[Piglit] [PATCH] msaa: add test to verify alpha-to-coverage together with alpha-to-one

Ilia Mirkin imirkin at alum.mit.edu
Fri Aug 4 16:44:25 UTC 2017


On Fri, Aug 4, 2017 at 12:40 PM, Brian Paul <brianp at vmware.com> wrote:
> On 08/04/2017 10:13 AM, Ilia Mirkin wrote:
>>
>> On Fri, Aug 4, 2017 at 11:36 AM, Brian Paul <brianp at vmware.com> wrote:
>>>
>>> On 08/04/2017 09:30 AM, Brian Paul wrote:
>>>>
>>>>
>>>> On 08/04/2017 08:51 AM, Roland Scheidegger wrote:
>>>>>
>>>>>
>>>>> Am 04.08.2017 um 16:43 schrieb Brian Paul:
>>>>>>
>>>>>>
>>>>>> On 08/04/2017 08:32 AM, sroland at vmware.com wrote:
>>>>>>>
>>>>>>>
>>>>>>> From: Roland Scheidegger <sroland at vmware.com>
>>>>>>>
>>>>>>> I believe the sole reason alpha-to-one exists at all is that it can
>>>>>>> be
>>>>>>> used together with alpha-to-coverage (otherwise why not just output
>>>>>>> alpha
>>>>>>> 1.0 in the fs in the first place...).
>>>>>>> Therefore, it seems quite appropriate to test this together with
>>>>>>> alpha-to-coverage.
>>>>>>>
>>>>>>> Note: the test is a simple copy of sample-alpha-to-coverage with only
>>>>>>> trivial adjustments. I don't think the common code used actually
>>>>>>> handles
>>>>>>> this case correctly for the reference image (as it's the same as when
>>>>>>> just using alpha-to-coverage, which can't be right), but I haven't
>>>>>>> looked
>>>>>>> at it long enough to be able to fix it.
>>>>>>> It is, however, good enough to pass on my nvidia card, albeit some
>>>>>>> mesa
>>>>>>> drivers trying to emulate this feature are definitely broken (I'm
>>>>>>> looking
>>>>>>> at you radeons...).
>>>>>>> ---
>>>>>>>     tests/all.py                                       |   5 +
>>>>>>>     .../ext_framebuffer_multisample/CMakeLists.gl.txt  |   2 +
>>>>>>>     .../sample-alpha-to-coverage-one.cpp               | 162
>>>>>>> +++++++++++++++++++++
>>>>>>>     3 files changed, 169 insertions(+)
>>>>>>>     create mode 100644
>>>>>>>
>>>>>>> tests/spec/ext_framebuffer_multisample/sample-alpha-to-coverage-one.cpp
>>>>>>>
>>>>>>> diff --git a/tests/all.py b/tests/all.py
>>>>>>> index fc00d99..bc5b9e0 100644
>>>>>>> --- a/tests/all.py
>>>>>>> +++ b/tests/all.py
>>>>>>> @@ -3028,6 +3028,11 @@ with profile.test_list.group_manager(
>>>>>>>                    sample_count, buffer_type],
>>>>>>>                   'sample-alpha-to-coverage {} {}'.format(
>>>>>>>                       sample_count, buffer_type))
>>>>>>> +
>>>>>>> g(['ext_framebuffer_multisample-sample-alpha-to-coverage-one',
>>>>>>> +               sample_count, buffer_type],
>>>>>>> +              'sample-alpha-to-coverage-one {} {}'.format(
>>>>>>> +                  sample_count, buffer_type))
>>>>>>> +
>>>>>>>
>>>>>>>             for test in ['line-smooth', 'point-smooth',
>>>>>>> 'polygon-smooth',
>>>>>>>                          'sample-alpha-to-one',
>>>>>>> diff --git a/tests/spec/ext_framebuffer_multisample/CMakeLists.gl.txt
>>>>>>> b/tests/spec/ext_framebuffer_multisample/CMakeLists.gl.txt
>>>>>>> index 6841267..e93c609 100644
>>>>>>> --- a/tests/spec/ext_framebuffer_multisample/CMakeLists.gl.txt
>>>>>>> +++ b/tests/spec/ext_framebuffer_multisample/CMakeLists.gl.txt
>>>>>>> @@ -62,6 +62,8 @@ piglit_add_executable
>>>>>>> (ext_framebuffer_multisample-samples samples.c)
>>>>>>>     piglit_add_executable
>>>>>>> (ext_framebuffer_multisample-sample-coverage
>>>>>>> common.cpp sample-coverage.cpp)
>>>>>>>     piglit_add_executable
