This patch adds a fast path to the GetByValue bytecode op that bypasses
a ton of things *if* a set of assumptions hold:
- The property key must be a non-negative Int32
- The base object must not interfere with indexed property access
- The property key must already be present as an own property
- The existing value must not have any accessors defined
If this holds (which it should in the common case), we can poke directly
at the indexed property storage and save a boatload of time.
10% speed-up on the entire Kraken benchmark :^)
(including: 31% speed-up on Kraken/audio-dft.js)
(including: 23% speed-up on Kraken/stanford-crypto-aes.js)
This function must return true if the object may intercept and customize
access to indexed properties (properties where the property name is a
non-negative integer.)
This will be used to implement fast path optimizations for array-like
accesses in subsequent commits.
The "dependency" lines really belong to the main port entry, it doesn't
make sense logically to represent them separately and handling them
together will also allow easier dependency management later on. This
commit greatly simplifies the port database parsing to facilitate this,
and removes the -d option from the command line. Instead, ports are
listed with their dependencies, if they have any.
The strings will get deduplicated when actually turned into
PrimitiveString objects at runtime anyway, and keeping the string
tables deduplicated was actually wasting a lot of time.
4.4% speed-up on Kraken/stanford-crypto-ccm.js :^)
Color::lightened() and Color::darkened() multiply the color values, so
they don't work for black. So for simplicity, we use a semi-transparent
white or black instead.
The following snippet would cause "i" to be incremented twice(!):
let a = []
let i = 0
a[++i] += 0
This patch solves the issue by remembering the base object and property
name for computed MemberExpression LHS in codegen. We the store the
result of the assignment to the same object and property (instead of
computing the LHS again).
3 new passes on test262. :^)
When the spec says to call "! TrimString", we should use MUST instead
of TRY. (We were previously using TRY in order to propagate OOM errors,
but we don't care about such OOMs anymore.)
Alter the card dimensions to be 80x110px with a corner radius of 7px.
This is inspired by the dimensions of physical playing cards. It gives
12px of padding between the illustration and the card's border.
Move the card letter and symbol closer to the edge to make space.
Adjust the Club symbol to have the same dimensions as the other symbols.
The program now terminates with an error if the command passed to
`-exec` is not terminated with a semicolon.
This commit also ensures that the argument containing the terminating
semicolon must be 1 byte long. Previously, any argument whose first
byte was a semicolon was treated as a valid terminator.
This behaves identically to the `-exec` option but prompts the user
for confirmation before executing the specified command.
A command is executed if a line beginning with 'y' or 'Y' is entered
by the user. This matches the behavior of `find` on Linux and FreeBSD
when using the POSIX locale.
Previously, we would divide the widget width and height by the tile size
and round up, which did not result in enough tiles to cover the entire
widget. Although this calculation is correct if you starting drawing
tiles in the top left corner, we have an additional offset to account
for.
Now, we take the number of tiles that fit in the widget completely and
pad it with 2 tiles to account for the partial left/right and top/bottom
sides. An additional tile is added to account for the iterator
translating by width / 2, which rounds down again.
The resulting tile rects are always intersected with the widget
dimensions, so even if we're generating more tile coordinates than
strictly necessary, we're not performing the actual download or draw
operations.
The `operator++` of the spiraling tile iterator was repeating the first
coordinates (`0, 0`) instead of moving to the next tile on the first
iteration. Swapping the move and check ensures we get to the end of the
iterator, fixing gray tiles that would sometimes pop up in the lower
right.
Since we never return from `operator++` without setting a valid
position, we can drop `current_x` and `current_y` and just use the
`Gfx::Point<T>` directly.
Since we divide the width and height of the downscaled tiles by 2,
bilinear blending is identical to box sampling and should be preferred
since it's the simpler one of the two algorithms.
Renaming the DeprecatedString version of this function to
deprecated_tag_name. A FlyString is used here as we often need to
perform equality checks here, and the HTMLParser already has tag_name as
a FlyString.
Remove a FIXME while we're at it - we were already following the spec
there, and we still are :^)
Which pretty much needs to be done together due to the amount of places
where they are compared together.
This also involves porting over StackOfOpenElements over to FlyString
from DeprecatedFly string to prevent a gazillion calls to
`.to_deprecated_fly_string` calls in HTMLParser.
Renaming the DeprecatedString version of this function to
Element::get_deprecated_attribute.
