This tool can quickly check all nanobench tests including svgs
and skps (<5 minutes for CPU, ~30 minutes for GPU) and find
significant performance regressions without much noise.
This tool is not only faster (lower latency to get regression
alerts), but also more sensitive compared to our k-means and
step-fitting bots (especially for changes that only affect very
few benches). It may still miss some regressions, but the
regressions reported should be valid with very high probability.
Bug: skia:
Change-Id: I02115e6c5ab630e4c56b2087ffeb5cae1d4a618e
Reviewed-on: https://skia-review.googlesource.com/50060
Commit-Queue: Yuqian Li <liyuqian@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Change-Id: Ice135dd4b6316b791276eefd522be6bd90b8b5eb
Reviewed-on: https://skia-review.googlesource.com/53261
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Stan Iliev <stani@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
So we can benchmark with given width/height dimensions.
Bug: skia:
Change-Id: I881c9f054cdcecc3e0b2bfb600afd778e1c57d95
Reviewed-on: https://skia-review.googlesource.com/51246
Reviewed-by: Mike Reed <reed@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
This reverts commit 8b059bd946.
Reason for revert: <INSERT REASONING HERE>
Original change's description:
> Revert "Update GrBitmapTextureMaker for handling mipped requests"
>
> This reverts commit d1935c16e8.
>
> Reason for revert: breaking lots of GMs, especially on gles
>
> Original change's description:
> > Update GrBitmapTextureMaker for handling mipped requests
> >
> > Specifically this updates the case when we are requesting to use mip
> > maps but there is already an unmipped version in the cache. Previously
> > we just grabbed the unmipped.
> >
> > Now we will create a new mipped resource. Upload the cpu data to all
> > the levels besides the base, copy the base level on GPU from the
> > original resource to the mipped one. Then the mipped resource will
> > take over the originals unique key.
> >
> > Bug: skia:
> > Change-Id: I38e9725c93280dc2460a0be8a7a229e7f20e1614
> > Reviewed-on: https://skia-review.googlesource.com/43840
> > Commit-Queue: Greg Daniel <egdaniel@google.com>
> > Reviewed-by: Robert Phillips <robertphillips@google.com>
>
> TBR=egdaniel@google.com,robertphillips@google.com,brianosman@google.com
>
> Change-Id: Id82e8b6e8ab69e46ff018bb07ae5d1f6ea8d7e76
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:
> Reviewed-on: https://skia-review.googlesource.com/52901
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Greg Daniel <egdaniel@google.com>
TBR=egdaniel@google.com,robertphillips@google.com,brianosman@google.com
Bug: skia:
Change-Id: Ie5b9553aa67def6f9c5a61f9b6b065b9fce3ff76
Reviewed-on: https://skia-review.googlesource.com/53240
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
No-Try: true
Change-Id: I449c0fd435b233e9c9538c17295ef18348380658
Reviewed-on: https://skia-review.googlesource.com/53120
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
One of the nice bits of constexpr is that an expression is constexpr if
there exists any set of argument values that make it constant. It
doesn't have to be constant for _all_ argument values.
This means that this expression is constexpr:
condition ? constexpr_value
: []{ arbitrary non-constexpr code; }();
... it's constant when condition is true.
We can use this to rewrite SkASSERT(condition) as
( (condition) ? (void)0
: []{ SK_ABORT(#condition); }() )
Both sides of the ?: are void, and when condition is true at compile
time the right hand side disappears completely. In C++11 constexpr
functions we just have to use the comma operator to jam SkASSERT()
into the order of evaluation:
constexpr uint32_t foo(int x, int y) {
return SkASSERT(x > y),
x - y;
}
Change-Id: I21878d14fb2af76d93591d2ae229460ee825cfde
Reviewed-on: https://skia-review.googlesource.com/52663
Reviewed-by: Ben Wagner <bungeman@google.com>
Reviewed-by: Trent Apted <tapted@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit 9f8d4d36b5.
Bug: skia:
Change-Id: I8d7c1df24d8b13b94404f3d9ba69a1ab55ee00c0
Reviewed-on: https://skia-review.googlesource.com/52920
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This reverts commit d1935c16e8.
Reason for revert: breaking lots of GMs, especially on gles
Original change's description:
> Update GrBitmapTextureMaker for handling mipped requests
>
> Specifically this updates the case when we are requesting to use mip
> maps but there is already an unmipped version in the cache. Previously
> we just grabbed the unmipped.
