xref: /third_party/vk-gl-cts/doc/testspecs/GLES31/functional.shaders.sample_variables.txt (revision e5c31af7)

-------------------------------------------------------------------------
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.
-------------------------------------------------------------------------
    Sample variable tests

Tests:
 + dEQP-GLES31.functional.shaders.sample_variables.*

Includes:
 + Rendering to default framebuffer, multi- and single sample
   textures and renderbuffers
 + Built-in variables:
    - gl_NumSamples
    - gl_MaxSamples
    - gl_SampleID
    - gl_SamplePos
    - gl_SampleMaskIn
 + Writing to gl_SampleMask
 + Interaction with OES_sample_shading

Excludes:
 + Rendering to every supported render buffer sample count / format
 + Rendering to every supported multisample texture sample count / format

Description:

num_samples.* cases test gl_NumSamples built-in value with different render
target configurations. max_samples.* cases test gl_MaxSamples built-in variable.
Values are verified with verifier fragment shader.

sample_id.* cases verify gl_SampleID values. A quad is drawn with a shader
program. With texture targets, each sample encodes its gl_SampleID to the
fragment output value and then texture contents are verified. With default
framebuffer and renderbuffer targets, render area is partitioned to a grid so
that in each grid cell only the fragment shader invocations with a certain
gl_SampleID output a non-zero green value. Result image is then verified to
contain only non-zero green values. The test with framebuffer and renderbuffer
thus only guarantees that for each pixel there exist a fragment shader
invocation for every valid gl_SampleID value, but not that a matching fragment
shader invocation is unique.

sample_pos.correctness.* cases verify gl_SamplePosition built-in variable
contains the position in which varyings were interpolated. Cases render a quad
with a verifier fragment shader. Fragment shader compares gl_SamplePosition
values against a per-sample interpolated screen-space location (in pixels)
varying. gl_SamplePosition should be equal to the fractional part of the
interpolated varying value.

sample_pos.distribution.* cases verify the distribution of gl_SamplePositions
within a pixel. The tests render a quad with a fragment shader that encodes the
gl_SamplePosition value and value legality (range check) to the color channels.
With texture render targets, the texture contents are read and positions are
verified for legality, uniqueness and distribution. With framebuffer and
renderbuffer render targets only the legality and distribution are verified.
Distribution is verified by averaging sample position values (texture) or by
resolving with an implementation defined method (framebuffer, renderbuffer).
Averaged/resolved value should be reasonably close to the pixel center. Too
large a bias away from the expected value will result in a quality warning.

sample_mask_in.* cases test reading gl_SampleMaskIn built-in variable and
sample_mask_in.* cases write values to gl_SampleMask built-in variable. Sample
mask operations are multisample operations and should not have effect with
single sample render targets. *_per_pixel variations test value when fragment
shader is executed once per pixel, and *_per_sample when shader is executed per
sample. *_per_two_samples variations use OES_sample_shading, with sample
shading value set to 50%, i.e. there should be AT LEAST one invocation per two
samples.

sample_mask_in.sample_mask cases test gl_SampleMaskIn that with multisampled
render targets gl_SampleMaskIn does not contain any bits not set in
GL_SAMPLE_MASK. With single sample render targets gl_SampleMaskIn should not
depend on GL_SAMPLE_MASK value.

sample_mask_in.bit_count cases test gl_SampleMaskIn contains only a certain
number of set bits. In the per pixel variation, gl_SampleMaskIn should contain
[1, numSamples] set bits. In per sample cases, the gl_SampleMaskIn should only
contain one set bit. In per two samples cases, there should be [1, numSamples/2]
bits (all but one invocation gets one sample, one invocation gets the other
samples).

sample_mask_in.bits_unique_per_sample cases tests that the single set bit in
gl_SampleMaskIn is unique. sample_mask_in.bits_unique_per_two_samples cases test
that each bit is set only in a single gl_SampleMaskIn bitset and that number of
distinct gl_SampleMaskIn bitsets matches the number guaranteed by
OES_sample_shading. Tests render to a multisample texture, encode
gl_SampleMaskIn to the color channel, and then read back the texture contents.

sample_mask.discard_half tests discard half of the samples by writing 0xAAAA...
to gl_SampleMask output variable. With single sample targets, writing to
gl_SampleMask should have no effect. With multisample targets the result image
color intensity should decrease. With multisample targets with sample count = 1,
all fragments should be discarded.

sample_mask.inverse cases set gl_SampleMask to the inverse of current
SAMPLE_MASK. With multisample render targets, all fragments should be discarded.
With single sample render targets, writing to gl_SampleMask should have no
effect.