[Piglit] [PATCH] gl-1.5-vertex-buffer-offsets: test unusual vertex offsets/strides
Chris Forbes
chrisf at ijw.co.nz
Thu Jul 2 16:43:49 PDT 2015
Has someone reached out to the Flockers devs and let them know what
they are doing is bogus? I would, but don't have a contact.
- Chris
On Tue, Jun 23, 2015 at 11:51 PM, Jose Fonseca <jfonseca at vmware.com> wrote:
> On 22/06/15 15:45, Brian Paul wrote:
>>
>> The draw-vertices tests exercises unusual vertex sizes and strides,
>> but not with interleaved arrays.
>>
>> This test creates a VBO with interleaved vertex positions and colors.
>> The colors are positioned at offsets like 9, 10 and 11 bytes from the
>> start of the vertex. Unusual strides between vertices are tested too.
>>
>> Exercises a bug in Gallium's u_vbuf code where the driver was passed
>> pipe_vertex_element::src_offset values which weren't a multiple of
>> four, even when PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY
>> returned 1.
>>
>> v2: add more comments, specifically describing Flockers. Return "warn"
>> if a test fails, not "fail".
>>
>> Reviewed-by: Marek Olšák <marek.olsak at amd.com>
>> ---
>> tests/all.py | 1 +
>> tests/spec/gl-1.5/CMakeLists.gl.txt | 1 +
>> tests/spec/gl-1.5/vertex-buffer-offsets.c | 149
>> ++++++++++++++++++++++++++++++
>> 3 files changed, 151 insertions(+)
>> create mode 100644 tests/spec/gl-1.5/vertex-buffer-offsets.c
>>
>> diff --git a/tests/all.py b/tests/all.py
>> index 3e91055..6c4fb41 100644
>> --- a/tests/all.py
>> +++ b/tests/all.py
>> @@ -1049,6 +1049,7 @@ with profile.group_manager(
>> 'normal3b3s-invariance-byte', run_concurrent=False)
>> g(['gl-1.5-normal3b3s-invariance', 'GL_SHORT'],
>> 'normal3b3s-invariance-short', run_concurrent=False)
>> + g(['gl-1.5-vertex-buffer-offsets'], 'vertex-buffer-offsets')
>>
>> with profile.group_manager(
>> PiglitGLTest,
>> diff --git a/tests/spec/gl-1.5/CMakeLists.gl.txt
>> b/tests/spec/gl-1.5/CMakeLists.gl.txt
>> index 8dcd95d..f10c6cb 100644
>> --- a/tests/spec/gl-1.5/CMakeLists.gl.txt
>> +++ b/tests/spec/gl-1.5/CMakeLists.gl.txt
>> @@ -10,5 +10,6 @@ link_libraries (
>> )
>>
>> piglit_add_executable (gl-1.5-normal3b3s-invariance
>> normal3b3s-invariance.c)
>> +piglit_add_executable (gl-1.5-vertex-buffer-offsets
>> vertex-buffer-offsets.c)
>>
>> # vim: ft=cmake:
>> diff --git a/tests/spec/gl-1.5/vertex-buffer-offsets.c
>> b/tests/spec/gl-1.5/vertex-buffer-offsets.c
>> new file mode 100644
>> index 0000000..1b9580d
>> --- /dev/null
>> +++ b/tests/spec/gl-1.5/vertex-buffer-offsets.c
>> @@ -0,0 +1,149 @@
>> +/*
>> + * Copyright 2015 VMware, Inc.
>> + *
>> + * 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.
>> + */
>> +
>> +/*
>> + * Test interleaved vertex arrays with unusual element offsets and
>> strides.
>> + *
>> + * The game Flockers (from Steam) uses some unusual vertex arrays.
>> + * For example: glVertexAttribPointerARB(index = 1, size = 3, type =
>> GL_FLOAT,
>> + * normalized = GL_FALSE, stride = 87, pointer = 0x4b). Note that the
>> + * offset to the float[3] attribute is 75 (0x4b) bytes and the stride
>> between
>> + * vertices is 87 bytes.
>> + *
>> + * According to the OpenGL specification, OpenGL 1.5, page 33:
>> + * "Clients must align data elements consistent with the requirements of
>> the
>> + * client platform, with an additional base-level requirement that an
>> offset
>> + * within a buffer to a datum comprising N basic machine units be a
>> multiple
>> + * of N."
>> + *
>> + * However, the spec does not say what might happen if that requirement
>> is
>> + * not met. There is no language about raising a GL error or undefined
>> + * behavior.
>> + *
>> + * This test exercises float[3] attributes at unusual offsets/strides.