>
> Now we will create a new mipped resource. Upload the cpu data to all
> the levels besides the base, copy the base level on GPU from the
> original resource to the mipped one. Then the mipped resource will
> take over the originals unique key.
>
> Bug: skia:
> Change-Id: I38e9725c93280dc2460a0be8a7a229e7f20e1614
> Reviewed-on: https://skia-review.googlesource.com/43840
> Commit-Queue: Greg Daniel <egdaniel@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
TBR=egdaniel@google.com,robertphillips@google.com,brianosman@google.com
Change-Id: Id82e8b6e8ab69e46ff018bb07ae5d1f6ea8d7e76
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/52901
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
ANGLE now returns GL_HALF_FLOAT as the GL_IMPLEMENTATION_READ_TYPE for
ES3 contexts.
BUG=765953
CQ_INCLUDE_TRYBOTS=skia.primary:Perf-Win10-MSVC-AlphaR2-GPU-RadeonR9M470X-x86_64-Debug-ANGLE,Perf-Win10-MSVC-Golo-GPU-QuadroP400-x86_64-Debug-ANGLE,Perf-Win10-MSVC-NUC5i7RYH-GPU-IntelIris6100-x86_64-Debug-ANGLE,Perf-Win10-MSVC-NUC6i5SYK-GPU-IntelIris540-x86_64-Debug-ANGLE,Perf-Win10-MSVC-NUCD34010WYKH-GPU-IntelHD4400-x86_64-Debug-ANGLE,Perf-Win10-MSVC-ShuttleC-GPU-GTX960-x86_64-Debug-ANGLE,Test-Win10-MSVC-AlphaR2-GPU-RadeonR9M470X-x86_64-Debug-ANGLE,Test-Win10-MSVC-Golo-GPU-QuadroP400-x86_64-Debug-ANGLE,Test-Win10-MSVC-NUC5i7RYH-GPU-IntelIris6100-x86_64-Debug-ANGLE,Test-Win10-MSVC-NUC6i5SYK-GPU-IntelIris540-x86_64-Debug-ANGLE,Test-Win10-MSVC-NUCD34010WYKH-GPU-IntelHD4400-x86_64-Debug-ANGLE,Test-Win10-MSVC-ShuttleC-GPU-GTX960-x86_64-Debug-ANGLE
Change-Id: I556aecc040f0274b769a3b551b7032adf4ab035a
Reviewed-on: https://skia-review.googlesource.com/52523
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Change-Id: Ie9afd2de2bfb9bcd04b5f8b7f319a24025520bef
Reviewed-on: https://skia-review.googlesource.com/52400
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Bug: 769134
readByteArray can fail (due to not having enough available or due to the
wrong alignment). If it does, do not return an uninitialized block of
memory.
Further, drop the initial size check, which is covered by readByteArray.
Add a test.
Change-Id: Ia101697c5bb1ca3ae3df1795f37a74b2f602797d
Reviewed-on: https://skia-review.googlesource.com/52742
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
Make Chinese fling sample closer to Android test.
Fix a bug in GrDrawOpAtlas::compact(), where the atlas
generation wasn't incremented so the client didn't know it changed.
Add some debug info for GrDrawOpAtlas.
Add a comment in GrDrawOpAtlas::setLastUseTokenBulk.
Change-Id: I79192a017870541a79731b1a22f665ec5deeff09
Reviewed-on: https://skia-review.googlesource.com/52761
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Specifically this updates the case when we are requesting to use mip
maps but there is already an unmipped version in the cache. Previously
we just grabbed the unmipped.
Now we will create a new mipped resource. Upload the cpu data to all
the levels besides the base, copy the base level on GPU from the
original resource to the mipped one. Then the mipped resource will
take over the originals unique key.
Bug: skia:
Change-Id: I38e9725c93280dc2460a0be8a7a229e7f20e1614
Reviewed-on: https://skia-review.googlesource.com/43840
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
It's too inexact as-is. If the caller wants tolerance they can do
their own with knowledge of the pixel grid. The homogeneous math
is stable with infinities so it's really unnecessary here.
Bug: skia:7073
Change-Id: I4dc34ad96b859a138714b6d4f8804fec4f89f17a
Reviewed-on: https://skia-review.googlesource.com/51182
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Bug: skia:
Change-Id: I9302491c983fae5e2e84edc4ee135a917b099a42
Reviewed-on: https://skia-review.googlesource.com/52360
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
This is something I came up with while writing _lowp.cpp.
This should all be a logical no-op, but there are some code generation
changes. I'm not exactly sure why.
