New issue
Advanced search Search tips

Issue 722979 link

Starred by 1 user
Status: WontFix
Owner:
agrieve@chromium.org
Closed: May 2017
Cc:
sunn...@chromium.org
agrieve@chromium.org
estevenson@chromium.org
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

17kb regression in resource_sizes (MonochromePublic.apk) at 471361:471361

Project Member Reported by agrieve@chromium.org, May 16 2017 
Caused by: Reland of gpu: GPU service scheduler
https://codereview.chromium.org/2876913003

Comment 1 by agrieve@chromium.org, May 16 2017 

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=722979

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg6uD17gsM


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

Android Builder
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, May 16 2017 


=== BISECT JOB RESULTS ===
Bisect failed for unknown reasons

Please contact the team (see below) and report the error.


Bisect Details
  Configuration: android_nexus7_perf_bisect
  Benchmark    : resource_sizes
  Metric       : MonochromePublic.apk_Specifics/normalized apk size


To Run This Test
  src/build/android/resource_sizes.py --chromium-output-directory {CHROMIUM_OUTPUT_DIR} --chartjson {CHROMIUM_OUTPUT_DIR}/apks/MonochromePublic.apk

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8979409325240274688

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5893554079006720


| 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 Speed>Bisection.  Thank you!
Project Member

Comment 5 by 42576172...@developer.gserviceaccount.com, May 16 2017 


=== BISECT JOB RESULTS ===
Bisect failed for unknown reasons

Please contact the team (see below) and report the error.


Bisect Details
  Configuration: android_nexus7_perf_bisect
  Benchmark    : resource_sizes
  Metric       : MonochromePublic.apk_Specifics/normalized apk size


To Run This Test
  src/build/android/resource_sizes.py --chromium-output-directory {CHROMIUM_OUTPUT_DIR} --chartjson {CHROMIUM_OUTPUT_DIR}/apks/MonochromePublic.apk

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8979397597290455504

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5893554079006720


| 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 Speed>Bisection.  Thank you!

Comment 6 by estevenson@chromium.org, May 24 2017

Cc: sunn...@chromium.org
Looking at the CL I'd imagine this size jump is expected. sunnyps@, see the size diff below and check if anything is unexpected or could be reduced.

Section Sizes (Total=12.7kb (12956 bytes)):
    .bss: 0 bytes (0 bytes) (not included in totals)
    .data: 0 bytes (0 bytes) (0.0%)
    .data.rel.ro: 0 bytes (0 bytes) (0.0%)
    .data.rel.ro.local: 16 bytes (16 bytes) (0.1%)
    .rodata: 256 bytes (256 bytes) (2.0%)
    .text: 12.4kb (12684 bytes) (97.9%)

160 symbols added (+), 53 changed (~), 77 removed (-), 381386 unchanged (not shown)
2 paths added, 0 removed, 19 changed

Showing 290 symbols (290 unique) with total pss: 12536 bytes
.text=12.0kb     .rodata=256 bytes  .data*=16 bytes   .bss=3 bytes    total=12.2kb
Number of unique paths: 25

