[Piglit] [PATCH] max-texture-size-level: test per-level max texture size error checking
Brian Paul
brianp at vmware.com
Fri Sep 7 07:35:19 PDT 2012
On 09/06/2012 05:13 PM, Vadim Girlin wrote:
> On Thu, 2012-09-06 at 16:56 -0600, Brian Paul wrote:
>> On 09/06/2012 04:42 PM, Vadim Girlin wrote:
>>> On Thu, 2012-09-06 at 08:01 -0600, Brian Paul wrote:
>>>> On 09/05/2012 06:26 PM, Vadim Girlin wrote:
>>>>> On Tue, 2012-09-04 at 17:05 -0600, Brian Paul wrote:
>>>>>> ---
>>>>>> tests/all.tests | 1 +
>>>>>> tests/texturing/CMakeLists.gl.txt | 1 +
>>>>>> tests/texturing/max-texture-size-level.c | 87 ++++++++++++++++++++++++++++++
>>>>>> 3 files changed, 89 insertions(+), 0 deletions(-)
>>>>>> create mode 100644 tests/texturing/max-texture-size-level.c
>>>>>>
>>>>>> diff --git a/tests/all.tests b/tests/all.tests
>>>>>> index 5b1c1cc..154e1e5 100644
>>>>>> --- a/tests/all.tests
>>>>>> +++ b/tests/all.tests
>>>>>> @@ -734,6 +734,7 @@ add_plain_test(texturing, 'lodclamp-between')
>>>>>> add_plain_test(texturing, 'lodclamp-between-max')
>>>>>> add_plain_test(texturing, 'mipmap-setup')
>>>>>> add_plain_test(texturing, 'max-texture-size')
>>>>>> +add_concurrent_test(texturing, 'max-texture-size-level')
>>>>>> add_concurrent_test(texturing, 'proxy-texture')
>>>>>> add_plain_test(texturing, 'rg-draw-pixels')
>>>>>> add_plain_test(texturing, 'rg-teximage-01')
>>>>>> diff --git a/tests/texturing/CMakeLists.gl.txt b/tests/texturing/CMakeLists.gl.txt
>>>>>> index 4fa0f7f..c439273 100644
>>>>>> --- a/tests/texturing/CMakeLists.gl.txt
>>>>>> +++ b/tests/texturing/CMakeLists.gl.txt
>>>>>> @@ -45,6 +45,7 @@ piglit_add_executable (lodclamp lodclamp.c)
>>>>>> piglit_add_executable (lodclamp-between lodclamp-between.c)
>>>>>> piglit_add_executable (lodclamp-between-max lodclamp-between-max.c)
>>>>>> piglit_add_executable (max-texture-size max-texture-size.c)
>>>>>> +piglit_add_executable (max-texture-size-level max-texture-size-level.c)
>>>>>> piglit_add_executable (mipmap-setup mipmap-setup.c)
>>>>>> piglit_add_executable (proxy-texture proxy-texture.c)
>>>>>> piglit_add_executable (rg-draw-pixels rg-draw-pixels.c)
>>>>>> diff --git a/tests/texturing/max-texture-size-level.c b/tests/texturing/max-texture-size-level.c
>>>>>> new file mode 100644
>>>>>> index 0000000..390e862
>>>>>> --- /dev/null
>>>>>> +++ b/tests/texturing/max-texture-size-level.c
>>>>>> @@ -0,0 +1,87 @@
>>>>>> +/*
>>>>>> + * Copyright 2012 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 glTexImage with an image too large for the given mipmap level.
>>>>>> + *
>>>>>> + * Page 157 of the OpenGL 2.1 spec says:
>>>>>> + *
>>>>>> + * "In a similar fashion, the maximum allowable width of a one- or
>>>>>> + * two- dimensional texture image, and the maximum allowable height of a
>>>>>> + * two- dimensional texture image, must be at least 2k−lod + 2bt for
>>>>>> + * image arrays of level 0 through k, where k is the log base 2 of MAX
>>>>>> + * TEXTURE SIZE. The maximum allowable width and height of a cube map
>>>>>> + * texture must be the same, and must be at least 2k−lod + 2bt for image
>>>>>> + * arrays level 0 through k, where k is the log base 2 of MAX CUBE MAP
>>>>>> + * TEXTURE SIZE."
