Issue metadata
Sign in to add a comment
|
1.1%-3.3% regression in memory.top_10_mobile at 424717:424762 |
||||||||||||||||||||
Issue descriptionSee the link to graphs below.
,
Oct 13 2016
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8998902626694018736
,
Oct 13 2016
=== Auto-CCing suspected CL author skia-deps-roller@chromium.org === Hi skia-deps-roller@chromium.org, the bisect results pointed to your CL below as possibly causing a regression. Please have a look at this info and see whether your CL be related. ===== BISECT JOB RESULTS ===== Status: completed ===== SUSPECTED CL(s) ===== Subject : Roll src/third_party/skia/ 5772eaa91..7ab96e921 (2 commits). Author : skia-deps-roller Commit description: https://chromium.googlesource.com/skia.git/+log/5772eaa91d6a..7ab96e92196d $ git log 5772eaa91..7ab96e921 --date=short --no-merges --format='%ad %ae %s' 2016-10-12 senorblanco GrTessellator: make inverse fill types more sane. 2016-10-12 ethannicholas Turned on SkSL->GLSL compiler GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003 CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel TBR=jvanverth@google.com Review-Url: https://codereview.chromium.org/2419523002 Cr-Commit-Position: refs/heads/master@{#424733} Commit : ce86adf7be9ee2f397d9a1a093c0379ba6351a88 Date : Wed Oct 12 14:47:03 2016 ===== TESTED REVISIONS ===== Revision Mean Std Dev N Good? chromium@424732 40086467 367618 12 good chromium@424732,skia@7ab96e9219 39603139 917167 12 good chromium@424733 40712216 451512 12 bad <-- chromium@424734 40576877 509220 12 bad chromium@424736 40948376 596900 8 bad chromium@424739 40651416 297381 8 bad Bisect job ran on: android_nexus5_perf_bisect Bug ID: 655673 Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests memory.top_10_mobile_stress Test Metric: memory:chrome:all_processes:reported_by_os:system_memory:native_heap:proportional_resident_size_avg/foreground/http_en_m_wikipedia_org_wiki_Science Relative Change: 1.61% Score: 99.8 Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5_perf_bisect/builds/4240 Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998902626694018736 Not what you expected? We'll investigate and get back to you! https://chromeperf.appspot.com/bad_bisect?try_job_id=5791598137311232 | O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq | X | for more information addressing perf regression bugs. For feedback, | / \ | file a bug with component Tests>AutoBisect. Thank you!
,
Oct 14 2016
,
Oct 18 2016
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8998454975041296848
,
Oct 18 2016
Started bisect from here: https://chromeperf.appspot.com/report?sid=9ed3ae8162fbda8cb382f6dcdce1c1fa9ff5489215c620cf6f825f013f778088&start_rev=424402&end_rev=425624
,
Oct 18 2016
=== Auto-CCing suspected CL author ethannicholas@google.com === Hi ethannicholas@google.com, the bisect results pointed to your CL below as possibly causing a regression. Please have a look at this info and see whether your CL be related. ===== BISECT JOB RESULTS ===== Status: completed ===== SUSPECTED CL(s) ===== Subject : Turned on SkSL->GLSL compiler Author : ethannicholas Commit description: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003 Committed: https://skia.googlesource.com/skia/+/9b0fe3d125f237d9884732a48414fa85fc71b4e3 Committed: https://skia.googlesource.com/skia/+/b12b3c6908c62c908b3680be01e3b5bfd30de310 Committed: https://skia.googlesource.com/skia/+/f008b0a59f45c0d4bea3e66faf3b01805009ec89 Committed: https://skia.googlesource.com/skia/+/08b2ccf398e2b81bc05d2c105837e5419899469b Committed: https://skia.googlesource.com/skia/+/dcfe6dba4a335e50e86ff68e3252065d4197432c Committed: https://skia.googlesource.com/skia/+/ccb1dd8f267f9d7fe7c9d0ce222ebc81b41853b3 Review-Url: https://codereview.chromium.org/2288033003 Commit : 5961bc9278a00e56dacdd9408d0744b5a0a3b493 Date : Wed Oct 12 13:39:56 2016 ===== TESTED REVISIONS ===== Revision Mean Std Dev N Good? chromium@424722 26484154 85258.7 5 good chromium@424732 26420256 18559.6 5 good chromium@424732,skia@5961bc9278 27078483 84000.5 5 bad <-- chromium@424732,skia@7ab96e9219 27085651 28830.6 5 bad chromium@424733 27125382 85877.5 5 bad chromium@424734 27161632 52489.4 5 bad chromium@424735 27107155 48580.1 5 bad chromium@424737 27065581 48526.1 5 bad chromium@424742 27130502 45700.7 5 bad chromium@424761 27089133 35791.7 5 bad Bisect job ran on: android_nexus5_perf_bisect Bug ID: 655673 Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=load.search.yandex system_health.memory_mobile Test Metric: memory:chrome:all_processes:reported_by_os:system_memory:native_heap:proportional_resident_size_avg/load_search/load_search_yandex Relative Change: 2.28% Score: 99.9 Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5_perf_bisect/builds/4254 Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998454975041296848 Not what you expected? We'll investigate and get back to you! https://chromeperf.appspot.com/bad_bisect?try_job_id=5888346436403200 | O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq | X | for more information addressing perf regression bugs. For feedback, | / \ | file a bug with component Tests>AutoBisect. Thank you!