Index, Running Total, Section@Address, PSS
------------------------------------------------------------
+ 0)       1044 (8.3%)  t@0x1808808  1044    gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::RunNextTask
+ 1)       1760 (14.0%) t@0x18080bc  716     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::deque::__add_back_capacity
+ 2)       2318 (18.5%) t@0x188dfb0  558     gpu/ipc/service/gpu_channel.cc
               std::__ndk1::vector::emplace
+ 3)       2838 (22.6%) t@0x1807e4c  520     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::deque::__add_front_capacity
+ 4)       3350 (26.7%) t@0x18078ac  512     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::TryScheduleSequence
+ 5)       3858 (30.8%) t@0x1808fc0  508     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::vector::insert
- 6)       3380 (27.0%) t@0x188b0a6  -478    gpu/ipc/service/gpu_channel.cc
               std::__ndk1::__hash_table::__node_insert_unique
+ 7)       3808 (30.4%) t@0x188ebcc  428     gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannel::GpuChannel
+ 8)       4224 (33.7%) t@0x1808c94  416     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::ScheduleTask
+ 9)       4624 (36.9%) t@0x24e7c58  400     content/renderer/render_thread_impl.cc
               content::CreateOffscreenContext
- 10)      4224 (33.7%) t@0x24e4b38  -400    content/renderer/render_thread_impl.cc
               content::CreateOffscreenContext
- 11)      3840 (30.6%) t@0x188bbfc  -384    gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannel::GpuChannel
+ 12)      4216 (33.6%) t@0x87d7e0   376     gpu/ipc/client/command_buffer_proxy_impl.cc
               gpu::CommandBufferProxyImpl::Create
- 13)      3840 (30.6%) t@0x87d7a4   -376    gpu/ipc/client/command_buffer_proxy_impl.cc
               gpu::CommandBufferProxyImpl::Create
+ 14)      4206 (33.6%) t@0x188d920  366     gpu/ipc/service/gpu_channel.cc
               std::__ndk1::vector::emplace
+ 15)      4570 (36.5%) t@0x188dc48  364     gpu/ipc/service/gpu_channel.cc
               std::__ndk1::vector::emplace
~ 16)      4890 (39.0%) t@0x188e7ac  320     gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannelMessageFilter::OnMessageReceived
+ 17)      5206 (41.5%) t@0x1808e84  316     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::__split_buffer::push_back
+ 18)      5514 (44.0%) t@0x1807028  308     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Sequence::~Sequence
+ 19)      5806 (46.3%) t@0x188b674  292     gpu/ipc/service/gpu_channel.cc
               std::__ndk1::vector::assign
+ 20)      6092 (48.6%) t@0x188f212  286     gpu/ipc/service/gpu_channel_manager.cc
               gpu::GpuChannelManager::GpuChannelManager
- 21)      5818 (46.4%) t@0x188c302  -274    gpu/ipc/service/gpu_channel_manager.cc
               gpu::GpuChannelManager::GpuChannelManager
+ 22)      6086 (48.5%) t@0x18077a0  268     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::RebuildSchedulingQueue
+ 23)      6350 (50.7%) t@0x180752c  264     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::remove
+ 24)      6582 (52.5%) t@0x180863c  232     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::ShouldYield
+ 25)      6794 (54.2%) t@0x1807c4c  212     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::__split_buffer::push_back
+ 26)      7006 (55.9%) t@0x1807d78  212     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::__split_buffer::push_back
+ 27)      7214 (57.5%) t@0x1808408  208     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::__push_heap_front
+ 28)      7422 (59.2%) t@0x1807aac  208     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::__split_buffer::push_front
+ 29)      7630 (60.9%) t@0x1807b7c  208     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::__split_buffer::push_front
+ 30)      7838 (62.5%) t@0x180765c  208     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::vector::__push_back_slow_path
+ 31)      8046 (64.2%) t@0xf5bc80   208     services/ui/public/cpp/gpu/context_provider_command_buffer.cc
               ui::ContextProviderCommandBuffer::ContextProviderCommandBuffer
- 32)      7838 (62.5%) t@0xf5bc0c   -208    services/ui/public/cpp/gpu/context_provider_command_buffer.cc
               ui::ContextProviderCommandBuffer::ContextProviderCommandBuffer
+ 33)      8038 (64.1%) t@0x18091bc  200     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::CreateSequence
- 34)      7840 (62.5%) t@0x188afe0  -198    gpu/ipc/service/gpu_channel.cc
               std::__ndk1::__hash_table::__rehash
+ 35)      8028 (64.0%) t@0x18073f0  188     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::vector::__push_back_slow_path
- 36)      7844 (62.6%) t@0x1888384  -184    gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannelMessageQueue::GpuChannelMessageQueue
+ 37)      8026 (64.0%) t@0x188a980  182     gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannelMessageQueue::GpuChannelMessageQueue
+ 38)      8206 (65.5%) t@0x1807280  180     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Sequence::BeginTask
~ 39)      8385 (66.9%) r@Group      179     {{no path}}
               ** merge strings (count=6)
+ 40)      8555 (68.2%) t@0x188d546  170     gpu/ipc/service/gpu_channel.cc
               base::internal::flat_tree::erase
~ 41)      8391 (66.9%) t@0x188d70c  -164    gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannel::OnDestroyCommandBuffer
+ 42)      8551 (68.2%) t@Group      160     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::~Scheduler (count=2)
