This entails a few parts:
1. Isolating tools/skpbench/ to be used by the perf task.
2. Adding skpbench to BUILD_PRODUCTS_ISOLATE_WHITELIST
3. Add recipe commands to run skpbench. While it is conceptually similar
to nanobench, it doesn't support images nor svgs, only skps, so we don't
need to copy those to devices or bring them in via CIPD.
4. Add recipe commands to parse skpbench output and upload to Perf.
5. Update gen_tasks.go to handle skpbench jobs and tasks.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4551
Change-Id: I09bf0b078067279e138e95eca9316e55539ec735
Reviewed-on: https://skia-review.googlesource.com/4551
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
- Appends the clock and unit to the result type (e.g. "accum_cpu_ms").
- Removes the "options" key.
- Adds "bench_type" and "source_type" keys to the properties.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4737
Change-Id: I2fd27eac4f42f443131fef6b1774f20e60571cd5
Reviewed-on: https://skia-review.googlesource.com/4737
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Use anisotropic scale to show bug in verylargebitmap test.
BUG=chromium:664615
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4774
Change-Id: If92eab6d26b9fb506670412a80df259f99db2a21
Reviewed-on: https://skia-review.googlesource.com/4774
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This file has been renamed SkOSPath.cpp and users no longer refer to the
old name. Remove this now empty file and its build rule.
Change-Id: I5041b61e749cc86f491b6c99b6744d7a0d4f9d1c
Reviewed-on: https://skia-review.googlesource.com/4766
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Ben Wagner <bungeman@google.com>
Replaces with switch statements that will produce a compiler warning if a type is added and helper functions are not updated.
This also removes several unused helper functions for these enum types.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4486
Change-Id: Iee6d8e5095dc4ba2488a8928bf60f8e903cd3320
Reviewed-on: https://skia-review.googlesource.com/4486
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Running is_clang.py is the current long-poll in `gn gen` time.
We can avoid it trivially in a few situations:
- We always use Clang on Android, iOS, and Mac.
- If cc and cxx are clang and clang++, it's Clang.
This cuts `gn gen` time from 80ms to 20ms on my laptop.
(Did you know gn has a --tracelog=trace.log option for creating Chrome-tracing-compatible traces? Pretty neat.)
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4746
Change-Id: Ic81d221675e8309cc9942bb9e62243d86658b02d
Reviewed-on: https://skia-review.googlesource.com/4746
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Ben Wagner <bungeman@google.com>
Here first just the simplest, constant-color shaders.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4743
Change-Id: I92c6523660e21a1e2aa353524570230282ba5dfe
Reviewed-on: https://skia-review.googlesource.com/4743
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This should fix Build-Mac-Clang-arm64-Debug-GN_iOS_NoBuildbot.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4755
Change-Id: Ie49ce5642fb2d373102c9309074e13ee3035a569
Reviewed-on: https://skia-review.googlesource.com/4755
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit 238b820369.
Reason for revert: breaks 32-bit bots, iOS bots, Google3 roll, -ASAN bot.
Original change's description:
> Always build the ANGLE test code. Always build ANGLE on windows and linux.
>
> Make ANGLE test code independent of having ANGLE lib. Make ANGLE test code not include EGL headers.
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4040
>
> DOCS_PREVIEW= https://skia.org/?cl=4040
>
> Change-Id: I7b857e9785246743f53fb969647b1162ce7419ab
> Reviewed-on: https://skia-review.googlesource.com/4040
> Commit-Queue: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Mike Klein <mtklein@chromium.org>
>
TBR=mtklein@chromium.org,mtklein@google.com,bsalomon@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I19bab8c93baebf032f8a4cefbedfe7359317e806
Reviewed-on: https://skia-review.googlesource.com/4758
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
If these are not marked static (or SK_FORCE_INLINE) they become extremely dangerous to use from files built with different optimization flags than the baseline. The One Definition Rule becomes trivially easy to violate, and the linker can easily pick a CPU-specific version of the function.
TBR=reed@google.com
BUG=chromium:664864
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4752
Change-Id: If3044a32229db1a1e3a8b83aace6047d92b36d05
Reviewed-on: https://skia-review.googlesource.com/4752
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
These are so easy we might as well...
I did a quick check of relevant-looking defines:
- GYP defined WITH_SIMD, but it looks like that's already defined (by jconfig.h?);
- GYP defined RGBX_FILLER_0XFF, but that affects only x86/x86-64;
- GYP defined STRICT_MEMORY_ACCESS, which does nothing;
- GYP defined MOTION_JPEG_SUPPORTED, which does nothing (and we'd probably not care anyway).
BUG=skia:5875
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4745
Change-Id: Ib1f28d354630be472c4d9648d5ade74a452a9e24
Reviewed-on: https://skia-review.googlesource.com/4745
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
These libraries aren't configured in Fuchsia and would need some GN
reworking to enable. We can wire this up in Fuchsia if/when this
capability when needed.
Change-Id: Iec2d339da07f44516592d6114563a8a4fb59d952
Reviewed-on: https://skia-review.googlesource.com/4744
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
It has been incorrectly interpreting its SkColor as sRGB all the time. Now, we plumb through the destintation color space and some scratch space, letting it decide how to interpret its SkColor later when it knows about the dst color space. The scratch space is blitter scoped, which lets this be thread safe (this is much like SkShader::Context).
This only corrects the gamma transformation for now. I've kept my previous TODO about gamut transformation. Everything assumes sRGB gamut for now.
Shaders will get the same treatement in this pipeline.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4725
Change-Id: I55b0c7d5db9ad8d7dcdd6295c9dac61d10aeaed4
Reviewed-on: https://skia-review.googlesource.com/4725
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit c51a3a4e3c.
