---
tests/shaders/shader_runner.c | 13 ++---
1 file changed, 10 insertions(+), 3 deletions(-)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index 44aa5da37..a2a1d43fb 100644
--- a/tests/shaders/shader_runner.c
+++ b/tests/shaders/shader_runner.c
@@ -1143,13
Right now, each time any deqp-like profile ask for the case list, it
calls to generate the caselist. As by default it doesn't pass any
parameter, it creates all the possible caselist. And there are some
test sets (like khr_gles) that calls this method several time. On slow
machines this means a
Without this commit, generating the case list is called without any
parameter by default, so it would generate all the caselist, even if
you don't need it, or if you have just called the method before (like
in khr_gles where you call several time gen_caselist).
For the latter an alternative would
Since commit "tests: use meta profiles"
(400dcd5cf18ac51ab8e9681d8ecab4132986efbd) some test sets are not
represented anymore as python scripts. So first, it would be good to
mention it, and second it would be good to mention that it would be
better to run the test sets using the set name, not
Since commit "tests: use meta profiles"
(400dcd5cf18ac51ab8e9681d8ecab4132986efbd) some test sets are not
represented anymore as python scripts. So first, it would be good to
mention it, and second it would be good to mention that it would be
better to run the test sets using the set name, not
Current workaround assumes that glMapBuffer and glMapBufferRange
'access' bitfield uses the same values, when it is not.
---
.../shaders/shader_runner_gles_workarounds.h | 19 +--
1 file changed, 17 insertions(+), 2 deletions(-)
diff --git
Alejandro Piñeiro (2):
shader_runner: fix error checking on MapBuffer gles workaround
shader_runner: get proper access bitfield when calling MapBufferRange
.../shaders/shader_runner_gles_workarounds.h | 21 ---
1 file changed, 18 insertions(+), 3 deletions(-)
--
2.19.1
---
tests/shaders/shader_runner_gles_workarounds.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tests/shaders/shader_runner_gles_workarounds.h
b/tests/shaders/shader_runner_gles_workarounds.h
index d3ab77506..77956f6b9 100644
---
There are some GL extensions/features, like shader subroutines, that
doesn't have their equivalent on SPIR-V, so those tests should be
skipped on SPIR-V mode. We reuse the code that checks if a given
SPIR-V extension equivalent to a GL extension is supported, by setting
the mapping to a string
This only affects the case where SPIR-V shaders are run, specially
when using tests coming from other extensions.
In several cases, OpenGL supports an extension with a given
functionality. And such functionality it is not supported on core
SPIR-V but on a equivalent SPV functionality.
But it is
glslang doesn't like those spaces, failing to generate SPIR-V.
Additionally this improves readibility of the generated shader
(although this is subjective).
---
.../templates/gen_extensions_defined/disabled.glsl.mako | 2 +-
.../templates/gen_extensions_defined/enabled.glsl.mako| 2
If we are generating SPIR-V tests using already existing tests, we
would need both programs.
Note that we are not removing the option to configure where those
binaries are on the script using envvars, as that can be handy if you
want to use glslang from master compiled by yourself, and having
SPIRV YES/ONLY assumes that the test includes the SPIR-V shader, and
that you should use it. This commit checks that this really happens,
and prints a proper error message if that is not the case.
Without this commit, a SPIRV YES/ONLY test that by mistake didn't
include the [spirv xx] sections
From: Neil Roberts
Adds a script based on gen_builtin_uniform_tests.py which directly
generates the equivalent SPIR-V assembly to test with the ARB_gl_spirv
extension. The main reason to do this rather than converting the GLSL
tests with glslang is that not all of the GLSL builtins have direct
From: Antia Puentes
Adds an option to CMake which can be enabled to generate additional
shader_tests for ARB_gl_spirv by translating shader_tests written for
other specifications to SPIR-V. Specifically, uses the gen_gl_spirv.py
script to translate the shader_tests under the 'tests' directory,
From: Nicolai Hähnle
This will be used by shader_runner to check whether SPIRV object files
are up-to-date.
---
tests/util/piglit-util.c | 34 ++
tests/util/piglit-util.h | 1 +
2 files changed, 35 insertions(+)
diff --git a/tests/util/piglit-util.c
From: Arcady Goldmints-Orlov
The difference is that we explicitly specify locations for inputs and outputs in
component-layout tests.
The current version of glslang doesn't handle a mixture of explicit and implicit
locations for inputs and outputs, so make all locations explicit.
