xref: /aosp_15_r20/external/deqp/doc/testspecs/GLES3/performance.txt (revision 35238bce31c2a825756842865a792f8cf7f89930)

-------------------------------------------------------------------------
drawElements Quality Program Test Specification
-----------------------------------------------

Copyright 2014 The Android Open Source Project

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

     http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-------------------------------------------------------------------------
    Performance tests

This test specification describes general techniques and methodologies used
in performance test cases.


Measuring performance:

Performance test cases must satisfy following conditions:

 + Result should represent the performance of the tested feature or use case.
   - Use of any other features should be kept at minimum.
   - Architecture-specific optimizations should not affect result unless they
     target the feature being tested.
   - Measurement overhead should be minimized.

 + Hardware must be utilized to the maximum.
   - In most cases total throughput is more important than latency.
   - Latency is measured where it matters.
   - Test cases should behave like other graphics-intensive applications on
     that platform. This usually means that test cases should render multiple
	 frames and utilize platform-defined standard mechanisms to display each
	 frame on screen.

 + Test result should be stable across test runs.
   - Final result should be a function of results from multiple iterations if
     performance is expected to vary between iterations (due to undeterministic
     scheduling for example).
   - Simple average may not always be the right function, often stability of
     the performance is more important from user experience perspective.

 + Test result values should be meaningful.
   - Result should be meaningful without knowing the exact implementation
     details of the test case.
   - Good example: Millions of pixels or vertices with shader X per second
   - Bad example: Frames per second

 + Results can be compared across different implementations and configurations.
   - If possible, configuration changes (viewport size for example) should not
	 affect the test result.
   - Where configuration may affect the result, test log should include the
	 configuration details.


Test output:

Test cases will log at least following variables, if available:

 * Viewport size
 * Color, depth, stencil and multisample bits
 * Number of draw calls
 * Shader program source
 * Automatic calibration values
 * Individual iteration times (if iteration count is reasonable)
 * Minimum and maximum iteration times
 * Average iteration time
 * Minimum and maximum computed performance
 * Average computed performance


Shader execution tests:

Shader execution tests measure the performance of single pair of vertex and
fragment shaders, optionally combined with fixed-function per-fragment
operations such as blending.

Each iteration (frame) renders N grids of quads. Each grid, drawn with single
draw call, fills the screen entirely without any overlap between quads.
The number of quads depends on targeted vertex load. Test cases targeting
fragment shaders only use a single quad. N (number of times screen is filled)
is chosen either automatically to target certain iteration time or specified
by the test case explicitly.

Test cases that measure fragment-side performance must make sure fragments
are not discarded early due to Z-culling or any other optimization. The
preferred way to do this is to enable simple additive blending. The extra
cost of that is one read from the framebuffer and per-channel saturating
additions.

The result is MPix/s, MVert/s or weighted sum of both depending on test
case type.