>>>>>> + */
>>>>>> +
>>>>>> +#include "piglit-util-gl-common.h"
>>>>>> +
>>>>>> +PIGLIT_GL_TEST_MAIN(100, 100, GLUT_RGBA | GLUT_DOUBLE)
>>>>>> +
>>>>>> +enum piglit_result
>>>>>> +piglit_display(void)
>>>>>> +{
>>>>>> + /* no op */
>>>>>> + return PIGLIT_PASS;
>>>>>> +}
>>>>>> +
>>>>>> +
>>>>>> +void
>>>>>> +piglit_init(int argc, char **argv)
>>>>>> +{
>>>>>> + GLint maxSize;
>>>>>> + GLuint tex;
>>>>>> + bool pass;
>>>>>> +
>>>>>> + glGetIntegerv(GL_MAX_TEXTURE_SIZE,&maxSize);
>>>>>> + printf("GL_MAX_TEXTURE_SIZE = %d\n", maxSize);
>>>>>> +
>>>>>> + glGenTextures(1,&tex);
>>>>>> + glBindTexture(GL_TEXTURE_2D, tex);
>>>>>> +
>>>>>> + /* The max texture size should be OK for mipmap level zero. */
>>>>>> + glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA,
>>>>>> + maxSize, maxSize, 0,
>>>>>> + GL_RGBA, GL_UNSIGNED_BYTE, NULL);
>>>>>> + pass = piglit_check_gl_error(GL_NO_ERROR);
>>>>>
>>>>> I think we can also expect OUT_OF_MEMORY here, is it correct?
>>>>
>>>> It's unlikely.
>>>>
>>>> FWIW, GL_OUT_OF_MEMORY can theoretically be generated by almost any GL
>>>> call.
>>>>
>>>> -Brian
>>>
>>> MAX_TEXTURE_SIZE is 16384 for my radeon 5750, so this test is trying to
>>> create the texture consuming 16384 x 16384 x 4bytes + mipmap levels =
>>> 1.3GB. There is 0.5GB of onboard VRAM. I guess the probability of
>>> OUT_OF_MEMORY in my case is close to 100%.
>>>
>>> I don't think that the driver should return lower value for
>>> MAX_TEXTURE_SIZE - it's possible to create 16384x1 texture, or 1x16384,
>>> or even 16384x16384 in some other format. As far as I understand it,
>>> MAX_TEXTURE_SIZE doesn't give any guarantees about possibility of using
>>> MAX_TEXTURE_SIZE x MAX_TEXTURE_SIZE texture in any format, it just tells
>>> that sizes greater than MAX_TEXTURE_SIZE aren't allowed.
>>>
>>> Here is a fragment of FAQ from opengl.org:
>>> http://www.opengl.org/archives/resources/faq/technical/texture.htm#text0120
>>>
>>>> You can obtain an estimate of the maximum texture size your
>>>> implementation supports with the following call:
>>>>
>>>> GLint texSize;
>>>> glGetIntegerv(GL_MAX_TEXTURE_SIZE,&texSize);
>>>>
>>>> ...
>>>>
>>>> This is only an estimate, because the glGet*() function doesn't know
>>>> what format, internalformat, type, and other parameters you'll be
>>>> using for any given texture. OpenGL 1.1 and greater solves this
>>>> problem by allowing texture proxy.
>>>
>>> I'm sure that this test will always fail on my hw due to OOM.
>>
>> OK, I see your point.
>>
>> I think we can fix the test by testing a 2D texture that's maxSize x 1
>> (and 1 x maxSize) or so. I'll post a patch in a bit.
>>
>> Could you try the test as-is? I'm just curious to see if it does
>> indeed fail as you expect.
>>
>
> Yes, it fails:
>
>> GL_MAX_TEXTURE_SIZE = 16384
>> radeon: Failed to allocate a buffer:
>> radeon: size : 1431656448 bytes
>> radeon: alignment : 32768 bytes
>> radeon: domains : 4
>> radeon: Failed to allocate a buffer:
>> radeon: size : 1431656448 bytes
>> radeon: alignment : 32768 bytes
>> radeon: domains : 4
>> Unexpected GL error: GL_OUT_OF_MEMORY 0x505
>> (Error at /home/vg/build/video/64_piglit/tests/texturing/max-texture-size-level.c:68)
>> PIGLIT: {'result': 'fail' }
>
> Thanks for looking into that issue.
>
> PS. Probably some related bugs were fixed recently, but some time ago
> "max-texture-size" due to similar problems usually killed my apps (with
> OOM killer) when I was running piglit, that's why I'd like to avoid such
> problems.
Yeah, that test still has some issues.
But Mesa/gallium is missing some stuff too. A while ago I added a
ctx->Const.MaxTextureMbytes field to limit the max texture size. But
there's no way for a gallium driver to report such a value up to Mesa
(we may need a new PIPE_CAP_).
I'll try to look into that when I get some time.
-Brian
More information about the Piglit
mailing list