---
From: Neil Roberts
The test signatures now have an additional member called
template_spirv. If this is not None then it is a template which can be
used to generate an equivalent line of SPIR-V assembly.
---
generated_tests/builtin_function.py | 547 ++--
1 file changed,
From ARB_gl_spirv spec:
"Changes to Chapter 3 of the OpenGL Shading Language 4.50 Specification
(Basics)
After describing the compatibility profile rules, add:
"Compatibility-profile features are not available when
generating SPIR-V.""
With this commit, tests that
From: Arcady Goldmints-Orlov
The difference with
arb_compute_shader/execution/basic-ssbo.shader_test is that we
explicitly assign a binding to the SSBO. Without a binding number
specified, glslang implicitly assigns a binding number of 1 to the
SSBO, while Mesa assigns a binding number of 0. To
So when executing shader tests, they will be executed with -spirv.
So for example:
./piglit run tests/shader.py -t ARB_explicit_uniform_location --spirv
results/explicit_uniform_location-spirv
Will run all the shader tests for that extension using -spirv.
---
framework/options.py | 1
---
tests/shaders/shader_runner.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index 927926a35..7c6810099 100644
--- a/tests/shaders/shader_runner.c
+++ b/tests/shaders/shader_runner.c
@@ -850,6 +850,8 @@ static enum
Hänle
Signed-off-by: Alejandro Piñeiro
Signed-off-by: Neil Roberts
Signed-off-by: Antia Puentes
---
generated_tests/gen_gl_spirv_tests.py | 1424 +
1 file changed, 1424 insertions(+)
create mode 100644 generated_tests/gen_gl_spirv_tests.py
diff --git a/generated_tests
wrote the first
version of the file.
Several people have contributed:
Signed-off-by: Nicolai Hänle
Signed-off-by: Antia Puentes
Signed-off-by: Alejandro Piñeiro
Signed-off-by: Neil Roberts
Signed-off-by: Eleni Maria Stea
Signed-off-by: Arcady Goldmints-Orlov
---
tests/gl_spirv.exclude | 305
From: Nicolai Hähnle
---
.gitignore | 1 +
1 file changed, 1 insertion(+)
diff --git a/.gitignore b/.gitignore
index 0f4c952c2..b8008d0a2 100644
--- a/.gitignore
+++ b/.gitignore
@@ -3,6 +3,7 @@
*.tar.bz2
*.zip
*~
+*.spv
.ninja_log
build.ninja
--
2.19.1
tests that
even having SPIR-V included, we want them to run on GLSL mode.
Obviously it is a mistake to use "-spirv" and "-glsl" at the same
time.
v2: update on the code and commit message due the addition of the
"-glsl" command line option,
---
.../{atomic-uint-fs.shader_runner => atomic-uint-fs.shader_test} | 0
1 file changed, 0 insertions(+), 0 deletions(-)
rename tests/spec/arb_gl_spirv/execution/uniform/{atomic-uint-fs.shader_runner
=> atomic-uint-fs.shader_test} (100%)
diff --git
From: Neil Roberts
This is forbidden in Vulkan but it is available in GLSL and the SPIR-V
spec allows it. It isn’t mentioned in the ARB_gl_spirv spec but it
seems likely that it’s intended to work.
---
.../execution/pixel-center-half.shader_test | 123 +
From: Neil Roberts
Using OriginLowerLeft isn’t allowed on Vulkan but explicitly allowed
by ARB_gl_spirv so it’s probably worth testing.
---
.../execution/origin-lower-left.shader_test | 191 +
.../execution/origin-upper-left.shader_test | 192 ++
2 files
From: Antia Puentes
---
...ared-uint-atomic-mem-functions.shader_test | 262 ++
1 file changed, 262 insertions(+)
create mode 100644
tests/spec/arb_gl_spirv/execution/shared-uint-atomic-mem-functions.shader_test
diff --git
From: Antia Puentes
---
.../execution/atomic-image.shader_test| 425 ++
1 file changed, 425 insertions(+)
create mode 100644 tests/spec/arb_gl_spirv/execution/atomic-image.shader_test
diff --git a/tests/spec/arb_gl_spirv/execution/atomic-image.shader_test
From: Neil Roberts
This value is different from InstanceID in Vulkan because it shouldn’t
include the baseinstance.