,
Oct 19 2016
Note this CL seems to cause about ~800KB regression in overall pss, android graphics, private dirty and native heap. See: https://chrome-health.googleplex.com/health-plan/android-chrome/memory/nexus5/?from_commit=424665&to_commit=424948 You should be able to reproduce the regression with the test command above (and try --output-format=html).
,
Oct 21 2016
Should have a fix for this soon.
,
Oct 27 2016
Any updates with the fix?
,
Oct 27 2016
In the process of landing the patch in Skia as we speak, should be rolled into Chromium today.
,
Oct 27 2016
The following revision refers to this bug: https://skia.googlesource.com/skia.git/+/cffaa70896fa5bc6c7bf98abbaafb1a755b49762 commit cffaa70896fa5bc6c7bf98abbaafb1a755b49762 Author: ethannicholas <ethannicholas@google.com> Date: Thu Oct 27 15:15:50 2016 Reduced skslc memory consumption The big change here is smarter generic type handling which allows us to keep far fewer entries in the core symboltable. This also comments out a number of OpenGL builtin functions which Skia does not use and is unlikely to in the future. BUG= 655673 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2442063002 Review-Url: https://codereview.chromium.org/2442063002 [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/src/sksl/SkSLContext.h [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/src/sksl/SkSLIRGenerator.cpp [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/src/sksl/SkSLParser.cpp [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/src/sksl/SkSLToken.h [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/src/sksl/ir/SkSLFunctionCall.h [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/src/sksl/ir/SkSLFunctionDeclaration.h [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/src/sksl/ir/SkSLType.h [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/src/sksl/sksl.include [modify] https://crrev.com/cffaa70896fa5bc6c7bf98abbaafb1a755b49762/tests/SkSLErrorTest.cpp
,
Oct 27 2016
The following revision refers to this bug: https://skia.googlesource.com/skia.git/+/56380524d903f27627a75e2e1189463999725008 commit 56380524d903f27627a75e2e1189463999725008 Author: benjaminwagner <benjaminwagner@google.com> Date: Thu Oct 27 16:08:06 2016 Revert of Reduced skslc memory consumption (patchset #3 id:50001 of https://codereview.chromium.org/2442063002/ ) Reason for revert: texelFetch removed, but is used in some shaders. Original issue's description: > Reduced skslc memory consumption > > The big change here is smarter generic type handling which allows us to > keep far fewer entries in the core symboltable. This also comments out > a number of OpenGL builtin functions which Skia does not use and is > unlikely to in the future. > BUG= 655673 > GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2442063002 > > Committed: https://skia.googlesource.com/skia/+/cffaa70896fa5bc6c7bf98abbaafb1a755b49762 TBR=reed@google.com,ethannicholas@google.com # Skipping CQ checks because original CL landed less than 1 days ago. NOPRESUBMIT=true NOTREECHECKS=true NOTRY=true BUG= 655673 Review-Url: https://codereview.chromium.org/2458723002 [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/src/sksl/SkSLContext.h [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/src/sksl/SkSLIRGenerator.cpp [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/src/sksl/SkSLParser.cpp [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/src/sksl/SkSLToken.h [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/src/sksl/ir/SkSLFunctionCall.h [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/src/sksl/ir/SkSLFunctionDeclaration.h [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/src/sksl/ir/SkSLType.h [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/src/sksl/sksl.include [modify] https://crrev.com/56380524d903f27627a75e2e1189463999725008/tests/SkSLErrorTest.cpp
,
