All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=811452

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=4186a92ccc2285995e23eb103c825bfccf851d1077b2b0edb5fce86a26ae5c10


Bot(s) for this bug's original alert(s):

chromium-rel-win7-gpu-intel
win-high-dpi

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/16972ff5840000

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/16972ff5840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14a7a603840000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/14a7a603840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/16a31143840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/16aed536440000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14a7463e440000

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/14a7463e440000

Roll skia/third_party/externals/skcms/ b117b4e7b..645741555 (1 commit) by skcms-skia-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com
https://skia.googlesource.com/skia/+/aee01c298519c9d08d7cc13fd7c4264d01f8630e

Roll src/third_party/skia/ 0b33cc42b..76546506d (5 commits) by skia-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com
https://chromium.googlesource.com/chromium/src/+/ed59c79fe42e87604b0d99006572b6d70a97311e

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/16aed536440000

Assigning to bsalomon, who was the sheriff on the skia rolls. Sorry this bug is confusing; it looks like one of the regressions didn't reproduce (removed from the bug), and one reproduced, but if you look at the graphs from the bisect in #9 (https://pinpoint-dot-chromeperf.appspot.com/job/14a7463e440000) you'll see:

1) The test broke at skcms roll https://skia.googlesource.com/skia/+/aee01c298519c9d08d7cc13fd7c4264d01f8630e
2) The test was fixed but performance also regressed at or before skia roll https://chromium.googlesource.com/chromium/src/+/ed59c79fe42e87604b0d99006572b6d70a97311e

Any ideas how to dig in further? Also adding smoothness.tough_filters_cases owner senorblanco

Greg, looking at the roll in #11 2), your change is the only one that I seems like it could potentially impacted this test. Do you think that is possible?

I suppose it possible, are there directions for running these tests locally in one of the links?

The following revision refers to this bug:
  https://skia.googlesource.com/skia/+/b78dd5d01eb16ae3cb9104ce8c0fa7e861431259

commit b78dd5d01eb16ae3cb9104ce8c0fa7e861431259
Author: Greg Daniel <egdaniel@google.com>
Date: Wed Mar 14 14:09:13 2018

Add back missing unique key checks when creating CachedBitmap/Image Proxies

Its possible that this could fix perf regression in Chrome.

Bug:  811452 
Change-Id: I2d4f7827092b361469586580f0c7c843ab2d5cec
Reviewed-on: https://skia-review.googlesource.com/114280
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>

[modify] https://crrev.com/b78dd5d01eb16ae3cb9104ce8c0fa7e861431259/src/gpu/SkGr.cpp

For future reference, note that the bug cites the "Filter Terrain" subtest, which you can run locally by opening this page: 

http://letmespellitoutforyou.com/samples/svg/filter_terrain.svg

Use Chrome's FPS counter to check for regressions.

Alternately, you can also run the telemetry test itself, something like 

telemetry/run_benchmark smoothness.tough_filters_cases

(See https://github.com/catapult-project/catapult/blob/master/telemetry/docs/run_benchmarks_locally.md)

The following revision refers to this bug:
  https://skia.googlesource.com/skia/+/4f57eb8002a0488cfbf08717b25af93fae37a39a

commit 4f57eb8002a0488cfbf08717b25af93fae37a39a
Author: Greg Daniel <egdaniel@google.com>
Date: Wed Mar 14 17:22:32 2018

Revert "Add back missing unique key checks when creating CachedBitmap/Image Proxies"

This reverts commit b78dd5d01eb16ae3cb9104ce8c0fa7e861431259.

Reason for revert: possibly breaking chrome roll

Original change's description:
> Add back missing unique key checks when creating CachedBitmap/Image Proxies
> 
> Its possible that this could fix perf regression in Chrome.
> 
> Bug:  811452 
> Change-Id: I2d4f7827092b361469586580f0c7c843ab2d5cec
> Reviewed-on: https://skia-review.googlesource.com/114280
> Commit-Queue: Greg Daniel <egdaniel@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>

TBR=egdaniel@google.com,bsalomon@google.com,robertphillips@google.com

Change-Id: I71646befd07bf28442ac3b9225c021cd141bf398
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug:  811452 
Reviewed-on: https://skia-review.googlesource.com/114422
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>

[modify] https://crrev.com/4f57eb8002a0488cfbf08717b25af93fae37a39a/src/gpu/SkGr.cpp

The following revision refers to this bug:
  https://skia.googlesource.com/skia/+/a421112ac4a9836b2dc910a55d3139de12a1654b

commit a421112ac4a9836b2dc910a55d3139de12a1654b
Author: Greg Daniel <egdaniel@google.com>
Date: Wed Mar 14 19:09:39 2018

Reland "Add back missing unique key checks when creating CachedBitmap/Image Proxies"

This reverts commit 4f57eb8002a0488cfbf08717b25af93fae37a39a.

Reason for revert: Doesn't look to be cause of chrome roll failures

Original change's description:
> Revert "Add back missing unique key checks when creating CachedBitmap/Image Proxies"
> 
> This reverts commit b78dd5d01eb16ae3cb9104ce8c0fa7e861431259.
> 
> Reason for revert: possibly breaking chrome roll
> 
> Original change's description:
> > Add back missing unique key checks when creating CachedBitmap/Image Proxies
> > 
> > Its possible that this could fix perf regression in Chrome.
> > 
> > Bug:  811452 
> > Change-Id: I2d4f7827092b361469586580f0c7c843ab2d5cec
> > Reviewed-on: https://skia-review.googlesource.com/114280
> > Commit-Queue: Greg Daniel <egdaniel@google.com>
> > Reviewed-by: Brian Salomon <bsalomon@google.com>
> > Reviewed-by: Robert Phillips <robertphillips@google.com>
> 
> TBR=egdaniel@google.com,bsalomon@google.com,robertphillips@google.com
> 
> Change-Id: I71646befd07bf28442ac3b9225c021cd141bf398
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug:  811452 
> Reviewed-on: https://skia-review.googlesource.com/114422
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Greg Daniel <egdaniel@google.com>

TBR=egdaniel@google.com,bsalomon@google.com,robertphillips@google.com

Change-Id: I957d66123bce150387a1ce6a80bcfc06c82d5fa1
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug:  811452 
Reviewed-on: https://skia-review.googlesource.com/114426
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>

[modify] https://crrev.com/a421112ac4a9836b2dc910a55d3139de12a1654b/src/gpu/SkGr.cpp

Looking at the graphs we've had three runs in a row all back at the original perf levels, so marking this as fixed.