This makes the difference between selector matching scaling on the ARM
Cortex-A9 and not, because the auto-derived `PartialEq` implementation
blows out the 32KB I-cache. With this change, there is a 2x improvement
in selector matching over sequential when using all 8 cores. (More work
needs to be done; this is a start.)
r? any DOM expert
<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/servo/servo/6308)
<!-- Reviewable:end -->
This makes the difference between selector matching scaling on the ARM
Cortex-A9 and not, because the auto-derived `PartialEq` implementation
blows out the 32KB I-cache. With this change, there is a 2x improvement
in selector matching over sequential when using all 8 cores. (More work
needs to be done; this is a start.)
The util component specified fnv and smallvec as dependencies and publicly
reexported both of them. Several other components utilized these reexports,
presumably because fnv and smallvec used to live in the tree so reexporting
made the transition easier.
These indirect dependencies through the util component are unnecessary.
This commit removes the fnv & smallvec crate reexports in the util component.
It exchange, it adds fnv & smallvec as dependencies to non-util components
wherever needed. Finally, it removes the fnv dependency from util as it is not
utilized anywhere in the util component.
This used to conflict with the util crate from the standard library, which
has long since been removed.
The import in layout has not been changed because of a conflict with the
util mod there.
Now `#[dom_struct]` also generates Reflectable impls, and there's another lint to ensure that a DOM struct only contains one bare DOM field (as the first field) or a Reflector.
A lot of this was generated by sed -- each autogenerated change has its own commit for easy review; these will be squashed later.
The security properties of SipHash are irrelevant for event listeners
and the creation of the random number generator was showing up high in
the profiles.
I converted them all with a few exceptions:
- Methods that were used by trait objects, since trait objects don't
work with `self` methods.
- Methods that take an &'b JSRef<'a, T> and return an &'b. In reality,
many (all?) could return an &'a instead, but this isn't allowed by the
Deref trait.
- Methods that internally rely on the same issue with Deref.
- I left out the traits involved in layout entirely, even though not all
of their methods suffer from one of the above problems.
There will probably be solutions to all of these problems in the future.
Replace &JSRef with JSRef in the bulk of the generated code. This will
remove a level of indirection throughout all DOM code.
This patch doesn't change methods implemented on JSRef<T> to take `self`
rather than `&self`, and it leaves a few other uses of &JSRef, but those
changes can be made incrementally.