Change-Id: Iaad36b5298b37fe26ebd375a147a48852f98e1e4
Reviewed-on: https://skia-review.googlesource.com/52003
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
After updating the transparency index code, the reader parameter to
SkGIFFrameContext's ctor is no longer needed. This patch removes it.
BUG=skia:7069
Change-Id: If129f825639e8d43d73794adca2de09785f56a3c
Reviewed-on: https://skia-review.googlesource.com/52602
Commit-Queue: Chris Blume <cblume@chromium.org>
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>
The lowp start_pipeline() always zeros, and with floats we always zero
when compiled as part of Skia, so this just makes the offline float
consistent with the others.
It's getting confusing to think about which code zeros and which
doesn't, and it'd be nicer to be able to rely on zeros.
This should change code generation only to the start_pipelines in
the .S files.
Change-Id: I1178b83c01e609e40dc7912d8d56df8e36eb339d
Reviewed-on: https://skia-review.googlesource.com/52001
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit 837c6c7c0c.
Bug: skia:
Change-Id: I1821f1b2b772c67f1b749692b398eb757d8073c9
Reviewed-on: https://skia-review.googlesource.com/52744
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
We look at t to create a mask in mask_2pt_conical_degenerates to be
applied later to the colors after the normal gradient stages have run.
But if t itself is NaN, that will wreak havoc in the normal gradient
stages. So in addition to building the mask to kill off degenerate
colors, let's also set degenerate t to zero, which should be a safe
value.
This fixes the fuzz mentioned in this bug.
BUG=skia:7078
Change-Id: I8301450c707bdbf941abd0339959f9e60d46d955
Reviewed-on: https://skia-review.googlesource.com/52763
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Bug: skia:7077
Change-Id: I5119a7d6c7cd2317129d40ae3e0997dfc9978c25
Reviewed-on: https://skia-review.googlesource.com/52664
Reviewed-by: Ravi Mistry <rmistry@google.com>
Commit-Queue: Eric Boren <borenet@google.com>
Change-Id: Id2ae524f765414e4c3d990066fcd3ebcea693657
Reviewed-on: https://skia-review.googlesource.com/52781
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Bug: skia:
Change-Id: I3ea8c8acf56d103a8f38e652b418239e0f517839
Reviewed-on: https://skia-review.googlesource.com/52720
Reviewed-by: Robert Phillips <robertphillips@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
All scan converters need an SkIRect clipBounds and a bool
containedInClip. However, we previously sent in an SkRegion
and a SkIRect*, and convert them into clipBounds and
containedInClip all over the places. This CL converts them
only once inside do_fill_path and change all args to SkIRect
and bool.
Bug: skia:
Change-Id: I05f1d76322942d8817860fd33991f7f7ce918e7c
Reviewed-on: https://skia-review.googlesource.com/52741
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
This reverts commit 742f3d02a1.
Reason for revert: Aaah!
Original change's description:
> Make threaded proxy generation MDB-friendly, and defer instantiation
>
> Replaces GrPrepareCallback with GrDeferredProxyUploader, stored directly
> on GrTextureProxy. Op lists now store a list of referenced proxies that
> are being generated by worker threads. At flush time, iterate over those
> proxies, and invoke their uploader.
>
> Lifetime of the uploader object is now tied to the proxy, but the ASAP
> upload function will free the proxy's uploader, if it's called.
>
> Bug: skia:
> Change-Id: Ieb2c6a805d19990012839a8e103c3ca5b8d3dfc6
> Reviewed-on: https://skia-review.googlesource.com/49904
> Commit-Queue: Brian Osman <brianosman@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
TBR=egdaniel@google.com,robertphillips@google.com,brianosman@google.com
Change-Id: I8f76a67044dc4159f903097d8b1ef19ffb48c730
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/52760
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Replaces GrPrepareCallback with GrDeferredProxyUploader, stored directly
on GrTextureProxy. Op lists now store a list of referenced proxies that
are being generated by worker threads. At flush time, iterate over those
proxies, and invoke their uploader.
Lifetime of the uploader object is now tied to the proxy, but the ASAP
upload function will free the proxy's uploader, if it's called.
Bug: skia:
Change-Id: Ieb2c6a805d19990012839a8e103c3ca5b8d3dfc6
Reviewed-on: https://skia-review.googlesource.com/49904
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
No-Try: true
Docs-Preview: https://skia.org/?cl=52163
Change-Id: I2bcd73bc7597219e4748c28e9120b5138a0eb3d1
Reviewed-on: https://skia-review.googlesource.com/52163
Reviewed-by: Eric Boren <borenet@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Several upcoming additions should go in here
Change-Id: I642f3c7cc36b1e6512ee0170640449e88a666d2c
Reviewed-on: https://skia-review.googlesource.com/52661
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
TBR=bsalomon@google.com
Bug: skia:
Change-Id: If12c3b44de76a2fed24dd527cb774fe5be270e8e
Reviewed-on: https://skia-review.googlesource.com/52260
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
The TSAN bots fail regularly with races on fConvexity. Very annoying.