While performing this rename, port over functions where it is trivial to
do so to the Optional<String> version of this function.
Similar to Element::has_attribute, ideally this would take a
`FlyString const&`, but NamedNodeMap::get_attribute_ns already takes a
StringView. To aid in porting away from DeprecatedString, just take a
StringView for now.
By loading only the fonts actually used on a page, we can often avoid
making a lot of unnecessary requests and style invalidations.
This change makes initial loading of apple.com much faster.
Fixes https://github.com/SerenityOS/serenity/issues/20747
A key press, which is an ASCII control character will no longer cause
TableView to begin editing.
This fixes an issue in Spreadsheet where navigating to a cell then
pressing escape would cause a that cell's text to be set to a
non-printable value. Pressing escape after navigating to a cell
now has no effect.
Bash eats the backslash in this format (similarly for W\ORD etc.).
Dr.POSIX doesn't specify this anywhere, but it's used all over the
place, so let's support it.
The information the user is most interested in is usually in the center,
so we should start loading tiles from the center and move outwards.
Since tiles are loaded in draw order, simply drawing them in this order
achieves the desired effect. The current center-outwards loading
algorithm is a basic spiral algorithm, but others may be evaluated
later.
- Convert to FlatPtr instead of doing pointer arithmetic on a too-large
pointer type in find_min_and_max_block_addresses(). This makes the
range more accurate.
- Untag possible cell pointers in add_possible_value() before doing the
rejection. Otherwise we end up rejecting most pointers since the tags
sit in the highest bits!
This fixes a crash when running the Speedometer benchmark.
Just using Vector::resize() meant that we allocated exact capacity
instead of leaving padding at the end. This patch adds a call to
grow_capacity() before resize(), which ensures that we grow with the
usual extra padding.
This change adds a check to discard pointers that are lower than the
minimum address of all allocated blocks or higher than the maximum
address of all blocks. By doing this we avoid executing plenty of set()
operations on the HashMap in the add_possible_value().
With this change gather_conservative_roots() run 10x times faster in
Speedometer React-Redux-TodoMVC test.
Also, re-order things to match. No behaviour changes.
This reveals quite a few properties that are missing here, or which we
implement somewhat incorrectly.
The following commit will port MIME types to String. Traits<String>
- used in Vector::contains_slow - can't compare String type with char*,
so we need to use StringView instead.
Previously this didn't cause issues because the default flex-factor is
0, but once we only store a flex-factor for FlexibleLength-type
GridSizes, this causes a crash.
There's probably a nicer way of doing this where we don't need to expand
the gray value into a full Vector3, but for now this is good enough.
Makes PDFs written by macOS 13.5.2's "Save as PDF..." feature show up.
We can use `ensure_capacity` for binding vectors if we know their sizes
in advance. This ensures that binding vectors aren't reallocated during
the `function_declaration_instantiation` execution.
With this change, `try_grow_capacity()` and `shrink_to_fit()` are no
longer visible in the `function_declaration_instantiation()` profiles
when running React-Redux-TodoMVC from Speedometer.
If any of binding pattern entry's name is expession
`contains_expression()` should return true.
For example:
```js
function evalInComputedPropertyKey(
{[eval("var x = 'inner'")]: ignored}
) {}
```
`contains_expression()` should return true for the binding param in
this function.
We should initialize jump targets when constructing the jump instruction
instead of doing it later. This was already the case in all construction
sites but one. This first patch converts all those sites to pass final
targets to the constructor directly.
This makes multiple levels of quote actually use different quotation
marks, instead of always the first available pair of them.
Each Layout::Node remembers what the quote-nesting level was before its
content was evaluated, so that we can re-use this number in
`apply_style()`. This is a bit hacky, since we end up converting the
`content` value into a string twice.
`StyleProperties::content()` now takes an initial quote-nesting level,
and returns the final level after that content.
This allows any effects of `content` (eg quotes and counters) to happen
in the right order.
To make it work there are a couple of other changes needed:
- Skip nodes generated by ::before when constructing button layout.
- When in a flex parent, don't merge an inline text node into a previous
one that is generated from a pseudo-element.
This change makes LibJS correctly report a syntax error when a unary
expression is followed by exponentiation, as the spec requires.
Apparently this is due to that expression being ambiguous ordering.