Oct 27 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/7a8aa74e0d0c01bdd56048b6bfe33e94bc1688d1 commit 7a8aa74e0d0c01bdd56048b6bfe33e94bc1688d1 Author: skia-deps-roller <skia-deps-roller@chromium.org> Date: Thu Oct 27 17:27:53 2016 Roll src/third_party/skia/ 96b38426b..56380524d (8 commits). https://chromium.googlesource.com/skia.git/+log/96b38426b625..56380524d903 $ git log 96b38426b..56380524d --date=short --no-merges --format='%ad %ae %s' 2016-10-27 benjaminwagner Revert of Reduced skslc memory consumption (patchset #3 id:50001 of https://codereview.chromium.org/2442063002/ ) 2016-10-27 liyuqian Do not skip fractional y for SkAAClip 2016-10-27 scroggo Revert "Always use a color table with 256 colors" 2016-10-27 egdaniel Default vulkan sdk to VULKAN_SDK env var in GN 2016-10-27 scroggo Fix decoding GIF to 565 2016-10-27 ethannicholas Reduced skslc memory consumption 2016-10-27 fmalita Avoid separate allocation of SkImageCacherator 2016-10-27 borenet recipes: Fix missing patch_set property BUG= 655673 , 659883 , 659972 , 655673 Documentation for the AutoRoller is here: https://skia.googlesource.com/buildbot/+/master/autoroll/README.md If the roll is causing failures, see: http://www.chromium.org/developers/tree-sheriffs/sheriff-details-chromium#TOC-Failures-due-to-DEPS-rolls CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel TBR=reed@google.com Review-Url: https://codereview.chromium.org/2455053003 Cr-Commit-Position: refs/heads/master@{#428068} [modify] https://crrev.com/7a8aa74e0d0c01bdd56048b6bfe33e94bc1688d1/DEPS
,
Oct 28 2016
The following revision refers to this bug: https://skia.googlesource.com/skia.git/+/471e89405b71f04e07ae2887bde061185e262c81 commit 471e89405b71f04e07ae2887bde061185e262c81 Author: ethannicholas <ethannicholas@google.com> Date: Fri Oct 28 16:02:46 2016 Reduced skslc memory consumption The big change here is smarter generic type handling which allows us to keep far fewer entries in the core symboltable. This also comments out a number of OpenGL builtin functions which Skia does not use and is unlikely to in the future. BUG= 655673 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2442063002 Committed: https://skia.googlesource.com/skia/+/cffaa70896fa5bc6c7bf98abbaafb1a755b49762 Review-Url: https://codereview.chromium.org/2442063002 [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/src/sksl/SkSLContext.h [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/src/sksl/SkSLIRGenerator.cpp [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/src/sksl/SkSLParser.cpp [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/src/sksl/SkSLToken.h [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/src/sksl/ir/SkSLFunctionCall.h [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/src/sksl/ir/SkSLFunctionDeclaration.h [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/src/sksl/ir/SkSLType.h [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/src/sksl/sksl.include [modify] https://crrev.com/471e89405b71f04e07ae2887bde061185e262c81/tests/SkSLErrorTest.cpp
,
Oct 31 2016
Looks like #15 rolled into chromium at r428405, some memory measurements did drop around that change, but some new large regressions in graphics appeared. Those are being tracked in issue 660780 where a bisect is running to confirm if it's indeed due to this change.
,
Oct 31 2016
,
Nov 1 2016
Adding Ben as an FYI
,
Nov 1 2016
I don't seem to have permission to view 660780 -- did we determine whether this new regression is related to my changes or not?
,
Nov 1 2016
Ethan - try using your chromium.org login to view 660780. We're having trouble bisecting on the other bug.
,
Nov 7 2016
Do we feel the memory usage is sufficiently improved to close this?
,
Nov 8 2016
I do think it's fine (metrics have recovered on all main {product, device} configurations). I would still like to investigate issue 660780, though. Will have a look in a couple of days and re-open if necessary.
|
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by lanwei@google.com
, Oct 13 2016