We used to have this very same problem with SkPath::fFirstDirection
until we made it atomic. This does the same to fConvexity.
This makes the field as lightly atomic as possible, with all operations
using a relaxed memory order. The value of fConvexity isn't guarding
any other non-atomic memory or implying any other writes have happened
so I don't think we need anything beyond relaxed here.
Change-Id: I0da1f892dc2b7072d692ce8b460fb1862aebef77
Reviewed-on: https://skia-review.googlesource.com/52180
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit 88757dacd4.
Reason for revert: Still seems to be failing Chromium "telemetry_perf_unittests (with patch) on Android" on android_n5x_swarming_rel.
Original change's description:
> guard old apis for querying byte-size of a bitmap/imageinfo/pixmap
>
> Now with legacy behavior for allocpixels
>
> This was reverted, so the current CL is a "fix" on top of ...
> https://skia-review.googlesource.com/c/skia/+/50980
>
> Related update to Chrome (in preparation for this change)
> https://chromium-review.googlesource.com/c/chromium/src/+/685719
>
> Bug: skia:
> Change-Id: I4b370ee7e95083ab27421f008132219c9c7b86e9
> Reviewed-on: https://skia-review.googlesource.com/51341
> Reviewed-by: Florin Malita <fmalita@chromium.org>
> Commit-Queue: Mike Reed <reed@google.com>
TBR=fmalita@chromium.org,reed@google.com
Change-Id: I827a0ca1d1e3909e648fde3342cdb8601d34da8d
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/52381
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Bug: skia:3550
Change-Id: I00f83f40b30549cbfcb65e6a39afba6355d62299
Reviewed-on: https://skia-review.googlesource.com/52420
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Other browsers (including old Chrome) treat invalid palette indices as
transparent for gifs. And there are gifs in the wild which rely on this.
As an example, if the palette only has 64 entries (0-63) then index 64
is treated as transparent.
BUG=skia:7069
Change-Id: I15e8919a953387506c9ac5945c3ae6a2b90189ab
Reviewed-on: https://skia-review.googlesource.com/51100
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Chris Blume <cblume@google.com>
Bug: b/65290323
If a webp file is truncated such that no rows can be decoded,
WebPIDecGetRGB does not initialize its "last_y" parameter. We use
rowsDecoded (passed as last_y) to determine which remaining rows to
fill.
Check the return value of WebPIDecGetRGB. If it fails (returns null),
or rowsDecoded is <= 0 (matching Chromium's check), return
kInvalidInput, since there is nothing to draw.
Note that this is a change in behavior for Android. Previously we
would decode an empty webp to just a transparent/black rectangle,
whereas now we simply fail. I think this is a change for the better.
Add a test which truncates a file to have 0 rows available and attempts
to decode it. msan verifies that we no longer depend on the
uninitialized value.
Stop attempting to test decoding subsets from an incomplete webp (in
CodecTest.cpp). Unless we have decoded the portion covered by the
subset, this will fail.
Remove test images inc0.webp (from both dm/ and colorspace/) and
inc1.webp. These just decode to transparent rectangles. Replace them
with inc2.webp and inc3.webp, which decode part of the image and then
have to fill with transparent.
Change-Id: I64d40be91c574b45963f9a43d8dd8f4929dd2939
Reviewed-on: https://skia-review.googlesource.com/50303
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: James Zern <jzern@google.com>
Bug: skia:
Change-Id: Id6808734df2335c521d4a563122ceeca906b7601
Reviewed-on: https://skia-review.googlesource.com/52002
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This is a reland of https://skia-review.googlesource.com/c/skia/+/42940
which was reverted since its parent CL had to be reverted. The parent
has since relanded.
Bug: skia:
Change-Id: Ibf4bf7ce0332e4faea9a6e95a40d1c1f07fb0e75
Reviewed-on: https://skia-review.googlesource.com/51244
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Bug: skia:
Change-Id: I55641aa4bef5a7ac863e3aae3d2902ef408f0384
Reviewed-on: https://skia-review.googlesource.com/52121
Commit-Queue: Eric Boren <borenet@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
