and when a new
instance variable is set on an object, that object "transitions" to a new shape
in the shape tree. Each shape has an ID that is used for caching. The shape
structure is independent of class, so objects of different types can have the
same shape.
For example:
```ruby
class Foo
def initialize
# Starts with shape id 0
@a = 1 # transitions to shape id 1
@b = 1 # transitions to shape id 2
end
end
class Bar
def initialize
# Starts with shape id 0
@a = 1 # transitions to shape id 1
@b = 1 # transitions to shape id 2
end
end
foo = Foo.new # `foo` has shape id 2
bar = Bar.new # `bar` has shape id 2
```
Both `foo` and `bar` instances have the same shape because they both set
instance variables of the same name in the same order.
This technique can help to improve inline cache hits as well as generate more
efficient machine code in JIT compilers.
This commit also adds some methods for debugging shapes on objects. See
`RubyVM::Shape` for more details.
For more context on Object Shapes, see [Feature: #18776]
Co-Authored-By: Aaron Patterson <tenderlove@ruby-lang.org>
Co-Authored-By: Eileen M. Uchitelle <eileencodes@gmail.com>
Co-Authored-By: John Hawthorn <john@hawthorn.email>
As of fbaac837cf, when we were performing
a safe call (`o&.x=`) with a conditional assign (`||= 1`) and discarding
the result the stack would end up in a bad state due to a missing pop.
This commit fixes that by adjusting the target label of the branchnil to
be before a pop in that case (as was previously done in the
non-conditional assignment case).
A common pattern when the block is an explicit parameter is to branch
based on the block parameter instead of using `block_given?`, for
example `block.call if block`.
This commit checks in the peephole optimizer for that case and uses the
getblockparamproxy optimization, which avoids allocating a proc for
simple cases, whenever a getblockparam instruction is followed
immediately by branchif or branchunless.
./miniruby --dump=insns -e 'def foo(&block); 123 if block; end'
== disasm: #<ISeq:foo@-e:1 (1,0)-(1,34)> (catch: FALSE)
local table (size: 1, argc: 0 [opts: 0, rest: -1, post: 0, block: 0, kw: -1@-1, kwrest: -1])
[ 1] block@0<Block>
0000 getblockparamproxy block@0, 0 ( 1)[LiCa]
0003 branchunless 8
0005 putobject 123
0007 leave [Re]
0008 putnil
0009 leave [Re]
Tabs were expanded because the file did not have any tab indentation in unedited lines.
Please update your editor config, and use misc/expand_tabs.rb in the pre-commit hook.
Previously YARV bytecode implemented constant caching by having a pair
of instructions, opt_getinlinecache and opt_setinlinecache, wrapping a
series of getconstant calls (with putobject providing supporting
arguments).
This commit replaces that pattern with a new instruction,
opt_getconstant_path, handling both getting/setting the inline cache and
fetching the constant on a cache miss.
This is implemented by storing the full constant path as a
null-terminated array of IDs inside of the IC structure. idNULL is used
to signal an absolute constant reference.
$ ./miniruby --dump=insns -e '::Foo::Bar::Baz'
== disasm: #<ISeq:<main>@-e:1 (1,0)-(1,13)> (catch: FALSE)
0000 opt_getconstant_path <ic:0 ::Foo::Bar::Baz> ( 1)[Li]
0002 leave
The motivation for this is that we had increasingly found the need to
disassemble the instructions between the opt_getinlinecache and
opt_setinlinecache in order to determine the constant we are fetching,
or otherwise store metadata.
This disassembly was done:
* In opt_setinlinecache, to register the IC against the constant names
it is using for granular invalidation.
* In rb_iseq_free, to unregister the IC from the invalidation table.
* In YJIT to find the position of a opt_getinlinecache instruction to
invalidate it when the cache is populated
* In YJIT to register the constant names being used for invalidation.
With this change we no longe need disassemly for these (in fact
rb_iseq_each is now unused), as the list of constant names being
referenced is held in the IC. This should also make it possible to make
more optimizations in the future.
This may also reduce the size of iseqs, as previously each segment
required 32 bytes (on 64-bit platforms) for each constant segment. This
implementation only stores one ID per-segment.
There should be no significant performance change between this and the
previous implementation. Previously opt_getinlinecache was a "leaf"
instruction, but it included a jump (almost always to a separate cache
line). Now opt_getconstant_path is a non-leaf (it may
raise/autoload/call const_missing) but it does not jump. These seem to
even out.
catch_excep_t is a field that exists for MJIT. In the process of
rewriting MJIT in Ruby, I added API to convert 1/0 of _Bool to
true/false, and it seemed confusing and hard to maintain if you
don't use _Bool for *_p fields.
There's no point in making a copy of an array just to expand it. Saves
an unnecessary array allocation in the multiple assignment case, with
a 35-84% improvement in affected cases in benchmark/masgn.yml.