>>>>>>> (ext_framebuffer_multisample-sample-alpha-to-coverage common.cpp
>>>>>>>                    draw-buffers-common.cpp
>>>>>>> sample-alpha-to-coverage.cpp)
>>>>>>> +piglit_add_executable
>>>>>>> (ext_framebuffer_multisample-sample-alpha-to-coverage-one common.cpp
>>>>>>> +               draw-buffers-common.cpp
>>>>>>> sample-alpha-to-coverage-one.cpp)
>>>>>>>     piglit_add_executable
>>>>>>> (ext_framebuffer_multisample-sample-alpha-to-one common.cpp
>>>>>>>                    draw-buffers-common.cpp sample-alpha-to-one.cpp)
>>>>>>>     piglit_add_executable (ext_framebuffer_multisample-turn-on-off
>>>>>>> common.cpp turn-on-off.cpp)
>>>>>>> diff --git
>>>>>>>
>>>>>>>
>>>>>>> a/tests/spec/ext_framebuffer_multisample/sample-alpha-to-coverage-one.cpp
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> b/tests/spec/ext_framebuffer_multisample/sample-alpha-to-coverage-one.cpp
>>>>>>>
>>>>>>> new file mode 100644
>>>>>>> index 0000000..887893a
>>>>>>> --- /dev/null
>>>>>>> +++
>>>>>>>
>>>>>>>
>>>>>>> b/tests/spec/ext_framebuffer_multisample/sample-alpha-to-coverage-one.cpp
>>>>>>>
>>>>>>> @@ -0,0 +1,162 @@
>>>>>>> +/*
>>>>>>> + * Copyright © 2012 Intel Corporation
>>>>>>> + *
>>>>>>> + * Permission is hereby granted, free of charge, to any person
>>>>>>> obtaining a
>>>>>>> + * copy of this software and associated documentation files (the
>>>>>>> "Software"),
>>>>>>> + * to deal in the Software without restriction, including without
>>>>>>> limitation
>>>>>>> + * the rights to use, copy, modify, merge, publish, distribute,
>>>>>>> sublicense,
>>>>>>> + * and/or sell copies of the Software, and to permit persons to
>>>>>>> whom the
>>>>>>> + * Software is furnished to do so, subject to the following
>>>>>>> conditions:
>>>>>>> + *
>>>>>>> + * The above copyright notice and this permission notice (including
>>>>>>> the next
>>>>>>> + * paragraph) shall be included in all copies or substantial
>>>>>>> portions
>>>>>>> of the
>>>>>>> + * Software.
>>>>>>> + *
>>>>>>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>>>>>>> EXPRESS OR
>>>>>>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>>>>>>> MERCHANTABILITY,
>>>>>>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO
>>>>>>> EVENT
>>>>>>> SHALL
>>>>>>> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES
>>>>>>> OR OTHER
>>>>>>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
>>>>>>> ARISING
>>>>>>> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
>>>>>>> OTHER DEALINGS
>>>>>>> + * IN THE SOFTWARE.
>>>>>>> + */
>>>>>>> +
>>>>>>> +#include "draw-buffers-common.h"
>>>>>>> +
>>>>>>> +/**
>>>>>>> + * \file sample-alpha-to-coverage.cpp
>>>>>>> + *
>>>>>>> + * Verify sample alpha to coverage plus alpha to one with
>>>>>>> multisample
>>>>>>> FBO
>>>>>>> + *
>>>>>>> + * When rendering to multisample FBO, fragment's alpha value should
>>>>>>> be
>>>>>>> + * used to determine the coverage value, and then the alpha value
>>>>>>> set
>>>>>>> to 1.
>>>>>>> + *
>>>>>>> + * This test operates by drawing a pattern in multisample FBO to
>>>>>>> generate
>>>>>>> + * reference and test image. Reference image is drawn to right half
>>>>>>> of window
>>>>>>> + * system frame buffer and test image to left half.
>>>>>>> + *
>>>>>>> + * Compute the expected color / depth values.
>>>>>>> + *
>>>>>>> + * Probe color / depth buffer blitted to downsampled FBO
>>>>>>> (resolve_fbo) and
>>>>>>> + * compare against expected color values.