---
.../execution/vs-instance-id.shader_test | 165 ++
1 file changed, 165 insertions(+)
create mode 100644
From: Neil Roberts
These opcodes are not available via GLSL so it seems worth testing
them separately. This involves generating NaN values which apparently
isn’t required by the spec at least on GLSL. The tests therefore try
to calculate the square root of -1 and if check if it actually
From: Neil Roberts
---
.../tcs-patch-vertices-in.shader_test | 373 ++
1 file changed, 373 insertions(+)
create mode 100644
tests/spec/arb_gl_spirv/execution/tcs-patch-vertices-in.shader_test
diff --git
From: Neil Roberts
This is just like “draw arrays instanced” but has an extra parameter
to set the baseinstance.
---
tests/shaders/shader_runner.c | 10 ++
1 file changed, 10 insertions(+)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index
From: Antia Puentes
Tests glGetProgramResourceIndex(), glGetProgramLocation() and
glGetProgramLocationIndex().
---
tests/shaders/shader_runner.c | 97 +--
1 file changed, 92 insertions(+), 5 deletions(-)
diff --git a/tests/shaders/shader_runner.c
Testing different things, from very basic types, to something more
complex like arrays of blocks where just one member is captured.
All of them are VS, as it is the most common use case.
---
.../execution/xfb/vs_aoa.shader_test | 131 +
.../execution/xfb/vs_block.shader_test
From: Antia Puentes
---
.../program/max-length-xfb-query.shader_test | 79 +++
1 file changed, 79 insertions(+)
create mode 100644
tests/spec/arb_gl_spirv/execution/queries/program/max-length-xfb-query.shader_test
diff --git
---
.../execution/simple-geometry.shader_test | 241 ++
1 file changed, 241 insertions(+)
create mode 100644
tests/spec/arb_gl_spirv/execution/simple-geometry.shader_test
diff --git a/tests/spec/arb_gl_spirv/execution/simple-geometry.shader_test
From: Neil Roberts
There is now an [elements] section which can contain a list of indices
as well as a draw elements command to draw with it.
---
tests/shaders/shader_runner.c | 99 +++
1 file changed, 99 insertions(+)
diff --git a/tests/shaders/shader_runner.c
From: Neil Roberts
This value is different from Vulkan because it should include the
firstvertex and basevertex parameters.
---
.../execution/vs-vertex-id.shader_test| 192 ++
1 file changed, 192 insertions(+)
create mode 100644
This commit adds query objects for GL_PRIMITIVES_GENERATED and
GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN, and only for xfb
drawing. Those query objects are created implicitly during xfb
drawing.
We don't add general query object support for two reasons:
* Right now only xfb tests would be using
From: Eleni Maria Stea
---
.../execution/simple-function.shader_test | 98 +++
.../linker/simple-function-call.shader_test | 112 ++
2 files changed, 210 insertions(+)
create mode 100644
tests/spec/arb_gl_spirv/execution/simple-function.shader_test
create
From: Neil Roberts
This is copied from
arb_enhanced_layouts/execution/block-member-locations
---
...o-block-member-layout-location.shader_test | 257 ++
1 file changed, 257 insertions(+)
create mode 100644
=UNIFORM_BLOCK_NAME_LENGTH, ..) -> params
= 1
GetActiveAttrib(..) -> length = 0, name = ""
GetAttribLocation(.., name) -> -1
GetFragDataLocation(.., name) -> -1
GetFragDataIndex(.., name) -> -1"
v2: tweak the commit message to explain better what
---
.../array_of_input_output_block.shader_test | 171 ++
1 file changed, 171 insertions(+)
create mode 100644
tests/spec/arb_gl_spirv/execution/array_of_input_output_block.shader_test
diff --git
a/tests/spec/arb_gl_spirv/execution/array_of_input_output_block.shader_test
Limited to capture doubles and floats, for simplicity. This patch
introduces the following [test] commands:
* "expected buffer ": it allocates a expected buffer of the
given in bytes. It frees any previous one if already
allocated.
* "expected buffer float ": sets
at the
For now just:
* Simple
* Simple with array.
In general ARB_gl_spirv implementation didn't need to add too many
specifics. Those two were enough to catch the minor things, and enough
to cover the barebone testing. For a more detailed testing it is more
practical to borrow the extensive testing
From: Antia Puentes
---
tests/shaders/shader_runner.c | 372 ++
1 file changed, 283 insertions(+), 89 deletions(-)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index 7d24c50bc..97cb89079 100644
--- a/tests/shaders/shader_runner.c
From: Neil Roberts
This tests handling of the OpTypeRuntimeArray opcode.
