These unused comparison operators are the only users of
<functional> in SkRefCnt.h, for std::less. <functional>
is an expensive header to compile, and SkRefCnt.h is popular,
so it helps to cut dependencies like this.
Mostly we just need to add #include <functional> in a few
places that were picking it up via SkRefCnt.h.
In SkPixmapPriv.h, it looked simpler to template the argument,
since everything was inline anyway.
Change-Id: I7c125bb26a04199847357c729a1b178256c6ef8d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/236942
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
<ostream> is one of the more expensive headers to include
and that's amplified by SkRefCnt.h's popularity.
We've been including <ostream> for sk_sp's operator<<. That's only
used by Chromium and while we could just sprinkle in a bunch of .get()
calls and remove operator<<, when I started going through and actually
doing that I got the feeling I was making things pointlessly harder to
read and write, and wanted to find a way to make it actually work.
My next instinct was to template it without mentioning ostreams,
template <typename OS, typename T>
auto operator<<(OS& os, const sk_sp<T>& sp) -> decltype(os << sp.get()) {
return os << sp.get();
}
but that makes this operator<< ambiguous with some other templated operator<<
in GTest. They got in first, so they win...
So ultimately, switch <ostream> to <iosfwd>. Anyone using our
operator<<() presumably has <ostream> included already, and the #include
cost for <iosfwd> is small enough that I don't think we'll mind keeping
this around indefinitely.
To repro, look at before/after of -ftime-trace:
~/chromium/src/third_party/llvm-build/Release+Asserts/bin/clang++ -I. -Os -c src/core/SkCanvas.cpp -ftime-trace
I have tested locally that Chromium builds with this change.
Change-Id: I9decc2e65b5cc8fd07d8106a5eff81901aedd7d5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/237190
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Current strategy: everything from the top
Things to look at first are the manual changes:
- added tools/rewrite_includes.py
- removed -Idirectives from BUILD.gn
- various compile.sh simplifications
- tweak tools/embed_resources.py
- update gn/find_headers.py to write paths from the top
- update gn/gn_to_bp.py SkUserConfig.h layout
so that #include "include/config/SkUserConfig.h" always
gets the header we want.
No-Presubmit: true
Change-Id: I73a4b181654e0e38d229bc456c0d0854bae3363e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209706
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Hal Canary <halcanary@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
SkRefCntBase is an implementation detail of SkRefCnt, so let's not
mention it in SkContext_Compute.h. While I'm here, I notice SkNVRefCnt
would work fine too.
No one else calls internal_dispose_restore_refcnt_to_1(), so we can
inline it. While here, I notice it's resetting the ref count to 1 even
in release builds. I'm not sure if that is/can be optimized away, but
in any case I think we can wrap with #ifdef SK_DEBUG, if only to make it
clear that it's only there to support the assert in the destructor.
I've removed validate(). Most of the places it's called read pretty
weird, and I think suggest some sort of class-specific validation than
just checking that we're holding a ref. SkWeakRefCnt::validate() isn't
called anywhere.
There were few users of getRefCnt() outside SkRefCnt itself, so I
removed the rest and made it private.
I've added a few this-> to self calls while at it.
Change-Id: I98be06677a6e8b8e66f44cbb17d14e38b0f39d38
Reviewed-on: https://skia-review.googlesource.com/c/167160
Auto-Submit: Mike Klein <mtklein@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
This reverts commit 0fb1ee98cf.
Reason for revert: looks like this increased size by ~8K.
Original change's description:
> replace SkNVRefCnt with SkRefCnt
>
> SkNVRefCnt trades a small amount of code size (vtable) and runtime
> (vptr) memory usage for a larger amount of code size (templating). It
> was written back in a time when all we were really thinking about was
> runtime memory usage, so I'm curious to see where performance, code
> size, and memory usage all move if it's removed.
>
> Looking at the types I've changed here, my guess is that performance and
> memory usage will be basically unchanged, and that code size will drop a
> bit. Nothing else it's nicer to have only one ref-counting base class.