>>>>>>> + *
>>>>>>> + * Author: Anuj Phogat <anuj.phogat at gmail.com>
>>>>>>> + */
>>>>>>> +
>>>>>>> +PIGLIT_GL_TEST_CONFIG_BEGIN
>>>>>>> +
>>>>>>> +    config.supports_gl_compat_version = 10;
>>>>>>> +
>>>>>>> +    config.window_width = 512;
>>>>>>> +    config.window_height = 256;
>>>>>>> +    config.window_visual = PIGLIT_GL_VISUAL_DOUBLE |
>>>>>>> PIGLIT_GL_VISUAL_RGBA;
>>>>>>> +
>>>>>>> +PIGLIT_GL_TEST_CONFIG_END
>>>>>>> +
>>>>>>> +static GLenum buffer_to_test;
>>>>>>> +
>>>>>>> +void
>>>>>>> +print_usage_and_exit(char *prog_name)
>>>>>>> +{
>>>>>>> +    printf("Usage: %s <num_samples> <test_type>\n"
>>>>>>> +           "  where <test_type> is one of:\n"
>>>>>>> +           "    color\n"
>>>>>>> +           "    depth\n",
>>>>>>> +           prog_name);
>>>>>>> +    piglit_report_result(PIGLIT_FAIL);
>>>>>>> +}
>>>>>>> +
>>>>>>> +void
>>>>>>> +piglit_init(int argc, char **argv)
>>>>>>> +{
>>>>>>> +    const int num_attachments = 1;
>>>>>>> +    int samples;
>>>>>>> +
>>>>>>> +    piglit_require_gl_version(21);
>>>>>>> +    piglit_require_extension("GL_ARB_framebuffer_object");
>>>>>>> +    piglit_require_extension("GL_ARB_vertex_array_object");
>>>>>>> +
>>>>>>> +    if (argc < 3)
>>>>>>> +        print_usage_and_exit(argv[0]);
>>>>>>> +    {
>>>>>>> +        char *endptr = NULL;
>>>>>>> +        samples = strtol(argv[1], &endptr, 0);
>>>>>>> +        if (endptr != argv[1] + strlen(argv[1]))
>>>>>>> +            print_usage_and_exit(argv[0]);
>>>>>>> +    }
>>>>>>> +    if (strcmp(argv[2], "color") == 0) {
>>>>>>> +        buffer_to_test = GL_COLOR_BUFFER_BIT;
>>>>>>> +    } else if (strcmp(argv[2], "depth") == 0) {
>>>>>>> +        buffer_to_test = GL_DEPTH_BUFFER_BIT;
>>>>>>> +        glEnable(GL_DEPTH_TEST);
>>>>>>> +        glDepthFunc(GL_ALWAYS);
>>>>>>> +    } else
>>>>>>> +        print_usage_and_exit(argv[0]);
>>>>>>> +
>>>>>>> +    int pattern_width = piglit_width / 2;
>>>>>>> +    int pattern_height = piglit_height / num_attachments;
>>>>>>> +
>>>>>>> +    piglit_ortho_projection(pattern_width,
>>>>>>> +                pattern_height,
>>>>>>> +                GL_TRUE);
>>>>>>> +
>>>>>>> +    /* Skip the test if samples > GL_MAX_SAMPLES */
>>>>>>> +    GLint max_samples;
>>>>>>> +    glGetIntegerv(GL_MAX_SAMPLES, &max_samples);
>>>>>>> +
>>>>>>> +    if (samples > max_samples)
>>>>>>> +        piglit_report_result(PIGLIT_SKIP);
>>>>>>> +
>>>>>>> +    ms_fbo_and_draw_buffers_setup(samples,
>>>>>>> +                      pattern_width,
>>>>>>> +                      pattern_height,
>>>>>>> +                      num_attachments,
>>>>>>> +                      buffer_to_test,
>>>>>>> +                      GL_RGBA);
>>>>>>> +    shader_compile(true /* sample_alpha_to_coverage */,
>>>>>>> +               false /* dual_src_blend */,
>>>>>>> +               true /* frag_out_zero_write */);
>>>>>>> +}
>>>>>>> +
>>>>>>> +enum piglit_result
>>>>>>> +piglit_display()
>>>>>>> +{
>>>>>>> +    bool pass = true;
>>>>>>> +    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, piglit_winsys_fbo);
>>>>>>> +    glClearColor(0.0, 0.0, 0.0, 1.0);
>>>>>>> +    glClear(GL_COLOR_BUFFER_BIT);
>>>>>>> +    allocate_data_arrays();
>>>>>>> +
>>>>>>> +    /* Reference image drawn here doesn't represent an expected
>>>>>>> image.
>>>>>>> +     * Reference image is drawn only to visualize the image
>>>>>>> difference
>>>>>>> +     * caused by enabling sample_alpha_to_coverage in test image.
