GTest perf regressions #79958
Labels
No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset Browser
Interest
Asset Browser Project Overview
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
EEVEE & Viewport
Interest
Freestyle
Interest
Geometry Nodes
Interest
Grease Pencil
Interest
ID Management
Interest
Images & Movies
Interest
Import Export
Interest
Line Art
Interest
Masking
Interest
Metal
Interest
Modeling
Interest
Modifiers
Interest
Motion Tracking
Interest
Nodes & Physics
Interest
OpenGL
Interest
Overlay
Interest
Overrides
Interest
Performance
Interest
Physics
Interest
Pipeline, Assets & IO
Interest
Platforms, Builds & Tests
Interest
Python API
Interest
Render & Cycles
Interest
Render Pipeline
Interest
Sculpt, Paint & Texture
Interest
Text Editor
Interest
Translations
Interest
Triaging
Interest
Undo
Interest
USD
Interest
User Interface
Interest
UV Editing
Interest
VFX & Video
Interest
Video Sequencer
Interest
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
Legacy
Blender 2.8 Project
Legacy
Milestone 1: Basic, Local Asset Browser
Legacy
OpenGL Error
Meta
Good First Issue
Meta
Papercut
Meta
Retrospective
Meta
Security
Module
Animation & Rigging
Module
Core
Module
Development Management
Module
EEVEE & Viewport
Module
Grease Pencil
Module
Modeling
Module
Nodes & Physics
Module
Pipeline, Assets & IO
Module
Platforms, Builds & Tests
Module
Python API
Module
Render & Cycles
Module
Sculpt, Paint & Texture
Module
Triaging
Module
User Interface
Module
VFX & Video
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Priority
High
Priority
Low
Priority
Normal
Priority
Unbreak Now!
Status
Archived
Status
Confirmed
Status
Duplicate
Status
Needs Info from Developers
Status
Needs Information from User
Status
Needs Triage
Status
Resolved
Type
Bug
Type
Design
Type
Known Issue
Type
Patch
Type
Report
Type
To Do
No Milestone
No project
No Assignees
4 Participants
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: blender/blender#79958
Loading…
Reference in New Issue
No description provided.
Delete Branch "%!s(<nil>)"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Recently we have restructured the infrastructure around gtest a little bit, mostly due to a handful of tests linking "all of blender" leading to long link times for the initial builds leading to complaints from developers.
In an effort to increase developer productivity we solved this by putting all those tests into
blender_test
so we only had to pay those costs once.Which was great! and things would still be great if we stopped there, but we didn't.
I have been somewhat unavailable lately, so I have missed what the reasoning behind some of these choices may have been. but the following things have happened:
1 - The cheaper tests, that did not link "all of blender" have been moved into
blender_test
as well, meaning if you have a previously standalone test that tested a bit shuffle (or something else independent) that would link quickly into it's own executable, you now have to pay the "all of blender" link costs, leading to reduced developer productivity for most developers.1a - Now that all tests live in this mega binary, the startup time in the debugger while it resolves symbols has put a damper on things as well (@howardt reported ~10 seconds for debugger startup for what used to be instantly) once more not great for developer productivity
2 - Rather than a single invocation for the test binary to run all tests we now call the test binary once per test, with > 1000 tests even small costs start to add up quickly, starting a process is an inherently expensive activity, starting a process that has a large amount of static initializers (cycles/USD/alembic/mantaflow being the main offenders here, I did validate in a profiler that static initializers during CRT startup are one of the main contributors there, this is not "a hunch") just makes it worse, randomly selected example
this test took 2 ms to actually run the test, but required 420ms (for debug, release it is about ~120ms) if you include process startup/tear-down time, now that times a 1000+ tests.. not good
So changes we did with good intentions to improve developer productivity have had a tiny bit of the exact opposite effect.
I don't have an easy solution here, but things cannot stay the way they are right now
Changed status from 'Needs Triage' to: 'Confirmed'
Added subscribers: @howardt, @LazyDodo, @dr.sybren
Added subscriber: @Raimund58
This issue was referenced by blender/cycles@10b8f6b7c0
This issue was referenced by
532ac1163c
Added subscriber: @brecht
Changed status from 'Confirmed' to: 'Resolved'
This should be resolved now.