+ 43)      8707 (69.5%) t@0x180853c  156     gpu/command_buffer/service/scheduler.cc
               base::internal::flat_tree::erase
+ 44)      8863 (70.7%) t@0x180876c  156     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::SyncTokenFenceReleased
+ 45)      8999 (71.8%) t@0x188efb8  136     gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannel::HandleMessage
+ 46)      9127 (72.8%) t@0x1808388  128     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Sequence::ScheduleTask
+ 47)      9247 (73.8%) t@0x1808c1c  120     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::ContinueTask
+ 48)      9367 (74.7%) t@0x1807334  120     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Sequence::FinishTask
+ 49)      9483 (75.6%) t@0x180772c  116     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::__push_heap_back
+ 50)      9587 (76.5%) t@0x1808054  104     gpu/command_buffer/service/scheduler.cc
               std::__ndk1::deque::push_front
+ 51)      9691 (77.3%) t@0x188d8b6  104     gpu/ipc/service/gpu_channel.cc
               std::__ndk1::vector::__swap_out_circular_buffer
+ 52)      9791 (78.1%) t@0x188ae2c  100     gpu/ipc/service/gpu_channel.cc
               base::BindOnce
+ 53)      9891 (78.9%) t@0x1806fc4  100     gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::SchedulingState::AsValue const
+ 54)      9987 (79.7%) t@0x1806f10  96      gpu/command_buffer/service/scheduler.cc
               base::internal::Invoker::Run
~ 55)     10083 (80.4%) t@0x1893358  96      gpu/ipc/service/gpu_command_buffer_stub.cc
               gpu::GpuCommandBufferStub::Initialize
~ 56)     10179 (81.2%) t@0x16e62f8  96      services/ui/gpu/gpu_service.cc
               ui::GpuService::InitializeWithHost
- 57)     10085 (80.4%) t@0x1888f28  -94     gpu/ipc/service/gpu_channel.cc
               std::__ndk1::__hash_table::find
+ 58)     10169 (81.1%) t@0x1806f70  84      gpu/command_buffer/service/scheduler.cc
               base::internal::Invoker::Run
~ 59)     10253 (81.8%) t@0x188e1dc  84      gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannel::OnCreateCommandBuffer
+ 60)     10337 (82.5%) t@0x188a7c4  84      gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannelMessageFilter::GpuChannelMessageFilter
- 61)     10253 (81.8%) t@0x188bd9a  -84     gpu/ipc/service/gpu_channel.cc
               std::__ndk1::deque::pop_front
+ 62)     10333 (82.4%) t@0x1808e34  80      gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::DestroySequence
+ 63)     10411 (83.0%) t@0x76d28c   78      gpu/command_buffer/common/scheduling_priority.cc
               gpu::SchedulingPriorityToString
+ 64)     10485 (83.6%) t@0x188da8c  74      gpu/ipc/service/gpu_channel.cc
               base::flat_map::operator[]
+ 65)     10559 (84.2%) t@0x188e3da  74      gpu/ipc/service/gpu_channel.cc
               base::internal::Invoker::RunOnce
- 66)     10487 (83.7%) t@0x1888198  -72     gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannelMessageFilter::GpuChannelMessageFilter
+ 67)     10559 (84.2%) t@0x1808724  72      gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::EnableSequence
~ 68)     10627 (84.8%) t@0x188e5f8  68      gpu/ipc/service/gpu_channel.cc
               gpu::GpuChannelMessageQueue::FinishMessageProcessing
+ 69)     10695 (85.3%) t@0x18085f8  68      gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::DisableSequence
+ 70)     10763 (85.9%) t@0x18073ac  68      gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Scheduler
+ 71)     10831 (86.4%) t@0x180723c  68      gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Sequence::UpdateSchedulingState
+ 72)     10899 (86.9%) t@0x188d824  68      gpu/ipc/service/gpu_channel.cc
               std::__ndk1::vector::__move_range
+ 73)     10963 (87.5%) t@0x18074ec  64      gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Sequence::AddReleaseFence
+ 74)     11027 (88.0%) t@0x18074ac  64      gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Sequence::AddWaitFence
+ 75)     11091 (88.5%) t@0x18071fc  64      gpu/command_buffer/service/scheduler.cc
               gpu::Scheduler::Sequence::IsRunnable const
+ 76)     11155 (89.0%) t@0x180c288  64      gpu/command_buffer/service/sync_point_manager.cc
               gpu::SyncPointManager::GetSyncTokenReleaseSequenceId
+ 77)     11217 (89.5%) t@0x188d688  62      gpu/ipc/service/gpu_channel.cc
               base::internal::flat_tree::find const


agrieve@ - one really weird thing is that there was somehow an extra android res .xml file added.

<       968  2001-01-01 00:00   res/color/bookmark_drawer_text_color.xml

Only 1kb but still probably worth looking into.
diff_results.txt
50.8 KB View Download

Comment 7 by estevenson@chromium.org, May 24 2017 

+ please to c#6 :).

Comment 8 by sunn...@chromium.org, May 25 2017 

Nothing looks out of place. I guess RunNextTask and TryScheduleSequence are big because the vector and heap calls are inlined into them? The other things are from deque, vector, flat_map, etc. and I'm not sure what we can do about that.

Comment 9 by sunn...@chromium.org, May 25 2017 

Oh one more thing, we'll be removing some code as a result of this work. The GpuChannelMessageQueue class and some methods on GpuChannel will go away so maybe ultimately it'll be a wash?

Comment 10 by estevenson@chromium.org, May 25 2017

Cc: estevenson@chromium.org
Status: WontFix (was: Assigned)
Sounds good to me, we can close this one then. It would be great if you commented here once some of those changes have landed so we can verify. Thanks!
►

Sign in to add a comment