• Show log

    Commit

  • Hash : d25c9d7f
    Author : Jamie Madill
    Date : 2021-07-07T12:15:12

    Add perf tests runner script.
    
    This script will be responsible for running the perf tests
    multiple times to try and stabilize measurements. We'll use it on
    the bots instead of just running the perf tests directly.
    
    Because the script invokes the binary multiple times, this slows
    down execution. Most significantly on Android, where we now need
    to use 20 shards, up from 6.
    
    Also marks one test as flaky on OpenGL.
    
    Bug: angleproject:6090
    Change-Id: I5280035cb0bdb290a68dc6961a384eaf4b40dd4b
    Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3011422
    Commit-Queue: Jamie Madill <jmadill@chromium.org>
    Reviewed-by: Cody Northrop <cnorthrop@google.com>
    Reviewed-by: Tim Van Patten <timvp@google.com>
    

  • Properties

  • Git HTTP https://git.kmx.io/kc3-lang/angle.git
    Git SSH git@git.kmx.io:kc3-lang/angle.git
    Public access ? public
    Description

    A conformant OpenGL ES implementation for Windows, Mac, Linux, iOS and Android.

    Homepage

    Github

    Users
    thodg_m kc3_lang_org thodg_w www_kmx_io thodg thodg_l
    Tags

  • README.md

  • ANGLE Test Harness

    The ANGLE test harness is a harness around GoogleTest that provides functionality similar to the Chromium test harness. It features:

    • splitting a test set into shards
    • catching and reporting crashes and timeouts
    • outputting to the Chromium JSON test results format
    • multi-process execution

    Command-Line Arguments

    The ANGLE test harness accepts all standard GoogleTest arguments. The harness also accepts the following additional command-line arguments:

    • --batch-size limits the number of tests to run in each batch
    • --batch-timeout limits the amount of time spent in each batch
    • --bot-mode enables multi-process execution and test batching
    • --debug-test-groups dumps the test config categories when using bot-mode
    • --filter-file allows passing a larger gtest_filter via a file
    • --histogram-json-file outputs a formatted JSON file for perf dashboards
    • --max-processes limits the number of simuntaneous processes
    • --results-directory specifies a directory to write test results to
    • --results-file specifies a location for the JSON test result output
    • --shard-count and --shard-index control the test sharding
    • --test-timeout limits the amount of time spent in each test
    • --flaky-retries allows for tests to fail a fixed number of times and still pass
    • --disable-crash-handler forces off OS-level crash handling
    • --isolated-outdir specifies a test artifacts directory
    • --max-failures specifies a count of failures after which the harness early exits.

    --isolated-script-test-output and --isolated-script-perf-test-output mirror --results-file and --histogram-json-file respectively.

    As well as the custom command-line arguments we support a few standard GoogleTest arguments:

    • gtest_filter works as it normally does with GoogleTest
    • gtest_also_run_disabled_tests works as it normally does as well

    Other GoogleTest arguments are not supported although they may work.

    Implementation Notes

    • The test harness only requires angle_common and angle_util.
    • It does not depend on any Chromium browser code. This allows us to compile on other non-Clang platforms.
    • It uses rapidjson to read and write JSON files.
    • Test execution is not currently deterministic in multi-process mode.

    Normal Mode vs Bot Mode

    The test runner has two main modes of operation: normal and bot mode.

    During normal mode:

    • Tests are executed single-process and single-thread.
    • The test runner executes the GoogleTest Run function.
    • We use a TestEventListener to record test results for our output JSON file.
    • A watchdog thread will force a fast exit if no test results get recorded after a timeout.
    • Crashes are handled via ANGLE’s test crash handling code.

    During bot mode:

    • Tests are run in multiple processes depending on the system processor count.
    • A server process records the child processes’ stdout and stderr.
    • The server terminates a child process if there’s no progress after a timeout.
    • The server sorts work into batches according to the back-end configuration.
    • This prevents driver errors from using multiple back-ends in the same process.
    • Batches are striped to help split up slow groups of tests.
    • The server passes test batches to child processes via a gtest_filter file.
    • Bot mode does not work on Android or Fuchsia.

    See the source code for more details: TestSuite.h and TestSuite.cpp.

    Potential Areas of Improvement

    • Deterministic test execution.
    • Using sockets to communicate with test children. Similar to dEQP’s test harness.
    • Closer integration with ANGLE’s test expectations and system config libraries.
    • Supporting a GoogleTest-free integration.