This change, when combined with https://skia-review.googlesource.com/c/buildbot/+/77701
should cut down on the duplicated work of extracting the large
toolchains from CIPD.
Since the Isolate* steps can be cached (i.e. are idempotent)
they will only run about 1/week (unless updated) and
all subsequent tasks (primarily Builds) will go much faster.
We estimate the overhead on Build-Debian-Android to go from
about 90s (which was more time than the actual build) to
about 10s. Build-Win-Vulkan's overhead will improve from
about 180s to about 35s (1/3 of which is uploading to isolate).
Other CIPD assets could be handled in a similar fashion;
the ones here are the biggest offenders and the lowest
hanging fruit. Doing this to other assets (e.g. clang_win)
would have minimal improvements (<10s).
There are other tasks with large amounts of overhead
(e.g. Build-Mac-Android, Build-Win-Android, Build-WASM)
but none of those are depended on by any Tests, so any
speed-ups would have less wide-reaching-impact, at the
cost of using more Isolate cache/diskspace.
See https://docs.google.com/document/d/1DFlcpqg7XqEPE5oYT1V3so2ih2285heS5w3mPT-GMBA/edit#
for more information.
Bug: skia:
Change-Id: I40dd87fe72c3d49292762a09dad6df0dfbe78f61
Reviewed-on: https://skia-review.googlesource.com/77560
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
This has the side effect of using the bots in the new GCE project as
well.
Bug: skia:7278
Change-Id: Ie14c93d6e3d12ccbfb679089bc50bca482fbf605
Reviewed-on: https://skia-review.googlesource.com/76261
Commit-Queue: Eric Boren <borenet@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
Bug: skia:
Change-Id: Iefef7d617e58de2b3be2e27aac075f822641e4ce
Reviewed-on: https://skia-review.googlesource.com/77641
Reviewed-by: Stephan Altmueller <stephana@google.com>
Commit-Queue: Eric Boren <borenet@google.com>
I want to compare this with the IntelBayTrail and if the results are
similar, replace those bots.
No-Try: true
Change-Id: Ib5476fe91dc446182cd1b37e93fe17962dcf961a
Reviewed-on: https://skia-review.googlesource.com/76900
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Change-Id: Iee28f684bd6fa541e36f677ee4261e637cbd4611
Reviewed-on: https://skia-review.googlesource.com/77201
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
For internal hardware, it tends not to work - they work at one
clock speed, despite advertising others.
Bug: skia:
NOTRY=true
Change-Id: I10bf0fc1ab4d60bfbc2eefcef5b42ceab9e3f435
Reviewed-on: https://skia-review.googlesource.com/76720
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Also drive-by cleanup of "Win8-MSVC-ShuttleB", none of which currently
exist.
No-Try: true
Change-Id: Ide28481939b7ec2a0733ab07673379c951123f5d
Reviewed-on: https://skia-review.googlesource.com/75361
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Bug: skia:
Change-Id: I36cb94877d513fc81c211b0e58b5c4be0451ac91
Reviewed-on: https://skia-review.googlesource.com/74601
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Bug: skia:
Change-Id: I391cbc6cb2bf2ae88af0612964f4265788c8e771
Reviewed-on: https://skia-review.googlesource.com/74600
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:
NOTRY=true
Change-Id: I1a1755dd03f2e6ebd8d9b2c9235cca8eb34f04ad
Reviewed-on: https://skia-review.googlesource.com/75280
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
This also makes *sure* the CPU frequency we set the
device to actually "takes". Along the way, I learned
if scaling_max_freq is < the frequency we set, the
scaling_max_freq will be used instead, which was
happening to the PixelCs and AndroidOnes.
As a result, this may make those two Test- configs faster.
Bug: skia:
Change-Id: I10c98d37e296a19e1cf67bfe7269bb59cdd912d5
Reviewed-on: https://skia-review.googlesource.com/74360
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
These will be replaced by the Nexus 5x.
Bug: skia:7309
No-Try: true
Change-Id: I2a56a494203f2af41f16dcfd55ebe1ca28e9e939
Reviewed-on: https://skia-review.googlesource.com/73881
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
Bug: skia:
Change-Id: I994f67c3043306d7fa612feb03f8fbe8d7bf4c91
Reviewed-on: https://skia-review.googlesource.com/73760
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Change-Id: I667e0b8206461df797f0f5e481fe78a3f928481a
Reviewed-on: https://skia-review.googlesource.com/73261
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Will remove after transition to new image.
Change-Id: I4254643aa1279b6e2046d2cdef9ff1d481f85531
Reviewed-on: https://skia-review.googlesource.com/73260
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
The old code made the wrong assumptions about premultiplication.
There are three relevant steps here for decoding a webp frame:
1 tell libwebp to decode
2 colorXform the result (sometimes)
3 blend with the prior frame (sometimes)
Rearrange the code to premultiply at the blend step, in a linear space.
If the client wants unpremul, the blend step will unpremul after.
If there is no blending, the colorXform (if any) will premultiply.
If only step 1 is necessary, let libwebp premultiply.
