| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
|
|
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
|
|
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
|
|
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
|
|
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
|
|
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
|
|
CC: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
|
|
Fixes spurious failure on vc4, which only does unorm depth.
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
|
|
Back in the day it was called MAX_VARYING_FLOATS, which is an alias
for MAX_VARYING_COMPONENTS. Fixes a bunch of varying-components tests
skipping on vc4.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
|
|
|
with SSO
|
|
Cc: Ian Romanick <ian.d.romanick@intel.com>
|
|
Test results:
Nvidia GeForce 840M - NVIDIA 352.41: fail
Using the component qualifier on a vertex input array seems to uncover a
nasty bug on the Nvidia driver. The explict location ends up getting changed
for some reason.
|
|
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
|
|
Piglit has a number of mechanisms for fast skipping, some for GL
features, some for platform support, and some other things. It puts a
nice little message in stdout to let the developer know why it skipped,
but currently that message includes something to the effect of "This is
an unhandled exception", which it isn't it's being handled
appropriately. This patch fixes that.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
|
|
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Serge Martin <edb+piglit@sigluy.net>
|
|
clCreatProgramWithBinary expects a "const unsigned char **", but the code
passes an "unsighed char **". This is a particular kind of messiness
that is hard to fix, so just cast when passing into
clCreateProgramWithBinary.
Thanks to idr and Matt for help getting this right.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Serge Martin <edb+piglit@sigluy.net>
|
|
These test the cases where gl_FragCoord and gl_PointCoord are
allowed to be declared invariant.
Signed-off-by: Lars Hamre <chemecse@gmail.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
|
|
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewed-by: Serge Martin <edb+piglit@sigluy.net>
|
|
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewed-by: Serge Martin <edb+piglit@sigluy.net>
|
|
This allows data to be set for arbitrary array sized attributes in
shader runner.
For example to set mat2x3[2]:
attname[0]/mat2x3/3/1 attname[0]/mat2x3/3/2 attname[1]/mat2x3/3/1 attname[1]/mat2x3/3/2
The syntax has been extended so the recommended type to specify in the
[vertex data] header line is the GLSL corresponding one, not just
"float", "double", "int" or "uint".
In any case, the extended syntax is backward compatible so it is still
possible to specify a vec3 attribute like:
attname/float/3
In addition to the now recommended format:
attname/vec3/3
v2: Added more verbose comments and documentation as requested by
Alejandro Piñeiro.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Signed-off-by: Andres Gomez <agomez@igalia.com>
|
|
We apparently have some generated parser tests for ES 3.x extensions
that use compute shaders. These didn't work because glslparsertest
didn't actually support these features.
For example, spec/oes_sample_variables/preprocessor/enabled-es.comp
failed as soon as I enabled the OES_sample_variables extension.
v2: Use 40, not 43 for tessellation versions (thanks Ilia!)
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
|
|
In piglit when a test generates subtests we treat the "worst" subtest as
the test status, and in most cases this works as expected. There is at
least one case where this is not correct, and that is the case of crash.
There are a couple of reasons that crash should not be masked. One is
that it is generated by the framework when the test binary hits an
assert or segfaults, or any number of similar cases. The second is that
it may mean that all of the subtests did not run, as such we don't want
the status to be masked by the "worst" subtest which would be fail at
worst.
This is maybe not the strictly correct way to handle this. Probably the
better solution is to include the overall result in the max call if the
overall result is not NOTRUN. But this fixes the problem and that can be
done later.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Tested-by: Brian Paul <brianp@vmware.com>
|
|
Currently a test can crash, but if some subtests have a better status
then the status will be the worst of the subtests. This isn't correct
behavior because 'crash' is a framework generated status, not a status
that the runner itself generates.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
reviewed-by: Brian Paul <brianp@vmware.com>
Tested-by: Brian Paul <brianp@vmware.com>
|
|
vs-non-uniform-control-flow-packing.c:91:2: error: use of undeclared identifier 'uint'; did you mean 'rint'?