---
.../execution/ssbo/unsized-array.shader_test | 119 ++
1 file changed, 119 insertions(+)
create mode 100644
tests/spec/arb_gl_spirv/execution/ssbo/unsized-array.shader_test
diff --git
Arrays of arrays of ubo/ssbo doesn't need to be supported on Vulkan,
but, from ARB_gl_spirv spec:
"Changes of system features, relative to Vulkan:
. Vulkan does not allow multi-dimensional arrays of resources like
UBOs and SSBOs in its SPIR-V environment spec. SPIR-V
From: Antia Puentes
Added tests to check that the ARB_program_interface_query queries
involving names behave as described in the ARB_gl_spirv extension.
The SPIR-V used do not contain names. As names are optional in SPIR-V
this is useful to test the behaviour of the name related queries when
no
The test includes two ssbos, with the same content but different
layouts, in order to ensure that works in both cases.
Note that std140 and std430 layouts are used on the base GLSL shader
used to generate the SPIR-V shaders. SPIR-V shaders doesn't include
info about the layout. From ARB_gl_spirv
From: Neil Roberts
Adds two tests, one which has an explicit offset in GLSL using the
layout(offset) decoration on an interface member, and another which
has a struct embedded in an interface. On SPIR-V it is possible to set
an explicit offset even within this embedded struct (note that this
From: Antia Puentes
---
tests/shaders/shader_runner.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index 624817b59..d786963bd 100644
--- a/tests/shaders/shader_runner.c
+++ b/tests/shaders/shader_runner.c
@@ -393,6 +393,7 @@
From: Antia Puentes
Tests glGetProgramResourceName().
---
tests/shaders/shader_runner.c | 74 ++-
1 file changed, 73 insertions(+), 1 deletion(-)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index 51c31c9d8..893a99422 100644
---
For tests that triggered the lowering nir_split_vars to split arrays,
we found some crashes on ARB_gl_spirv if we didn't take into account
the explicit array stride on the splitted type. This is basically a
regression catching test.
This test is based on
Those include:
* Simple test with just one ubo
* Two different ubo on the same stage
* Same ubo used on two different stages
* Array and array using indirect index of ubos
Finally this also includes a test for a shader that was crashing on a
early version of the SPIR-V support.
Right now the main reason is to check that we are expanding uniforms
for arrays and array of arrays of structs properly. This would be even
more important with UBO/SSBO tests.
The other reason is to check that hidden uniforms are really
hidden. As far as we know, current arb_gl_spirv tests
From: Antia Puentes
Tests glGetProgramInterfaceiv().
---
tests/shaders/shader_runner.c | 30 +-
1 file changed, 29 insertions(+), 1 deletion(-)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index 331221449..51c31c9d8 100644
---
ARB_gl_spirv requires that the SPIR-V includes the explicit matrix and
array stride. GLSL layouts are mapped in SPIR-V with explicit values
for offset, matrix stride and array stride, but as far as it respect
alignments and minimum offsets, the strides doesn't need to be exactly
what OpenGL GLSL
From: Antia Puentes
The SPIR-V used does not contain names. As names are optional in
SPIR-V this is useful to test the behaviour of the name related
queries when no name reflection is available.
---
.../program/max-length-queries.shader_test| 211 +
Specifically the ubo contains some basic types, a struct type, with
array and matrices members.
v2:
* Add some layer of complexity (to include some cases mentioned by
Timothy on the mesa-dev review)
* No expected final substractions/additions value of content, but
check for real
id results for the
query (Antia Puentes).
Signed-off-by: Alejandro Piñeiro
Signed-off-by: Antia Puentes
---
tests/shaders/shader_runner.c | 80 +++
1 file changed, 80 insertions(+)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index
ches 59-64: as mentioned, minor clean-ups and fixes.
[1] https://gitlab.freedesktop.org/mesa/mesa/merge_requests/178
Alejandro Piñeiro (26):
shader_runner: add support for glGetProgram queries
arb_gl_spirv: add GL_ACTIVE_UNIFORMS checks
shader_runner: add force_no_names mode
arb_gl_spirv: add s
r the NULL terminator when
returning the name length and glGetProgramResourceiv does, and
improved the error message (Antia Puentes).