Reason for revert: had to revert this in the copy constructor. Might as well revert here too for consistency.
Original change's description:
> Annotate benign race on SkPath::fConvexity.
>
> We're tired of this making our TSAN bot flaky, and equally tired of trying to fix it.
>
> BUG=skia:5776
>
> CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-Clang-GCE-CPU-AVX2-x86_64-Release-TSAN-Trybot
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4691
>
> Change-Id: Ib6005282c514a6d785fd4fffe5387cbb1caccbe1
> Reviewed-on: https://skia-review.googlesource.com/4691
> Commit-Queue: Mike Klein <mtklein@chromium.org>
> Reviewed-by: Ben Wagner <bungeman@google.com>
>
TBR=mtklein@chromium.org,bungeman@google.com,reed@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I906b1e017471d9e00fc58402136497459d881f72
Reviewed-on: https://skia-review.googlesource.com/4740
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
This reverts commit f978f12af5.
Reason for revert: slowing the TSAN bot down to the point of timeouts.
Original change's description:
> Annotate SkPath::fCovexity benign race in the other constructor too.
>
> I overlooked the copy constructor last time.
>
> BUG=skia:5776
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4730
>
> Change-Id: I30bc89e1472dd48badf57664cfae8899f44bca9e
> Reviewed-on: https://skia-review.googlesource.com/4730
> Reviewed-by: Mike Klein <mtklein@chromium.org>
> Commit-Queue: Mike Klein <mtklein@chromium.org>
>
TBR=mtklein@chromium.org,reed@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I6e34b1fc80fd98387d79521ce24f4545b990eb25
Reviewed-on: https://skia-review.googlesource.com/4739
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
There is support for all features of SkColorSpace_A2B.
Tests for these functionality were adapted from
the XYZ xform, plus a CLUT-specific test was added.
Shared functions used by both SkColorSpaceXform_XYZ and SkColorSpaceXform_A2B
have been moved into a shared header.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2449243003
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Review-Url: https://codereview.chromium.org/2449243003
Extract SkPackedID and its strongly typed subclasses SkPackedGlyphID and
SkPackedUnicharID out of SkGlyph. This simplifies the code handling
these types, as well as making it clearer that we wouuld eventually like
to get away from this scheme.
Changes SkScalerContext::getPath to take SkPackedGlyphID.
Changes SkScalerContext::generatePath to take SkGlyphID.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4722
Change-Id: I365c0c618b7ae0d348272155fac7761a69faa920
Reviewed-on: https://skia-review.googlesource.com/4722
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Herb Derby <herb@google.com>
If we don't snap, we could have end up with many tiny line segments
wihtin a single row of pixels. That combined with our partial alpha
rounding code (https://codereview.chromium.org/2483523002/) could
accumlate so much error that violates our alpha <= 256 assert.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4724
Change-Id: I58b02cffc8a4dc9f541e194d61a2f6fef7b85c97
Reviewed-on: https://skia-review.googlesource.com/4724
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
(1) Compile tool with GN.
(2) Rename tool to colorspaceinfo.
(3) Support both images and icc profiles as input.
(4) Print out information on transfer fn in addition to gamut.
TODO: Output graphs of the three transfer fn curves.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4721
Change-Id: I2e75139685ea41446d3ae6f9803c8068ea05661a
Reviewed-on: https://skia-review.googlesource.com/4721
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Brian Osman <brianosman@google.com>
This sets the stage for using the Proxy's/RenderTargetContext's ID above the flush and the RenderTarget's/GrGpuResource's below the flush.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4650
Change-Id: I9f1e6b00c02a0691d90b58c49e1d8c60684884c1
Reviewed-on: https://skia-review.googlesource.com/4650
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This bug chromium:662780 exists after our original fix (https://codereview.chromium.org/2477393002/) because this path (added in unit test) is calling blitAntiRect rather than blitAntiH when the path is drifted across the boundary. (The quadratic edge drifts across the boundary after an update and sets a dX=0 line segment which triggers blitAntiRect.)
Note that I didn't assert for the dLeft = dRite = 0 case because the left/right there won't drift after the SkTMin/SkTMax in line 964/966.
Theoretically we can revert the relaxation in https://codereview.chromium.org/2477393002/ (that's only a relaxation for analytic AA, not supersampled AA). However, consider that the initial landing of analytic AA is so painful, I decide to revert that relaxation only after our successful landing...
BUG=chromium:662780, chromium:662862
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2482193004
Review-Url: https://codereview.chromium.org/2482193004
When we opt into Lua, this builds SampleLua into SampleApp, and the lua_app and lua_pictures tools.
I've tested this builds with and without skia_use_system_lua on my Mac laptop and Linux desktop.
I've made lua_pictures.cpp's flags static to avoid conflicts with flags in SkCommonFlags.cpp.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4699
DOCS_PREVIEW= https://skia.org/?cl=4699
Change-Id: I8176fd51d8a38746e7d730cfcce66da42b9a015a
Reviewed-on: https://skia-review.googlesource.com/4699
Reviewed-by: Hal Canary <halcanary@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Looking at SkSmallAlloc hasn't left me terribly impressed. I think we can replace it with something a lot simpler to work with.
That simpler thing's core would be something like SkFixedAlloc, which allocates objects out of a fixed sized buffer, and cleans them up when done.
If needed, we can wrap that with logic to try to allocate out of an SkFixedAlloc, falling back on mallc() when exhausted.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4658
Change-Id: I8d94156ddf98802e42ec0890cff0f06b21f073b0
Reviewed-on: https://skia-review.googlesource.com/4658
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Reed <reed@google.com>