>
> Change-Id: I7d56a2b9e2b9fb000ff97792159ea1ff4f5e6f13
> Reviewed-on: https://skia-review.googlesource.com/c/166203
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Mike Klein <mtklein@google.com>
TBR=mtklein@google.com,bsalomon@google.com,mtklein@chromium.org
Change-Id: Ibcfcc4b523c466a535bea5ffa30d0fe2574c5bd7
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/166360
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
SkNVRefCnt trades a small amount of code size (vtable) and runtime
(vptr) memory usage for a larger amount of code size (templating). It
was written back in a time when all we were really thinking about was
runtime memory usage, so I'm curious to see where performance, code
size, and memory usage all move if it's removed.
Looking at the types I've changed here, my guess is that performance and
memory usage will be basically unchanged, and that code size will drop a
bit. Nothing else it's nicer to have only one ref-counting base class.
Change-Id: I7d56a2b9e2b9fb000ff97792159ea1ff4f5e6f13
Reviewed-on: https://skia-review.googlesource.com/c/166203
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Update all users to sk_sp.
Change-Id: I6453b9456b9a8f9e2b756381797f1382ef9e6561
Reviewed-on: https://skia-review.googlesource.com/141052
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Herb Derby <herb@google.com>
This reverts commit 61582510ee.
Reason for revert: Chromium should now be fixed.
Original change's description:
> SkRefCnt, SkTypes: fix includes for clients
>
> Change-Id: I1cfdc03963eaab9687608974958d20411806cfeb
> Reviewed-on: https://skia-review.googlesource.com/136592
> Commit-Queue: Hal Canary <halcanary@google.com>
> Reviewed-by: Hal Canary <halcanary@google.com>
TBR=halcanary@google.com
Change-Id: I1a3588ed9e19ee5884f8a907abcfc2aa00cfe5e9
Reviewed-on: https://skia-review.googlesource.com/136700
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
Change-Id: Iaeb886c1c80ea1603dc86ed40c8cfffec333b456
Reviewed-on: https://skia-review.googlesource.com/136244
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Hal Canary <halcanary@google.com>
Change-Id: I345c83783c578f5ce25b4fc46c971c055e113cd0
Reviewed-on: https://skia-review.googlesource.com/134945
Auto-Submit: Hal Canary <halcanary@google.com>
Commit-Queue: Hal Canary <halcanary@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
The sk_sp class has been using the operator pointer to field as a c++98
version of explicit operator bool. This change updates this class to use
explicit operator bool. The one visible change is that the pointer to
field version isn't quite as explcit, requiring code changes for some
users.
Change-Id: Iddf8fb347b1d3ec33db1af08489c9fd885c9bf08
Reviewed-on: https://skia-review.googlesource.com/130380
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This is more consistent with our other SK_BUILD_FOR_... macros,
and less likely to collide with other preprocessor logic.
(Luckily, this was defined in public.bzl, so we can do this
all in one CL in the Skia repo.)
Change-Id: I5f232888288c9c53fad445545d983d0fb0b4add8
Reviewed-on: https://skia-review.googlesource.com/86940
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Change-Id: If524a02e916d711057d6f2102efb0404b1c0b988
Reviewed-on: https://skia-review.googlesource.com/10036
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
If SkRefCnt_SafeAssign is erroneously invoked simultaneously by two
threads on the same destination, a likely outcome is that the
reference count of an object other than the replaced dst will
be decremented. This results in an extra reference count decrement,
which usually means that the final reference count decrement will
be applied to a random location in an object that has already been
reallocated. This is exceedingly hard to debug.
We add a hack to detect such data races with sufficiently high
probability, so that such a data race bug should sometimes actually
generate bug reports that lend themselves to diagnosis.
We detect changes within a relatively larger range that normally
includes several (typically slow) memory fences. Not all such changes
would provoke a crash. Even if we decrement the wrong reference count,
there's a chance we would decrement a dead location. Thus, to
avoid potentially adding instability, we currently only log.
This change tries to minimize additional runtime overhead.
The macro is only expanded a few dozen times, so we do not worry too
much about code size.
Bug: b/31227650
Change-Id: Ia40c9ed2c4d0fa578ea682fbec4b71a2ef22a5d1
Reviewed-on: https://skia-review.googlesource.com/9994
Commit-Queue: Derek Sollenberger <djsollen@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
This function will always be useful. Don't make it look/feel like a
temporary hack.