Strangely this check does not seem to apply in the same way for '++' and
'--' for reasons that I don't fully understand. For example
```
let x = 5;
++x ** 2
```
Since `--5` and `++5` on it's own results in a syntax error anyway, it
seems we do not need to perform this exponentiation check in those
places.
Diff Tests:
+6 ✅ -6 ❌
This reduces the minimum size of a basic block from 4 KiB to 0 bytes.
With this change, memory usage at the end of Speedometer is 1.2 GiB,
down from 1.8 GiB.
ImageStyleValue has a visit_edges() method, although it is not a
GC-allocated object. This is necessary because it owns a GC-allocated
ImageRequest that we want to visit, instead of using JS::Handle, to
avoid leaks. In the future, we might want to make StyleValue be
GC-allocated.
For now, this change adds missing visit_edges() calls for objects that
own ImageStyleValue.
Instead of calling out to helper functions for flow control (and then
checking control flags on every iteration), we now simply inline those
ops in the interpreter loop directly.
If we don't have a local unwind context to handle the exception, we can
just return right away. This allows us to remove one check from the
inner loop.
Nuke all the per-instruction bounds checking when iterating instructions
by using raw pointers instead of indexing into a ReadonlyBytes.
The interpreter loop already checks that we're in-bounds anyway.
Using JS::Handle in WebEngineCustomData means that mutation observers
will live as long as VM while actually they should be deallocated as
soon as they are no longer used in a script that created them.
There is not need to use SafeFunction because
define_native_function or define_native_accessor will pass callback
forward to NativeFunction that uses HeapFunction to visit it.
This commit introduces 3 things:
- Support for the color type in HTMLInputElement itself
- A mechanism for handling non event loop blocking dialogs in Page
- The associated plumbing up to ViewImplementation
Frontends may add support for the color picker with the
ViewImplementation.on_request_color_picker function
This is 10-20% of a speed increase on platforms with fast I/O (Linux)
and not a slowdown on Serenity. Again, the file system layer is the
limit for us :^)
The internal reuse of FixedMemoryStream makes this straightforward.
There alread is one user of the new API, demonstrating the need for this
change beyond what I said out to use it for :^)
Since it will become a stream in a little bit, it should behave like all
non-trivial stream classes, who are not primarily intended to have
shared ownership to make closing behavior more predictable. Across all
uses of MappedFile, there is only one use case of shared mapped files in
LibVideo, which now uses the thin SharedMappedFile wrapper.
Previously, the same file would be counted more than once if its
containing directory was visited multiple times, or there were
multiple hard links pointing to it.
We now keep track of every visited inode to ensure that files aren't
evaluated multiple times. This matches the behavior of `du` on FreeBSD
and Linux.
Most shorthands can be reconstructed this way, using our generated
property data, so let's use them instead of manually implementing the
code.
Some of these were previously doing some form of error checking or
defaulting, but both of those are unnecessary. (And actually, would
crash if there wasn't a value available due to calling release_nonnull()
on a null RefPtr.) At this point, the CSS machinery has already made
sure each property has a value, and that the value is valid for that
property.
These three are the ones that ShorthandStyleValue uses to serialize
`grid`, so let's use them here.
The spec also mentions `grid-auto-*` properties as being set by `grid`,
but I'll leave that for someone who understands grid better.
When parsing these, <number> is allowed anywhere that would usually
allow a <length>, <length-percentage>, or <angle>. The spec is not
clear on exactly how this should work
(see https://github.com/w3c/svgwg/issues/792 ) so I'm using some
artistic license until things are clearer:
- If we expected a <length>, treat the <number> as pixels.
- If we expected an <angle>, treat the <number> as degrees.
- Only allow direct <number> tokens, not calc() or other functions.
From what I can tell this is what the spec *intended* but I may be very
wrong. In any case, telling the ParsingContext whether we're parsing
one of these attributes is a cleaner approach and more correct than
temporarily enabling quirks mode, which we did previously.
Before the completion_steps for timer were casted from JS::SafeFunction
to Function in HTML::Timer constructor, which is incorrect because then
callback's captured GC-allocated objects are not protected from being
deallocated. Let's modify HTML::Timer to use JS::HeapFunction for the
callback instead.
By using Core::Timer that accepts Function instead of JS::SafeFunction
in Platform::Timer does we fix memory leak caused by circular
dependency of timer's callback and timer itself.