Signed-off-by: Alejandro Piñeiro
---
tests/shaders/shader_runner.c | 392 ++
1 file changed, 300 insertions(+), 92 deletions(-)
diff --g
From: Antia Puentes
---
.../arb_enhanced_layouts/transform-feedback-layout-query-api.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
a/tests/spec/arb_enhanced_layouts/transform-feedback-layout-query-api.c
GLSL codepath uses piglit_build_simple_program_multiple_shaders, that
already calls glLinkProgram.
Minor thing, but makes debug the test more easy.
---
.../transform-feedback-layout-qualifiers.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git
Before this commit a piglit run like this:
./piglit run tests/all.py -t arb_enhanced_layouts-transform-feedback
../piglit_results/foo
Would include the layout-qualifiers tests, but not the layout-query
one.
---
tests/opengl.py | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
We found this issues while working with ARB_gl_spirv, but all those
are independent of it, so we are sending independently.
Alejandro Piñeiro (2):
arb_enhanced_layouts: don't call glLinkProgram twice
arb_enhanced_layouts: use consistent test names
Antia Puentes (1):
arb_enhanced_layouts
Because after all, you are not running GLSL shaders in that
case. SPIR-V requirements are OpenGL version >= 3.30 and having
ARB_gl_spirv available.
About why this is needed: most SPIR-V tests include the GLSL used to
generate the SPIR-V, so the SPIR-V could be regenerated easily if we
test needs
From: Neil Roberts
v2: imported to piglit from a example vkrunner examples branch, also
updated description on the top comment (Alejandro Piñeiro)
v3: use location 1 instead of location 0 for one of the tests (Lionel)
Reviewed-by: Lionel Landwerlin
---
.../block-layout
From: Neil Roberts
v2: (changes made by Alejandro Piñeiro)
* Imported to piglit from a example vkrunner examples branch, also
updated description on the top comment (Alejandro Piñeiro)
* Fixed the fact that the interface type was not being properly
assigned the Block OpDecorate
From: Neil Roberts
v2: imported to piglit from a example vkrunner examples branch, also
updated description on the top comment (Alejandro Piñeiro)
---
This tests are intended to test the patches at the following mesa MR:
https://gitlab.freedesktop.org/mesa/mesa/merge_requests/142
Although
It was using the same color for the base color, drawing color and
expected color.
As we are here, we also remove the debug names, as the test should
work without names.
---
.../execution/vs-ps-simple.shader_test| 23 +--
1 file changed, 6 insertions(+), 17 deletions(-)
On 08/11/18 13:24, Neil Roberts wrote:
> All but one of these are currently broken on anv.
>
> See: https://patchwork.freedesktop.org/series/40797/
FWIW, the first patch of that series is outdated, due all the changes on
master since then. A updated version was included on the ubo/ssbo
support
Equivalent to the GLSL tests added with the series "Test cases for
large GLSL local arrays". Quoting from the cover-letter of that
series:
"Drivers typically need special handling for local array storage
when arrays exceed a certain size, this adds a few test cases for
exercising some
, Alejandro Piñeiro wrote:
> Hi,
>
> this is the second version of the ubo/ssbo tests for
> ARB_gl_spirv. Differences compared with v1:
>
> * Rebased against today master.
>
> * Updated the pass checks on several tests. Instead of a silly
> addition/substract
For tests that triggered nir_split_vars to split arrays, we found some
crashes on ARB_gl_spirv if we didn't take into account the explicit
array stride on the splitted type. This is basically a regression
catching test.
This test is based on spec/glsl-1.40/uniform_buffer/fs-array-copy test
that
Equivalent to the ubo tests added with the commit "arb_gl_spirv: add
some simple ubo tests", but without the linker tests and the
regression-check test.
v2: no expected final substractions/additions value of content, but
check for real values (Timothy)
---
ARB_gl_spirv requires that the SPIR-V includes the explicit matrix
stride. In general layouts are mapping with explicit offset, matrix
stride and array stride. As far as it respect alignments and minimum
offsets, matrix_stride doesn't need to be exactly what OpenGL GLSL
computes, and in fact, the
From: Neil Roberts
This tests handling of the OpTypeRuntimeArray opcode.
---
.../execution/ssbo/unsized-array.shader_test | 119 +
1 file changed, 119 insertions(+)
create mode 100644
tests/spec/arb_gl_spirv/execution/ssbo/unsized-array.shader_test
diff --git
v2: no expected final substractions/additions value of content, but
check for real values (Timothy)
---
.../execution/ssbo/matrix/complex.shader_test | 702 +
1 file changed, 702 insertions(+)
create mode 100644
The test includes two ssbos, with the same content but different
layouts, in order to ensure that works in both cases.