This optimizes unbalanced multiple assignment cases such as:
```ruby
a.b, c.d = e, f, g
a.b, c.d, e.f = g, h
```
Previously, this would use:
```
newarray(3)
expandarray(2, 0)
newarray(2)
expandarray(3, 0)
```
These would both allocate arrays. This switches to opt_reverse
with either pop or putnil:
```
pop
opt_reverse(2)
putnil
opt_reverse(3)
```
This avoids an unnecessary array allocation, and results in a 35-76%
performance increase in these types of unbalanced cases (tested with
benchmark/masgn.yml).
This renames the reverse instruction to opt_reverse, since now it
is only added by the optimizer. Then it uses as a more general
form of swap. This optimizes multiple assignment in the popped
case with more than two elements.
An optimization for multiple assignment in the popped case to avoid
array allocation was lost in my fix to make multiple assignment follow
left-to-right evaluation (50c54d40a8).
Before, in the two element case, swap was used. Afterward, newarray(2)
and expandarray(2, 0) were used, which is the same as swap, with the
addition of an unnecessary allocation.
Because this issue is not specific to multiple assignment, and the
multiple assignment code is complex enough as it is, this updates
the peephole optimizer to do the newarray(2)/expandarray(2, 0) -> swap
conversion.
A more general optimization pass for
newarray(X)/expandarray(X, 0) -> reverse(X) will follow, but that
requires readding the reverse instruction.
rb_ary_tmp_new suggests that the array is temporary in some way, but
that's not true, it just creates an array that's hidden and not on the
transient heap. This commit renames it to rb_ary_hidden_new.
The RARRAY_LITERAL_FLAG was added in commit
5871ecf956 to improve CoW performance for
array literals by not keeping track of reference counts.
This commit reverts that commit and has an alternate implementation that
is more generic for all frozen arrays. Since frozen arrays cannot be
modified, we don't need to set the RARRAY_SHARED_ROOT_FLAG and we don't
need to do reference counting.
Array created as literals during iseq compilation don't need a
reference count since they can never be modified. The previous
implementation would mutate the hidden array's reference count,
causing copy-on-write invalidation.
This commit adds a RARRAY_LITERAL_FLAG for arrays created through
rb_ary_literal_new. Arrays created with this flag do not have reference
count stored and just assume they have infinite number of references.
Co-authored-by: Jean Boussier <jean.boussier@gmail.com>
At that commit, I fixed a wrong conditional expression that was always
true. However, that seemed to have caused a regression. [Bug #18906]
This change removes the condition to make the code always enabled.
It had been enabled until that commit, albeit unintentionally, and even
if it is enabled it only consumes a tiny bit of memory, so I believe it
is harmless. [Bug #18906]
We need to dump relative offsets for inline storage entries so that
loading iseqs as an array works as well. This commit also has some
minor refactoring to make computing relative ISE information easier.
This should fix the iseq dump / load as array tests we're seeing fail in
CI.
Co-Authored-By: John Hawthorn <john@hawthorn.email>
ISeqs loaded from binary were breaking because the storage partition
calculation had bugs in it. Specifically it couldn't take in to account
the case when inline storage was overallocated (for example when we
allocate inline storage for an instruction but peephole optimization
eliminates that instruction).
`RUBY_ISEQ_DUMP_DEBUG=to_binary make test-all` would break, and this
patch fixes it
This commit adds a bitfield to the iseq body that stores offsets inside
the iseq buffer that contain values we need to mark. We can use this
bitfield to mark objects instead of disassembling the instructions.
This commit also groups inline storage entries and adds a counter for
each entry. This allows us to iterate and mark each entry without
disassembling instructions
Since we have a bitfield and grouped inline caches, we can mark all
VALUE objects associated with instructions without actually
disassembling the instructions at mark time.
[Feature #18875] [ruby-core:109042]
... because insns_info_index could not be zero here. Also it adds an
invariant check for that.
This change will prevent the following warning of GCC 12.1
http://rubyci.s3.amazonaws.com/arch/ruby-master/log/20220613T000004Z.log.html.gz
```
compile.c:2230:39: warning: array subscript 2147483647 is outside array bounds of ‘struct iseq_insn_info_entry[2147483647]’ [-Warray-bounds]
2230 | insns_info[insns_info_index-1].line_no != adjust->line_no) {
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~
```
Use ISEQ_BODY macro to get the rb_iseq_constant_body of the ISeq. Using
this macro will make it easier for us to change the allocation strategy
of rb_iseq_constant_body when using Variable Width Allocation.
Jemma Issroff
yui-knk
S-H-GAMELINKS
Yusuke Endoh
Koichi Sasada
Aaron Patterson
git
John Hawthorn
Samuel Williams
HParker
Maple Ong
Nobuyoshi Nakada
Takashi Kokubun
Jeremy Evans
Peter Zhu
Aaron Patterson