This fixes an animated image that has an opaque frame 0 followed by a
frame with alpha that blends with it.
Add the test image that failed (https://mathiasbynens.be/demo/animated-webp)
The prior fix is in 42bae8faa4. It did
not properly handle the colorXform when there was no blending step.
Change-Id: I2b9d265ba162eaf7e55a106c8f79341826cee0d3
Reviewed-on: https://skia-review.googlesource.com/72281
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
One (Win7) GDI bot is enough.
Change-Id: I03312c21e7b6da2a50225fd5dbc50bc69a6bd4c2
Reviewed-on: https://skia-review.googlesource.com/72640
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
The bots that execute these jobs produce noisy perf numbers.
No-Try: true
Change-Id: Icb3a54e324e0823f680042838b6f883ff5937f93
Reviewed-on: https://skia-review.googlesource.com/72200
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Bug: skia:7305
Change-Id: Ifb270cba27daaef75d3930f990e19215a251ca28
Reviewed-on: https://skia-review.googlesource.com/71921
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
This isn't used and has become a maintenance burden.
Change-Id: I5f3af8f91e5c4f073fe4ea30e0a7f1f61efeea47
Reviewed-on: https://skia-review.googlesource.com/70640
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Change-Id: I15b76d5da795ee01eb7e403721beebf5f67d1bc4
Reviewed-on: https://skia-review.googlesource.com/71920
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Bug: skia:7305
Change-Id: I18d9656ca74e27ed2c2ac18c55d056901a4ac683
Reviewed-on: https://skia-review.googlesource.com/71721
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Bug: skia:
Change-Id: Ie56f848889cfc7331109aed997a85bc42e27f60e
Reviewed-on: https://skia-review.googlesource.com/70724
Commit-Queue: Eric Boren <borenet@google.com>
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
Bug: skia:7296
Change-Id: Id30df3ed2264679338913f275caa69b8e2278075
Reviewed-on: https://skia-review.googlesource.com/70661
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
This reverts commit 42bae8faa4.
Reason for revert: Breaking GMs. A more extensive fix is needed.
Original change's description:
> Fix webp bug compositing alpha frames on opaque
>
> select_xform_alpha is used to determine how the color transform should
> handle alpha values. In a similar way, we're using it here to determine
> whether to premultiply pixels before blending them. In this case, the
> source is unpremul, so we should be premultiplying them, but since we
> are compositing on an opaque frame, the dst must be opaque and
> select_xform_alpha returns kOpaque. As a result, we do not premultiply
> (and even hint to the transform that the pixels are opaque). Since this
> all applies to the pre-blended pixels, we should not care that the dst
> is opaque. So drop the call to select_xform_alpha and just use the alpha
> type of the source. This matches the comment on the lines above.
>
> Add the test image that failed (https://mathiasbynens.be/demo/animated-webp)
>
> Change-Id: Ibd13c1f067bdf369ce1c882d4f6057aadccfa313
> Reviewed-on: https://skia-review.googlesource.com/71560
> Commit-Queue: Leon Scroggins <scroggo@google.com>
> Reviewed-by: Mike Klein <mtklein@chromium.org>
TBR=mtklein@chromium.org,scroggo@google.com
Change-Id: I6f535ff9b773a93e02a0358b830291594a6e738c
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/71720
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
select_xform_alpha is used to determine how the color transform should
handle alpha values. In a similar way, we're using it here to determine
whether to premultiply pixels before blending them. In this case, the
source is unpremul, so we should be premultiplying them, but since we
are compositing on an opaque frame, the dst must be opaque and
select_xform_alpha returns kOpaque. As a result, we do not premultiply
(and even hint to the transform that the pixels are opaque). Since this
all applies to the pre-blended pixels, we should not care that the dst
is opaque. So drop the call to select_xform_alpha and just use the alpha
type of the source. This matches the comment on the lines above.
Add the test image that failed (https://mathiasbynens.be/demo/animated-webp)
Change-Id: Ibd13c1f067bdf369ce1c882d4f6057aadccfa313
Reviewed-on: https://skia-review.googlesource.com/71560
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Change-Id: I7df9d1a9118fbd9c545912a38af4f94276f20b0f
Reviewed-on: https://skia-review.googlesource.com/71521
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
These won't have the "Win" or "Win10" bits in them anymore.
Change-Id: I2917c4227efcac7c2169a111fdaf62fcd83ea94a
Reviewed-on: https://skia-review.googlesource.com/70800
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
- Replace build-time GDI support in all of Skia
with run-time GDI support only in DM.
- Make the GDI bots NativeFont bots paralelling
the other NativeFonts bots.
Change-Id: I424b20f6983d8a8ba8574650efefea2b8776bbe2
Reviewed-on: https://skia-review.googlesource.com/70721
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
readd [scaled_]tilemodes_npot to blacklist after it was removed last week
since it is failing randomly on bots.
Bug: skia:
Change-Id: Ieef97470c1ae4c960fa6efb7294e2ff203cc0bd6
Reviewed-on: https://skia-review.googlesource.com/70740
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
