The SkArenaAlloc destructor was showing up as hot in profiling,
especially on Linux. The reason is that it was being used
incorrectly: the size estimate was being used as the chunk size. It
turns out that the best performance seems to be achieved with no
initial allocations and a fixed chunk size of 16K, as the CPU path
renderer does.
Also, allocate the bisectors used for edge inversions on the
stack, not the heap. And remove some unused code.
BUG=skia:
Change-Id: I754531c753c9e602713bf2c8bb5a0eaf174bb962
Reviewed-on: https://skia-review.googlesource.com/8560
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Windows needs this as a shim to switch to the System V ABI and back.
Other platforms need it too, if only to make UBSAN happy about calling
functions through the correct function pointers.
One day maybe we can move the looping logic inside start_pipeline?
Change-Id: I47d9ef48752becc6c43fc052b12a540c157bcaaa
Reviewed-on: https://skia-review.googlesource.com/8542
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
It's unclear what this test should be -- or even if it's valid any more.
Change-Id: I62b5d99ad19c9b2813ef5a3c34c9b11991c01c5f
Reviewed-on: https://skia-review.googlesource.com/8494
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Just recently noticed we weren't doing this.
It seems to be triggering all the time.
Draws the same of course.
Change-Id: Ie3d869a05194a5d52d635643eba1c96c8a21ce67
Reviewed-on: https://skia-review.googlesource.com/8471
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
mad (Multiply-anD-Add) conveys the desired behavior better than fma...
we don't particularly care about the fusedness of the mad, only that
we're doing the fastest multiply-add available.
This also avoids conflicts with fma() that may be defined in <math.h>.
Change-Id: I9298ce9ff6c6aa5b7798ac446eea248ab035625f
Reviewed-on: https://skia-review.googlesource.com/8543
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
If there's a reflection in the matrix, the scales
could be negative.
BUG=chromium:682918
Change-Id: Ifde42c3bc8b1f2938009b47d3a8380cf871448e9
Reviewed-on: https://skia-review.googlesource.com/8541
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
No real change here.
Change-Id: I56449c292585038901d78902e6aeb68203e36351
Reviewed-on: https://skia-review.googlesource.com/8476
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
BUG=skia:
Change-Id: Ic4bce28f4bc45f73cbfcacc3630140d11fa9c41f
Reviewed-on: https://skia-review.googlesource.com/8530
Reviewed-by: Leon Scroggins <scroggo@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Matt Sarett <msarett@google.com>
This reverts commit 957cf45949.
Reason for revert: crashes
Original change's description:
> Turn on SkJumper except on Windows.
>
> ...only for our local testing. No Skia users should be affected yet.
>
> Change-Id: If1c46efe418458f747aafd6be8e2ae799c74b1c4
> Reviewed-on: https://skia-review.googlesource.com/8470
> Reviewed-by: Mike Klein <mtklein@chromium.org>
> Commit-Queue: Mike Klein <mtklein@chromium.org>
>
TBR=mtklein@chromium.org,herb@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: Id27f51c3b9356471b3178bbd63b021bd67a1d6fb
Reviewed-on: https://skia-review.googlesource.com/8477
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Corresponds with Android change
https://android-review.googlesource.com/#/c/326439/
"this version includes encoder and performance improvements"
Update build file
* Many files have been renamed from
src/<subdir>/<name>.c
to
src/<subdir>/<name>_<subdir>.c
* Build new files (*_msa.c, *_neon.c, predictor_enc.c)
This should fix issue 5876, which was caused by a compiler bug.
With the added NEON implementation, we will no longer trigger
the bug.
BUG=skia:5876
CQ_INCLUDE_TRYBOTS=skia.primary:Build-Mac-Clang-arm-Debug-iOS
Review-Url: https://codereview.chromium.org/2689283007
...only for our local testing. No Skia users should be affected yet.
Change-Id: If1c46efe418458f747aafd6be8e2ae799c74b1c4
Reviewed-on: https://skia-review.googlesource.com/8470
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Should produce sharper results than arbitrary fixed sizes.
Adds a new test to pathfill GM.
Was: https://skia-review.googlesource.com/c/8328/
BUG=chromium:682918, skia:6238
Change-Id: Ia62ea5ce6b4a5ac2b8b51d06d57dc951d6c340b8
Reviewed-on: https://skia-review.googlesource.com/8384
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
The multiply by 0x77800000 is quite slow when the input is denormalized.
We don't mind flushing those values (in the range of 1e-5) to zero.
Implement portable load_f16() / store_f16() too.
Change-Id: I125cff1c79ca71d9abe22ac7877136d86707cb56
Reviewed-on: https://skia-review.googlesource.com/8467
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
It's easier to work on SkJumper if everything funnels through run().
I don't anticipate huge benefit from compile() without JITing,
but it's something we can always put back if we find a need.
Change-Id: Id5256fd21495e8195cad1924dbad81856416d913
Reviewed-on: https://skia-review.googlesource.com/8468
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
This is a color filter to apply several contrast adjustments for users
with low vision, including inverting the colors (in either RGB or HSL
space), applying gamma correction, converting to grayscale, and increasing
the contrast.