>>>>>>> +     */
>>>>>>> +    if(buffer_to_test == GL_COLOR_BUFFER_BIT)
>>>>>>> +        draw_reference_image(true /* sample_alpha_to_coverage */,
>>>>>>> +                     true /* sample_alpha_to_one */);
>>>>>>> +
>>>>>>> +    draw_test_image(true /* sample_alpha_to_coverage */,
>>>>>>> +            true /* sample_alpha_to_one */);
>>>>>>> +
>>>>>>> +    pass = piglit_check_gl_error(GL_NO_ERROR) && pass;
>>>>>>> +
>>>>>>> +    /* Probe all the test image blitted to resolve fbo and compare
>>>>>>> with
>>>>>>> +     * expected color values.  This method of verification is
>>>>>>> appropriate
>>>>>>> +     * for tests with sample-alpha-to-coverage enabled. Possibility
>>>>>>> of
>>>>>>> +     * dithering effect when the coverage value is not a strict
>>>>>>> multiple
>>>>>>> +     * of 1 / num_samples makes image compare (test / reference
>>>>>>> image)
>>>>>>> +     * unsuitable for this test.
>>>>>>> +     */
>>>>>>> +    if(buffer_to_test == GL_COLOR_BUFFER_BIT)
>>>>>>> +        pass = probe_framebuffer_color() && pass;
>>>>>>> +    else if (buffer_to_test == GL_DEPTH_BUFFER_BIT)
>>>>>>> +        pass = probe_framebuffer_depth() && pass;
>>>>>>> +
>>>>>>> +    /* Free the memory allocated for data arrays */
>>>>>>> +    free_data_arrays();
>>>>>>> +
>>>>>>> +    if (!piglit_automatic && buffer_to_test == GL_COLOR_BUFFER_BIT)
>>>>>>> +        piglit_present_results();
>>>>>>> +
>>>>>>> +    return pass ? PIGLIT_PASS : PIGLIT_FAIL;
>>>>>>> +}
>>>>>>>
>>>>>>
>>>>>> I have a patch in my tree for testing the same thing.
>>>>>>
>>>>>> I've cleaned it up and attached it below.  What do you think?  Less
>>>>>> code.
>>>>>>
>>>>>
>>>>> Yes, much better, I'll drop this then. (I was actually wondering why
>>>>> the
>>>>> alpha-to-one and alpha-to-coverage ones were separate to begin with...)
>>>>
>>>>
>>>>
>>>> OK, I'll post my patch for review after I figure out the magic
>>>> incantation needed in all.py to pass the --alpha-to-one flag to the
>>>> test.  I'm nearly at my wits end here.  I may need Dylan to help me.
>>>>
>>>
>>> Dylan,
>>>
>>> I'm trying to modify this block (or add a similar block) at line 3047 of
>>> all.py:
>>>
>>>          for buffer_type in ('color', 'depth'):
>>>              g(['ext_framebuffer_multisample-sample-alpha-to-coverage',
>>>                 sample_count, buffer_type],
>>>                'sample-alpha-to-coverage {} {}'.format(
>>>                    sample_count, buffer_type))
>>>
>>> to pass the new --alpha-to-one argument to the test.  How do I do that?
>>> I've tried about a dozen things without luck so far.
>>>
>>> Honestly, the syntax for specifying tests in all.py seem more complicated
>>> than it should be.
>>
>>
>> That might be my bad - it was my suggestion. I think a comment at the
>> top would go a long way to clearing up any confusion.
>>
>> g(["foo-test-name", "--alpha-to-one"])
>>
>> should have the desired effect. (Assuming there's a "with .... as g"
>> block around it.)
>
>
> Yeah, I tried that earlier.  No go:
>
> Fatal Error: A test has already been assigned the name:
> spec at ext_framebuffer_multisample@sample-alpha-to-coverage 2 color
> Further, the two tests are not the same,
> The original test has this command:
> "/home/brian/piglit-2017/piglit/bin/ext_framebuffer_multisample-sample-alpha-to-coverage
> 2 color -auto -fbo"
> The new test has this command:
> "/home/brian/piglit-2017/piglit/bin/ext_framebuffer_multisample-sample-alpha-to-coverage
> --alpha-to-one 2 color -auto -fbo"
> Command exited with non-zero status 1
> 12.49user 1.53system 0:14.12elapsed 99%CPU (0avgtext+0avgdata
> 201484maxresident)k
> 0inputs+528outputs (2major+147526minor)pagefaults 0swaps
>
> And changing the test name to
> g(['ext_framebuffer_multisample-sample-alpha-to-coverage-alpha-to-one', ...)
> doesn't work either.

Right, so it's complaining about two tests having the same name.
Generically the way that generator (g) function works is it takes 2
arguments - one is a list of arguments starting with the binary name,
to pass to execve() or similar, and the second is the test name (which
by default is set to IIRC the concatenation of all arguments, but I
forget). You need to modify that second argument to include some
mention of the alpha-to-one thing so that its name is different from
the previously defined test.

Cheers,

  -ilia


More information about the Piglit mailing list