>> + * If a failure is detected we generate "warn" instead of "fail" since
>> + * according to the spec, the failure is allowed, but there are apps
>> (such
>> + * as Flockers) that will hit this issue.
>> + *
>> + * If a failure/warning is reported, the OpenGL implementor will have to
>> + * decide if conformance or app support is more important.
>> + *
>> + * Brian Paul
>> + */
>> +
>> +#include "piglit-util-gl.h"
>> +
>> +PIGLIT_GL_TEST_CONFIG_BEGIN
>> + config.supports_gl_compat_version = 15;
>> + config.window_visual = PIGLIT_GL_VISUAL_RGBA |
>> PIGLIT_GL_VISUAL_DOUBLE;
>> +PIGLIT_GL_TEST_CONFIG_END
>> +
>> +
>> +void
>> +piglit_init(int argc, char **argv)
>> +{
>> + /* nothing */
>> +}
>> +
>> +
>> +static bool
>> +test_offset_stride(int color_offset, int stride)
>> +{
>> + static const GLfloat vertex[4][2] = {
>> + { -1, -1 },
>> + { 1, -1 },
>> + { 1, 1 },
>> + { -1, 1 }
>> + };
>> + static const GLfloat color[4] = { 0.0, 1.0, 0.5, 1.0 };
>> + GLubyte buffer[1000];
>> + GLuint buf;
>> + int i, pos;
>> + bool p;
>> +
>> + assert(color_offset >= sizeof(vertex[0]));
>> + assert(stride >= color_offset + sizeof(color));
>> +
>> + pos = 0;
>> + for (i = 0; i < 4; i++) {
>> + /* copy vertex position into buffer */
>> + memcpy(buffer + pos, vertex[i], sizeof(vertex[i]));
>> +
>> + /* copy vertex color into buffer at unusual offset */
>> + memcpy(buffer + pos + color_offset, color, sizeof(color));
>> +
>> + pos += stride;
>> + }
>> + assert(pos <= sizeof(buffer));
>> +
>> + glGenBuffers(1, &buf);
>> + glBindBuffer(GL_ARRAY_BUFFER, buf);
>> + glBufferData(GL_ARRAY_BUFFER, sizeof(buffer),
>> + buffer, GL_STATIC_DRAW);
>> +
>> + glVertexPointer(2, GL_FLOAT, stride, (void *) 0);
>> + glColorPointer(4, GL_FLOAT, stride, (void *) (size_t)
>> color_offset);
>> + glEnable(GL_VERTEX_ARRAY);
>> + glEnable(GL_COLOR_ARRAY);
>> +
>> + glClear(GL_COLOR_BUFFER_BIT);
>> + glDrawArrays(GL_TRIANGLE_FAN, 0, 4);
>> +
>> + p = piglit_probe_rect_rgba(0, 0, piglit_width, piglit_height,
>> color);
>> +
>> + if (!p) {
>> + printf("Failure for color_offset %d, stride %d\n",
>> + color_offset, stride);
>> + }
>> +
>> + piglit_present_results();
>> +
>> + glDeleteBuffers(1, &buf);
>> +
>> + return p;
>> +}
>> +
>> +
>> +enum piglit_result
>> +piglit_display(void)
>> +{
>> + bool pass = true;
>> +
>> + /* test nice values */
>> + pass = test_offset_stride(8, 24) && pass;
>> + pass = test_offset_stride(12, 28) && pass;
>> +
>
>
> The unusual cases should probably be restricted to x86, as it might crash
> elsewhere (no the test, since it uses mempcy, but inside the OpenGL
> implementation). At least, a /* FIXME: skip on platforms that don't support
> unaligned floats */ so that somebody working on those platforms knows what
> needs to be done.
>
>> + /* test unusual offset */
>> + pass = test_offset_stride(9, 32) && pass;
>> +
>> + /* test unusual stride */
>> + pass = test_offset_stride(8, 27) && pass;
>> +
>> + /* test unusual offset, unusual stride */
>> + pass = test_offset_stride(9, 25) && pass;
>> + pass = test_offset_stride(10, 26) && pass;
>> + pass = test_offset_stride(11, 27) && pass;
>> +
>> + /* Report warn, not fail (see comments above) */
>> + return pass ? PIGLIT_PASS : PIGLIT_WARN;
>> +}
>
>
> I also wonder if we should fail instead of just warn on x86, or whatever
> platform (Windows or Linux) that Flockers runs.
>
> Either way
>
> Reviewed-by: Jose Fonseca <jfonseca at vmware.com>
>
> Jose
>
>
> _______________________________________________
> Piglit mailing list
> Piglit at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/piglit
More information about the Piglit
mailing list