BUG=skia:6235
CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD
Change-Id: Icb8f3e290932d8bcd9387fb1f39dd20767e15cf6
Reviewed-on: https://skia-review.googlesource.com/7460
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Reed <reed@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Change-Id: If9f73e712e429564fef58ccb838c212ec8d2e68c
Reviewed-on: https://skia-review.googlesource.com/8525
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
With at least some system fonts and apparently all fonts from data the
x-height and cap-height on at least 10.12.3 are always 0.6666 and .8888
of the ascent, respectively. The values from the 'OS/2' table appear
to be read, but then the values are overwritten. Work around this by
using the values from the 'OS/2' table when available.
This also removes fFUnitMatrix as it is no longer used.
BUG=skia:6203
Change-Id: Ib79d9d32bca77797de043499c854c35e5ef3aa4b
Reviewed-on: https://skia-review.googlesource.com/8452
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Mike Reed <reed@google.com>
BUG=skia:
Change-Id: I00b52e6de3b46b7d26c1df2aa63c521b6c7ece2d
Reviewed-on: https://skia-review.googlesource.com/8526
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Reed <reed@google.com>
SkJumper's looking promising enough that I want this further out of my face.
Change-Id: I6dbe71aeabe32f7f4258ba157460e6985733b0ce
Reviewed-on: https://skia-review.googlesource.com/8528
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
BUG=skia:
Change-Id: I4bc11042dd1dbf1eabd40af206027bc65acc3186
Reviewed-on: https://skia-review.googlesource.com/8444
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This may reduce the number of "why not GrTextureProxy" issues
Change-Id: I9e0e5042f5801ba9a933b697a380cb0cb54b4522
Reviewed-on: https://skia-review.googlesource.com/8510
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This is split out of https://skia-review.googlesource.com/c/7889/ (Remove SkSpecialImage's GrTexture-based ctors) to allow focusing on the caching/subsetting changes.
Change-Id: Ic2af7a0d03e88941ab5b6cdfcecbbdefe3578eda
Reviewed-on: https://skia-review.googlesource.com/8456
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
The previous approach re-inflated views on every frame simply because the FPS
number changed (which was passed in a JSON object and used to check whether
the old view matched the new one based on a tag on the view that was set by
the previous FPS string).
The new approach is not pretty (there's no reason to even re-populate the
views when the only thing that needs to happen is to set the appropriate
FPS info in the floating TextView), but at least this avoids the inflation
overhead of the previous approach.
BUG=skia:
Change-Id: Ide1c736c49663cfcf478653a7cdaa99c736f1783
Reviewed-on: https://skia-review.googlesource.com/8461
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Chet Haase <chet@google.com>
Easier to work on SkJumper with SkSplicer out of my hair.
Change-Id: I6217082c5b4f83c79ac231f23e42a92d2efd6446
Reviewed-on: https://skia-review.googlesource.com/8509
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Add a null-check on geometry processor creation.
Add a test which exercises it: AA tessellator, with a non-invertible
matrix and a fragment processor which needs local coords (e.g., linear
gradient).
BUG=691902
Change-Id: I005b893aed58d3ad2500c41501045ac94b0b4b95
Reviewed-on: https://skia-review.googlesource.com/8462
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Also refer to it directly rather than using the term "modulate"
Change-Id: Ifa44a4d46e1be11b567943f58ead24e38f10d03b
Reviewed-on: https://skia-review.googlesource.com/8488
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Instead of trying to genereate snippets of code that an work when
splicing and when interpreting, let's just make one snippet for each.
And then, let's add one more portable single float version!
Why build one when you can have three at thrice the price?
This doesn't make these new strategies active, but does start moving the
code in build_stages.py and SkSplicer_stages.cpp towards that.
Change-Id: I2197fa2a11e233f24da67e54a4f291135652b087
Reviewed-on: https://skia-review.googlesource.com/8453
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
As a soft rollout, this only affects the Linux NUCS
(on Ubuntu16).
BUG=skia:6206
Change-Id: Ic314e85159cd3c0d9e55bbdce412a8d61adebb33
Reviewed-on: https://skia-review.googlesource.com/8276
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
This will prevent us from clipping the gamut
to sRGB.
BUG=skia:
Change-Id: Ifc34369d96aa9dd92ae2af72aac1cfa17fdc4b94
Reviewed-on: https://skia-review.googlesource.com/8025
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>
BUG=skia:6242
Change-Id: I63cce6019ce854ba807b2a819ef4fe3214898bee
Reviewed-on: https://skia-review.googlesource.com/8344
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>
BUG=skia:
Change-Id: I4f3c6370b3ef4247aa446716c7c154899925d089
Reviewed-on: https://skia-review.googlesource.com/8442
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
This also fixes bin/coverage.
fuzz/coverage borrows heavily from bin/coverage.
BUG=skia:
Change-Id: I9e353d1f5ea3bca1d57d66b1c1ecabc6f9b23cee
Reviewed-on: https://skia-review.googlesource.com/8414
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