uint d0[2] = {0x0, 0x40000000};
^~~~
rint
/usr/include/math.h:482:15: note: 'rint' declared here
extern double rint(double);
^
Fixes: c956a3ff3243 ("arb_gpu_shader_fp64: Add tests to check non-uniform-control-flow writes")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
|
|
They are use to detect if a bug in i965 driver related to
operations working with doubles under non-uniform control flow
is still happening.
v2:
- Remove empty lines, fix GLSL versions and minor changes (Andrés)
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Reviewed-by: Andres Gomez <agomez@igalia.com>
|
|
This adds a handful simpler tests for ARB_cull_distance. Instead of the
uncessarily complicated math in the old tests, we just set the distances
based on gl_VertexID. That way we can reliably select which triangles to
cull and/or clip and where we clip them.
We also replace fs-cull-and-clip-distance-exceed-max.shader_test with a
C test so we can properly catch the specified compile or link error when
one of the array sizes or the sum of their sizes exceed the limits.
Reviewed-by: Dave Airlie <airlied@redhat.com>
|
|
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
|
|
I think we probably need to rework these tests and make sure
they are testing what they say.
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
Some of these now fail on r600 before I implement cull distances
|
|
These require GLSL 400 + ARB_shader_precision. They could instead
required GLSL 410. But only requiring GLSL 400 doesn't do what you
want.
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
|
|
Which can be generated by cmake and used by clang_check.
|
|
Also, removed 5 redundant tests replaced by
the generator.
Reviewed-by: Antia Puentes <apuentes@igalia.com>
Acked-by: Dylan Baker <baker.dylan.c@gmail.com>
Signed-off-by: Andres Gomez <agomez@igalia.com>
|
|
qbo.c: In function ‘run_subtest_and_present’:
qbo.c:269:2: warning: format not a string literal and no format arguments [-Wformat-security]
piglit_report_subtest_result(r, subtest_name);
^
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
|
|
The i965 driver has historically done integer multisample resolves with a
sort of integer averaging operation. The spec used to not really say
anything at all about integer formats when discussing multisampling. Since
the tests were written, the spec has been substantially clerified. The new
rule is that integer MSAA resolves are supposed to pick a single arbitrary
sample for each texel rather than trying to combine them in any way.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
|
|
We used to always use GL_TEXTURE_RECTANGLE which we may not actually want.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
|
|
This option allows the format of the output string to be modified by
passing a command line argument. This allows for specialized formats to
be printed for other uses than the original usage that print-commands
was designed for.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
|
|
This is mostly just restructuring code so that piglit-print-command
works like the other piglit-* commands, calling back into the same
module that the main piglit application does. This allows the addition
of the 'piglit print-cmds' for those who prefer that (or who use an
installed piglit and don't have piglit-*.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
|
|
Currently if a test name is passed that doesn't exist in the first
profile a stack trace will be generated. This isn't necessary, since
it's an expected behavior that reordering fails when asked to order
a test not contained in the profile. Instead just fail gracefully with
a helpful message.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
|
|
This test has an extra instance of it's name in the sanity profile when
compared to the all profile. Make them the same between the two
profiles.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
|
|
vs and fs are unused in this function.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
|
|
There are missing tests to check for readonly and writeonly memory qualifiers
on a named interface block. And Mesa is actually failing to apply these
qualifiers right now.
v2: (Ilia Mirkin) Put each test case into its own shader file.
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
|
|
|
|
|
|
clearbuffer-mixed-format regressed, because GL_RED_BITS doesn't have
anything to do with the read buffer.
|
|
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
|
|
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
|
|
In commit b0d05323e code was added to produce clearer error messages for
tests who's status changed from crash to fail and vice versa. There is a
not so subtle bug in that patch, it adds a "crash" element, but that
element should be an "error" element.
This patch fixes that bug.
cc: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Mark Janes <mark.a.janes@intel.com>
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com
|
|
v2: fix sizeof test
v3: OCL1.0 is OK with extension : enable
Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
Reviewed-by: Serge Martin <edb+piglit@sigluy.net>
|
|
This reverts commit 0d3742cde7a46446cdbc75c1444c32feb15baa80.
This breaks because the C and python layers fight for who does what.
|