Note that std140 and std430 layouts are used on the base GLSL shader
used to generate the SPIR-V shaders. SPIR-V shaders doesn't include
info about the layout. From ARB_gl_spirv
v2: no expected final substractions/additions value of content, but
check for real values (Timothy)
---
.../execution/ssbo/matrix/column-major.shader_test | 188 +
.../ssbo/matrix/column-vs-row.shader_test | 291 +
Similar to the already existing verify program_interface_query, but
with the glGetProgramiv queries.
Useful, for example, to verify the current number of active (via
GL_ACTIVE_UNIFORMS).
Note that now there are two ways to verify link success. With the
already existing "link succes" shader
Right now the main reason is to check that we are expanding uniforms
for arrays and array of arrays of structs properly. This would be even
more important with UBO/SSBO tests.
The other reason is to check that hidden uniforms are really
hidden. As far as we know, current arb_gl_spirv tests
For more info, see the long explanation at the commit "arb_gl_spirv:
add ubo matrix test with different matrix_stride"
v2: no expected final substractions/additions value of content, but
check for real values (Timothy)
---
.../array-different-array-stride-ubo.shader_test | 271
The series can be found on this branch:
https://github.com/Igalia/piglit/tree/arb_gl_spirv-series5-ubo-ssbo-v2
Alejandro Piñeiro (15):
shader_runner/spirv: log skip reason on ARB_gl_spirv extension check
shader_runner/spirv: check if test really includes spirv
shader_runner: add
SPIRV YES/ONLY assumes that the test includes the SPIR-V shader, and
that you should use it. This commit checks that this really happens,
and prints a proper error message.
Without this commit, a SPIRV YES/ONLY test that by mistake didn't
include the [spirv xx] sections would be just skipped with
Right now UBO data filling is based on using the name of the ubo and
their components, in order to get the block and uniform index. It also
uses it to set the binding (so it forces block_index and block_binding
to be equal).
Since ARB_shading_language_420pack it is possible to set a explicit
ubo
Specifically the ubo contains some basic types, a struct type, with
array and matrices members.
v2:
* Add some layer of complexity (to include some cases mentioned by
Timothy on the mesa-dev review)
* No expected final substractions/additions value of content, but
check for real
v2: no expected final substractions/additions value of content, but
check for real values (Timothy)
---
.../execution/ubo/matrix/column-major.shader_test | 176
.../execution/ubo/matrix/column-vs-row.shader_test | 303 +
From: Neil Roberts
Adds two tests, one which has an explicit offset in GLSL using the
layout(offset) decoration on an interface member, and another which
has a struct embedded in an interface. On SPIR-V it is possible to set
an explicit offset even within this embedded struct.
---
Add two linker tests and several simple execution tests for
ubo. Linker tests just checks for some basic linker errors (like
mistmatched ubo on different stages).
The execution tests include:
* Simple one with just one ubo
* Two different ubo on the same stage
* Same ubo used on
UBO test using several matrices on one block. All possible size
combination, mixing row and column major.
---
.../execution/ubo/matrix/complex.shader_test | 624 +
1 file changed, 624 insertions(+)
create mode 100644
---
tests/shaders/shader_runner.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/tests/shaders/shader_runner.c b/tests/shaders/shader_runner.c
index 6df8bd860..e5566ee7d 100644
--- a/tests/shaders/shader_runner.c
+++ b/tests/shaders/shader_runner.c
@@ -814,6 +814,8 @@ static enum
Right now UBO data filling is based on using the name of the ubo and
their components, in order to get the block and uniform index. It also
uses it to set the binding (so it forces block_index and block_binding
to be equal).
Since ARB_shading_language_420pack it is possible to set a explicit
ubo
On 21/09/18 00:13, Timothy Arceri wrote:
> On 16/9/18 2:22 am, Alejandro Piñeiro wrote:
>> For more info, see the long explanation at the commit "arb_gl_spirv:
>> add ubo matrix test with different matrix_stride"
>> ---
>> .../array-different
On 20/09/18 17:02, Ian Romanick wrote:
> On 09/15/2018 09:22 AM, Alejandro Piñeiro wrote:
>> Right now UBO data filling is based on using the name of the ubo and
>> their components, in order to get the block and uniform index. It also
>> uses it to set the binding (so
1 - 100 of 428 matches
Mail list logo