BUG=skia:
Change-Id: I6506d7d51dc3b25a7dbcea4ac273f51cf05f3a89
Reviewed-on: https://skia-review.googlesource.com/6330
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
It is no longer used.
Change-Id: Ie2f9a39a4295005cb39bdf2f8fc15542ee75d207
Reviewed-on: https://skia-review.googlesource.com/4386
Commit-Queue: Ben Wagner <bungeman@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
In some legacy situations users of SkRefCnt subclasses were keeping
the objects alive with a reference count of 0. Now that these users are
cleaned up, remove the hack which allowed such code to keep functioning.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3264
Change-Id: I22f63d87b6d995cad6326998284930ad9eaa2983
Reviewed-on: https://skia-review.googlesource.com/3264
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
SkLiteRecorder, a new SkCanvas, fills out SkLiteDL, a new SkDrawable.
This SkDrawable is a display list similar to SkRecord and SkBigPicture / SkRecordedDrawable, but with a few new design points inspired by Android and slimming paint:
1) SkLiteDL is structured as one big contiguous array rather than the two layer structure of SkRecord. This trades away flexibility and large-op-count performance for better data locality for small to medium size pictures.
2) We keep a global freelist of SkLiteDLs, both reusing the SkLiteDL struct itself and its contiguous byte array. This keeps the expected number of mallocs per display list allocation <1 (really, ~0) for cyclical use cases.
These two together mean recording is faster. Measuring against the code we use at head, SkLiteRecorder trends about ~3x faster across various size pictures, matching speed at 0 draws and beating the special-case 1-draw pictures we have today. (I.e. we won't need those special case implementations anymore, because they're slower than this new generic code.) This new strategy records 10 drawRects() in about the same time the old strategy took for 2.
This strategy stays the winner until at least 500 drawRect()s on my laptop, where I stopped checking.
A simpler alternative to freelisting is also possible (but not implemented here), where we allow the client to manually reset() an SkLiteDL for reuse when its refcnt is 1. That's essentially what we're doing with the freelist, except tracking what's available for reuse globally instead of making the client do it.
This code is not fully capable yet, but most of the key design points are there. The internal structure of SkLiteDL is the area I expect to be most volatile (anything involving Op), but its interface and the whole of SkLiteRecorder ought to be just about done.
You can run nanobench --match picture_overhead as a demo. Everything it exercises is fully fleshed out, so what it tests is an apples-to-apples comparison as far as recording costs go. I have not yet compared playback performance.
It should be simple to wrap this into an SkPicture subclass if we want.
I won't start proposing we replace anything old with anything new quite yet until I have more ducks in a row, but this does look pretty promising (similar to the SkRecord over old SkPicture change a couple years ago) and I'd like to land, experiment, iterate, especially with an eye toward Android.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2213333002
Review-Url: https://codereview.chromium.org/2213333002
This enables removing the more complicated atomic shims from SkAtomics.h.
TBR=reed
This doesn't actually change any API.
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-TSAN-Trybot,Test-Ubuntu-GCC-Golo-GPU-GT610-x86_64-Release-TSAN-Trybot
Review URL: https://codereview.chromium.org/1867863002
The C++ standard library uses the name "release" for the operation we call "detach".
Rewriting each "detach(" to "release(" brings us a step closer to using standard library types directly (e.g. std::unique_ptr instead of SkAutoTDelete).
This was a fairly blind transformation. There may have been unintentional conversions in here, but it's probably for the best to have everything uniformly say "release".
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1809733002
Review URL: https://codereview.chromium.org/1809733002
The 'element_type' typedef is to play nice with std::pointer_traits.
The full complement of operators and swap to match unique_ptr so that
sk_sp can be properly compared to nullptr and used with standard
containers.
Update to 'reset' so that calling 'unref' is the last operation.
This also adds tests for these changes, and sets the fPtr to nullptr
in debug for easier bug finding.
Review URL: https://codereview.chromium.org/1773453002
Previously, sk_sp::reset(T* t) did not release its own reference
if its internal pointer was the same as 't'. This leaks a reference.
Now always release the current reference when non-nullptr.
Review URL: https://codereview.chromium.org/1767983002
